diff --git a/Makefile.am b/Makefile.am
index f777f20c02953a750e90ac8e7767e82377d52cff..e46a9e3fbcd6bd0a4d675a31801856f0afde39e5 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -4,8 +4,8 @@ SUBDIRS =
DIST_SUBDIRS =
if BUILD_S_OMP
-SUBDIRS += sPuReMD
-DIST_SUBDIRS += sPuReMD
+SUBDIRS += PuReMD-GPU
+DIST_SUBDIRS += PuReMD-GPU
endif
if BUILD_MPI
SUBDIRS += PuReMD
diff --git a/PuReMD-GPU/Makefile.am b/PuReMD-GPU/Makefile.am
index 3ab7bdba493aa8a5d56c64a100239eeff99d4563..f57bed636093bf0d5a822a4c77d5861524705e26 100644
--- a/PuReMD-GPU/Makefile.am
+++ b/PuReMD-GPU/Makefile.am
@@ -1,7 +1,9 @@
ACLOCAL_AMFLAGS = -I ../m4
+if USE_CUDA
SUFFIXES = .cu
include ../cuda.am
+endif
AM_CFLAGS = -Wall -O3 -funroll-loops -fstrict-aliasing -m64
AM_CPPFLAGS =
@@ -18,21 +20,35 @@ NVCCFLAGS += --compiler-options "$(DEFS) -D__SM_35__ -O3 -funroll-loops -fstrict
#NVCCFLAGS += -Xcompiler -fPIC -dc
#NVCCFLAGS += --ptxas-options -v
-bin_PROGRAMS = bin/puremd-gpu
-bin_puremd_gpu_SOURCES = src/analyze.c src/print_utils.c \
- src/restart.c src/param.c src/pdb_tools.c src/box.c \
- src/lin_alg.c src/QEq.c src/allocate.c src/bond_orders.c \
+bin_PROGRAMS = bin/spuremd
+bin_spuremd_SOURCES = src/analyze.c src/print_utils.c \
+ src/restart.c src/tool_box.c src/control.c src/ffield.c \
+ src/geo_tools.c src/box.c \
+ src/lin_alg.c src/qeq.c src/allocate.c src/bond_orders.c \
src/forces.c src/four_body_interactions.c \
src/grid.c src/init_md.c src/integrate.c src/list.c \
src/lookup.c src/neighbors.c \
src/reset_utils.c src/single_body_interactions.c \
src/system_props.c src/three_body_interactions.c \
src/traj.c src/two_body_interactions.c src/vector.c \
- src/testmd.c \
- src/cuda_utils.cu src/cuda_copy.cu src/cuda_init.cu src/cuda_reduction.cu \
- src/cuda_center_mass.cu src/cuda_box.cu src/validation.cu \
+ src/testmd.c
+include_HEADERS = src/mytypes.h src/analyze.h src/print_utils.h \
+ src/restart.h src/tool_box.c src/control.h src/ffield.c \
+ src/geo_tools.h src/box.h \
+ src/lin_alg.h src/qeq.h src/allocate.h src/bond_orders.h \
+ src/forces.h src/four_body_interactions.h \
+ src/grid.h src/init_md.h src/integrate.h src/list.h \
+ src/lookup.h src/neighbors.h \
+ src/reset_utils.h src/single_body_interactions.h \
+ src/system_props.h src/three_body_interactions.h \
+ src/traj.h src/two_body_interactions.h src/vector.h
+
+if USE_CUDA
+bin_spuremd_SOURCES += src/cuda_utils.cu src/cuda_copy.cu \
+ src/cuda_init.cu src/cuda_reduction.cu \
+ src/cuda_center_mass.cu src/cuda_box.cu src/cuda_validation.cu \
src/cuda_allocate.cu src/cuda_bond_orders.cu \
- src/cuda_lin_alg.cu src/cuda_QEq.cu \
+ src/cuda_lin_alg.cu src/cuda_qeq.cu \
src/cuda_forces.cu src/cuda_four_body_interactions.cu \
src/cuda_grid.cu src/cuda_init_md.cu src/cuda_integrate.cu src/cuda_list.cu \
src/cuda_lookup.cu src/cuda_neighbors.cu \
@@ -40,19 +56,11 @@ bin_puremd_gpu_SOURCES = src/analyze.c src/print_utils.c \
src/cuda_system_props.cu src/cuda_three_body_interactions.cu \
src/cuda_two_body_interactions.cu src/cuda_environment.cu \
src/cuda_post_evolve.cu
-include_HEADERS = src/mytypes.h src/analyze.h src/print_utils.h \
- src/restart.h src/param.h src/pdb_tools.h src/box.h \
- src/lin_alg.h src/QEq.h src/allocate.h src/bond_orders.h \
- src/forces.h src/four_body_interactions.h \
- src/grid.h src/init_md.h src/integrate.h src/list.h \
- src/lookup.h src/neighbors.h \
- src/reset_utils.h src/single_body_interactions.h \
- src/system_props.h src/three_body_interactions.h \
- src/traj.h src/two_body_interactions.h src/vector.h \
- src/cuda_utils.h src/cuda_copy.h src/cuda_init.h src/cuda_reduction.h \
- src/cuda_center_mass.h src/cuda_box.h src/validation.h \
+include_HEADERS += src/cuda_utils.h src/cuda_copy.h \
+ src/cuda_init.h src/cuda_reduction.h \
+ src/cuda_center_mass.h src/cuda_box.h src/cuda_validation.h \
src/cuda_allocate.h src/cuda_bond_orders.h \
- src/cuda_lin_alg.h src/cuda_QEq.h \
+ src/cuda_lin_alg.h src/cuda_qeq.h \
src/cuda_forces.h src/cuda_four_body_interactions.h \
src/cuda_grid.h src/cuda_init_md.h src/cuda_integrate.h src/cuda_list.h \
src/cuda_lookup.h src/cuda_neighbors.h \
@@ -62,8 +70,16 @@ include_HEADERS = src/mytypes.h src/analyze.h src/print_utils.h \
src/cuda_post_evolve.h
# dummy source to cause C linking
-nodist_EXTRA_bin_puremd_gpu_SOURCES = src/dummy.c
+nodist_EXTRA_bin_spuremd_SOURCES = src/dummy.c
+
+endif
+
+
+bin_spuremd_CFLAGS = $(AM_CFLAGS) $(CFLAGS)
+bin_spuremd_CPPFLAGS = $(AM_CPPFLAGS) $(CPPFLAGS)
+bin_spuremd_LDFLAGS = $(AM_LDFLAGS) $(LDFLAGS)
-bin_puremd_gpu_CFLAGS = $(AM_CFLAGS) $(CFLAGS) $(CUDA_CFLAGS)
-bin_puremd_gpu_CPPFLAGS = $(AM_CPPFLAGS) $(CPPFLAGS)
-bin_puremd_gpu_LDFLAGS = $(AM_LDFLAGS) $(LDFLAGS) $(CUDA_LIBS)
+if USE_CUDA
+bin_spuremd_CFLAGS += $(CUDA_CFLAGS)
+bin_spuremd_LDFLAGS += $(CUDA_LIBS)
+endif
diff --git a/PuReMD-GPU/aclocal.m4 b/PuReMD-GPU/aclocal.m4
index 2e1d098d2159d3a3069bc44cc5e0942cb9e86070..d6bf5baa543b24a4c3d8f9fc06ee2020a1d3f9bf 100644
--- a/PuReMD-GPU/aclocal.m4
+++ b/PuReMD-GPU/aclocal.m4
@@ -1150,4 +1150,5 @@ AC_SUBST([am__tar])
AC_SUBST([am__untar])
]) # _AM_PROG_TAR
+m4_include([../m4/ax_compiler_vendor.m4])
m4_include([../m4/ax_cuda.m4])
diff --git a/PuReMD-GPU/configure.ac b/PuReMD-GPU/configure.ac
index 38c7cf737e44056c612a3b48f55708486ab78af4..5d0bca878282d6b7181c47328048ad590cd1afbf 100644
--- a/PuReMD-GPU/configure.ac
+++ b/PuReMD-GPU/configure.ac
@@ -53,47 +53,137 @@ AC_CHECK_TYPES([gzFile], [],
# Checks for library functions.
AC_FUNC_MALLOC
AC_FUNC_STRTOD
-AC_CHECK_FUNCS([memset pow sqrt])
+AC_CHECK_FUNCS([gettimeofday memset pow sqrt])
+
+# Check for compiler vendor
+AX_COMPILER_VENDOR
+if test "x$ax_cv_c_compiler_vendor" = "xgnu"; then
+ if test "x$BUILD_DEBUG" = "x"; then
+ CFLAGS="$CFLAGS -Wall -O3 -funroll-loops -fstrict-aliasing"
+ else
+ CFLAGS="$CFLAGS -Wall"
+ fi
+fi
+if test "x$ax_cv_c_compiler_vendor" = "xintel"; then
+ if test "x$BUILD_DEBUG" = "x"; then
+ CFLAGS="$CFLAGS -fast"
+ fi
+fi
-# Check for CUDA support.
-CONFIGURE_HEADLINE([ CUDA support ])
-AX_CUDA
-NVCCFLAGS=
-if test "BUILD_DEBUG" = "true"
+# Check for OpenMP support.
+if test "x$BUILD_OPENMP" = "xyes"; then
+ AC_OPENMP
+ if test "x${OPENMP_CFLAGS}" = "x"; then
+ AC_MSG_WARN([
+ -----------------------------------------------
+ Unable to find OpenMP support on this system.
+ Building a single-threaded version.
+ -----------------------------------------------])
+ else
+ # bug due to recent Intel compiler change (?)
+ if test "x$ax_cv_c_compiler_vendor" = "xintel"; then
+ OPENMP_CFLAGS="-qopenmp"
+ fi
+ AC_SUBST(AM_CFLAGS, "$OPENMP_CFLAGS")
+ AC_SUBST(AM_CPPFLAGS, "$OPENMP_CFLAGS")
+ fi
+fi
+
+if test "x$BUILD_SUPERLU_MT" != "x"
then
- NVCCFLAGS+=" -g -G"
+ CPPFLAGS="${CPPFLAGS} -I${BUILD_SUPERLU_MT}/include"
+ LDFLAGS="${LDFLAGS} -L${BUILD_SUPERLU_MT}/lib"
+ #TODO: implement better BLAS detection
+ LIBS="${LIBS} -lblas"
+# BLAS_FOUND_LIBS="yes"
+# AC_SEARCH_LIBS([dtrsv_], [blas blas_OPENMP],
+# [], [BLAS_FOUND_LIBS="no"], [])
+# AS_IF([test "x${BLAS_FOUND_LIBS}" != "xyes"],
+# [AC_MSG_ERROR([Unable to find BLAS library.])])
+ AC_CHECK_HEADERS([slu_mt_ddefs.h], [SUPERLU_MT_FOUND_HEADERS="yes"])
+ AS_IF([test "x${SUPERLU_MT_FOUND_HEADERS}" != "xyes"],
+ [AC_MSG_ERROR([Unable to find SuperLU MT headers.])])
+ SUPERLU_MT_FOUND_LIBS="yes"
+ #TODO: fix issue where multiple -l flags added, one for each call below
+ AC_SEARCH_LIBS([intMalloc], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([get_perm_c], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([pdgstrf_init], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp -lblas -lblas_OPENMP])
+ AC_SEARCH_LIBS([pdgstrf], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp -lblas -lblas_OPENMP])
+ AC_SEARCH_LIBS([pxgstrf_finalize], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp -lblas -lblas_OPENMP])
+ AC_SEARCH_LIBS([StatAlloc], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([StatInit], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([StatFree], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([Destroy_SuperNode_SCP], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AC_SEARCH_LIBS([Destroy_CompCol_NCP], [superlu_mt superlu_mt_OPENMP],
+ [], [SUPERLU_MT_FOUND_LIBS="no"], [-lgomp])
+ AS_IF([test "x${SUPERLU_MT_FOUND_LIBS}" != "xyes"],
+ [AC_MSG_ERROR([Unable to find SuperLU MT library.])])
+ AC_DEFINE([HAVE_SUPERLU_MT], [1], [Define to 1 if you have SuperLU_MT support enabled.])
fi
-AC_DEFINE([HAVE_CUDA], [1], [Define to 1 if you have CUDA support enabled.])
-
-AC_CHECK_LIB([cuda], [cuGetErrorString])
-AC_CHECK_LIB([cudart], [cudaMalloc])
-AC_CHECK_LIB([cublas], [cublasDnrm2])
-AC_CHECK_LIB([cusparse], [cusparseCreateMatDescr])
-#AC_SEARCH_LIBS([cublasDaxpy], [cublas])
-#AC_SEARCH_LIBS([cublasDscal], [cublas])
-#AC_SEARCH_LIBS([cublasDdot], [cublas])
-#AC_SEARCH_LIBS([cudaThreadSynchronize], [cudart])
-#AC_SEARCH_LIBS([cudaGetLastError], [cudart])
-#AC_SEARCH_LIBS([cusparseCreateMatDescr], [cusparse])
-#AC_SEARCH_LIBS([cusparseSetMatType], [cusparse])
-#AC_SEARCH_LIBS([cusparseSetMatIndexBase], [cusparse])
+
+# Check for CUDA support.
+if test "x$BUILD_GPU" != "x"
+then
+ CONFIGURE_HEADLINE([ CUDA support ])
+ AX_CUDA
+ NVCCFLAGS=
+ if test "BUILD_DEBUG" = "true"
+ then
+ NVCCFLAGS+=" -g -G"
+ fi
+ AC_DEFINE([HAVE_CUDA], [1], [Define to 1 if you have CUDA support enabled.])
+
+ AC_CHECK_LIB([cuda], [cuGetErrorString])
+ AC_CHECK_LIB([cudart], [cudaMalloc])
+ AC_CHECK_LIB([cublas], [cublasDnrm2])
+ AC_CHECK_LIB([cusparse], [cusparseCreateMatDescr])
+# AC_SEARCH_LIBS([cublasDaxpy], [cublas])
+# AC_SEARCH_LIBS([cublasDscal], [cublas])
+# AC_SEARCH_LIBS([cublasDdot], [cublas])
+# AC_SEARCH_LIBS([cudaThreadSynchronize], [cudart])
+# AC_SEARCH_LIBS([cudaGetLastError], [cudart])
+# AC_SEARCH_LIBS([cusparseCreateMatDescr], [cusparse])
+# AC_SEARCH_LIBS([cusparseSetMatType], [cusparse])
+# AC_SEARCH_LIBS([cusparseSetMatIndexBase], [cusparse])
#
-#AC_SEARCH_LIBS([cublasDnrm2], [cublas],
-# [CUBLAS_FOUND_LIBS="yes"], [CUBLAS_FOUND_LIBS="no"], [-lcublas])
-#AS_IF([test "x${CUBLAS_FOUND_LIBS}" != "xyes"],
-# [AC_MSG_ERROR([Unable to find CUBLAS library.])])
+# AC_SEARCH_LIBS([cublasDnrm2], [cublas],
+# [CUBLAS_FOUND_LIBS="yes"], [CUBLAS_FOUND_LIBS="no"], [-lcublas])
+# AS_IF([test "x${CUBLAS_FOUND_LIBS}" != "xyes"],
+# [AC_MSG_ERROR([Unable to find CUBLAS library.])])
#
-#AC_SEARCH_LIBS([cusparseSetMatType], [cusparse],
-# [CUSPARSE_FOUND_LIBS="yes"], [CUSPARSE_FOUND_LIBS="no"], [-lcusparse])
-#AS_IF([test "x${CUSPARSE_FOUND_LIBS}" != "xyes"],
-# [AC_MSG_ERROR([Unable to find CUSPARSE library.])])
+# AC_SEARCH_LIBS([cusparseSetMatType], [cusparse],
+# [CUSPARSE_FOUND_LIBS="yes"], [CUSPARSE_FOUND_LIBS="no"], [-lcusparse])
+# AS_IF([test "x${CUSPARSE_FOUND_LIBS}" != "xyes"],
+# [AC_MSG_ERROR([Unable to find CUSPARSE library.])])
#
-#AC_CHECK_TYPES([cublasHandle_t], [],
-# [AC_MSG_FAILURE([cublasHandle_t type not found in cublas.h], [1])], [#include<cublas_v2.h>])
-#AC_CHECK_TYPES([cusparseHandle_t], [],
-# [AC_MSG_FAILURE([cusparseHandle_t type not found in cusparse.h], [1])], [#include<cusparse_v2.h>])
-#AC_CHECK_TYPES([cusparseMatDescr_t], [],
-# [AC_MSG_FAILURE([cusparseMatDescr_t type not found in cusparse.h], [1])], [#include<cusparse_v2.h>])
+# AC_CHECK_TYPES([cublasHandle_t], [],
+# [AC_MSG_FAILURE([cublasHandle_t type not found in cublas.h], [1])], [#include<cublas_v2.h>])
+# AC_CHECK_TYPES([cusparseHandle_t], [],
+# [AC_MSG_FAILURE([cusparseHandle_t type not found in cusparse.h], [1])], [#include<cusparse_v2.h>])
+# AC_CHECK_TYPES([cusparseMatDescr_t], [],
+# [AC_MSG_FAILURE([cusparseMatDescr_t type not found in cusparse.h], [1])], [#include<cusparse_v2.h>])
+else
+ AM_CONDITIONAL(USE_CUDA, test "x" = "xyes")
+fi
+
+if test "x$BUILD_DEBUG" != "x"
+then
+ CFLAGS="${CFLAGS} ${DEBUG_FLAGS}"
+fi
+
+if test "x$BUILD_GPROF" != "x"
+then
+ CFLAGS="${CFLAGS} ${GPROF_FLAGS}"
+fi
if test "BUILD_PROF" = "true"
then
diff --git a/PuReMD-GPU/src/QEq.c b/PuReMD-GPU/src/QEq.c
deleted file mode 100644
index 8cc638ea90dcc25f86d33f275b162c8e531d82bb..0000000000000000000000000000000000000000
--- a/PuReMD-GPU/src/QEq.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
-
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
- Hasan Metin Aktulga, haktulga@cs.purdue.edu
- Ananth Y Grama, ayg@cs.purdue.edu
-
- This program is free software; you can redistribute it and/or
- modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
- the License, or (at your option) any later version.
-
- This program is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
- See the GNU General Public License for more details:
- <http://www.gnu.org/licenses/>.
- ----------------------------------------------------------------------*/
-
-#include "QEq.h"
-
-#include "allocate.h"
-#include "lin_alg.h"
-#include "list.h"
-#include "print_utils.h"
-#include "index_utils.h"
-#include "system_props.h"
-
-#include "sort.h"
-
-
-int compare_matrix_entry(const void *v1, const void *v2)
-{
- return ((sparse_matrix_entry *)v1)->j - ((sparse_matrix_entry *)v2)->j;
-}
-
-
-void Sort_Matrix_Rows( sparse_matrix *A )
-{
- int i, si, ei;
-
- for( i = 0; i < A->n; ++i ) {
- si = A->start[i];
- ei = A->start[i+1];
- qsort( &(A->entries[si]), ei - si,
- sizeof(sparse_matrix_entry), compare_matrix_entry );
- }
-}
-
-
-void Calculate_Droptol( sparse_matrix *A, real *droptol, real dtol )
-{
- int i, j, k;
- real val;
-
- /* init droptol to 0 */
- for( i = 0; i < A->n; ++i )
- droptol[i] = 0;
-
- /* calculate sqaure of the norm of each row */
- for( i = 0; i < A->n; ++i ) {
- for( k = A->start[i]; k < A->start[i+1]-1; ++k ) {
- j = A->entries[k].j;
- val = A->entries[k].val;
-
- droptol[i] += val*val;
- droptol[j] += val*val;
- }
-
- val = A->entries[k].val; // diagonal entry
- droptol[i] += val*val;
- }
-
- /* calculate local droptol for each row */
- //fprintf( stderr, "droptol: " );
- for( i = 0; i < A->n; ++i ) {
- //fprintf( stderr, "%f-->", droptol[i] );
- droptol[i] = SQRT( droptol[i] ) * dtol;
- //fprintf( stderr, "%f ", droptol[i] );
- }
- //fprintf( stderr, "\n" );
-}
-
-
-int Estimate_LU_Fill( sparse_matrix *A, real *droptol )
-{
- int i, j, pj;
- int fillin;
- real val;
-
- fillin = 0;
-
- //fprintf( stderr, "n: %d\n", A->n );
- for( i = 0; i < A->n; ++i )
- for( pj = A->start[i]; pj < A->start[i+1]-1; ++pj ){
- j = A->entries[pj].j;
- val = A->entries[pj].val;
- //fprintf( stderr, "i: %d, j: %d", i, j );
-
- if( fabs(val) > droptol[i] )
- ++fillin;
- }
-
- return fillin + A->n;
-}
-
-
-void ICHOLT( sparse_matrix *A, real *droptol,
- sparse_matrix *L, sparse_matrix *U )
-{
- sparse_matrix_entry tmp[1000];
- int i, j, pj, k1, k2, tmptop, Ltop;
- real val;
- int *Utop;
-
- Utop = (int*) malloc((A->n+1) * sizeof(int));
-
- // clear variables
- Ltop = 0;
- tmptop = 0;
- for( i = 0; i <= A->n; ++i )
- L->start[i] = U->start[i] = 0;
-
- for( i = 0; i < A->n; ++i )
- Utop[i] = 0;
-
- //fprintf( stderr, "n: %d\n", A->n );
- for( i = 0; i < A->n; ++i ){
- L->start[i] = Ltop;
- tmptop = 0;
-
- for( pj = A->start[i]; pj < A->start[i+1]-1; ++pj ){
- j = A->entries[pj].j;
- val = A->entries[pj].val;
- //fprintf( stderr, "i: %d, j: %d", i, j );
-
- if( fabs(val) > droptol[i] ){
- k1 = 0;
- k2 = L->start[j];
- while( k1 < tmptop && k2 < L->start[j+1] ){
- if( tmp[k1].j < L->entries[k2].j )
- ++k1;
- else if( tmp[k1].j > L->entries[k2].j )
- ++k2;
- else
- val -= (tmp[k1++].val * L->entries[k2++].val);
- }
-
- // L matrix is lower triangular,
- // so right before the start of next row comes jth diagonal
- val /= L->entries[L->start[j+1]-1].val;
-
- tmp[tmptop].j = j;
- tmp[tmptop].val = val;
- ++tmptop;
- }
- //fprintf( stderr, " -- done\n" );
- }
-
- // compute the ith diagonal in L
- // sanity check
- if( A->entries[pj].j != i ) {
- fprintf( stderr, "i=%d, badly built A matrix!\n", i );
- exit(999);
- }
-
- val = A->entries[pj].val;
- for( k1 = 0; k1 < tmptop; ++k1 )
- val -= (tmp[k1].val * tmp[k1].val);
-
- tmp[tmptop].j = i;
- tmp[tmptop].val = SQRT(val);
-
- // apply the dropping rule once again
- //fprintf( stderr, "row%d: tmptop: %d\n", i, tmptop );
- //for( k1 = 0; k1<= tmptop; ++k1 )
- // fprintf( stderr, "%d(%f) ", tmp[k1].j, tmp[k1].val );
- //fprintf( stderr, "\n" );
- //fprintf( stderr, "row(%d): droptol=%.4f\n", i+1, droptol[i] );
- for( k1 = 0; k1 < tmptop; ++k1 )
- if( fabs(tmp[k1].val) > droptol[i] / tmp[tmptop].val ){
- L->entries[Ltop].j = tmp[k1].j;
- L->entries[Ltop].val = tmp[k1].val;
- U->start[tmp[k1].j+1]++;
- ++Ltop;
- //fprintf( stderr, "%d(%.4f) ", tmp[k1].j+1, tmp[k1].val );
- }
- // keep the diagonal in any case
- L->entries[Ltop].j = tmp[k1].j;
- L->entries[Ltop].val = tmp[k1].val;
- ++Ltop;
- //fprintf( stderr, "%d(%.4f)\n", tmp[k1].j+1, tmp[k1].val );
- }
-
- L->start[i] = Ltop;
- //fprintf( stderr, "nnz(L): %d, max: %d\n", Ltop, L->n * 50 );
-
- for( i = 1; i <= U->n; ++i )
- Utop[i] = U->start[i] = U->start[i] + U->start[i-1] + 1;
-
- for( i = 0; i < L->n; ++i )
- for( pj = L->start[i]; pj < L->start[i+1]; ++pj ){
- j = L->entries[pj].j;
- U->entries[Utop[j]].j = i;
- U->entries[Utop[j]].val = L->entries[pj].val;
- Utop[j]++;
- }
-
- //fprintf( stderr, "nnz(U): %d, max: %d\n", Utop[U->n], U->n * 50 );
-}
-
-
-void Init_MatVec( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list *far_nbrs )
-{
- int i, fillin;
- real s_tmp, t_tmp;
- //char fname[100];
-
- if(control->refactor > 0 &&
- ((data->step-data->prev_steps)%control->refactor==0 || workspace->L.entries==NULL))
- {
- //Print_Linear_System( system, control, workspace, data->step );
- Sort_Matrix_Rows( &workspace->H );
-
- //fprintf( stderr, "H matrix sorted\n" );
-
- Calculate_Droptol( &workspace->H, workspace->droptol, control->droptol );
- //fprintf( stderr, "drop tolerances calculated\n" );
-
- if( workspace->L.entries == NULL )
- {
- fillin = Estimate_LU_Fill( &workspace->H, workspace->droptol );
-
-#ifdef __DEBUG_CUDA__
- fprintf( stderr, "fillin = %d\n", fillin );
-#endif
-
- if( Allocate_Matrix( &(workspace->L), far_nbrs->n, fillin ) == 0 ||
- Allocate_Matrix( &(workspace->U), far_nbrs->n, fillin ) == 0 )
- {
- fprintf( stderr, "not enough memory for LU matrices. terminating.\n" );
- exit(INSUFFICIENT_SPACE);
- }
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "fillin = %d\n", fillin );
- fprintf( stderr, "allocated memory: L = U = %ldMB\n",
- fillin * sizeof(sparse_matrix_entry) / (1024*1024) );
-#endif
- }
-
- ICHOLT( &workspace->H, workspace->droptol, &workspace->L, &workspace->U );
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "icholt-" );
- //sprintf( fname, "%s.L%d.out", control->sim_name, data->step );
- //Print_Sparse_Matrix2( workspace->L, fname );
- //Print_Sparse_Matrix( U );
-#endif
- }
-
- /* extrapolation for s & t */
- for( i = 0; i < system->N; ++i ) {
- // no extrapolation
- //s_tmp = workspace->s[0][i];
- //t_tmp = workspace->t[0][i];
-
- // linear
- //s_tmp = 2 * workspace->s[0][i] - workspace->s[1][i];
- //t_tmp = 2 * workspace->t[0][i] - workspace->t[1][i];
-
- // quadratic
- //s_tmp = workspace->s[2][i] + 3 * (workspace->s[0][i]-workspace->s[1][i]);
- t_tmp = workspace->t[index_wkspace_sys(2,i,system->N)] + 3*(workspace->t[index_wkspace_sys(0,i,system->N)]-workspace->t[index_wkspace_sys(1,i,system->N)]);
-
- // cubic
- s_tmp = 4 * (workspace->s[index_wkspace_sys(0,i,system->N)] + workspace->s[index_wkspace_sys(2,i,system->N)]) -
- (6 * workspace->s[index_wkspace_sys(1,i,system->N)] + workspace->s[index_wkspace_sys(3,i,system->N)] );
- //t_tmp = 4 * (workspace->t[0][i] + workspace->t[2][i]) -
- // (6 * workspace->t[1][i] + workspace->t[3][i] );
-
- // 4th order
- //s_tmp = 5 * (workspace->s[0][i] - workspace->s[3][i]) +
- // 10 * (-workspace->s[1][i] + workspace->s[2][i] ) + workspace->s[4][i];
- //t_tmp = 5 * (workspace->t[0][i] - workspace->t[3][i]) +
- // 10 * (-workspace->t[1][i] + workspace->t[2][i] ) + workspace->t[4][i];
-
- workspace->s[index_wkspace_sys(4,i,system->N)] = workspace->s[index_wkspace_sys(3,i,system->N)];
- workspace->s[index_wkspace_sys(3,i,system->N)] = workspace->s[index_wkspace_sys(2,i,system->N)];
- workspace->s[index_wkspace_sys(2,i,system->N)] = workspace->s[index_wkspace_sys(1,i,system->N)];
- workspace->s[index_wkspace_sys(1,i,system->N)] = workspace->s[index_wkspace_sys(0,i,system->N)];
- workspace->s[index_wkspace_sys(0,i,system->N)] = s_tmp;
-
- workspace->t[index_wkspace_sys(4,i,system->N)] = workspace->t[index_wkspace_sys(3,i,system->N)];
- workspace->t[index_wkspace_sys(3,i,system->N)] = workspace->t[index_wkspace_sys(2,i,system->N)];
- workspace->t[index_wkspace_sys(2,i,system->N)] = workspace->t[index_wkspace_sys(1,i,system->N)];
- workspace->t[index_wkspace_sys(1,i,system->N)] = workspace->t[index_wkspace_sys(0,i,system->N)];
- workspace->t[index_wkspace_sys(0,i,system->N)] = t_tmp;
- }
-}
-
-
-void Calculate_Charges( reax_system *system, static_storage *workspace )
-{
- int i;
- real u, s_sum, t_sum;
-
- s_sum = t_sum = 0.;
- for( i = 0; i < system->N; ++i ) {
- s_sum += workspace->s[index_wkspace_sys(0,i,system->N)];
- t_sum += workspace->t[index_wkspace_sys(0,i,system->N)];
- }
-
- u = s_sum / t_sum;
-
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, "Host --->s %13.2f, t %13.f, u %13.2f \n", s_sum, t_sum, u );
-#endif
-
- for( i = 0; i < system->N; ++i )
- {
- system->atoms[i].q = workspace->s[index_wkspace_sys(0,i,system->N)] - u * workspace->t[index_wkspace_sys(0,i,system->N)];
- }
-}
-
-
-void QEq( reax_system *system, control_params *control, simulation_data *data,
- static_storage *workspace, list *far_nbrs,
- output_controls *out_control )
-{
- int matvecs;
-
- //real t_start, t_elapsed;
-
- //t_start = Get_Time ();
- Init_MatVec( system, control, data, workspace, far_nbrs );
- //t_elapsed = Get_Timing_Info ( t_start );
-
- //fprintf (stderr, " CPU Init_MatVec timing ----> %f \n", t_elapsed );
-
- //if( data->step % 10 == 0 )
- // Print_Linear_System( system, control, workspace, far_nbrs, data->step );
-
- //t_start = Get_Time ( );
- matvecs = GMRES( workspace, &workspace->H,
- workspace->b_s, control->q_err, &workspace->s[0], out_control->log, system );
- matvecs += GMRES( workspace, &workspace->H,
- workspace->b_t, control->q_err, &workspace->t[0], out_control->log, system );
- //t_elapsed = Get_Timing_Info ( t_start );
-
- //fprintf (stderr, " CPU GMRES timing ---> %f \n", t_elapsed );
-
- //matvecs = GMRES_HouseHolder( workspace, workspace->H,
- // workspace->b_s, control->q_err, workspace->s[0], out_control->log );
- //matvecs += GMRES_HouseHolder( workspace, workspace->H,
- // workspace->b_t, control->q_err, workspace->t[0], out_control->log );
-
- //matvecs = PGMRES( workspace, &workspace->H, workspace->b_s, control->q_err,
- // &workspace->L, &workspace->U, &workspace->s[index_wkspace_sys(0,0,system->N)], out_control->log, system );
- //matvecs += PGMRES( workspace, &workspace->H, workspace->b_t, control->q_err,
- // &workspace->L, &workspace->U, &workspace->t[index_wkspace_sys(0,0,system->N)], out_control->log, system );
-
- //matvecs=PCG( workspace, workspace->H, workspace->b_s, control->q_err,
- // workspace->L, workspace->U, workspace->s[0], out_control->log ) + 1;
- ///matvecs+=PCG( workspace, workspace->H, workspace->b_t, control->q_err,
- // workspace->L, workspace->U, workspace->t[0], out_control->log ) + 1;
-
- //matvecs = CG( workspace, workspace->H,
- // workspace->b_s, control->q_err, workspace->s[0], out_control->log ) + 1;
- //matvecs += CG( workspace, workspace->H,
- // workspace->b_t, control->q_err, workspace->t[0], out_control->log ) + 1;
-
- //matvecs = SDM( workspace, workspace->H,
- // workspace->b_s, control->q_err, workspace->s[0], out_control->log ) + 1;
- //matvecs += SDM( workspace, workspace->H,
- // workspace->b_t, control->q_err, workspace->t[0], out_control->log ) + 1;
-
- //fprintf (stderr, " GMRES done with iterations %d \n", matvecs );
-
- data->timing.matvecs += matvecs;
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "linsolve-" );
-#endif
-
- Calculate_Charges( system, workspace );
- //fprintf( stderr, "%d %.9f %.9f %.9f %.9f %.9f %.9f\n",
- // data->step,
- // workspace->s[0][0], workspace->t[0][0],
- // workspace->s[0][1], workspace->t[0][1],
- // workspace->s[0][2], workspace->t[0][2] );
- // if( data->step == control->nsteps )
- //Print_Charges( system, control, workspace, data->step );
-}
diff --git a/PuReMD-GPU/src/QEq.h b/PuReMD-GPU/src/QEq.h
deleted file mode 100644
index 31dfbf61ba05ec79d32313c3ab648eb259f183f2..0000000000000000000000000000000000000000
--- a/PuReMD-GPU/src/QEq.h
+++ /dev/null
@@ -1,61 +0,0 @@
-/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
-
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
- Hasan Metin Aktulga, haktulga@cs.purdue.edu
- Ananth Y Grama, ayg@cs.purdue.edu
-
- This program is free software; you can redistribute it and/or
- modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
- the License, or (at your option) any later version.
-
- This program is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
- See the GNU General Public License for more details:
- <http://www.gnu.org/licenses/>.
- ----------------------------------------------------------------------*/
-
-#ifndef __QEq_H_
-#define __QEq_H_
-
-#include "mytypes.h"
-
-
-void QEq( reax_system*, control_params*, simulation_data*, static_storage*,
- list*, output_controls* );
-
-
-static inline HOST_DEVICE void swap(sparse_matrix_entry *array, int index1, int index2)
-{
- sparse_matrix_entry temp = array[index1];
- array[index1] = array[index2];
- array[index2] = temp;
-}
-
-
-static inline HOST_DEVICE void quick_sort(sparse_matrix_entry *array, int start, int end)
-{
- int i = start;
- int k = end;
-
- if (end - start >= 1)
- {
- int pivot = array[start].j;
-
- while (k > i)
- {
- while ((array[i].j <= pivot) && (i <= end) && (k > i)) i++;
- while ((array[k].j > pivot) && (k >= start) && (k >= i)) k--;
- if (k > i) swap(array, i, k);
- }
- swap(array, start, k);
- quick_sort(array, start, k - 1);
- quick_sort(array, k + 1, end);
- }
-}
-
-
-#endif
diff --git a/PuReMD-GPU/src/allocate.c b/PuReMD-GPU/src/allocate.c
index 65f0eb2a872673259d508f17fc0da43530a7426f..bbae7dce50b8fdf60e5a0d8c459e50ca36dd67cb 100644
--- a/PuReMD-GPU/src/allocate.c
+++ b/PuReMD-GPU/src/allocate.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -21,103 +22,148 @@
#include "allocate.h"
#include "list.h"
+#include "tool_box.h"
+
+/* allocate space for atoms */
+int PreAllocate_Space( reax_system *system, control_params *control,
+ static_storage *workspace )
+{
+ int i;
+
+ system->atoms = (reax_atom*) scalloc( system->N,
+ sizeof(reax_atom), "atoms" );
+ workspace->orig_id = (int*) scalloc( system->N,
+ sizeof(int), "orid_id" );
+
+ /* space for keeping restriction info, if any */
+ if ( control->restrict_bonds )
+ {
+ workspace->restricted = (int*) scalloc( system->N,
+ sizeof(int), "restricted_atoms" );
+
+ workspace->restricted_list = (int*) scalloc( system->N,
+ sizeof(int), "restricted_list" );
+
+ workspace->restricted_list = (int*) scalloc( MAX_RESTRICT * system->N,
+ sizeof(int), "restricted_list[i]" );
+ }
+
+ return SUCCESS;
+}
void Reallocate_Neighbor_List( list *far_nbrs, int n, int num_intrs )
{
Delete_List( far_nbrs );
- if(!Make_List( n, num_intrs, TYP_FAR_NEIGHBOR, far_nbrs ))
+
+ if (!Make_List( n, num_intrs, TYP_FAR_NEIGHBOR, far_nbrs ))
{
fprintf(stderr, "Problem in initializing far nbrs list. Terminating!\n");
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
#if defined(DEBUG_FOCUS)
fprintf( stderr, "num_far = %d, far_nbrs = %d -> reallocating!\n",
- num_intrs, far_nbrs->num_intrs );
- fprintf( stderr, "memory allocated: far_nbrs = %ldMB\n",
- num_intrs * sizeof(far_neighbor_data) / (1024*1024) );
+ num_intrs, far_nbrs->num_intrs );
+ fprintf( stderr, "memory allocated: far_nbrs = %ldMB\n",
+ num_intrs * sizeof(far_neighbor_data) / (1024 * 1024) );
#endif
}
-HOST int Allocate_Matrix( sparse_matrix *H, int n, int m )
+int Allocate_Matrix( sparse_matrix *pH, int n, int m )
{
+ sparse_matrix *H;
+
+ if ( (pH = (sparse_matrix*) malloc(sizeof(sparse_matrix))) == NULL )
+ {
+ return FAILURE;
+ }
+
+ H = pH;
H->n = n;
H->m = m;
- if( (H->start = (int*) malloc(sizeof(int) * n+1)) == NULL )
- return 0;
-
- if( (H->end = (int*) malloc(sizeof(int) * n+1)) == NULL )
- return 0;
- if( (H->entries =
- (sparse_matrix_entry*) malloc(sizeof(sparse_matrix_entry)*m)) == NULL )
- return 0;
+ if ( (H->start = (unsigned int*) malloc(sizeof(int) * (n + 1))) == NULL
+ || (H->j = (unsigned int*) malloc(sizeof(int) * m)) == NULL
+ || (H->val = (real*) malloc(sizeof(real) * m)) == NULL )
+ {
+ return FAILURE;
+ }
- return 1;
+ return SUCCESS;
}
void Deallocate_Matrix( sparse_matrix *H )
{
free(H->start);
- free(H->entries);
- free(H->end);
+ free(H->j);
+ free(H->val);
+ free(H);
}
int Reallocate_Matrix( sparse_matrix *H, int n, int m, char *name )
{
Deallocate_Matrix( H );
- if( !Allocate_Matrix( H, n, m ) ) {
+
+ if ( Allocate_Matrix( H, n, m ) == FAILURE )
+ {
fprintf(stderr, "not enough space for %s matrix. terminating!\n", name);
- exit( 1 );
+ exit( INSUFFICIENT_MEMORY );
}
#if defined(DEBUG_FOCUS)
fprintf( stderr, "reallocating %s matrix, n = %d, m = %d\n",
- name, n, m );
- fprintf( stderr, "memory allocated: %s = %ldMB\n",
- name, m * sizeof(sparse_matrix_entry) / (1024*1024) );
+ name, n, m );
+ fprintf( stderr, "memory allocated: %s = %ldMB\n",
+ name, m * sizeof(sparse_matrix_entry) / (1024 * 1024) );
#endif
- return 1;
+
+ return SUCCESS;
}
-int Allocate_HBond_List( int n, int num_h, int *h_index, int *hb_top,
- list *hbonds )
+int Allocate_HBond_List( int n, int num_h, int *h_index, int *hb_top,
+ list *hbonds )
{
int i, num_hbonds;
num_hbonds = 0;
/* find starting indexes for each H and the total number of hbonds */
- for( i = 1; i < n; ++i )
- hb_top[i] += hb_top[i-1];
- num_hbonds = hb_top[n-1];
+ for ( i = 1; i < n; ++i )
+ {
+ hb_top[i] += hb_top[i - 1];
+ }
+ num_hbonds = hb_top[n - 1];
- if( !Make_List(num_h, num_hbonds, TYP_HBOND, hbonds ) )
+ if ( !Make_List(num_h, num_hbonds, TYP_HBOND, hbonds ) )
{
fprintf( stderr, "not enough space for hbonds list. terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
- for( i = 0; i < n; ++i )
- if( h_index[i] == 0 ){
- Set_Start_Index( 0, 0, hbonds );
- Set_End_Index( 0, 0, hbonds );
+ for ( i = 0; i < n; ++i )
+ {
+ if ( h_index[i] == 0 )
+ {
+ Set_Start_Index( 0, 0, hbonds );
+ Set_End_Index( 0, 0, hbonds );
}
- else if( h_index[i] > 0 ){
- Set_Start_Index( h_index[i], hb_top[i-1], hbonds );
- Set_End_Index( h_index[i], hb_top[i-1], hbonds );
+ else if ( h_index[i] > 0 )
+ {
+ Set_Start_Index( h_index[i], hb_top[i - 1], hbonds );
+ Set_End_Index( h_index[i], hb_top[i - 1], hbonds );
}
+ }
#if defined(DEBUG_FOCUS)
fprintf( stderr, "allocating hbonds - num_hbonds: %d\n", num_hbonds );
- fprintf( stderr, "memory allocated: hbonds = %ldMB\n",
- num_hbonds * sizeof(hbond_data) / (1024*1024) );
+ fprintf( stderr, "memory allocated: hbonds = %ldMB\n",
+ num_hbonds * sizeof(hbond_data) / (1024 * 1024) );
#endif
- return 1;
+ return SUCCESS;
}
@@ -129,10 +175,14 @@ int Reallocate_HBonds_List( int n, int num_h, int *h_index, list *hbonds )
#if defined(DEBUG_FOCUS)
fprintf( stderr, "reallocating hbonds\n" );
#endif
- hb_top = (int *)calloc( n, sizeof(int) );
- for( i = 0; i < n; ++i )
- if( h_index[i] >= 0 )
- hb_top[i] = MAX(Num_Entries(h_index[i],hbonds)*SAFE_HBONDS, MIN_HBONDS);
+ hb_top = calloc( n, sizeof(int) );
+ for ( i = 0; i < n; ++i )
+ {
+ if ( h_index[i] >= 0 )
+ {
+ hb_top[i] = MAX(Num_Entries(h_index[i], hbonds) * SAFE_HBONDS, MIN_HBONDS);
+ }
+ }
Delete_List( hbonds );
@@ -140,7 +190,7 @@ int Reallocate_HBonds_List( int n, int num_h, int *h_index, list *hbonds )
free( hb_top );
- return 1;
+ return SUCCESS;
}
@@ -150,29 +200,32 @@ int Allocate_Bond_List( int n, int *bond_top, list *bonds )
num_bonds = 0;
/* find starting indexes for each atom and the total number of bonds */
- for( i = 1; i < n; ++i )
- bond_top[i] += bond_top[i-1];
- num_bonds = bond_top[n-1];
+ for ( i = 1; i < n; ++i )
+ {
+ bond_top[i] += bond_top[i - 1];
+ }
+ num_bonds = bond_top[n - 1];
- if( !Make_List(n, num_bonds, TYP_BOND, bonds ) )
+ if ( !Make_List(n, num_bonds, TYP_BOND, bonds ) )
{
fprintf( stderr, "not enough space for bonds list. terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
- Set_Start_Index( 0, 0, bonds );
- Set_End_Index( 0, 0, bonds );
- for( i = 1; i < n; ++i ) {
- Set_Start_Index( i, bond_top[i-1], bonds );
- Set_End_Index( i, bond_top[i-1], bonds );
+ Set_Start_Index( 0, 0, bonds );
+ Set_End_Index( 0, 0, bonds );
+ for ( i = 1; i < n; ++i )
+ {
+ Set_Start_Index( i, bond_top[i - 1], bonds );
+ Set_End_Index( i, bond_top[i - 1], bonds );
}
#if defined(DEBUG_FOCUS)
fprintf( stderr, "allocating bonds - num_bonds: %d\n", num_bonds );
- fprintf( stderr, "memory allocated: bonds = %ldMB\n",
- num_bonds * sizeof(bond_data) / (1024*1024) );
+ fprintf( stderr, "memory allocated: bonds = %ldMB\n",
+ num_bonds * sizeof(bond_data) / (1024 * 1024) );
#endif
- return 1;
+ return SUCCESS;
}
@@ -184,9 +237,10 @@ int Reallocate_Bonds_List( int n, list *bonds, int *num_bonds, int *est_3body )
#if defined(DEBUG_FOCUS)
fprintf( stderr, "reallocating bonds\n" );
#endif
- bond_top = (int *)calloc( n, sizeof(int) );
+ bond_top = calloc( n, sizeof(int) );
*est_3body = 0;
- for( i = 0; i < n; ++i ){
+ for ( i = 0; i < n; ++i )
+ {
*est_3body += SQR( Num_Entries( i, bonds ) );
bond_top[i] = MAX( Num_Entries( i, bonds ) * 2, MIN_BONDS );
}
@@ -194,17 +248,18 @@ int Reallocate_Bonds_List( int n, list *bonds, int *num_bonds, int *est_3body )
Delete_List( bonds );
Allocate_Bond_List( n, bond_top, bonds );
- *num_bonds = bond_top[n-1];
+ *num_bonds = bond_top[n - 1];
free( bond_top );
- return 1;
+ return SUCCESS;
}
-void Reallocate( reax_system *system, static_storage *workspace, list **lists,
- int nbr_flag )
+void Reallocate( reax_system *system, static_storage *workspace, list **lists,
+ int nbr_flag )
{
+ int i, j, k;
int num_bonds, est_3body;
reallocate_data *realloc;
grid *g;
@@ -212,70 +267,75 @@ void Reallocate( reax_system *system, static_storage *workspace, list **lists,
realloc = &(workspace->realloc);
g = &(system->g);
- if( realloc->num_far > 0 && nbr_flag ) {
- fprintf (stderr, " Reallocating neighbors \n");
- Reallocate_Neighbor_List( (*lists)+FAR_NBRS,
- system->N, realloc->num_far * SAFE_ZONE );
+ if ( realloc->num_far > 0 && nbr_flag )
+ {
+ Reallocate_Neighbor_List( (*lists) + FAR_NBRS,
+ system->N, realloc->num_far * SAFE_ZONE );
realloc->num_far = -1;
}
- if( realloc->Htop > 0 ){
- fprintf (stderr, " Reallocating Matrix \n");
- Reallocate_Matrix(&(workspace->H), system->N, realloc->Htop*SAFE_ZONE,"H");
+ if ( realloc->Htop > 0 )
+ {
+ Reallocate_Matrix(workspace->H, system->N, realloc->Htop * SAFE_ZONE, "H");
realloc->Htop = -1;
- Deallocate_Matrix( &workspace->L );
- Deallocate_Matrix( &workspace->U );
+ Deallocate_Matrix( workspace->L );
+ Deallocate_Matrix( workspace->U );
+ workspace->L = NULL;
+ workspace->U = NULL;
}
- if( realloc->hbonds > 0 ){
- fprintf (stderr, " Reallocating hbonds \n");
+ if ( realloc->hbonds > 0 )
+ {
Reallocate_HBonds_List(system->N, workspace->num_H, workspace->hbond_index,
- (*lists)+HBONDS );
+ (*lists) + HBONDS );
realloc->hbonds = -1;
}
num_bonds = est_3body = -1;
- if( realloc->bonds > 0 ){
- fprintf (stderr, " Reallocating bonds \n");
- Reallocate_Bonds_List( system->N, (*lists)+BONDS, &num_bonds, &est_3body );
+ if ( realloc->bonds > 0 )
+ {
+ Reallocate_Bonds_List( system->N, (*lists) + BONDS, &num_bonds, &est_3body );
realloc->bonds = -1;
realloc->num_3body = MAX( realloc->num_3body, est_3body );
}
- if( realloc->num_3body > 0 ) {
- fprintf (stderr, " Reallocating 3Body \n");
- Delete_List( (*lists)+THREE_BODIES );
+ if ( realloc->num_3body > 0 )
+ {
+ Delete_List( (*lists) + THREE_BODIES );
- if( num_bonds == -1 )
- num_bonds = ((*lists)+BONDS)->num_intrs;
+ if ( num_bonds == -1 )
+ num_bonds = ((*lists) + BONDS)->num_intrs;
realloc->num_3body *= SAFE_ZONE;
- if( !Make_List( num_bonds, realloc->num_3body,
- TYP_THREE_BODY, (*lists)+THREE_BODIES ) )
+ if ( !Make_List( num_bonds, realloc->num_3body,
+ TYP_THREE_BODY, (*lists) + THREE_BODIES ) )
{
fprintf( stderr, "Problem in initializing angles list. Terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
realloc->num_3body = -1;
#if defined(DEBUG_FOCUS)
fprintf( stderr, "reallocating 3 bodies\n" );
fprintf( stderr, "reallocated - num_bonds: %d\n", num_bonds );
fprintf( stderr, "reallocated - num_3body: %d\n", realloc->num_3body );
- fprintf( stderr, "reallocated 3body memory: %ldMB\n",
- realloc->num_3body*sizeof(three_body_interaction_data)/
- (1024*1024) );
+ fprintf( stderr, "reallocated 3body memory: %ldMB\n",
+ realloc->num_3body * sizeof(three_body_interaction_data) /
+ (1024 * 1024) );
#endif
}
- if( realloc->gcell_atoms > -1 ){
+ if ( realloc->gcell_atoms > -1 )
+ {
#if defined(DEBUG_FOCUS)
fprintf(stderr, "reallocating gcell: g->max_atoms: %d\n", g->max_atoms);
#endif
free (g->atoms);
- g->atoms = (int *) calloc ( g->ncell[0]*g->ncell[1]*g->ncell[2],
- sizeof (int) * workspace->realloc.gcell_atoms);
+ g->atoms = (int *) calloc( g->ncell[0]*g->ncell[1]*g->ncell[2],
+ sizeof(int) * workspace->realloc.gcell_atoms );
+ realloc->gcell_atoms = -1;
+
realloc->gcell_atoms = -1;
}
}
diff --git a/PuReMD-GPU/src/allocate.h b/PuReMD-GPU/src/allocate.h
index b03ed80b34f153b9929ccaa80bc5c27fbf6ce540..72f724dac852c626e7c70018bba95c5f21dc5b51 100644
--- a/PuReMD-GPU/src/allocate.h
+++ b/PuReMD-GPU/src/allocate.h
@@ -28,9 +28,12 @@
extern "C" {
#endif
+int PreAllocate_Space( reax_system*, control_params*, static_storage* );
+
void Reallocate( reax_system*, static_storage*, list**, int );
int Allocate_Matrix( sparse_matrix*, int, int );
+
void Deallocate_Matrix( sparse_matrix *);
int Allocate_HBond_List( int, int, int*, int*, list* );
diff --git a/PuReMD-GPU/src/analyze.c b/PuReMD-GPU/src/analyze.c
index 8eef938a372add29eb044846b12c77eddd47aad0..014eea8f46093e1a381e0a5e44241cfdc3d7b719 100644
--- a/PuReMD-GPU/src/analyze.c
+++ b/PuReMD-GPU/src/analyze.c
@@ -772,17 +772,16 @@ void Calculate_Drift( reax_system *system, control_params *control,
&(system->box), driftvec );
if ( fabs( driftvec[0] ) >= system->box.box_norms[0] / 2.0 - 2.0 ||
- fabs( driftvec[0] ) >= system->box.box_norms[0] / 2.0 - 2.0 ||
- fabs( driftvec[0] ) >= system->box.box_norms[0] / 2.0 - 2.0 )
+ fabs( driftvec[1] ) >= system->box.box_norms[1] / 2.0 - 2.0 ||
+ fabs( driftvec[2] ) >= system->box.box_norms[2] / 2.0 - 2.0 )
{
/* the atom has moved almost half the box size.
exclude it from further drift computations as it might have an
improper contribution due to periodic boudnaries. */
+ workspace->x_old[i][0] = -999999999.0;
+ workspace->x_old[i][1] = -999999999.0;
+ workspace->x_old[i][2] = -999999999.0;
- //TODO -- check this one. may be not initializing this properly
- //workspace->x_old[i][0] = workspace->x_old[i][2] = workspace->x_old[i][2] = -999999999999.0;
- workspace->x_old[i][0] = workspace->x_old[i][2] = workspace->x_old[i][2] = -999999999.0;
- //TODO -- check this one. may be not initializing this properly
continue;
}
diff --git a/PuReMD-GPU/src/box.c b/PuReMD-GPU/src/box.c
index e42395c5556042493c0f707879772d84e9d18658..a7911fda49b1dccbbfccf24ae2fa445c608e0d5e 100644
--- a/PuReMD-GPU/src/box.c
+++ b/PuReMD-GPU/src/box.c
@@ -1,58 +1,189 @@
/*----------------------------------------------------------------------
- PuReMD-GPu - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "box.h"
+
+#include "tool_box.h"
#include "vector.h"
-void Init_Box_From_CRYST(real a, real b, real c,
- real alpha, real beta, real gamma,
+void Make_Consistent( simulation_box* box )
+{
+ real one_vol;
+
+ box->volume =
+ box->box[0][0] * (box->box[1][1] * box->box[2][2] -
+ box->box[2][1] * box->box[2][1]) +
+ box->box[0][1] * (box->box[2][0] * box->box[1][2] -
+ box->box[1][0] * box->box[2][2]) +
+ box->box[0][2] * (box->box[1][0] * box->box[2][1] -
+ box->box[2][0] * box->box[1][1]);
+
+ one_vol = 1.0 / box->volume;
+
+ box->box_inv[0][0] = (box->box[1][1] * box->box[2][2] -
+ box->box[1][2] * box->box[2][1]) * one_vol;
+ box->box_inv[0][1] = (box->box[0][2] * box->box[2][1] -
+ box->box[0][1] * box->box[2][2]) * one_vol;
+ box->box_inv[0][2] = (box->box[0][1] * box->box[1][2] -
+ box->box[0][2] * box->box[1][1]) * one_vol;
+
+ box->box_inv[1][0] = (box->box[1][2] * box->box[2][0] -
+ box->box[1][0] * box->box[2][2]) * one_vol;
+ box->box_inv[1][1] = (box->box[0][0] * box->box[2][2] -
+ box->box[0][2] * box->box[2][0]) * one_vol;
+ box->box_inv[1][2] = (box->box[0][2] * box->box[1][0] -
+ box->box[0][0] * box->box[1][2]) * one_vol;
+
+ box->box_inv[2][0] = (box->box[1][0] * box->box[2][1] -
+ box->box[1][1] * box->box[2][0]) * one_vol;
+ box->box_inv[2][1] = (box->box[0][1] * box->box[2][0] -
+ box->box[0][0] * box->box[2][1]) * one_vol;
+ box->box_inv[2][2] = (box->box[0][0] * box->box[1][1] -
+ box->box[0][1] * box->box[1][0]) * one_vol;
+
+ box->box_norms[0] = SQRT( SQR(box->box[0][0]) +
+ SQR(box->box[0][1]) +
+ SQR(box->box[0][2]) );
+ box->box_norms[1] = SQRT( SQR(box->box[1][0]) +
+ SQR(box->box[1][1]) +
+ SQR(box->box[1][2]) );
+ box->box_norms[2] = SQRT( SQR(box->box[2][0]) +
+ SQR(box->box[2][1]) +
+ SQR(box->box[2][2]) );
+
+ box->trans[0][0] = box->box[0][0] / box->box_norms[0];
+ box->trans[0][1] = box->box[1][0] / box->box_norms[0];
+ box->trans[0][2] = box->box[2][0] / box->box_norms[0];
+
+ box->trans[1][0] = box->box[0][1] / box->box_norms[1];
+ box->trans[1][1] = box->box[1][1] / box->box_norms[1];
+ box->trans[1][2] = box->box[2][1] / box->box_norms[1];
+
+ box->trans[2][0] = box->box[0][2] / box->box_norms[2];
+ box->trans[2][1] = box->box[1][2] / box->box_norms[2];
+ box->trans[2][2] = box->box[2][2] / box->box_norms[2];
+
+ one_vol = box->box_norms[0] * box->box_norms[1] * box->box_norms[2] * one_vol;
+
+ box->trans_inv[0][0] = (box->trans[1][1] * box->trans[2][2] -
+ box->trans[1][2] * box->trans[2][1]) * one_vol;
+ box->trans_inv[0][1] = (box->trans[0][2] * box->trans[2][1] -
+ box->trans[0][1] * box->trans[2][2]) * one_vol;
+ box->trans_inv[0][2] = (box->trans[0][1] * box->trans[1][2] -
+ box->trans[0][2] * box->trans[1][1]) * one_vol;
+
+ box->trans_inv[1][0] = (box->trans[1][2] * box->trans[2][0] -
+ box->trans[1][0] * box->trans[2][2]) * one_vol;
+ box->trans_inv[1][1] = (box->trans[0][0] * box->trans[2][2] -
+ box->trans[0][2] * box->trans[2][0]) * one_vol;
+ box->trans_inv[1][2] = (box->trans[0][2] * box->trans[1][0] -
+ box->trans[0][0] * box->trans[1][2]) * one_vol;
+
+ box->trans_inv[2][0] = (box->trans[1][0] * box->trans[2][1] -
+ box->trans[1][1] * box->trans[2][0]) * one_vol;
+ box->trans_inv[2][1] = (box->trans[0][1] * box->trans[2][0] -
+ box->trans[0][0] * box->trans[2][1]) * one_vol;
+ box->trans_inv[2][2] = (box->trans[0][0] * box->trans[1][1] -
+ box->trans[0][1] * box->trans[1][0]) * one_vol;
+
+// for (i=0; i < 3; i++)
+// {
+// for (j=0; j < 3; j++)
+// fprintf(stderr,"%lf\t",box->trans[i][j]);
+// fprintf(stderr,"\n");
+// }
+// fprintf(stderr,"\n");
+// for (i=0; i < 3; i++)
+// {
+// for (j=0; j < 3; j++)
+// fprintf(stderr,"%lf\t",box->trans_inv[i][j]);
+// fprintf(stderr,"\n");
+// }
+
+ box->g[0][0] = box->box[0][0] * box->box[0][0] +
+ box->box[0][1] * box->box[0][1] +
+ box->box[0][2] * box->box[0][2];
+ box->g[1][0] =
+ box->g[0][1] = box->box[0][0] * box->box[1][0] +
+ box->box[0][1] * box->box[1][1] +
+ box->box[0][2] * box->box[1][2];
+ box->g[2][0] =
+ box->g[0][2] = box->box[0][0] * box->box[2][0] +
+ box->box[0][1] * box->box[2][1] +
+ box->box[0][2] * box->box[2][2];
+
+ box->g[1][1] = box->box[1][0] * box->box[1][0] +
+ box->box[1][1] * box->box[1][1] +
+ box->box[1][2] * box->box[1][2];
+ box->g[1][2] =
+ box->g[2][1] = box->box[1][0] * box->box[2][0] +
+ box->box[1][1] * box->box[2][1] +
+ box->box[1][2] * box->box[2][2];
+
+ box->g[2][2] = box->box[2][0] * box->box[2][0] +
+ box->box[2][1] * box->box[2][1] +
+ box->box[2][2] * box->box[2][2];
+
+ // These proportions are only used for isotropic_NPT!
+ box->side_prop[0] = box->box[0][0] / box->box[0][0];
+ box->side_prop[1] = box->box[1][1] / box->box[0][0];
+ box->side_prop[2] = box->box[2][2] / box->box[0][0];
+}
+
+
+/* setup the simulation box */
+void Setup_Box( real a, real b, real c, real alpha, real beta, real gamma,
simulation_box* box )
{
double c_alpha, c_beta, c_gamma, s_gamma, zi;
- c_alpha = cos(DEG2RAD(alpha));
- c_beta = cos(DEG2RAD(beta));
- c_gamma = cos(DEG2RAD(gamma));
- s_gamma = sin(DEG2RAD(gamma));
+ if ( IS_NAN_REAL(a) || IS_NAN_REAL(b) || IS_NAN_REAL(c)
+ || IS_NAN_REAL(alpha) || IS_NAN_REAL(beta) || IS_NAN_REAL(gamma) )
+ {
+ fprintf( stderr, "Invalid simulation box boundaries for big box (NaN). Terminating...\n" );
+ exit( INVALID_INPUT );
+ }
- zi = (c_alpha - c_beta * c_gamma)/s_gamma;
+ c_alpha = COS(DEG2RAD(alpha));
+ c_beta = COS(DEG2RAD(beta));
+ c_gamma = COS(DEG2RAD(gamma));
+ s_gamma = SIN(DEG2RAD(gamma));
+ zi = (c_alpha - c_beta * c_gamma) / s_gamma;
- box->box[0][0] = a;
- box->box[0][1] = 0.0;
+ box->box[0][0] = a;
+ box->box[0][1] = 0.0;
box->box[0][2] = 0.0;
-
- box->box[1][0] = b * c_gamma;
- box->box[1][1] = b * s_gamma;
- box->box[1][2] = 0.0;
-
+ box->box[1][0] = b * c_gamma;
+ box->box[1][1] = b * s_gamma;
+ box->box[1][2] = 0.0;
box->box[2][0] = c * c_beta;
box->box[2][1] = c * zi;
box->box[2][2] = c * SQRT(1.0 - SQR(c_beta) - SQR(zi));
+#if defined(DEBUG)
+ fprintf( stderr, "box is %8.2f x %8.2f x %8.2f\n",
+ box->box[0][0], box->box[1][1], box->box[2][2] );
+#endif
Make_Consistent( box );
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "box is %8.2f x %8.2f x %8.2f\n",
- box->box[0][0], box->box[1][1], box->box[2][2] );
-#endif
}
@@ -60,8 +191,8 @@ void Update_Box( rtensor box_tensor, simulation_box* box )
{
int i, j;
- for (i=0; i < 3; i++)
- for (j=0; j < 3; j++)
+ for (i = 0; i < 3; i++)
+ for (j = 0; j < 3; j++)
box->box[i][j] = box_tensor[i][j];
Make_Consistent( box );
@@ -70,200 +201,37 @@ void Update_Box( rtensor box_tensor, simulation_box* box )
void Update_Box_Isotropic( simulation_box *box, real mu )
{
- /*box->box[0][0] =
+ /*box->box[0][0] =
POW( V_new / ( box->side_prop[1] * box->side_prop[2] ), 1.0/3.0 );
- box->box[1][1] = box->box[0][0] * box->side_prop[1];
- box->box[2][2] = box->box[0][0] * box->side_prop[2];
- */
+ box->box[1][1] = box->box[0][0] * box->side_prop[1];
+ box->box[2][2] = box->box[0][0] * box->side_prop[2];
+ */
rtensor_Copy( box->old_box, box->box );
box->box[0][0] *= mu;
box->box[1][1] *= mu;
box->box[2][2] *= mu;
- box->volume = box->box[0][0]*box->box[1][1]*box->box[2][2];
+ box->volume = box->box[0][0] * box->box[1][1] * box->box[2][2];
Make_Consistent(box/*, periodic*/);
}
void Update_Box_SemiIsotropic( simulation_box *box, rvec mu )
{
- /*box->box[0][0] =
+ /*box->box[0][0] =
POW( V_new / ( box->side_prop[1] * box->side_prop[2] ), 1.0/3.0 );
- box->box[1][1] = box->box[0][0] * box->side_prop[1];
- box->box[2][2] = box->box[0][0] * box->side_prop[2]; */
+ box->box[1][1] = box->box[0][0] * box->side_prop[1];
+ box->box[2][2] = box->box[0][0] * box->side_prop[2]; */
rtensor_Copy( box->old_box, box->box );
box->box[0][0] *= mu[0];
box->box[1][1] *= mu[1];
box->box[2][2] *= mu[2];
- box->volume = box->box[0][0]*box->box[1][1]*box->box[2][2];
+ box->volume = box->box[0][0] * box->box[1][1] * box->box[2][2];
Make_Consistent(box);
}
-void Make_Consistent(simulation_box* box)
-{
- real one_vol;
-
- box->volume =
- box->box[0][0] * (box->box[1][1]*box->box[2][2] -
- box->box[2][1]*box->box[2][1]) +
- box->box[0][1] * (box->box[2][0]*box->box[1][2] -
- box->box[1][0]*box->box[2][2]) +
- box->box[0][2] * (box->box[1][0]*box->box[2][1] -
- box->box[2][0]*box->box[1][1]);
-
- one_vol = 1.0/box->volume;
-
- box->box_inv[0][0] = (box->box[1][1]*box->box[2][2] -
- box->box[1][2]*box->box[2][1]) * one_vol;
- box->box_inv[0][1] = (box->box[0][2]*box->box[2][1] -
- box->box[0][1]*box->box[2][2]) * one_vol;
- box->box_inv[0][2] = (box->box[0][1]*box->box[1][2] -
- box->box[0][2]*box->box[1][1]) * one_vol;
-
- box->box_inv[1][0] = (box->box[1][2]*box->box[2][0] -
- box->box[1][0]*box->box[2][2]) * one_vol;
- box->box_inv[1][1] = (box->box[0][0]*box->box[2][2] -
- box->box[0][2]*box->box[2][0]) * one_vol;
- box->box_inv[1][2] = (box->box[0][2]*box->box[1][0] -
- box->box[0][0]*box->box[1][2]) * one_vol;
-
- box->box_inv[2][0] = (box->box[1][0]*box->box[2][1] -
- box->box[1][1]*box->box[2][0]) * one_vol;
- box->box_inv[2][1] = (box->box[0][1]*box->box[2][0] -
- box->box[0][0]*box->box[2][1]) * one_vol;
- box->box_inv[2][2] = (box->box[0][0]*box->box[1][1] -
- box->box[0][1]*box->box[1][0]) * one_vol;
-
- box->box_norms[0] = SQRT( SQR(box->box[0][0]) +
- SQR(box->box[0][1]) +
- SQR(box->box[0][2]) );
- box->box_norms[1] = SQRT( SQR(box->box[1][0]) +
- SQR(box->box[1][1]) +
- SQR(box->box[1][2]) );
- box->box_norms[2] = SQRT( SQR(box->box[2][0]) +
- SQR(box->box[2][1]) +
- SQR(box->box[2][2]) );
-
- box->trans[0][0] = box->box[0][0]/box->box_norms[0];
- box->trans[0][1] = box->box[1][0]/box->box_norms[0];
- box->trans[0][2] = box->box[2][0]/box->box_norms[0];
-
- box->trans[1][0] = box->box[0][1]/box->box_norms[1];
- box->trans[1][1] = box->box[1][1]/box->box_norms[1];
- box->trans[1][2] = box->box[2][1]/box->box_norms[1];
-
- box->trans[2][0] = box->box[0][2]/box->box_norms[2];
- box->trans[2][1] = box->box[1][2]/box->box_norms[2];
- box->trans[2][2] = box->box[2][2]/box->box_norms[2];
-
- one_vol = box->box_norms[0]*box->box_norms[1]*box->box_norms[2]*one_vol;
-
- box->trans_inv[0][0] = (box->trans[1][1]*box->trans[2][2] -
- box->trans[1][2]*box->trans[2][1]) * one_vol;
- box->trans_inv[0][1] = (box->trans[0][2]*box->trans[2][1] -
- box->trans[0][1]*box->trans[2][2]) * one_vol;
- box->trans_inv[0][2] = (box->trans[0][1]*box->trans[1][2] -
- box->trans[0][2]*box->trans[1][1]) * one_vol;
-
- box->trans_inv[1][0] = (box->trans[1][2]*box->trans[2][0] -
- box->trans[1][0]*box->trans[2][2]) * one_vol;
- box->trans_inv[1][1] = (box->trans[0][0]*box->trans[2][2] -
- box->trans[0][2]*box->trans[2][0]) * one_vol;
- box->trans_inv[1][2] = (box->trans[0][2]*box->trans[1][0] -
- box->trans[0][0]*box->trans[1][2]) * one_vol;
-
- box->trans_inv[2][0] = (box->trans[1][0]*box->trans[2][1] -
- box->trans[1][1]*box->trans[2][0]) * one_vol;
- box->trans_inv[2][1] = (box->trans[0][1]*box->trans[2][0] -
- box->trans[0][0]*box->trans[2][1]) * one_vol;
- box->trans_inv[2][2] = (box->trans[0][0]*box->trans[1][1] -
- box->trans[0][1]*box->trans[1][0]) * one_vol;
-
- // for (i=0; i < 3; i++)
- // {
- // for (j=0; j < 3; j++)
- // fprintf(stderr,"%lf\t",box->trans[i][j]);
- // fprintf(stderr,"\n");
- // }
- // fprintf(stderr,"\n");
- // for (i=0; i < 3; i++)
- // {
- // for (j=0; j < 3; j++)
- // fprintf(stderr,"%lf\t",box->trans_inv[i][j]);
- // fprintf(stderr,"\n");
- // }
-
-
- box->g[0][0] = box->box[0][0] * box->box[0][0] +
- box->box[0][1] * box->box[0][1] +
- box->box[0][2] * box->box[0][2];
- box->g[1][0] =
- box->g[0][1] = box->box[0][0] * box->box[1][0] +
- box->box[0][1] * box->box[1][1] +
- box->box[0][2] * box->box[1][2];
- box->g[2][0] =
- box->g[0][2] = box->box[0][0] * box->box[2][0] +
- box->box[0][1] * box->box[2][1] +
- box->box[0][2] * box->box[2][2];
-
- box->g[1][1] = box->box[1][0] * box->box[1][0] +
- box->box[1][1] * box->box[1][1] +
- box->box[1][2] * box->box[1][2];
- box->g[1][2] =
- box->g[2][1] = box->box[1][0] * box->box[2][0] +
- box->box[1][1] * box->box[2][1] +
- box->box[1][2] * box->box[2][2];
-
- box->g[2][2] = box->box[2][0] * box->box[2][0] +
- box->box[2][1] * box->box[2][1] +
- box->box[2][2] * box->box[2][2];
-
- // These proportions are only used for isotropic_NPT!
- box->side_prop[0] = box->box[0][0] / box->box[0][0];
- box->side_prop[1] = box->box[1][1] / box->box[0][0];
- box->side_prop[2] = box->box[2][2] / box->box[0][0];
-}
-
-
-void Transform( rvec x1, simulation_box *box, char flag, rvec x2 )
-{
- int i, j;
- real tmp;
-
- // printf(">x1: (%lf, %lf, %lf)\n",x1[0],x1[1],x1[2]);
-
- if (flag > 0) {
- for (i=0; i < 3; i++) {
- tmp = 0.0;
- for (j=0; j < 3; j++)
- tmp += box->trans[i][j]*x1[j];
- x2[i] = tmp;
- }
- }
- else {
- for (i=0; i < 3; i++) {
- tmp = 0.0;
- for (j=0; j < 3; j++)
- tmp += box->trans_inv[i][j]*x1[j];
- x2[i] = tmp;
- }
- }
- // printf(">x2: (%lf, %lf, %lf)\n", x2[0], x2[1], x2[2]);
-}
-
-
-void Transform_to_UnitBox( rvec x1, simulation_box *box, char flag, rvec x2 )
-{
- Transform( x1, box, flag, x2 );
-
- x2[0] /= box->box_norms[0];
- x2[1] /= box->box_norms[1];
- x2[2] /= box->box_norms[2];
-}
-
-
void Distance_on_T3_Gen( rvec x1, rvec x2, simulation_box* box, rvec r )
{
rvec xa, xb, ra;
@@ -301,12 +269,12 @@ void Inc_on_T3_Gen( rvec x, rvec dx, simulation_box* box )
real Metric_Product( rvec x1, rvec x2, simulation_box* box )
{
int i, j;
- real dist=0.0, tmp;
+ real dist = 0.0, tmp;
- for( i = 0; i < 3; i++ )
+ for ( i = 0; i < 3; i++ )
{
tmp = 0.0;
- for( j = 0; j < 3; j++ )
+ for ( j = 0; j < 3; j++ )
tmp += box->g[i][j] * x2[j];
dist += x1[i] * tmp;
}
@@ -315,12 +283,59 @@ real Metric_Product( rvec x1, rvec x2, simulation_box* box )
}
-/* Determines if the distance between x1 and x2 is < vlist_cut.
+int Are_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
+ real cutoff, far_neighbor_data *data )
+{
+ real norm_sqr, d, tmp;
+ int i;
+
+ norm_sqr = 0;
+
+ for ( i = 0; i < 3; i++ )
+ {
+ d = x2[i] - x1[i];
+ tmp = SQR(d);
+
+ if ( tmp >= SQR( box->box_norms[i] / 2.0 ) )
+ {
+ if ( x2[i] > x1[i] )
+ {
+ d -= box->box_norms[i];
+ data->rel_box[i] = -1;
+ }
+ else
+ {
+ d += box->box_norms[i];
+ data->rel_box[i] = +1;
+ }
+
+ data->dvec[i] = d;
+ norm_sqr += SQR(d);
+ }
+ else
+ {
+ data->dvec[i] = d;
+ norm_sqr += tmp;
+ data->rel_box[i] = 0;
+ }
+ }
+
+ if ( norm_sqr <= SQR(cutoff) )
+ {
+ data->d = sqrt(norm_sqr);
+ return TRUE;
+ }
+
+ return FALSE;
+}
+
+
+/* Determines if the distance between x1 and x2 is < vlist_cut.
If so, this neighborhood is added to the list of far neighbors.
Periodic boundary conditions do not apply. */
-void Get_NonPeriodic_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
- control_params *control,
- far_neighbor_data *new_nbrs, int *count )
+void Get_NonPeriodic_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
+ control_params *control,
+ far_neighbor_data *new_nbrs, int *count )
{
real norm_sqr;
@@ -328,7 +343,8 @@ void Get_NonPeriodic_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
norm_sqr = rvec_Norm_Sqr( new_nbrs[0].dvec );
- if( norm_sqr <= SQR( control->vlist_cut ) ) {
+ if ( norm_sqr <= SQR( control->vlist_cut ) )
+ {
*count = 1;
new_nbrs[0].d = SQRT( norm_sqr );
@@ -341,11 +357,11 @@ void Get_NonPeriodic_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
/* Finds periodic neighbors in a 'big_box'. Here 'big_box' means:
the current simulation box has all dimensions > 2 *vlist_cut.
- If the periodic distance between x1 and x2 is than vlist_cut, this
+ If the periodic distance between x1 and x2 is than vlist_cut, this
neighborhood is added to the list of far neighbors. */
-void Get_Periodic_Far_Neighbors_Big_Box( rvec x1, rvec x2, simulation_box *box,
- control_params *control,
- far_neighbor_data *periodic_nbrs,
+void Get_Periodic_Far_Neighbors_Big_Box( rvec x1, rvec x2, simulation_box *box,
+ control_params *control,
+ far_neighbor_data *periodic_nbrs,
int *count )
{
real norm_sqr, d, tmp;
@@ -353,19 +369,23 @@ void Get_Periodic_Far_Neighbors_Big_Box( rvec x1, rvec x2, simulation_box *box,
norm_sqr = 0;
- for( i = 0; i < 3; i++ ) {
+ for ( i = 0; i < 3; i++ )
+ {
d = x2[i] - x1[i];
tmp = SQR(d);
// fprintf(out,"Inside Sq_Distance_on_T3, %d, %lf, %lf\n",
// i,tmp,SQR(box->box_norms[i]/2.0));
- if( tmp >= SQR( box->box_norms[i] / 2.0 ) ) {
- if( x2[i] > x1[i] ) {
+ if ( tmp >= SQR( box->box_norms[i] / 2.0 ) )
+ {
+ if ( x2[i] > x1[i] )
+ {
d -= box->box_norms[i];
periodic_nbrs[0].rel_box[i] = -1;
// periodic_nbrs[0].ext_factor[i] = +1;
}
- else {
+ else
+ {
d += box->box_norms[i];
periodic_nbrs[0].rel_box[i] = +1;
// periodic_nbrs[0].ext_factor[i] = -1;
@@ -374,15 +394,17 @@ void Get_Periodic_Far_Neighbors_Big_Box( rvec x1, rvec x2, simulation_box *box,
periodic_nbrs[0].dvec[i] = d;
norm_sqr += SQR(d);
}
- else {
+ else
+ {
periodic_nbrs[0].dvec[i] = d;
norm_sqr += tmp;
periodic_nbrs[0].rel_box[i] = 0;
// periodic_nbrs[0].ext_factor[i] = 0;
- }
+ }
}
- if( norm_sqr <= SQR( control->vlist_cut ) ) {
+ if ( norm_sqr <= SQR( control->vlist_cut ) )
+ {
*count = 1;
periodic_nbrs[0].d = SQRT( norm_sqr );
}
@@ -390,16 +412,16 @@ void Get_Periodic_Far_Neighbors_Big_Box( rvec x1, rvec x2, simulation_box *box,
}
-/* Finds all periodic far neighborhoods between x1 and x2
+/* Finds all periodic far neighborhoods between x1 and x2
((dist(x1, x2') < vlist_cut, periodic images of x2 are also considered).
Here the box is 'small' meaning that at least one dimension is < 2*vlist_cut.
-IMPORTANT: This part might need some improvement. In NPT, the simulation box
-might get too small (such as <5 A!). In this case we have to consider the
-periodic images of x2 that are two boxs away!!!
- */
+ IMPORTANT: This part might need some improvement. In NPT, the simulation box
+ might get too small (such as <5 A!). In this case we have to consider the
+ periodic images of x2 that are two boxs away!!!
+*/
void Get_Periodic_Far_Neighbors_Small_Box( rvec x1, rvec x2, simulation_box *box,
- control_params *control,
- far_neighbor_data *periodic_nbrs,
+ control_params *control,
+ far_neighbor_data *periodic_nbrs,
int *count )
{
int i, j, k;
@@ -418,14 +440,18 @@ void Get_Periodic_Far_Neighbors_Small_Box( rvec x1, rvec x2, simulation_box *box
imax, jmax, kmax ); */
- for( i = -imax; i <= imax; ++i )
- if(fabs(d_i=((x2[0]+i*box->box_norms[0])-x1[0]))<=control->vlist_cut) {
- for( j = -jmax; j <= jmax; ++j )
- if(fabs(d_j=((x2[1]+j*box->box_norms[1])-x1[1]))<=control->vlist_cut) {
- for( k = -kmax; k <= kmax; ++k )
- if(fabs(d_k=((x2[2]+k*box->box_norms[2])-x1[2]))<=control->vlist_cut) {
+ for ( i = -imax; i <= imax; ++i )
+ if (fabs(d_i = ((x2[0] + i * box->box_norms[0]) - x1[0])) <= control->vlist_cut)
+ {
+ for ( j = -jmax; j <= jmax; ++j )
+ if (fabs(d_j = ((x2[1] + j * box->box_norms[1]) - x1[1])) <= control->vlist_cut)
+ {
+ for ( k = -kmax; k <= kmax; ++k )
+ if (fabs(d_k = ((x2[2] + k * box->box_norms[2]) - x1[2])) <= control->vlist_cut)
+ {
sqr_norm = SQR(d_i) + SQR(d_j) + SQR(d_k);
- if( sqr_norm <= SQR(control->vlist_cut) ) {
+ if ( sqr_norm <= SQR(control->vlist_cut) )
+ {
periodic_nbrs[ *count ].d = SQRT( sqr_norm );
periodic_nbrs[ *count ].dvec[0] = d_i;
@@ -466,21 +492,21 @@ void Get_Periodic_Far_Neighbors_Small_Box( rvec x1, rvec x2, simulation_box *box
/* Returns the mapping for the neighbor box pointed by (ix,iy,iz) */
/*int Get_Nbr_Box( simulation_box *box, int ix, int iy, int iz )
- {
- return (9 * ix + 3 * iy + iz + 13);
-// 13 is to handle negative indexes properly
+{
+ return (9 * ix + 3 * iy + iz + 13);
+ // 13 is to handle negative indexes properly
}*/
/* Returns total pressure vector for the neighbor box pointed by (ix,iy,iz) */
/*rvec Get_Nbr_Box_Press( simulation_box *box, int ix, int iy, int iz )
- {
+{
int map;
- map = 9 * ix + 3 * iy + iz + 13;
-// 13 is to adjust -1,-1,-1 correspond to index 0
+ map = 9 * ix + 3 * iy + iz + 13;
+ // 13 is to adjust -1,-1,-1 correspond to index 0
-return box->nbr_box_press[map];
+ return box->nbr_box_press[map];
}*/
@@ -489,53 +515,53 @@ return box->nbr_box_press[map];
{
int map;
- map = 9 * ix + 3 * iy + iz + 13;
-// 13 is to adjust -1,-1,-1 correspond to index 0
+ map = 9 * ix + 3 * iy + iz + 13;
+ // 13 is to adjust -1,-1,-1 correspond to index 0
-rvec_Add( box->nbr_box_press[map], v );
+ rvec_Add( box->nbr_box_press[map], v );
}*/
/* Increments the total pressure vector for the neighbor box mapped to 'map' */
/*void Inc_Nbr_Box_Press( simulation_box *box, int map, rvec v )
- {
+{
rvec_Add( box->nbr_box_press[map], v );
- }*/
+}*/
-void Print_Box_Information( simulation_box* box, FILE *out )
+void Print_Box( simulation_box* box, FILE *out )
{
int i, j;
fprintf( out, "box: {" );
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
{
fprintf( out, "{" );
- for( j = 0; j < 3; ++j )
+ for ( j = 0; j < 3; ++j )
fprintf( out, "%8.3f ", box->box[i][j] );
fprintf( out, "}" );
}
fprintf( out, "}\n" );
- fprintf( out, "V: %8.3f\tdims: {%8.3f, %8.3f, %8.3f}\n",
- box->volume,
- box->box_norms[0], box->box_norms[1], box->box_norms[2] );
+ fprintf( out, "V: %8.3f\tdims: {%8.3f, %8.3f, %8.3f}\n",
+ box->volume,
+ box->box_norms[0], box->box_norms[1], box->box_norms[2] );
fprintf( out, "box_trans: {" );
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
{
fprintf( out, "{" );
- for( j = 0; j < 3; ++j )
+ for ( j = 0; j < 3; ++j )
fprintf( out, "%8.3f ", box->trans[i][j] );
fprintf( out, "}" );
}
fprintf( out, "}\n" );
fprintf( out, "box_trinv: {" );
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
{
fprintf( out, "{" );
- for( j = 0; j < 3; ++j )
+ for ( j = 0; j < 3; ++j )
fprintf( out, "%8.3f ", box->trans_inv[i][j] );
fprintf( out, "}" );
}
diff --git a/PuReMD-GPU/src/box.h b/PuReMD-GPU/src/box.h
index 418aa6208a81fb05ee56ff09afc6ff76751f75c9..84f8371becafbc86180e75b7b8a189c299c82493 100644
--- a/PuReMD-GPU/src/box.h
+++ b/PuReMD-GPU/src/box.h
@@ -25,9 +25,7 @@
#include "mytypes.h"
-/* Initializes box from CRYST1 line of PDB */
-void Init_Box_From_CRYST(real, real, real, real, real, real,
- simulation_box*/*, int*/);
+void Setup_Box( real, real, real, real, real, real, simulation_box* );
/* Initializes box from box rtensor */
void Update_Box(rtensor, simulation_box* /*, int*/);
@@ -36,14 +34,9 @@ void Update_Box_SemiIsotropic( simulation_box*, rvec /*, int*/ );
/* Computes all the transformations,
metric and other quantities from box rtensor */
-void Make_Consistent(simulation_box*/*, int*/ );
-
-/* Applies transformation to and from
- Cartesian to Triclinic coordinates based on flag */
-/* Use -1 flag for Cartesian -> Triclinic and +1 for otherway */
-void Transform( rvec, simulation_box*, char, rvec );
-void Transform_to_UnitBox( rvec, simulation_box*, char, rvec );
+void Make_Consistent( simulation_box* );
+int Are_Far_Neighbors( rvec, rvec, simulation_box*, real, far_neighbor_data* );
void Get_NonPeriodic_Far_Neighbors( rvec, rvec, simulation_box*,
control_params*, far_neighbor_data*, int* );
void Get_Periodic_Far_Neighbors_Big_Box( rvec, rvec, simulation_box*,
@@ -59,10 +52,6 @@ void Inc_Nbr_Box_Press( simulation_box*, int, int, int, rvec );*/
/* These functions assume that the coordinates are in triclinic system */
/* this function returns cartesian norm but triclinic distance vector */
-real Metric_Product( rvec, rvec, simulation_box* );
-
-void Print_Box_Information( simulation_box*, FILE* );
-
static inline HOST_DEVICE real Sq_Distance_on_T3( rvec x1, rvec x2, simulation_box* box, rvec r)
{
@@ -113,5 +102,9 @@ static inline HOST_DEVICE void Inc_on_T3( rvec x, rvec dx, simulation_box *box )
}
}
+real Metric_Product( rvec, rvec, simulation_box* );
+
+void Print_Box( simulation_box*, FILE* );
+
#endif
diff --git a/PuReMD-GPU/src/control.c b/PuReMD-GPU/src/control.c
new file mode 100644
index 0000000000000000000000000000000000000000..41f744969f1615ba621d85db852d998e92719b86
--- /dev/null
+++ b/PuReMD-GPU/src/control.c
@@ -0,0 +1,560 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#include <ctype.h>
+
+#include "control.h"
+#include "traj.h"
+#include "tool_box.h"
+
+
+char Read_Control_File( FILE* fp, reax_system *system, control_params* control,
+ output_controls *out_control )
+{
+ char *s, **tmp;
+ int c, i;
+ real val;
+ int ival;
+
+ /* assign default values */
+ strcpy( control->sim_name, "default.sim" );
+
+ control->restart = 0;
+ out_control->restart_format = WRITE_BINARY;
+ out_control->restart_freq = 0;
+ strcpy( control->restart_from, "default.res" );
+ out_control->restart_freq = 0;
+ control->random_vel = 0;
+
+ control->reposition_atoms = 0;
+
+ control->ensemble = NVE;
+ control->nsteps = 0;
+ control->dt = 0.25;
+
+ control->geo_format = PDB;
+ control->restrict_bonds = 0;
+
+ control->periodic_boundaries = 1;
+ control->periodic_images[0] = 0;
+ control->periodic_images[1] = 0;
+ control->periodic_images[2] = 0;
+
+ control->reneighbor = 1;
+ control->vlist_cut = 0;
+ control->nbr_cut = 4.;
+ control->r_cut = 10.;
+ control->r_sp_cut = 10.;
+ control->max_far_nbrs = 1000;
+ control->bo_cut = 0.01;
+ control->thb_cut = 0.001;
+ control->hb_cut = 7.50;
+
+ control->tabulate = 0;
+
+ control->qeq_solver_type = GMRES_S;
+ control->qeq_solver_q_err = 0.000001;
+ control->qeq_domain_sparsify_enabled = FALSE;
+ control->qeq_domain_sparsity = 1.0;
+ control->pre_comp_type = ICHOLT_PC;
+ control->pre_comp_sweeps = 3;
+ control->pre_comp_refactor = 100;
+ control->pre_comp_droptol = 0.01;
+ control->pre_app_type = TRI_SOLVE_PA;
+ control->pre_app_jacobi_iters = 50;
+
+ control->T_init = 0.;
+ control->T_final = 300.;
+ control->Tau_T = 1.0;
+ control->T_mode = 0.;
+ control->T_rate = 1.;
+ control->T_freq = 1.;
+
+ control->P[0] = 0.000101325;
+ control->P[1] = 0.000101325;
+ control->P[2] = 0.000101325;
+ control->Tau_P[0] = 500.0;
+ control->Tau_P[1] = 500.0;
+ control->Tau_P[2] = 500.0;
+ control->Tau_PT = 500.0;
+ control->compressibility = 1.0;
+ control->press_mode = 0;
+
+ control->remove_CoM_vel = 25;
+
+ out_control->debug_level = 0;
+ out_control->energy_update_freq = 10;
+
+ out_control->write_steps = 100;
+ out_control->traj_compress = 0;
+ out_control->write = fprintf;
+ out_control->traj_format = 0;
+ out_control->write_header =
+ (int (*)( reax_system*, control_params*,
+ static_storage*, void* )) Write_Custom_Header;
+ out_control->append_traj_frame =
+ (int (*)( reax_system*, control_params*, simulation_data*,
+ static_storage*, list **, void* )) Append_Custom_Frame;
+
+ strcpy( out_control->traj_title, "default_title" );
+ out_control->atom_format = 0;
+ out_control->bond_info = 0;
+ out_control->angle_info = 0;
+
+ control->molec_anal = NO_ANALYSIS;
+ control->freq_molec_anal = 0;
+ control->bg_cut = 0.3;
+ control->num_ignored = 0;
+ memset( control->ignore, 0, sizeof(int)*MAX_ATOM_TYPES );
+
+ control->dipole_anal = 0;
+ control->freq_dipole_anal = 0;
+
+ control->diffusion_coef = 0;
+ control->freq_diffusion_coef = 0;
+ control->restrict_type = 0;
+
+ /* memory allocations */
+ s = (char*) malloc(sizeof(char) * MAX_LINE);
+ tmp = (char**) malloc(sizeof(char*)*MAX_TOKENS);
+ for (i = 0; i < MAX_TOKENS; i++)
+ tmp[i] = (char*) malloc(sizeof(char) * MAX_LINE);
+
+ /* read control parameters file */
+ while (fgets(s, MAX_LINE, fp))
+ {
+ c = Tokenize(s, &tmp);
+
+ if ( strcmp(tmp[0], "simulation_name") == 0 )
+ {
+ strcpy( control->sim_name, tmp[1] );
+ }
+ //else if( strcmp(tmp[0], "restart") == 0 ) {
+ // ival = atoi(tmp[1]);
+ // control->restart = ival;
+ //}
+ else if ( strcmp(tmp[0], "restart_format") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->restart_format = ival;
+ }
+ else if ( strcmp(tmp[0], "restart_freq") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->restart_freq = ival;
+ }
+ else if ( strcmp(tmp[0], "random_vel") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->random_vel = ival;
+ }
+ else if ( strcmp(tmp[0], "reposition_atoms") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->reposition_atoms = ival;
+ }
+ else if ( strcmp(tmp[0], "ensemble_type") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->ensemble = ival;
+ }
+ else if ( strcmp(tmp[0], "nsteps") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->nsteps = ival;
+ }
+ else if ( strcmp(tmp[0], "dt") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->dt = val * 1.e-3; // convert dt from fs to ps!
+ }
+ else if ( strcmp(tmp[0], "periodic_boundaries") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->periodic_boundaries = ival;
+ }
+ else if ( strcmp(tmp[0], "periodic_images") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->periodic_images[0] = ival;
+ ival = atoi(tmp[2]);
+ control->periodic_images[1] = ival;
+ ival = atoi(tmp[3]);
+ control->periodic_images[2] = ival;
+ }
+ else if ( strcmp(tmp[0], "geo_format") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->geo_format = ival;
+ }
+ else if ( strcmp(tmp[0], "restrict_bonds") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->restrict_bonds = ival;
+ }
+ else if ( strcmp(tmp[0], "tabulate_long_range") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->tabulate = ival;
+ }
+ else if ( strcmp(tmp[0], "reneighbor") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->reneighbor = ival;
+ }
+ else if ( strcmp(tmp[0], "vlist_buffer") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->vlist_cut = val;
+ }
+ else if ( strcmp(tmp[0], "nbrhood_cutoff") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->nbr_cut = val;
+ }
+ else if ( strcmp(tmp[0], "thb_cutoff") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->thb_cut = val;
+ }
+ else if ( strcmp(tmp[0], "hbond_cutoff") == 0 )
+ {
+ val = atof( tmp[1] );
+ control->hb_cut = val;
+ }
+ else if ( strcmp(tmp[0], "qeq_solver_type") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->qeq_solver_type = ival;
+ }
+ else if ( strcmp(tmp[0], "qeq_solver_q_err") == 0 )
+ {
+ val = atof( tmp[1] );
+ control->qeq_solver_q_err = val;
+ }
+ else if ( strcmp(tmp[0], "qeq_domain_sparsity") == 0 )
+ {
+ val = atof( tmp[1] );
+ control->qeq_domain_sparsity = val;
+ control->qeq_domain_sparsify_enabled = TRUE;
+ }
+ else if ( strcmp(tmp[0], "pre_comp_type") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->pre_comp_type = ival;
+ }
+ else if ( strcmp(tmp[0], "pre_comp_refactor") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->pre_comp_refactor = ival;
+ }
+ else if ( strcmp(tmp[0], "pre_comp_droptol") == 0 )
+ {
+ val = atof( tmp[1] );
+ control->pre_comp_droptol = val;
+ }
+ else if ( strcmp(tmp[0], "pre_comp_sweeps") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->pre_comp_sweeps = ival;
+ }
+ else if ( strcmp(tmp[0], "pre_app_type") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->pre_app_type = ival;
+ }
+ else if ( strcmp(tmp[0], "pre_app_jacobi_iters") == 0 )
+ {
+ ival = atoi( tmp[1] );
+ control->pre_app_jacobi_iters = ival;
+ }
+ else if ( strcmp(tmp[0], "temp_init") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->T_init = val;
+
+ if ( control->T_init < 0.001 )
+ control->T_init = 0.001;
+ }
+ else if ( strcmp(tmp[0], "temp_final") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->T_final = val;
+
+ if ( control->T_final < 0.1 )
+ control->T_final = 0.1;
+ }
+ else if ( strcmp(tmp[0], "t_mass") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->Tau_T = val * 1.e-3; // convert t_mass from fs to ps
+ }
+ else if ( strcmp(tmp[0], "t_mode") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->T_mode = ival;
+ }
+ else if ( strcmp(tmp[0], "t_rate") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->T_rate = val;
+ }
+ else if ( strcmp(tmp[0], "t_freq") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->T_freq = val;
+ }
+ else if ( strcmp(tmp[0], "pressure") == 0 )
+ {
+ if ( control->ensemble == iNPT )
+ {
+ val = atof(tmp[1]);
+ control->P[0] = control->P[1] = control->P[2] = val;
+ }
+ else if ( control->ensemble == sNPT )
+ {
+ val = atof(tmp[1]);
+ control->P[0] = val;
+
+ val = atof(tmp[2]);
+ control->P[1] = val;
+
+ val = atof(tmp[3]);
+ control->P[2] = val;
+ }
+ }
+ else if ( strcmp(tmp[0], "p_mass") == 0 )
+ {
+ if ( control->ensemble == iNPT )
+ {
+ val = atof(tmp[1]);
+ control->Tau_P[0] = val * 1.e-3; // convert p_mass from fs to ps
+ }
+ else if ( control->ensemble == sNPT )
+ {
+ val = atof(tmp[1]);
+ control->Tau_P[0] = val * 1.e-3; // convert p_mass from fs to ps
+
+ val = atof(tmp[2]);
+ control->Tau_P[1] = val * 1.e-3; // convert p_mass from fs to ps
+
+ val = atof(tmp[3]);
+ control->Tau_P[2] = val * 1.e-3; // convert p_mass from fs to ps
+ }
+ }
+ else if ( strcmp(tmp[0], "pt_mass") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->Tau_PT = val * 1.e-3; // convert pt_mass from fs to ps
+ }
+ else if ( strcmp(tmp[0], "compress") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->compressibility = val;
+ }
+ else if ( strcmp(tmp[0], "press_mode") == 0 )
+ {
+ val = atoi(tmp[1]);
+ control->press_mode = val;
+ }
+ else if ( strcmp(tmp[0], "remove_CoM_vel") == 0 )
+ {
+ val = atoi(tmp[1]);
+ control->remove_CoM_vel = val;
+ }
+ else if ( strcmp(tmp[0], "debug_level") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->debug_level = ival;
+ }
+ else if ( strcmp(tmp[0], "energy_update_freq") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->energy_update_freq = ival;
+ }
+ else if ( strcmp(tmp[0], "write_freq") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->write_steps = ival;
+ }
+ else if ( strcmp(tmp[0], "traj_compress") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->traj_compress = ival;
+
+ if ( out_control->traj_compress )
+ out_control->write = (int (*)(FILE *, const char *, ...)) gzprintf;
+ else out_control->write = fprintf;
+ }
+ else if ( strcmp(tmp[0], "traj_format") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->traj_format = ival;
+
+ if ( out_control->traj_format == 0 )
+ {
+ out_control->write_header =
+ (int (*)( reax_system*, control_params*,
+ static_storage*, void* )) Write_Custom_Header;
+ out_control->append_traj_frame =
+ (int (*)(reax_system*, control_params*, simulation_data*,
+ static_storage*, list **, void*)) Append_Custom_Frame;
+ }
+ else if ( out_control->traj_format == 1 )
+ {
+ out_control->write_header =
+ (int (*)( reax_system*, control_params*,
+ static_storage*, void* )) Write_xyz_Header;
+ out_control->append_traj_frame =
+ (int (*)( reax_system*, control_params*, simulation_data*,
+ static_storage*, list **, void* )) Append_xyz_Frame;
+ }
+ }
+ else if ( strcmp(tmp[0], "traj_title") == 0 )
+ {
+ strcpy( out_control->traj_title, tmp[1] );
+ }
+ else if ( strcmp(tmp[0], "atom_info") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->atom_format += ival * 4;
+ }
+ else if ( strcmp(tmp[0], "atom_velocities") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->atom_format += ival * 2;
+ }
+ else if ( strcmp(tmp[0], "atom_forces") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->atom_format += ival * 1;
+ }
+ else if ( strcmp(tmp[0], "bond_info") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->bond_info = ival;
+ }
+ else if ( strcmp(tmp[0], "angle_info") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ out_control->angle_info = ival;
+ }
+ else if ( strcmp(tmp[0], "test_forces") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ }
+ else if ( strcmp(tmp[0], "molec_anal") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->molec_anal = ival;
+ }
+ else if ( strcmp(tmp[0], "freq_molec_anal") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->freq_molec_anal = ival;
+ }
+ else if ( strcmp(tmp[0], "bond_graph_cutoff") == 0 )
+ {
+ val = atof(tmp[1]);
+ control->bg_cut = val;
+ }
+ else if ( strcmp(tmp[0], "ignore") == 0 )
+ {
+ control->num_ignored = atoi(tmp[1]);
+ for ( i = 0; i < control->num_ignored; ++i )
+ control->ignore[atoi(tmp[i + 2])] = 1;
+ }
+ else if ( strcmp(tmp[0], "dipole_anal") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->dipole_anal = ival;
+ }
+ else if ( strcmp(tmp[0], "freq_dipole_anal") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->freq_dipole_anal = ival;
+ }
+ else if ( strcmp(tmp[0], "diffusion_coef") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->diffusion_coef = ival;
+ }
+ else if ( strcmp(tmp[0], "freq_diffusion_coef") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->freq_diffusion_coef = ival;
+ }
+ else if ( strcmp(tmp[0], "restrict_type") == 0 )
+ {
+ ival = atoi(tmp[1]);
+ control->restrict_type = ival;
+ }
+ else
+ {
+ fprintf( stderr, "WARNING: unknown parameter %s\n", tmp[0] );
+ exit( UNKNOWN_OPTION );
+ }
+ }
+
+ if (ferror(fp))
+ {
+ fprintf(stderr, "Error reading control file. Terminating.\n");
+ exit( INVALID_INPUT );
+ }
+
+ /* determine target T */
+ if ( control->T_mode == 0 )
+ control->T = control->T_final;
+ else control->T = control->T_init;
+
+
+ /* near neighbor and far neighbor cutoffs */
+ control->bo_cut = 0.01 * system->reaxprm.gp.l[29];
+ control->r_low = system->reaxprm.gp.l[11];
+ control->r_cut = system->reaxprm.gp.l[12];
+ control->r_sp_cut = control->r_cut * control->qeq_domain_sparsity;
+ control->vlist_cut += control->r_cut;
+
+ system->g.cell_size = control->vlist_cut / 2.;
+ for ( i = 0; i < 3; ++i )
+ {
+ system->g.spread[i] = 2;
+ }
+
+ /* free memory allocations at the top */
+ for ( i = 0; i < MAX_TOKENS; i++ )
+ {
+ free( tmp[i] );
+ }
+ free( tmp );
+ free( s );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr,
+ "en=%d steps=%d dt=%.5f opt=%d T=%.5f P=%.5f %.5f %.5f\n",
+ control->ensemble, control->nsteps, control->dt, control->tabulate,
+ control->T, control->P[0], control->P[1], control->P[2] );
+
+ fprintf(stderr, "control file read\n" );
+#endif
+
+ return SUCCESS;
+}
diff --git a/PuReMD-GPU/src/control.h b/PuReMD-GPU/src/control.h
new file mode 100644
index 0000000000000000000000000000000000000000..66d0dde7b4901d7a7b42512414328a8e6b256d83
--- /dev/null
+++ b/PuReMD-GPU/src/control.h
@@ -0,0 +1,29 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#ifndef __CONTROL_H_
+#define __CONTROL_H_
+
+#include "mytypes.h"
+
+char Read_Control_File( FILE*, reax_system*, control_params*, output_controls* );
+
+#endif
diff --git a/PuReMD-GPU/src/cuda_forces.cu b/PuReMD-GPU/src/cuda_forces.cu
index bf277b391ce0df0c5336ea0a0653b6863ca14fec..754668c9d9be6601aa9e9a649eb920899e39833c 100644
--- a/PuReMD-GPU/src/cuda_forces.cu
+++ b/PuReMD-GPU/src/cuda_forces.cu
@@ -36,10 +36,10 @@
#include "cuda_three_body_interactions.h"
#include "cuda_four_body_interactions.h"
#include "cuda_list.h"
-#include "cuda_QEq.h"
+#include "cuda_qeq.h"
#include "cuda_reduction.h"
#include "cuda_system_props.h"
-#include "validation.h"
+#include "cuda_validation.h"
#include "cudaProfiler.h"
diff --git a/PuReMD-GPU/src/cuda_init_md.cu b/PuReMD-GPU/src/cuda_init_md.cu
index 1a205506e4c5ff767e02398a3859f838818c1e1a..f0252a2f564a035d7317ea0698712318ee96df5e 100644
--- a/PuReMD-GPU/src/cuda_init_md.cu
+++ b/PuReMD-GPU/src/cuda_init_md.cu
@@ -49,7 +49,7 @@
#include "cuda_reduction.h"
#include "cuda_reset_utils.h"
#include "cuda_system_props.h"
-#include "validation.h"
+#include "cuda_validation.h"
void Cuda_Init_System( reax_system *system, control_params *control,
@@ -116,7 +116,9 @@ void Cuda_Init_Simulation_Data( reax_system *system, control_params *control,
Reset_Simulation_Data( data );
if( !control->restart )
+ {
data->step = data->prev_steps = 0;
+ }
switch( control->ensemble ) {
case NVE:
@@ -124,16 +126,18 @@ void Cuda_Init_Simulation_Data( reax_system *system, control_params *control,
*Evolve = Cuda_Velocity_Verlet_NVE;
break;
-
case NVT:
data->N_f = 3 * system->N + 1;
//control->Tau_T = 100 * data->N_f * K_B * control->T_final;
- if( !control->restart || (control->restart && control->random_vel) ) {
+
+ if( !control->restart || (control->restart && control->random_vel) )
+ {
data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
data->N_f * K_B * control->T );
data->therm.v_xi = data->therm.G_xi * control->dt;
data->therm.v_xi_old = 0;
data->therm.xi = 0;
+
#if defined(DEBUG_FOCUS)
fprintf( stderr, "init_md: G_xi=%f Tau_T=%f E_kin=%f N_f=%f v_xi=%f\n",
data->therm.G_xi, control->Tau_T, data->E_Kin,
@@ -144,12 +148,13 @@ void Cuda_Init_Simulation_Data( reax_system *system, control_params *control,
*Evolve = Cuda_Velocity_Verlet_Nose_Hoover_NVT_Klein;
break;
-
case NPT: // Anisotropic NPT
fprintf( stderr, "THIS OPTION IS NOT YET IMPLEMENTED! TERMINATING...\n" );
exit( UNKNOWN_OPTION );
data->N_f = 3 * system->N + 9;
- if( !control->restart ) {
+
+ if( !control->restart )
+ {
data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
data->N_f * K_B * control->T );
data->therm.v_xi = data->therm.G_xi * control->dt;
@@ -160,7 +165,6 @@ void Cuda_Init_Simulation_Data( reax_system *system, control_params *control,
*Evolve = Velocity_Verlet_Berendsen_Isotropic_NPT;
break;
-
case sNPT: // Semi-Isotropic NPT
fprintf( stderr, "THIS OPTION IS NOT YET IMPLEMENTED! TERMINATING...\n" );
exit( UNKNOWN_OPTION );
@@ -168,7 +172,6 @@ void Cuda_Init_Simulation_Data( reax_system *system, control_params *control,
*Evolve = Velocity_Verlet_Berendsen_SemiIsotropic_NPT;
break;
-
case iNPT: // Isotropic NPT
fprintf( stderr, "THIS OPTION IS NOT YET IMPLEMENTED! TERMINATING...\n" );
exit( UNKNOWN_OPTION );
diff --git a/PuReMD-GPU/src/cuda_integrate.cu b/PuReMD-GPU/src/cuda_integrate.cu
index cba0b79c39b4f9b66e5b506d11dcffb81adc488d..ab4d203139e4a8235d9035ab566fbdca7ded5c82 100644
--- a/PuReMD-GPU/src/cuda_integrate.cu
+++ b/PuReMD-GPU/src/cuda_integrate.cu
@@ -36,10 +36,10 @@
#include "cuda_forces.h"
#include "cuda_grid.h"
#include "cuda_neighbors.h"
-#include "cuda_QEq.h"
+#include "cuda_qeq.h"
#include "cuda_reset_utils.h"
#include "cuda_system_props.h"
-#include "validation.h"
+#include "cuda_validation.h"
GLOBAL void Cuda_Velocity_Verlet_NVE_atoms1 (reax_atom *atoms,
diff --git a/PuReMD-GPU/src/cuda_neighbors.cu b/PuReMD-GPU/src/cuda_neighbors.cu
index 876b6b9913e4d825e0cc8be5a2fc1d092c56d9f8..5cfe03dea5f5c314a7dfb60ef3d4df2e6c8f61ae 100644
--- a/PuReMD-GPU/src/cuda_neighbors.cu
+++ b/PuReMD-GPU/src/cuda_neighbors.cu
@@ -265,11 +265,10 @@ GLOBAL void k_Generate_Neighbor_Lists ( reax_atom *sys_atoms,
nbr_atoms = &(atoms [index_grid_atoms (x, y, z, 0, &g) ]);
max = top [index_grid_3d(x, y, z, &g)];
- for (m = 0; m < max; m++) {
+ for (m = 0; m < max; m++)
+ {
atom2 = nbr_atoms[m];
- //nbr_data = & ( far_nbrs.select.far_nbr_list[atom1 * g.max_cuda_nbrs + num_far] );
-
//CHANGE ORIGINAL
/*
if (atom1 > atom2) {
diff --git a/PuReMD-GPU/src/cuda_QEq.cu b/PuReMD-GPU/src/cuda_qeq.cu
similarity index 99%
rename from PuReMD-GPU/src/cuda_QEq.cu
rename to PuReMD-GPU/src/cuda_qeq.cu
index 033945338aa76aa4c02909f90d2ca3eb1dacad58..21c2e334be0d6ec46d0a42f4c4f8a44dfc0ebffc 100644
--- a/PuReMD-GPU/src/cuda_QEq.cu
+++ b/PuReMD-GPU/src/cuda_qeq.cu
@@ -18,14 +18,15 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#include "cuda_QEq.h"
+#include "cuda_qeq.h"
-#include "QEq.h"
+#include "qeq.h"
#include "allocate.h"
#include "lin_alg.h"
#include "list.h"
#include "print_utils.h"
#include "index_utils.h"
+#include "sort.h"
#include "system_props.h"
#include "cuda_copy.h"
@@ -33,9 +34,7 @@
#include "cuda_utils.h"
#include "cuda_lin_alg.h"
#include "cuda_reduction.h"
-
-#include "sort.h"
-#include "validation.h"
+#include "cuda_validation.h"
GLOBAL void Cuda_Sort_Matrix_Rows( sparse_matrix A )
diff --git a/PuReMD-GPU/src/cuda_QEq.h b/PuReMD-GPU/src/cuda_qeq.h
similarity index 100%
rename from PuReMD-GPU/src/cuda_QEq.h
rename to PuReMD-GPU/src/cuda_qeq.h
diff --git a/PuReMD-GPU/src/cuda_utils.cu b/PuReMD-GPU/src/cuda_utils.cu
index 1efcf28aa432f563749e49c68c67cc6b132e711e..6867857a4d58771f571fbb6810efbf926677f80f 100644
--- a/PuReMD-GPU/src/cuda_utils.cu
+++ b/PuReMD-GPU/src/cuda_utils.cu
@@ -29,7 +29,7 @@ cusparseMatDescr_t matdescriptor;
void cuda_malloc( void **ptr, int size, int memset, int err_code )
{
- cudaError_t retVal = cudaSuccess;
+ cudaError_t retVal;
//fprintf (stderr, "&ptr --. %ld \n", &ptr);
//fprintf (stderr, "ptr --> %ld \n", ptr );
@@ -45,7 +45,8 @@ void cuda_malloc( void **ptr, int size, int memset, int err_code )
//fprintf (stderr, "&ptr --. %ld \n", &ptr);
//fprintf (stderr, "ptr --> %ld \n", ptr );
- if ( memset ) {
+ if ( memset )
+ {
retVal = cudaMemset( *ptr, 0, size );
if ( retVal != cudaSuccess )
{
@@ -59,8 +60,12 @@ void cuda_malloc( void **ptr, int size, int memset, int err_code )
void cuda_free( void *ptr, int err_code )
{
- cudaError_t retVal = cudaSuccess;
- if (!ptr) return;
+ cudaError_t retVal;
+
+ if (!ptr)
+ {
+ return;
+ }
retVal = cudaFree( ptr );
@@ -75,9 +80,10 @@ void cuda_free( void *ptr, int err_code )
void cuda_memset( void *ptr, int data, size_t count, int err_code )
{
- cudaError_t retVal = cudaSuccess;
+ cudaError_t retVal;
retVal = cudaMemset( ptr, data, count );
+
if (retVal != cudaSuccess) {
fprintf( stderr, "ptr passed is %ld, value: %ld \n", ptr, &ptr );
fprintf( stderr, " size to memset: %d \n", count );
@@ -91,7 +97,7 @@ void cuda_memset( void *ptr, int data, size_t count, int err_code )
void copy_host_device( void *host, void *dev, int size, enum cudaMemcpyKind dir, int resid )
{
- cudaError_t retVal = cudaErrorNotReady;
+ cudaError_t retVal;
if ( dir == cudaMemcpyHostToDevice )
{
@@ -112,9 +118,10 @@ void copy_host_device( void *host, void *dev, int size, enum cudaMemcpyKind dir,
void copy_device( void *dest, void *src, int size, int resid )
{
- cudaError_t retVal = cudaErrorNotReady;
+ cudaError_t retVal;
retVal = cudaMemcpy( dest, src, size, cudaMemcpyDeviceToDevice );
+
if ( retVal != cudaSuccess )
{
fprintf( stderr, "could not copy resource %d from host to device: reason %d \n",
@@ -134,6 +141,7 @@ void compute_blocks( int *blocks, int *block_size, int count )
void compute_nearest_pow_2( int blocks, int *result )
{
int power = 1;
+
while (power < blocks)
{
power *= 2;
@@ -146,7 +154,9 @@ void compute_nearest_pow_2( int blocks, int *result )
void print_device_mem_usage( )
{
size_t total, free;
+
cudaMemGetInfo( &free, &total );
+
if ( cudaGetLastError() != cudaSuccess )
{
fprintf( stderr, "Error on the memory call \n" );
diff --git a/PuReMD-GPU/src/validation.cu b/PuReMD-GPU/src/cuda_validation.cu
similarity index 99%
rename from PuReMD-GPU/src/validation.cu
rename to PuReMD-GPU/src/cuda_validation.cu
index 21cd2145e689621ee0b3827889b106ed7c05af7f..b5348eba0871017a7225fc311e98d95531136792 100644
--- a/PuReMD-GPU/src/validation.cu
+++ b/PuReMD-GPU/src/cuda_validation.cu
@@ -18,13 +18,13 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#include "validation.h"
+#include "cuda_validation.h"
-#include "cuda_utils.h"
+#include "index_utils.h"
#include "list.h"
-
#include "sort.h"
-#include "index_utils.h"
+
+#include "cuda_utils.h"
int check_zero (real p1, real p2)
diff --git a/PuReMD-GPU/src/validation.h b/PuReMD-GPU/src/cuda_validation.h
similarity index 100%
rename from PuReMD-GPU/src/validation.h
rename to PuReMD-GPU/src/cuda_validation.h
diff --git a/PuReMD-GPU/src/param.c b/PuReMD-GPU/src/ffield.c
similarity index 58%
rename from PuReMD-GPU/src/param.c
rename to PuReMD-GPU/src/ffield.c
index 42e9ef612ec6c81d593f74acd796a7f564f96896..a5377e6f2b75dc6c6bec586d36517127bca40d50 100644
--- a/PuReMD-GPU/src/param.c
+++ b/PuReMD-GPU/src/ffield.c
@@ -1,9 +1,10 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
@@ -18,85 +19,10 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#include "param.h"
-#include "traj.h"
-#include "ctype.h"
-
-
-int Get_Atom_Type( reax_interaction *reaxprm, char *s )
-{
- int i;
-
- for ( i = 0; i < reaxprm->num_atom_types; ++i )
- if ( !strcmp( reaxprm->sbp[i].name, s ) )
- return i;
-
- fprintf( stderr, "Unknown atom type %s. Terminating...\n", s );
- exit( UNKNOWN_ATOM_TYPE_ERR );
-}
-
-
-int Tokenize(char* s, char*** tok)
-{
- char test[MAX_LINE];
- char *sep = "\t \n!=";
- char *word;
- int count = 0;
-
- strncpy( test, s, MAX_LINE );
-
- // fprintf( stderr, "|%s|\n", test );
-
- for ( word = strtok(test, sep); word; word = strtok(NULL, sep) )
- {
- strncpy( (*tok)[count], word, MAX_LINE );
- count++;
- }
-
- return count;
-}
-
-
-/* Initialize Taper params */
-void Init_Taper( control_params *control )
-{
- real d1, d7;
- real swa, swa2, swa3;
- real swb, swb2, swb3;
-
- swa = control->r_low;
- swb = control->r_cut;
-
- if ( fabs( swa ) > 0.01 )
- fprintf( stderr, "Warning: non-zero value for lower Taper-radius cutoff\n" );
-
- if ( swb < 0 )
- {
- fprintf( stderr, "Negative value for upper Taper-radius cutoff\n" );
- exit( INVALID_INPUT );
- }
- else if ( swb < 5 )
- fprintf( stderr, "Warning: low value for upper Taper-radius cutoff:%f\n",
- swb );
-
- d1 = swb - swa;
- d7 = POW( d1, 7.0 );
- swa2 = SQR( swa );
- swa3 = CUBE( swa );
- swb2 = SQR( swb );
- swb3 = CUBE( swb );
-
- control->Tap7 = 20.0 / d7;
- control->Tap6 = -70.0 * (swa + swb) / d7;
- control->Tap5 = 84.0 * (swa2 + 3.0 * swa * swb + swb2) / d7;
- control->Tap4 = -35.0 * (swa3 + 9.0 * swa2 * swb + 9.0 * swa * swb2 + swb3 ) / d7;
- control->Tap3 = 140.0 * (swa3 * swb + 3.0 * swa2 * swb2 + swa * swb3 ) / d7;
- control->Tap2 = -210.0 * (swa3 * swb2 + swa2 * swb3) / d7;
- control->Tap1 = 140.0 * swa3 * swb3 / d7;
- control->Tap0 = (-35.0 * swa3 * swb2 * swb2 + 21.0 * swa2 * swb3 * swb2 +
- 7.0 * swa * swb3 * swb3 + swb3 * swb3 * swb ) / d7;
-}
+#include <ctype.h>
+#include "ffield.h"
+#include "tool_box.h"
char Read_Force_Field( FILE* fp, reax_interaction* reax )
@@ -106,20 +32,20 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
char *tor_flag;
int c, i, j, k, l, m, n, o, p, cnt;
real val;
-
int __N;
int index1, index2;
s = (char*) malloc(sizeof(char) * MAX_LINE);
tmp = (char**) malloc(sizeof(char*)*MAX_TOKENS);
for (i = 0; i < MAX_TOKENS; i++)
+ {
tmp[i] = (char*) malloc(sizeof(char) * MAX_TOKEN_LEN);
+ }
/* reading first header comment */
fgets( s, MAX_LINE, fp );
-
/* line 2 is number of global parameters */
fgets( s, MAX_LINE, fp );
c = Tokenize( s, &tmp );
@@ -129,7 +55,7 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
if (n < 1)
{
fprintf( stderr, "WARNING: number of globals in ffield file is 0!\n" );
- return 1;
+ exit( INVALID_INPUT );
}
reax->gp.n_global = n;
@@ -146,20 +72,17 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
reax->gp.l[i] = val;
}
-
/* next line is number of atom types and some comments */
fgets( s, MAX_LINE, fp );
c = Tokenize( s, &tmp );
reax->num_atom_types = atoi(tmp[0]);
__N = reax->num_atom_types;
-
/* 3 lines of comments */
fgets(s, MAX_LINE, fp);
fgets(s, MAX_LINE, fp);
fgets(s, MAX_LINE, fp);
-
/* Allocating structures in reax_interaction */
reax->sbp = (single_body_parameters*)
calloc( reax->num_atom_types, sizeof(single_body_parameters) );
@@ -194,7 +117,9 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
c = Tokenize( s, &tmp );
for ( j = 0; j < strlen( tmp[0] ); ++j )
+ {
reax->sbp[i].name[j] = toupper( tmp[0][j] );
+ }
val = atof(tmp[1]);
reax->sbp[i].r_s = val;
@@ -281,6 +206,7 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
if ( reax->sbp[i].gamma_w > 0.5 ) // Shielding vdWaals
{
if ( reax->gp.vdw_type != 0 && reax->gp.vdw_type != 3 )
+ {
fprintf( stderr, "Warning: inconsistent vdWaals-parameters\n" \
"Force field parameters for element %s\n" \
"indicate inner wall+shielding, but earlier\n" \
@@ -288,9 +214,11 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
"This may cause division-by-zero errors.\n" \
"Keeping vdWaals-setting for earlier atoms.\n",
reax->sbp[i].name );
+ }
else
{
reax->gp.vdw_type = 3;
+
#if defined(DEBUG)
fprintf( stderr, "vdWaals type for element %s: Shielding+inner-wall",
reax->sbp[i].name );
@@ -300,6 +228,7 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
else // No shielding vdWaals parameters present
{
if ( reax->gp.vdw_type != 0 && reax->gp.vdw_type != 2 )
+ {
fprintf( stderr, "Warning: inconsistent vdWaals-parameters\n" \
"Force field parameters for element %s\n" \
"indicate inner wall without shielding, but earlier\n" \
@@ -307,9 +236,11 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
"This may cause division-by-zero errors.\n" \
"Keeping vdWaals-setting for earlier atoms.\n",
reax->sbp[i].name );
+ }
else
{
reax->gp.vdw_type = 2;
+
#if defined(DEBUG)
fprintf( stderr, "vdWaals type for element%s: No Shielding,inner-wall",
reax->sbp[i].name );
@@ -348,7 +279,6 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
}
}
-
/* next line is number of two body combination and some comments */
fgets(s, MAX_LINE, fp);
c = Tokenize(s, &tmp);
@@ -430,6 +360,7 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
/* calculating combination rules and filling up remaining fields. */
for (i = 0; i < reax->num_atom_types; i++)
+ {
for (j = i; j < reax->num_atom_types; j++)
{
index1 = i * __N + j;
@@ -450,7 +381,6 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
reax->tbp[index2].r_pp = 0.5 *
(reax->sbp[j].r_pi_pi + reax->sbp[i].r_pi_pi);
-
reax->tbp[index1].p_boc3 =
sqrt(reax->sbp[i].b_o_132 *
reax->sbp[j].b_o_132);
@@ -472,7 +402,6 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
sqrt(reax->sbp[j].b_o_133 *
reax->sbp[i].b_o_133);
-
reax->tbp[index1].D =
sqrt(reax->sbp[i].epsilon *
reax->sbp[j].epsilon);
@@ -505,9 +434,8 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
reax->tbp[index2].gamma =
POW(reax->sbp[j].gamma *
reax->sbp[i].gamma, -1.5);
-
}
-
+ }
/* next line is number of 2-body offdiagonal combinations and some comments */
/* these are two body offdiagonal terms that are different from the
@@ -572,7 +500,6 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
}
}
-
/* 3-body parameters -
supports multi-well potentials (upto MAX_3BODY_PARAM in mytypes.h) */
/* clear entries first */
@@ -636,7 +563,6 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
}
}
-
/* 4-body parameters are entered in compact form. i.e. 0-X-Y-0
correspond to any type of pair of atoms in 1 and 4
position. However, explicit X-Y-Z-W takes precedence over the
@@ -647,13 +573,19 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
/* clear all entries first */
for ( i = 0; i < reax->num_atom_types; ++i )
+ {
for ( j = 0; j < reax->num_atom_types; ++j )
+ {
for ( k = 0; k < reax->num_atom_types; ++k )
+ {
for ( m = 0; m < reax->num_atom_types; ++m )
{
reax->fbp[i * __N * __N * __N + j * __N * __N + k * __N + m].cnt = 0;
tor_flag[i * __N * __N * __N + j * __N * __N + k * __N + m] = 0;
}
+ }
+ }
+ }
/* next line is number of 4-body params and some comments */
fgets( s, MAX_LINE, fp );
@@ -714,7 +646,9 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
else /* This means the entry is of the form 0-X-Y-0 */
{
if ( k < reax->num_atom_types && m < reax->num_atom_types )
+ {
for ( p = 0; p < reax->num_atom_types; p++ )
+ {
for ( o = 0; o < reax->num_atom_types; o++ )
{
index1 = p * __N * __N * __N + k * __N * __N + m * __N + o;
@@ -743,11 +677,12 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
reax->fbp[index2].prm[0].p_cot1 = atof(tmp[8]);
}
}
+ }
+ }
}
}
-
/* next line is number of hydrogen bond params and some comments */
fgets( s, MAX_LINE, fp );
c = Tokenize( s, &tmp );
@@ -781,14 +716,14 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
}
}
-
/* deallocate helper storage */
for ( i = 0; i < MAX_TOKENS; i++ )
+ {
free( tmp[i] );
+ }
free( tmp );
free( s );
-
/* deallocate tor_flag */
free( tor_flag );
@@ -796,497 +731,5 @@ char Read_Force_Field( FILE* fp, reax_interaction* reax )
fprintf( stderr, "force field read\n" );
#endif
- return 0;
-}
-
-
-char Read_Control_File( FILE* fp, reax_system *system, control_params* control,
- output_controls *out_control )
-{
- char *s, **tmp;
- int c, i;
- real val;
- int ival;
-
- /* assign default values */
- strcpy( control->sim_name, "default.sim" );
-
- control->restart = 0;
- out_control->restart_format = 1;
- out_control->restart_freq = 0;
- strcpy( control->restart_from, "default.res" );
- out_control->restart_freq = 0;
- control->random_vel = 0;
-
- control->reposition_atoms = 0;
-
- control->ensemble = 0;
- control->nsteps = 0;
- control->dt = 0.25;
-
- control->geo_format = 1;
- control->restrict_bonds = 0;
-
- control->periodic_boundaries = 1;
- control->periodic_images[0] = 0;
- control->periodic_images[1] = 0;
- control->periodic_images[2] = 0;
-
- control->reneighbor = 1;
- control->vlist_cut = 0;
- control->nbr_cut = 4.;
- control->r_cut = 10;
- control->max_far_nbrs = 1000;
- control->bo_cut = 0.01;
- control->thb_cut = 0.001;
- control->hb_cut = 7.50;
-
- control->q_err = 0.000001;
- control->tabulate = 0;
- //TODO
- control->refactor = 100;
- //TODO -- change this to 5.
-
- control->droptol = 0.01;
-
- control->T_init = 0.;
- control->T_final = 300.;
- control->Tau_T = 1.0;
- control->T_mode = 0.;
- control->T_rate = 1.;
- control->T_freq = 1.;
-
- control->P[0] = 0.000101325;
- control->P[1] = 0.000101325;
- control->P[2] = 0.000101325;
- control->Tau_P[0] = 500.0;
- control->Tau_P[1] = 500.0;
- control->Tau_P[2] = 500.0;
- control->Tau_PT = 500.0;
- control->compressibility = 1.0;
- control->press_mode = 0;
-
- control->remove_CoM_vel = 25;
-
- out_control->debug_level = 0;
- out_control->energy_update_freq = 10;
-
- out_control->write_steps = 100;
- out_control->traj_compress = 0;
- out_control->write = fprintf;
- out_control->traj_format = 0;
- out_control->write_header =
- (int (*)( reax_system*, control_params*,
- static_storage*, void* )) Write_Custom_Header;
- out_control->append_traj_frame =
- (int (*)( reax_system*, control_params*, simulation_data*,
- static_storage*, list **, void* )) Append_Custom_Frame;
-
- strcpy( out_control->traj_title, "default_title" );
- out_control->atom_format = 0;
- out_control->bond_info = 0;
- out_control->angle_info = 0;
-
- control->molec_anal = 0;
- control->freq_molec_anal = 0;
- control->bg_cut = 0.3;
- control->num_ignored = 0;
- memset( control->ignore, 0, sizeof(int)*MAX_ATOM_TYPES );
-
- control->dipole_anal = 0;
- control->freq_dipole_anal = 0;
-
- control->diffusion_coef = 0;
- control->freq_diffusion_coef = 0;
- control->restrict_type = 0;
-
- /* memory allocations */
- s = (char*) malloc(sizeof(char) * MAX_LINE);
- tmp = (char**) malloc(sizeof(char*)*MAX_TOKENS);
- for (i = 0; i < MAX_TOKENS; i++)
- tmp[i] = (char*) malloc(sizeof(char) * MAX_LINE);
-
- /* read control parameters file */
- while (!feof(fp))
- {
- fgets(s, MAX_LINE, fp);
- c = Tokenize(s, &tmp);
-
- if ( strcmp(tmp[0], "simulation_name") == 0 )
- {
- strcpy( control->sim_name, tmp[1] );
- }
- //else if( strcmp(tmp[0], "restart") == 0 ) {
- // ival = atoi(tmp[1]);
- // control->restart = ival;
- //}
- else if ( strcmp(tmp[0], "restart_format") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->restart_format = ival;
- }
- else if ( strcmp(tmp[0], "restart_freq") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->restart_freq = ival;
- }
- else if ( strcmp(tmp[0], "random_vel") == 0 )
- {
- ival = atoi(tmp[1]);
- control->random_vel = ival;
- }
- else if ( strcmp(tmp[0], "reposition_atoms") == 0 )
- {
- ival = atoi(tmp[1]);
- control->reposition_atoms = ival;
- }
- else if ( strcmp(tmp[0], "ensemble_type") == 0 )
- {
- ival = atoi(tmp[1]);
- control->ensemble = ival;
- }
- else if ( strcmp(tmp[0], "nsteps") == 0 )
- {
- ival = atoi(tmp[1]);
- control->nsteps = ival;
- }
- else if ( strcmp(tmp[0], "dt") == 0 )
- {
- val = atof(tmp[1]);
- control->dt = val * 1.e-3; // convert dt from fs to ps!
- }
- else if ( strcmp(tmp[0], "periodic_boundaries") == 0 )
- {
- ival = atoi( tmp[1] );
- control->periodic_boundaries = ival;
- }
- else if ( strcmp(tmp[0], "periodic_images") == 0 )
- {
- ival = atoi(tmp[1]);
- control->periodic_images[0] = ival;
- ival = atoi(tmp[2]);
- control->periodic_images[1] = ival;
- ival = atoi(tmp[3]);
- control->periodic_images[2] = ival;
- }
- else if ( strcmp(tmp[0], "geo_format") == 0 )
- {
- ival = atoi( tmp[1] );
- control->geo_format = ival;
- }
- else if ( strcmp(tmp[0], "restrict_bonds") == 0 )
- {
- ival = atoi( tmp[1] );
- control->restrict_bonds = ival;
- }
- else if ( strcmp(tmp[0], "tabulate_long_range") == 0 )
- {
- ival = atoi( tmp[1] );
- control->tabulate = ival;
- }
- else if ( strcmp(tmp[0], "reneighbor") == 0 )
- {
- ival = atoi( tmp[1] );
- control->reneighbor = ival;
- }
- else if ( strcmp(tmp[0], "vlist_buffer") == 0 )
- {
- val = atof(tmp[1]);
- control->vlist_cut = val;
- }
- else if ( strcmp(tmp[0], "nbrhood_cutoff") == 0 )
- {
- val = atof(tmp[1]);
- control->nbr_cut = val;
- }
- else if ( strcmp(tmp[0], "thb_cutoff") == 0 )
- {
- val = atof(tmp[1]);
- control->thb_cut = val;
- }
- else if ( strcmp(tmp[0], "hbond_cutoff") == 0 )
- {
- val = atof( tmp[1] );
- control->hb_cut = val;
- }
- else if ( strcmp(tmp[0], "q_err") == 0 )
- {
- val = atof( tmp[1] );
- control->q_err = val;
- }
- else if ( strcmp(tmp[0], "ilu_refactor") == 0 )
- {
- ival = atoi( tmp[1] );
- control->refactor = ival;
- }
- else if ( strcmp(tmp[0], "ilu_droptol") == 0 )
- {
- val = atof( tmp[1] );
- control->droptol = val;
- }
- else if ( strcmp(tmp[0], "temp_init") == 0 )
- {
- val = atof(tmp[1]);
- control->T_init = val;
-
- if ( control->T_init < 0.001 )
- control->T_init = 0.001;
- }
- else if ( strcmp(tmp[0], "temp_final") == 0 )
- {
- val = atof(tmp[1]);
- control->T_final = val;
-
- if ( control->T_final < 0.1 )
- control->T_final = 0.1;
- }
- else if ( strcmp(tmp[0], "t_mass") == 0 )
- {
- val = atof(tmp[1]);
- control->Tau_T = val * 1.e-3; // convert t_mass from fs to ps
- }
- else if ( strcmp(tmp[0], "t_mode") == 0 )
- {
- ival = atoi(tmp[1]);
- control->T_mode = ival;
- }
- else if ( strcmp(tmp[0], "t_rate") == 0 )
- {
- val = atof(tmp[1]);
- control->T_rate = val;
- }
- else if ( strcmp(tmp[0], "t_freq") == 0 )
- {
- val = atof(tmp[1]);
- control->T_freq = val;
- }
- else if ( strcmp(tmp[0], "pressure") == 0 )
- {
- if ( control->ensemble == iNPT )
- {
- val = atof(tmp[1]);
- control->P[0] = control->P[1] = control->P[2] = val;
- }
- else if ( control->ensemble == sNPT )
- {
- val = atof(tmp[1]);
- control->P[0] = val;
-
- val = atof(tmp[2]);
- control->P[1] = val;
-
- val = atof(tmp[3]);
- control->P[2] = val;
- }
- }
- else if ( strcmp(tmp[0], "p_mass") == 0 )
- {
- if ( control->ensemble == iNPT )
- {
- val = atof(tmp[1]);
- control->Tau_P[0] = val * 1.e-3; // convert p_mass from fs to ps
- }
- else if ( control->ensemble == sNPT )
- {
- val = atof(tmp[1]);
- control->Tau_P[0] = val * 1.e-3; // convert p_mass from fs to ps
-
- val = atof(tmp[2]);
- control->Tau_P[1] = val * 1.e-3; // convert p_mass from fs to ps
-
- val = atof(tmp[3]);
- control->Tau_P[2] = val * 1.e-3; // convert p_mass from fs to ps
- }
- }
- else if ( strcmp(tmp[0], "pt_mass") == 0 )
- {
- val = atof(tmp[1]);
- control->Tau_PT = val * 1.e-3; // convert pt_mass from fs to ps
- }
- else if ( strcmp(tmp[0], "compress") == 0 )
- {
- val = atof(tmp[1]);
- control->compressibility = val;
- }
- else if ( strcmp(tmp[0], "press_mode") == 0 )
- {
- val = atoi(tmp[1]);
- control->press_mode = val;
- }
- else if ( strcmp(tmp[0], "remove_CoM_vel") == 0 )
- {
- val = atoi(tmp[1]);
- control->remove_CoM_vel = val;
- }
- else if ( strcmp(tmp[0], "debug_level") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->debug_level = ival;
- }
- else if ( strcmp(tmp[0], "energy_update_freq") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->energy_update_freq = ival;
- }
- else if ( strcmp(tmp[0], "write_freq") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->write_steps = ival;
- }
- else if ( strcmp(tmp[0], "traj_compress") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->traj_compress = ival;
-
- if ( out_control->traj_compress )
- out_control->write = (int (*)(FILE *, const char *, ...)) gzprintf;
- else out_control->write = fprintf;
- }
- else if ( strcmp(tmp[0], "traj_format") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->traj_format = ival;
-
- if ( out_control->traj_format == 0 )
- {
- out_control->write_header =
- (int (*)( reax_system*, control_params*,
- static_storage*, void* )) Write_Custom_Header;
- out_control->append_traj_frame =
- (int (*)(reax_system*, control_params*, simulation_data*,
- static_storage*, list **, void*)) Append_Custom_Frame;
- }
- else if ( out_control->traj_format == 1 )
- {
- out_control->write_header =
- (int (*)( reax_system*, control_params*,
- static_storage*, void* )) Write_xyz_Header;
- out_control->append_traj_frame =
- (int (*)( reax_system*, control_params*, simulation_data*,
- static_storage*, list **, void* )) Append_xyz_Frame;
- }
- }
- else if ( strcmp(tmp[0], "traj_title") == 0 )
- {
- strcpy( out_control->traj_title, tmp[1] );
- }
- else if ( strcmp(tmp[0], "atom_info") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->atom_format += ival * 4;
- }
- else if ( strcmp(tmp[0], "atom_velocities") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->atom_format += ival * 2;
- }
- else if ( strcmp(tmp[0], "atom_forces") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->atom_format += ival * 1;
- }
- else if ( strcmp(tmp[0], "bond_info") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->bond_info = ival;
- }
- else if ( strcmp(tmp[0], "angle_info") == 0 )
- {
- ival = atoi(tmp[1]);
- out_control->angle_info = ival;
- }
- else if ( strcmp(tmp[0], "test_forces") == 0 )
- {
- ival = atoi(tmp[1]);
- }
- else if ( strcmp(tmp[0], "molec_anal") == 0 )
- {
- ival = atoi(tmp[1]);
- control->molec_anal = ival;
- }
- else if ( strcmp(tmp[0], "freq_molec_anal") == 0 )
- {
- ival = atoi(tmp[1]);
- control->freq_molec_anal = ival;
- }
- else if ( strcmp(tmp[0], "bond_graph_cutoff") == 0 )
- {
- val = atof(tmp[1]);
- control->bg_cut = val;
- }
- else if ( strcmp(tmp[0], "ignore") == 0 )
- {
- control->num_ignored = atoi(tmp[1]);
- for ( i = 0; i < control->num_ignored; ++i )
- control->ignore[atoi(tmp[i + 2])] = 1;
- }
- else if ( strcmp(tmp[0], "dipole_anal") == 0 )
- {
- ival = atoi(tmp[1]);
- control->dipole_anal = ival;
- }
- else if ( strcmp(tmp[0], "freq_dipole_anal") == 0 )
- {
- ival = atoi(tmp[1]);
- control->freq_dipole_anal = ival;
- }
- else if ( strcmp(tmp[0], "diffusion_coef") == 0 )
- {
- ival = atoi(tmp[1]);
- control->diffusion_coef = ival;
- }
- else if ( strcmp(tmp[0], "freq_diffusion_coef") == 0 )
- {
- ival = atoi(tmp[1]);
- control->freq_diffusion_coef = ival;
- }
- else if ( strcmp(tmp[0], "restrict_type") == 0 )
- {
- ival = atoi(tmp[1]);
- control->restrict_type = ival;
- }
- else
- {
- fprintf( stderr, "WARNING: unknown parameter %s\n", tmp[0] );
- exit( 15 );
- }
- }
-
-
- /* determine target T */
- if ( control->T_mode == 0 )
- control->T = control->T_final;
- else control->T = control->T_init;
-
-
- /* near neighbor and far neighbor cutoffs */
- control->bo_cut = 0.01 * system->reaxprm.gp.l[29];
- control->r_low = system->reaxprm.gp.l[11];
- control->r_cut = system->reaxprm.gp.l[12];
- control->vlist_cut += control->r_cut;
-
- system->g.cell_size = control->vlist_cut / 2.;
- for ( i = 0; i < 3; ++i )
- system->g.spread[i] = 2;
-
-
- /* Initialize Taper function */
- Init_Taper( control );
-
-
- /* free memory allocations at the top */
- for ( i = 0; i < MAX_TOKENS; i++ )
- free( tmp[i] );
- free( tmp );
- free( s );
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr,
- "en=%d steps=%d dt=%.5f opt=%d T=%.5f P=%.5f %.5f %.5f\n",
- control->ensemble, control->nsteps, control->dt, control->tabulate,
- control->T, control->P[0], control->P[1], control->P[2] );
-
- fprintf(stderr, "control file read\n" );
-#endif
- return 0;
+ return SUCCESS;
}
diff --git a/PuReMD-GPU/src/param.h b/PuReMD-GPU/src/ffield.h
similarity index 67%
rename from PuReMD-GPU/src/param.h
rename to PuReMD-GPU/src/ffield.h
index 2b24b056983233840966a8de29ce902ca6beb981..4aaf32a644861b069e8cf87e2eec68aadf4d3c84 100644
--- a/PuReMD-GPU/src/param.h
+++ b/PuReMD-GPU/src/ffield.h
@@ -1,9 +1,10 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
@@ -18,24 +19,10 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#ifndef __PARAM_H_
-#define __PARAM_H_
+#ifndef __FFIELD_H_
+#define __FFIELD_H_
#include "mytypes.h"
-
-#define MAX_LINE 1024
-#define MAX_TOKENS 20
-#define MAX_TOKEN_LEN 1024
-
-
-int Get_Atom_Type( reax_interaction*, char* );
-
-int Tokenize( char*, char*** );
-
char Read_Force_Field( FILE*, reax_interaction* );
-char Read_Control_File( FILE*, reax_system*, control_params*,
- output_controls* );
-
-
#endif
diff --git a/PuReMD-GPU/src/forces.c b/PuReMD-GPU/src/forces.c
index c95d4896e32f60e954d79b0b623520afb042e9ea..debe6ac171f281a9f09a77c2c17ec1d315bc7f9b 100644
--- a/PuReMD-GPU/src/forces.c
+++ b/PuReMD-GPU/src/forces.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -26,30 +27,31 @@
#include "two_body_interactions.h"
#include "three_body_interactions.h"
#include "four_body_interactions.h"
+#include "index_utils.h"
#include "list.h"
#include "print_utils.h"
+#include "qeq.h"
#include "system_props.h"
-#include "QEq.h"
+#include "tool_box.h"
#include "vector.h"
-#include "index_utils.h"
-void Dummy_Interaction( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+void Dummy_Interaction( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
}
void Init_Bonded_Force_Functions( control_params *control )
-{
+{
Interaction_Functions[0] = Calculate_Bond_Orders;
Interaction_Functions[1] = Bond_Energy; //*/Dummy_Interaction;
Interaction_Functions[2] = LonePair_OverUnder_Coordination_Energy;
//*/Dummy_Interaction;
Interaction_Functions[3] = Three_Body_Interactions; //*/Dummy_Interaction;
Interaction_Functions[4] = Four_Body_Interactions; //*/Dummy_Interaction;
- if( control->hb_cut > 0 )
+ if ( control->hb_cut > 0 )
Interaction_Functions[5] = Hydrogen_Bonds; //*/Dummy_Interaction;
else Interaction_Functions[5] = Dummy_Interaction;
Interaction_Functions[6] = Dummy_Interaction; //empty
@@ -59,127 +61,123 @@ void Init_Bonded_Force_Functions( control_params *control )
}
-void Compute_Bonded_Forces( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+void Compute_Bonded_Forces( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i;
- real t_start, t_elapsed;
+ // real t_start, t_end, t_elapsed;
#ifdef TEST_ENERGY
/* Mark beginning of a new timestep in each energy file */
- fprintf( out_control->ebond, "step: %d\n%6s%6s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "bo", "ebond", "total" );
- fprintf( out_control->elp, "step: %d\n%6s%12s%12s%12s\n",
- data->step, "atom", "nlp", "elp", "total" );
- fprintf( out_control->eov, "step: %d\n%6s%12s%12s\n",
- data->step, "atom", "eov", "total" );
- fprintf( out_control->eun, "step: %d\n%6s%12s%12s\n",
- data->step, "atom", "eun", "total" );
- fprintf( out_control->eval, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3",
- "angle", "bo(12)", "bo(23)", "eval", "epen", "total" );
- fprintf( out_control->epen, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3",
- "angle", "bo(12)", "bo(23)", "epen", "total" );
- fprintf( out_control->ecoa, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3",
- "angle", "bo(12)", "bo(23)", "ecoa", "total" );
- fprintf( out_control->ehb, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3",
- "r(23)", "angle", "bo(12)", "ehb", "total" );
- fprintf( out_control->etor, "step: %d\n%6s%6s%6s%6s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3", "atom4",
- "phi", "bo(23)", "etor", "total" );
+ fprintf( out_control->ebond, "step: %d\n%6s%6s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "bo", "ebond", "total" );
+ fprintf( out_control->elp, "step: %d\n%6s%12s%12s%12s\n",
+ data->step, "atom", "nlp", "elp", "total" );
+ fprintf( out_control->eov, "step: %d\n%6s%12s%12s\n",
+ data->step, "atom", "eov", "total" );
+ fprintf( out_control->eun, "step: %d\n%6s%12s%12s\n",
+ data->step, "atom", "eun", "total" );
+ fprintf( out_control->eval, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "atom3",
+ "angle", "bo(12)", "bo(23)", "eval", "epen", "total" );
+ fprintf( out_control->epen, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "atom3",
+ "angle", "bo(12)", "bo(23)", "epen", "total" );
+ fprintf( out_control->ecoa, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "atom3",
+ "angle", "bo(12)", "bo(23)", "ecoa", "total" );
+ fprintf( out_control->ehb, "step: %d\n%6s%6s%6s%12s%12s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "atom3",
+ "r(23)", "angle", "bo(12)", "ehb", "total" );
+ fprintf( out_control->etor, "step: %d\n%6s%6s%6s%6s%12s%12s%12s%12s\n",
+ data->step, "atom1", "atom2", "atom3", "atom4",
+ "phi", "bo(23)", "etor", "total" );
fprintf( out_control->econ, "step:%d\n%6s%6s%6s%6s%12s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "atom3", "atom4",
- "phi", "bo(12)", "bo(23)", "bo(34)", "econ", "total" );
-#endif
-
- /* Implement all the function calls as function pointers */
- for( i = 0; i < NO_OF_INTERACTIONS; i++ ) {
- //for( i = 0; i < 5; i++ ) {
- t_start = Get_Time ();
- (Interaction_Functions[i])(system, control, data, workspace,
- lists, out_control);
- t_elapsed = Get_Timing_Info ( t_start );
-
-#ifdef __DEBUG_CUDA__
- fprintf( stderr, "function %d tme %lf - \n", i, t_elapsed );
+ data->step, "atom1", "atom2", "atom3", "atom4",
+ "phi", "bo(12)", "bo(23)", "bo(34)", "econ", "total" );
#endif
+ /* Implement all the function calls as function pointers */
+ for ( i = 0; i < NO_OF_INTERACTIONS; i++ )
+ {
+ (Interaction_Functions[i])(system, control, data, workspace,
+ lists, out_control);
#if defined(DEBUG_FOCUS)
fprintf( stderr, "f%d-", i );
#endif
#ifdef TEST_FORCES
- (Print_Interactions[i])(system, control, data, workspace,
- lists, out_control);
+ (Print_Interactions[i])(system, control, data, workspace,
+ lists, out_control);
#endif
}
}
-void Compute_NonBonded_Forces( reax_system *system, control_params *control,
- simulation_data *data,static_storage *workspace,
- list** lists, output_controls *out_control )
+void Compute_NonBonded_Forces( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list** lists, output_controls *out_control )
{
real t_start, t_elapsed;
#ifdef TEST_ENERGY
fprintf( out_control->evdw, "step: %d\n%6s%6s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "r12", "evdw", "total" );
+ data->step, "atom1", "atom2", "r12", "evdw", "total" );
fprintf( out_control->ecou, "step: %d\n%6s%6s%12s%12s%12s%12s%12s\n",
- data->step, "atom1", "atom2", "r12", "q1", "q2", "ecou", "total" );
+ data->step, "atom1", "atom2", "r12", "q1", "q2", "ecou", "total" );
#endif
t_start = Get_Time( );
QEq( system, control, data, workspace, lists[FAR_NBRS], out_control );
t_elapsed = Get_Timing_Info( t_start );
data->timing.QEq += t_elapsed;
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "qeq - " );
#endif
if ( control->tabulate == 0)
+ {
vdW_Coulomb_Energy( system, control, data, workspace, lists, out_control );
+ }
else
- Tabulated_vdW_Coulomb_Energy( system, control, data, workspace,
- lists, out_control );
-
+ {
+ Tabulated_vdW_Coulomb_Energy( system, control, data, workspace,
+ lists, out_control );
+ }
#if defined(DEBUG_FOCUS)
fprintf( stderr, "nonb forces - " );
#endif
#ifdef TEST_FORCES
- Print_vdW_Coulomb_Forces( system, control, data, workspace,
- lists, out_control );
+ Print_vdW_Coulomb_Forces( system, control, data, workspace,
+ lists, out_control );
#endif
}
-/* This version of Compute_Total_Force computes forces from coefficients
+/* This version of Compute_Total_Force computes forces from coefficients
accumulated by all interaction functions. Saves enormous time & space! */
-void Compute_Total_Force( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists )
+void Compute_Total_Force( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists )
{
int i, pj;
list *bonds = (*lists) + BONDS;
- for( i = 0; i < system->N; ++i )
- for( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
- if( i < bonds->select.bond_list[pj].nbr ) {
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT)
+ for ( i = 0; i < system->N; ++i )
+ for ( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
+ if ( i < bonds->select.bond_list[pj].nbr )
+ {
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT)
Add_dBond_to_Forces( i, pj, system, data, workspace, lists );
- else
+ else
Add_dBond_to_Forces_NPT( i, pj, system, data, workspace, lists );
}
}
void Validate_Lists( static_storage *workspace, list **lists, int step, int n,
- int Hmax, int Htop, int num_bonds, int num_hbonds )
+ int Hmax, int Htop, int num_bonds, int num_hbonds )
{
int i, flag;
list *bonds, *hbonds;
@@ -188,92 +186,104 @@ void Validate_Lists( static_storage *workspace, list **lists, int step, int n,
hbonds = *lists + HBONDS;
/* far neighbors */
- if( Htop > Hmax * DANGER_ZONE ) {
+ if ( Htop > Hmax * DANGER_ZONE )
+ {
workspace->realloc.Htop = Htop;
- if( Htop > Hmax ) {
- fprintf( stderr,
- "step%d - ran out of space on H matrix: Htop=%d, max = %d",
- step, Htop, Hmax );
- exit(INSUFFICIENT_SPACE);
+ if ( Htop > Hmax )
+ {
+ fprintf( stderr,
+ "step%d - ran out of space on H matrix: Htop=%d, max = %d",
+ step, Htop, Hmax );
+ exit( INSUFFICIENT_MEMORY );
}
}
/* bond list */
flag = -1;
workspace->realloc.num_bonds = num_bonds;
- for( i = 0; i < n-1; ++i )
- if( End_Index(i, bonds) >= Start_Index(i+1, bonds)-2 ) {
+ for ( i = 0; i < n - 1; ++i )
+ if ( End_Index(i, bonds) >= Start_Index(i + 1, bonds) - 2 )
+ {
workspace->realloc.bonds = 1;
- if( End_Index(i, bonds) > Start_Index(i+1, bonds) )
+ if ( End_Index(i, bonds) > Start_Index(i + 1, bonds) )
flag = i;
}
- if( flag > -1 ) {
+ if ( flag > -1 )
+ {
fprintf( stderr, "step%d-bondchk failed: i=%d end(i)=%d str(i+1)=%d\n",
- step, flag, End_Index(flag,bonds), Start_Index(flag+1,bonds) );
- exit(INSUFFICIENT_SPACE);
- }
+ step, flag, End_Index(flag, bonds), Start_Index(flag + 1, bonds) );
+ exit( INSUFFICIENT_MEMORY );
+ }
- if( End_Index(i, bonds) >= bonds->num_intrs-2 ) {
+ if ( End_Index(i, bonds) >= bonds->num_intrs - 2 )
+ {
workspace->realloc.bonds = 1;
- if( End_Index(i, bonds) > bonds->num_intrs ) {
+ if ( End_Index(i, bonds) > bonds->num_intrs )
+ {
fprintf( stderr, "step%d-bondchk failed: i=%d end(i)=%d bond_end=%d\n",
- step, flag, End_Index(i,bonds), bonds->num_intrs );
- exit(INSUFFICIENT_SPACE);
+ step, flag, End_Index(i, bonds), bonds->num_intrs );
+ exit( INSUFFICIENT_MEMORY );
}
}
/* hbonds list */
- if( workspace->num_H > 0 ) {
+ if ( workspace->num_H > 0 )
+ {
flag = -1;
workspace->realloc.num_hbonds = num_hbonds;
- for( i = 0; i < workspace->num_H-1; ++i )
- if( Num_Entries(i, hbonds) >=
- (Start_Index(i+1, hbonds) - Start_Index(i, hbonds)) * DANGER_ZONE ) {
+ for ( i = 0; i < workspace->num_H - 1; ++i )
+ if ( Num_Entries(i, hbonds) >=
+ (Start_Index(i + 1, hbonds) - Start_Index(i, hbonds)) * DANGER_ZONE )
+ {
workspace->realloc.hbonds = 1;
- if( End_Index(i, hbonds) > Start_Index(i+1, hbonds) )
+ if ( End_Index(i, hbonds) > Start_Index(i + 1, hbonds) )
flag = i;
}
- if( flag > -1 ) {
+ if ( flag > -1 )
+ {
fprintf( stderr, "step%d-hbondchk failed: i=%d end(i)=%d str(i+1)=%d\n",
- step, flag, End_Index(flag,hbonds), Start_Index(flag+1,hbonds) );
- exit(INSUFFICIENT_SPACE);
+ step, flag, End_Index(flag, hbonds), Start_Index(flag + 1, hbonds) );
+ exit( INSUFFICIENT_MEMORY );
}
- if( Num_Entries(i,hbonds) >=
- (hbonds->num_intrs - Start_Index(i,hbonds)) * DANGER_ZONE ) {
+ if ( Num_Entries(i, hbonds) >=
+ (hbonds->num_intrs - Start_Index(i, hbonds)) * DANGER_ZONE )
+ {
workspace->realloc.hbonds = 1;
- if( End_Index(i, hbonds) > hbonds->num_intrs ) {
+ if ( End_Index(i, hbonds) > hbonds->num_intrs )
+ {
fprintf( stderr, "step%d-hbondchk failed: i=%d end(i)=%d hbondend=%d\n",
- step, flag, End_Index(i,hbonds), hbonds->num_intrs );
- exit(INSUFFICIENT_SPACE);
+ step, flag, End_Index(i, hbonds), hbonds->num_intrs );
+ exit( INSUFFICIENT_MEMORY );
}
}
}
}
-void Init_Forces( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control ) {
+void Init_Forces( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
+{
int i, j, pj;
int start_i, end_i;
int type_i, type_j;
- int Htop, btop_i, btop_j, num_bonds, num_hbonds;
+ int Htop, H_sp_top, btop_i, btop_j, num_bonds, num_hbonds;
int ihb, jhb, ihb_top, jhb_top;
- int flag;
+ int flag, flag_sp;
real r_ij, r2, self_coef;
real dr3gamij_1, dr3gamij_3, Tap;
//real val, dif, base;
real C12, C34, C56;
real Cln_BOp_s, Cln_BOp_pi, Cln_BOp_pi2;
real BO, BO_s, BO_pi, BO_pi2;
- real p_boc1, p_boc2;
- sparse_matrix *H;
+ real p_boc1, p_boc2;
+ sparse_matrix *H, *H_sp;
list *far_nbrs, *bonds, *hbonds;
single_body_parameters *sbp_i, *sbp_j;
two_body_parameters *twbp;
@@ -287,44 +297,67 @@ void Init_Forces( reax_system *system, control_params *control,
bonds = *lists + BONDS;
hbonds = *lists + HBONDS;
- H = &workspace->H;
+ H = workspace->H;
+ H_sp = workspace->H_sp;
Htop = 0;
+ H_sp_top = 0;
num_bonds = 0;
num_hbonds = 0;
btop_i = btop_j = 0;
p_boc1 = system->reaxprm.gp.l[0];
p_boc2 = system->reaxprm.gp.l[1];
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
atom_i = &(system->atoms[i]);
type_i = atom_i->type;
start_i = Start_Index(i, far_nbrs);
end_i = End_Index(i, far_nbrs);
H->start[i] = Htop;
+ H_sp->start[i] = H_sp_top;
btop_i = End_Index( i, bonds );
sbp_i = &(system->reaxprm.sbp[type_i]);
ihb = ihb_top = -1;
- if( control->hb_cut > 0 && (ihb=sbp_i->p_hbond) == 1 )
+ if ( control->hb_cut > 0 && (ihb = sbp_i->p_hbond) == 1 )
ihb_top = End_Index( workspace->hbond_index[i], hbonds );
- for( pj = start_i; pj < end_i; ++pj ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ {
nbr_pj = &( far_nbrs->select.far_nbr_list[pj] );
j = nbr_pj->nbr;
atom_j = &(system->atoms[j]);
flag = 0;
- if((data->step-data->prev_steps) % control->reneighbor == 0) {
- if( nbr_pj->d <= control->r_cut)
+ flag_sp = 0;
+ if ((data->step - data->prev_steps) % control->reneighbor == 0)
+ {
+ if ( nbr_pj->d <= control->r_cut )
+ {
flag = 1;
- else flag = 0;
+ if ( nbr_pj->d <= control->r_sp_cut )
+ {
+ flag_sp = 1;
+ }
+ }
+ else
+ {
+ flag = 0;
+ flag_sp = 0;
+ }
}
- else if((nbr_pj->d=Sq_Distance_on_T3(atom_i->x,atom_j->x,&(system->box),
- nbr_pj->dvec))<=SQR(control->r_cut)){
- nbr_pj->d = sqrt(nbr_pj->d);
+ else if ((nbr_pj->d = Sq_Distance_on_T3(atom_i->x, atom_j->x, &(system->box),
+ nbr_pj->dvec)) <= SQR(control->r_cut))
+ {
+ if ( nbr_pj->d <= SQR(control->r_sp_cut))
+ {
+ flag_sp = 1;
+ }
+ nbr_pj->d = SQRT( nbr_pj->d );
flag = 1;
}
- if( flag ){
+ if ( flag )
+ {
type_j = system->atoms[j].type;
r_ij = nbr_pj->d;
sbp_j = &(system->reaxprm.sbp[type_j]);
@@ -338,63 +371,79 @@ void Init_Forces( reax_system *system, control_params *control,
Tap = Tap * r_ij + control->Tap3;
Tap = Tap * r_ij + control->Tap2;
Tap = Tap * r_ij + control->Tap1;
- Tap = Tap * r_ij + control->Tap0;
+ Tap = Tap * r_ij + control->Tap0;
dr3gamij_1 = ( r_ij * r_ij * r_ij + twbp->gamma );
dr3gamij_3 = POW( dr3gamij_1 , 0.33333333333333 );
- H->entries[Htop].j = j;
- H->entries[Htop].val = self_coef * Tap * EV_to_KCALpMOL / dr3gamij_3;
+ H->j[Htop] = j;
+ H->val[Htop] = self_coef * Tap * EV_to_KCALpMOL / dr3gamij_3;
++Htop;
- /* hydrogen bond lists */
- if( control->hb_cut > 0 && (ihb==1 || ihb==2) &&
- nbr_pj->d <= control->hb_cut ) {
+ /* H_sp matrix entry */
+ if ( flag_sp )
+ {
+ H_sp->j[H_sp_top] = j;
+ H_sp->val[H_sp_top] = H->val[Htop - 1];
+ ++H_sp_top;
+ }
+
+ /* hydrogen bond lists */
+ if ( control->hb_cut > 0 && (ihb == 1 || ihb == 2) &&
+ nbr_pj->d <= control->hb_cut )
+ {
// fprintf( stderr, "%d %d\n", atom1, atom2 );
jhb = sbp_j->p_hbond;
- if( ihb == 1 && jhb == 2 ) {
+ if ( ihb == 1 && jhb == 2 )
+ {
hbonds->select.hbond_list[ihb_top].nbr = j;
hbonds->select.hbond_list[ihb_top].scl = 1;
hbonds->select.hbond_list[ihb_top].ptr = nbr_pj;
++ihb_top;
++num_hbonds;
}
- else if( ihb == 2 && jhb == 1 ) {
+ else if ( ihb == 2 && jhb == 1 )
+ {
jhb_top = End_Index( workspace->hbond_index[j], hbonds );
hbonds->select.hbond_list[jhb_top].nbr = i;
hbonds->select.hbond_list[jhb_top].scl = -1;
hbonds->select.hbond_list[jhb_top].ptr = nbr_pj;
- Set_End_Index( workspace->hbond_index[j], jhb_top+1, hbonds );
+ Set_End_Index( workspace->hbond_index[j], jhb_top + 1, hbonds );
++num_hbonds;
}
}
/* uncorrected bond orders */
- if( far_nbrs->select.far_nbr_list[pj].d <= control->nbr_cut ) {
+ if ( far_nbrs->select.far_nbr_list[pj].d <= control->nbr_cut )
+ {
r2 = SQR(r_ij);
- if( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0) {
+ if ( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0)
+ {
C12 = twbp->p_bo1 * POW( r_ij / twbp->r_s, twbp->p_bo2 );
BO_s = (1.0 + control->bo_cut) * EXP( C12 );
}
else BO_s = C12 = 0.0;
- if( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0) {
+ if ( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0)
+ {
C34 = twbp->p_bo3 * POW( r_ij / twbp->r_p, twbp->p_bo4 );
BO_pi = EXP( C34 );
}
else BO_pi = C34 = 0.0;
- if( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0) {
- C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
- BO_pi2= EXP( C56 );
+ if ( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0)
+ {
+ C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
+ BO_pi2 = EXP( C56 );
}
else BO_pi2 = C56 = 0.0;
/* Initially BO values are the uncorrected ones, page 1 */
BO = BO_s + BO_pi + BO_pi2;
- if( BO >= control->bo_cut ) {
+ if ( BO >= control->bo_cut )
+ {
num_bonds += 2;
/****** bonds i-j and j-i ******/
ibond = &( bonds->select.bond_list[btop_i] );
@@ -414,7 +463,7 @@ void Init_Forces( reax_system *system, control_params *control,
ibond->sym_index = btop_j;
jbond->sym_index = btop_i;
++btop_i;
- Set_End_Index( j, btop_j+1, bonds );
+ Set_End_Index( j, btop_j + 1, bonds );
bo_ij = &( ibond->bo_data );
bo_ji = &( jbond->bo_data );
@@ -428,22 +477,22 @@ void Init_Forces( reax_system *system, control_params *control,
Cln_BOp_pi = twbp->p_bo4 * C34 / r2;
Cln_BOp_pi2 = twbp->p_bo6 * C56 / r2;
- /* Only dln_BOp_xx wrt. dr_i is stored here, note that
+ /* Only dln_BOp_xx wrt. dr_i is stored here, note that
dln_BOp_xx/dr_i = -dln_BOp_xx/dr_j and all others are 0 */
- rvec_Scale(bo_ij->dln_BOp_s,-bo_ij->BO_s*Cln_BOp_s,ibond->dvec);
- rvec_Scale(bo_ij->dln_BOp_pi,-bo_ij->BO_pi*Cln_BOp_pi,ibond->dvec);
+ rvec_Scale(bo_ij->dln_BOp_s, -bo_ij->BO_s * Cln_BOp_s, ibond->dvec);
+ rvec_Scale(bo_ij->dln_BOp_pi, -bo_ij->BO_pi * Cln_BOp_pi, ibond->dvec);
rvec_Scale(bo_ij->dln_BOp_pi2,
- -bo_ij->BO_pi2*Cln_BOp_pi2,ibond->dvec);
+ -bo_ij->BO_pi2 * Cln_BOp_pi2, ibond->dvec);
rvec_Scale(bo_ji->dln_BOp_s, -1., bo_ij->dln_BOp_s);
rvec_Scale(bo_ji->dln_BOp_pi, -1., bo_ij->dln_BOp_pi );
rvec_Scale(bo_ji->dln_BOp_pi2, -1., bo_ij->dln_BOp_pi2 );
- /* Only dBOp wrt. dr_i is stored here, note that
+ /* Only dBOp wrt. dr_i is stored here, note that
dBOp/dr_i = -dBOp/dr_j and all others are 0 */
- rvec_Scale( bo_ij->dBOp,
- -(bo_ij->BO_s * Cln_BOp_s +
- bo_ij->BO_pi * Cln_BOp_pi +
- bo_ij->BO_pi2 * Cln_BOp_pi2), ibond->dvec );
+ rvec_Scale( bo_ij->dBOp,
+ -(bo_ij->BO_s * Cln_BOp_s +
+ bo_ij->BO_pi * Cln_BOp_pi +
+ bo_ij->BO_pi2 * Cln_BOp_pi2), ibond->dvec );
rvec_Scale( bo_ji->dBOp, -1., bo_ij->dBOp );
rvec_Add( workspace->dDeltap_self[i], bo_ij->dBOp );
@@ -461,79 +510,91 @@ void Init_Forces( reax_system *system, control_params *control,
/*fprintf( stderr, "%d %d %g %g %g\n",
i+1, j+1, bo_ij->BO, bo_ij->BO_pi, bo_ij->BO_pi2 );*/
- /*fprintf( stderr, "Cln_BOp_s: %f, pbo2: %f, C12:%f\n",
+ /*fprintf( stderr, "Cln_BOp_s: %f, pbo2: %f, C12:%f\n",
Cln_BOp_s, twbp->p_bo2, C12 );
- fprintf( stderr, "Cln_BOp_pi: %f, pbo4: %f, C34:%f\n",
+ fprintf( stderr, "Cln_BOp_pi: %f, pbo4: %f, C34:%f\n",
Cln_BOp_pi, twbp->p_bo4, C34 );
fprintf( stderr, "Cln_BOp_pi2: %f, pbo6: %f, C56:%f\n",
Cln_BOp_pi2, twbp->p_bo6, C56 );*/
/*fprintf(stderr, "pbo1: %f, pbo2:%f\n", twbp->p_bo1, twbp->p_bo2);
fprintf(stderr, "pbo3: %f, pbo4:%f\n", twbp->p_bo3, twbp->p_bo4);
fprintf(stderr, "pbo5: %f, pbo6:%f\n", twbp->p_bo5, twbp->p_bo6);
- fprintf( stderr, "r_s: %f, r_p: %f, r_pp: %f\n",
+ fprintf( stderr, "r_s: %f, r_p: %f, r_pp: %f\n",
twbp->r_s, twbp->r_p, twbp->r_pp );
fprintf( stderr, "C12: %g, C34:%g, C56:%g\n", C12, C34, C56 );*/
/*fprintf( stderr, "\tfactors: %g %g %g\n",
- -(bo_ij->BO_s * Cln_BOp_s + bo_ij->BO_pi * Cln_BOp_pi +
+ -(bo_ij->BO_s * Cln_BOp_s + bo_ij->BO_pi * Cln_BOp_pi +
bo_ij->BO_pi2 * Cln_BOp_pp),
-bo_ij->BO_pi * Cln_BOp_pi, -bo_ij->BO_pi2 * Cln_BOp_pi2 );*/
- /*fprintf( stderr, "dBOpi:\t[%g, %g, %g]\n",
+ /*fprintf( stderr, "dBOpi:\t[%g, %g, %g]\n",
bo_ij->dBOp[0], bo_ij->dBOp[1], bo_ij->dBOp[2] );
- fprintf( stderr, "dBOpi:\t[%g, %g, %g]\n",
- bo_ij->dln_BOp_pi[0], bo_ij->dln_BOp_pi[1],
+ fprintf( stderr, "dBOpi:\t[%g, %g, %g]\n",
+ bo_ij->dln_BOp_pi[0], bo_ij->dln_BOp_pi[1],
bo_ij->dln_BOp_pi[2] );
fprintf( stderr, "dBOpi2:\t[%g, %g, %g]\n\n",
- bo_ij->dln_BOp_pi2[0], bo_ij->dln_BOp_pi2[1],
+ bo_ij->dln_BOp_pi2[0], bo_ij->dln_BOp_pi2[1],
bo_ij->dln_BOp_pi2[2] );*/
- Set_End_Index( j, btop_j+1, bonds );
+ Set_End_Index( j, btop_j + 1, bonds );
}
}
}
}
- H->entries[Htop].j = i;
- H->entries[Htop].val = system->reaxprm.sbp[type_i].eta;
+ /* diagonal entry */
+ H->j[Htop] = i;
+ H->val[Htop] = system->reaxprm.sbp[type_i].eta;
++Htop;
+ /* diagonal entry */
+ H_sp->j[H_sp_top] = i;
+ H_sp->val[H_sp_top] = H->val[Htop - 1];
+ ++H_sp_top;
+
Set_End_Index( i, btop_i, bonds );
- if( ihb == 1 )
+ if ( ihb == 1 )
Set_End_Index( workspace->hbond_index[i], ihb_top, hbonds );
- //fprintf( stderr, "%d bonds start: %d, end: %d\n",
+ //fprintf( stderr, "%d bonds start: %d, end: %d\n",
// i, Start_Index( i, bonds ), End_Index( i, bonds ) );
}
+// printf("Htop = %d\n", Htop);
+// printf("H_sp_top = %d\n", H_sp_top);
+
// mark the end of j list
- H->start[i] = Htop;
+ H->start[i] = Htop;
+ H_sp->start[i] = H_sp_top;
/* validate lists - decide if reallocation is required! */
- Validate_Lists( workspace, lists,
- data->step, system->N, H->m, Htop, num_bonds, num_hbonds );
+ Validate_Lists( workspace, lists,
+ data->step, system->N, H->m, Htop, num_bonds, num_hbonds );
#if defined(DEBUG_FOCUS)
- fprintf( stderr, "step%d: Htop = %d, num_bonds = %d, num_hbonds = %d\n",
- data->step, Htop, num_bonds, num_hbonds );
+ fprintf( stderr, "step%d: Htop = %d, num_bonds = %d, num_hbonds = %d\n",
+ data->step, Htop, num_bonds, num_hbonds );
+
#endif
}
-void Init_Forces_Tab( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control ) {
+void Init_Forces_Tab( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
+{
int i, j, pj;
int start_i, end_i;
int type_i, type_j;
- int Htop, btop_i, btop_j, num_bonds, num_hbonds;
+ int Htop, H_sp_top, btop_i, btop_j, num_bonds, num_hbonds;
int tmin, tmax, r;
int ihb, jhb, ihb_top, jhb_top;
- int flag;
+ int flag, flag_sp;
real r_ij, r2, self_coef;
real val, dif, base;
real C12, C34, C56;
real Cln_BOp_s, Cln_BOp_pi, Cln_BOp_pi2;
real BO, BO_s, BO_pi, BO_pi2;
- real p_boc1, p_boc2;
- sparse_matrix *H;
+ real p_boc1, p_boc2;
+ sparse_matrix *H, *H_sp;
list *far_nbrs, *bonds, *hbonds;
single_body_parameters *sbp_i, *sbp_j;
two_body_parameters *twbp;
@@ -547,44 +608,67 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
bonds = *lists + BONDS;
hbonds = *lists + HBONDS;
- H = &workspace->H;
+ H = workspace->H;
+ H_sp = workspace->H_sp;
Htop = 0;
+ H_sp_top = 0;
num_bonds = 0;
num_hbonds = 0;
btop_i = btop_j = 0;
p_boc1 = system->reaxprm.gp.l[0];
p_boc2 = system->reaxprm.gp.l[1];
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
atom_i = &(system->atoms[i]);
type_i = atom_i->type;
start_i = Start_Index(i, far_nbrs);
end_i = End_Index(i, far_nbrs);
H->start[i] = Htop;
+ H_sp->start[i] = H_sp_top;
btop_i = End_Index( i, bonds );
sbp_i = &(system->reaxprm.sbp[type_i]);
ihb = ihb_top = -1;
- if( control->hb_cut > 0 && (ihb=sbp_i->p_hbond) == 1 )
+ if ( control->hb_cut > 0 && (ihb = sbp_i->p_hbond) == 1 )
ihb_top = End_Index( workspace->hbond_index[i], hbonds );
- for( pj = start_i; pj < end_i; ++pj ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ {
nbr_pj = &( far_nbrs->select.far_nbr_list[pj] );
j = nbr_pj->nbr;
atom_j = &(system->atoms[j]);
flag = 0;
- if((data->step-data->prev_steps) % control->reneighbor == 0) {
- if(nbr_pj->d <= control->r_cut)
+ flag_sp = 0;
+ if ((data->step - data->prev_steps) % control->reneighbor == 0)
+ {
+ if (nbr_pj->d <= control->r_cut)
+ {
flag = 1;
- else flag = 0;
+ if ( nbr_pj->d <= control->r_sp_cut )
+ {
+ flag_sp = 1;
+ }
+ }
+ else
+ {
+ flag = 0;
+ flag_sp = 0;
+ }
}
- else if((nbr_pj->d=Sq_Distance_on_T3(atom_i->x,atom_j->x,&(system->box),
- nbr_pj->dvec))<=SQR(control->r_cut)){
+ else if ((nbr_pj->d = Sq_Distance_on_T3(atom_i->x, atom_j->x, &(system->box),
+ nbr_pj->dvec)) <= SQR(control->r_cut))
+ {
+ if ( nbr_pj->d <= SQR(control->r_sp_cut))
+ {
+ flag_sp = 1;
+ }
nbr_pj->d = sqrt(nbr_pj->d);
flag = 1;
}
- if( flag ){
+ if ( flag )
+ {
type_j = system->atoms[j].type;
r_ij = nbr_pj->d;
sbp_j = &(system->reaxprm.sbp[type_j]);
@@ -596,65 +680,81 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
/* cubic spline interpolation */
r = (int)(r_ij * t->inv_dx);
- if( r == 0 ) ++r;
- base = (real)(r+1) * t->dx;
+ if ( r == 0 ) ++r;
+ base = (real)(r + 1) * t->dx;
dif = r_ij - base;
- val = ((t->ele[r].d*dif + t->ele[r].c)*dif + t->ele[r].b)*dif +
- t->ele[r].a;
+ val = ((t->ele[r].d * dif + t->ele[r].c) * dif + t->ele[r].b) * dif +
+ t->ele[r].a;
val *= EV_to_KCALpMOL / C_ele;
- H->entries[Htop].j = j;
- H->entries[Htop].val = self_coef * val;
+ H->j[Htop] = j;
+ H->val[Htop] = self_coef * val;
++Htop;
- /* hydrogen bond lists */
- if( control->hb_cut > 0 && (ihb==1 || ihb==2) &&
- nbr_pj->d <= control->hb_cut ) {
+ /* H_sp matrix entry */
+ if ( flag_sp )
+ {
+ H_sp->j[H_sp_top] = j;
+ H_sp->val[H_sp_top] = H->val[Htop - 1];
+ ++H_sp_top;
+ }
+
+ /* hydrogen bond lists */
+ if ( control->hb_cut > 0 && (ihb == 1 || ihb == 2) &&
+ nbr_pj->d <= control->hb_cut )
+ {
// fprintf( stderr, "%d %d\n", atom1, atom2 );
jhb = sbp_j->p_hbond;
- if( ihb == 1 && jhb == 2 ) {
+ if ( ihb == 1 && jhb == 2 )
+ {
hbonds->select.hbond_list[ihb_top].nbr = j;
hbonds->select.hbond_list[ihb_top].scl = 1;
hbonds->select.hbond_list[ihb_top].ptr = nbr_pj;
++ihb_top;
++num_hbonds;
}
- else if( ihb == 2 && jhb == 1 ) {
+ else if ( ihb == 2 && jhb == 1 )
+ {
jhb_top = End_Index( workspace->hbond_index[j], hbonds );
hbonds->select.hbond_list[jhb_top].nbr = i;
hbonds->select.hbond_list[jhb_top].scl = -1;
hbonds->select.hbond_list[jhb_top].ptr = nbr_pj;
- Set_End_Index( workspace->hbond_index[j], jhb_top+1, hbonds );
+ Set_End_Index( workspace->hbond_index[j], jhb_top + 1, hbonds );
++num_hbonds;
}
}
/* uncorrected bond orders */
- if( far_nbrs->select.far_nbr_list[pj].d <= control->nbr_cut ) {
+ if ( far_nbrs->select.far_nbr_list[pj].d <= control->nbr_cut )
+ {
r2 = SQR(r_ij);
- if( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0) {
+ if ( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0)
+ {
C12 = twbp->p_bo1 * POW( r_ij / twbp->r_s, twbp->p_bo2 );
BO_s = (1.0 + control->bo_cut) * EXP( C12 );
}
else BO_s = C12 = 0.0;
- if( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0) {
+ if ( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0)
+ {
C34 = twbp->p_bo3 * POW( r_ij / twbp->r_p, twbp->p_bo4 );
BO_pi = EXP( C34 );
}
else BO_pi = C34 = 0.0;
- if( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0) {
- C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
- BO_pi2= EXP( C56 );
+ if ( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0)
+ {
+ C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
+ BO_pi2 = EXP( C56 );
}
else BO_pi2 = C56 = 0.0;
/* Initially BO values are the uncorrected ones, page 1 */
BO = BO_s + BO_pi + BO_pi2;
- if( BO >= control->bo_cut ) {
+ if ( BO >= control->bo_cut )
+ {
num_bonds += 2;
/****** bonds i-j and j-i ******/
ibond = &( bonds->select.bond_list[btop_i] );
@@ -666,6 +766,7 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
ibond->d = r_ij;
jbond->d = r_ij;
rvec_Copy( ibond->dvec, nbr_pj->dvec );
+ //fprintf (stderr, " %f - %f - %f \n", nbr_pj->dvec[0], nbr_pj->dvec[1], nbr_pj->dvec[2]);
rvec_Scale( jbond->dvec, -1, nbr_pj->dvec );
ivec_Copy( ibond->rel_box, nbr_pj->rel_box );
ivec_Scale( jbond->rel_box, -1, nbr_pj->rel_box );
@@ -674,7 +775,7 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
ibond->sym_index = btop_j;
jbond->sym_index = btop_i;
++btop_i;
- Set_End_Index( j, btop_j+1, bonds );
+ Set_End_Index( j, btop_j + 1, bonds );
bo_ij = &( ibond->bo_data );
bo_ji = &( jbond->bo_data );
@@ -688,22 +789,22 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
Cln_BOp_pi = twbp->p_bo4 * C34 / r2;
Cln_BOp_pi2 = twbp->p_bo6 * C56 / r2;
- /* Only dln_BOp_xx wrt. dr_i is stored here, note that
+ /* Only dln_BOp_xx wrt. dr_i is stored here, note that
dln_BOp_xx/dr_i = -dln_BOp_xx/dr_j and all others are 0 */
- rvec_Scale(bo_ij->dln_BOp_s,-bo_ij->BO_s*Cln_BOp_s,ibond->dvec);
- rvec_Scale(bo_ij->dln_BOp_pi,-bo_ij->BO_pi*Cln_BOp_pi,ibond->dvec);
+ rvec_Scale(bo_ij->dln_BOp_s, -bo_ij->BO_s * Cln_BOp_s, ibond->dvec);
+ rvec_Scale(bo_ij->dln_BOp_pi, -bo_ij->BO_pi * Cln_BOp_pi, ibond->dvec);
rvec_Scale(bo_ij->dln_BOp_pi2,
- -bo_ij->BO_pi2*Cln_BOp_pi2,ibond->dvec);
+ -bo_ij->BO_pi2 * Cln_BOp_pi2, ibond->dvec);
rvec_Scale(bo_ji->dln_BOp_s, -1., bo_ij->dln_BOp_s);
rvec_Scale(bo_ji->dln_BOp_pi, -1., bo_ij->dln_BOp_pi );
rvec_Scale(bo_ji->dln_BOp_pi2, -1., bo_ij->dln_BOp_pi2 );
- /* Only dBOp wrt. dr_i is stored here, note that
+ /* Only dBOp wrt. dr_i is stored here, note that
dBOp/dr_i = -dBOp/dr_j and all others are 0 */
- rvec_Scale( bo_ij->dBOp,
- -(bo_ij->BO_s * Cln_BOp_s +
- bo_ij->BO_pi * Cln_BOp_pi +
- bo_ij->BO_pi2 * Cln_BOp_pi2), ibond->dvec );
+ rvec_Scale( bo_ij->dBOp,
+ -(bo_ij->BO_s * Cln_BOp_s +
+ bo_ij->BO_pi * Cln_BOp_pi +
+ bo_ij->BO_pi2 * Cln_BOp_pi2), ibond->dvec );
rvec_Scale( bo_ji->dBOp, -1., bo_ij->dBOp );
rvec_Add( workspace->dDeltap_self[i], bo_ij->dBOp );
@@ -718,30 +819,37 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
bo_ij->Cdbo = bo_ij->Cdbopi = bo_ij->Cdbopi2 = 0.0;
bo_ji->Cdbo = bo_ji->Cdbopi = bo_ji->Cdbopi2 = 0.0;
- Set_End_Index( j, btop_j+1, bonds );
+ Set_End_Index( j, btop_j + 1, bonds );
}
}
}
}
- H->entries[Htop].j = i;
- H->entries[Htop].val = system->reaxprm.sbp[type_i].eta;
+ /* diagonal entry */
+ H->j[Htop] = i;
+ H->val[Htop] = system->reaxprm.sbp[type_i].eta;
++Htop;
+ /* diagonal entry */
+ H_sp->j[H_sp_top] = i;
+ H_sp->val[H_sp_top] = H->val[Htop - 1];
+ ++H_sp_top;
+
Set_End_Index( i, btop_i, bonds );
- if( ihb == 1 )
+ if ( ihb == 1 )
Set_End_Index( workspace->hbond_index[i], ihb_top, hbonds );
}
// mark the end of j list
- H->start[i] = Htop;
+ H->start[i] = Htop;
+ H_sp->start[i] = H_sp_top;
/* validate lists - decide if reallocation is required! */
- Validate_Lists( workspace, lists,
- data->step, system->N, H->m, Htop, num_bonds, num_hbonds );
+ Validate_Lists( workspace, lists,
+ data->step, system->N, H->m, Htop, num_bonds, num_hbonds );
#if defined(DEBUG_FOCUS)
- fprintf( stderr, "step%d: Htop = %d, num_bonds = %d, num_hbonds = %d\n",
- data->step, Htop, num_bonds, num_hbonds );
+ fprintf( stderr, "step%d: Htop = %d, num_bonds = %d, num_hbonds = %d\n",
+ data->step, Htop, num_bonds, num_hbonds );
//Print_Bonds( system, bonds, "sbonds.out" );
//Print_Bond_List2( system, bonds, "sbonds.out" );
//Print_Sparse_Matrix2( H, "H.out" );
@@ -749,9 +857,10 @@ void Init_Forces_Tab( reax_system *system, control_params *control,
}
-void Estimate_Storage_Sizes( reax_system *system, control_params *control,
- list **lists, int *Htop, int *hb_top,
- int *bond_top, int *num_3body ) {
+void Estimate_Storage_Sizes( reax_system *system, control_params *control,
+ list **lists, int *Htop, int *hb_top,
+ int *bond_top, int *num_3body )
+{
int i, j, pj;
int start_i, end_i;
int type_i, type_j;
@@ -759,7 +868,7 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
real r_ij, r2;
real C12, C34, C56;
real BO, BO_s, BO_pi, BO_pi2;
- real p_boc1, p_boc2;
+ real p_boc1, p_boc2;
list *far_nbrs;
single_body_parameters *sbp_i, *sbp_j;
two_body_parameters *twbp;
@@ -770,7 +879,8 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
p_boc1 = system->reaxprm.gp.l[0];
p_boc2 = system->reaxprm.gp.l[1];
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
atom_i = &(system->atoms[i]);
type_i = atom_i->type;
start_i = Start_Index(i, far_nbrs);
@@ -778,7 +888,8 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
sbp_i = &(system->reaxprm.sbp[type_i]);
ihb = sbp_i->p_hbond;
- for( pj = start_i; pj < end_i; ++pj ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ {
nbr_pj = &( far_nbrs->select.far_nbr_list[pj] );
j = nbr_pj->nbr;
atom_j = &(system->atoms[j]);
@@ -786,46 +897,53 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
sbp_j = &(system->reaxprm.sbp[type_j]);
twbp = &(system->reaxprm.tbp[ index_tbp(type_i,type_j,system->reaxprm.num_atom_types) ]);
- if( nbr_pj->d <= control->r_cut ) {
+ if ( nbr_pj->d <= control->r_cut )
+ {
++(*Htop);
- /* hydrogen bond lists */
- if( control->hb_cut > 0.1 && (ihb==1 || ihb==2) &&
- nbr_pj->d <= control->hb_cut ) {
+ /* hydrogen bond lists */
+ if ( control->hb_cut > 0.1 && (ihb == 1 || ihb == 2) &&
+ nbr_pj->d <= control->hb_cut )
+ {
jhb = sbp_j->p_hbond;
- if( ihb == 1 && jhb == 2 )
+ if ( ihb == 1 && jhb == 2 )
++hb_top[i];
- else if( ihb == 2 && jhb == 1 )
+ else if ( ihb == 2 && jhb == 1 )
++hb_top[j];
}
/* uncorrected bond orders */
- if( nbr_pj->d <= control->nbr_cut ) {
+ if ( nbr_pj->d <= control->nbr_cut )
+ {
r_ij = nbr_pj->d;
r2 = SQR(r_ij);
- if( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0) {
+ if ( sbp_i->r_s > 0.0 && sbp_j->r_s > 0.0)
+ {
C12 = twbp->p_bo1 * POW( r_ij / twbp->r_s, twbp->p_bo2 );
BO_s = (1.0 + control->bo_cut) * EXP( C12 );
}
else BO_s = C12 = 0.0;
- if( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0) {
+ if ( sbp_i->r_pi > 0.0 && sbp_j->r_pi > 0.0)
+ {
C34 = twbp->p_bo3 * POW( r_ij / twbp->r_p, twbp->p_bo4 );
BO_pi = EXP( C34 );
}
else BO_pi = C34 = 0.0;
- if( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0) {
- C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
- BO_pi2= EXP( C56 );
+ if ( sbp_i->r_pi_pi > 0.0 && sbp_j->r_pi_pi > 0.0)
+ {
+ C56 = twbp->p_bo5 * POW( r_ij / twbp->r_pp, twbp->p_bo6 );
+ BO_pi2 = EXP( C56 );
}
else BO_pi2 = C56 = 0.0;
/* Initially BO values are the uncorrected ones, page 1 */
BO = BO_s + BO_pi + BO_pi2;
- if( BO >= control->bo_cut ) {
+ if ( BO >= control->bo_cut )
+ {
++bond_top[i];
++bond_top[j];
}
@@ -836,8 +954,8 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
*Htop += system->N;
*Htop *= SAFE_ZONE;
-
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
hb_top[i] = MAX( hb_top[i] * SAFE_HBONDS, MIN_HBONDS );
*num_3body += SQR(bond_top[i]);
bond_top[i] = MAX( bond_top[i] * 2, MIN_BONDS );
@@ -846,49 +964,40 @@ void Estimate_Storage_Sizes( reax_system *system, control_params *control,
}
-void Compute_Forces( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list** lists, output_controls *out_control )
+void Compute_Forces( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list** lists, output_controls *out_control )
{
real t_start, t_elapsed;
t_start = Get_Time( );
- if( !control->tabulate )
+ if ( !control->tabulate )
+ {
Init_Forces( system, control, data, workspace, lists, out_control );
- else Init_Forces_Tab( system, control, data, workspace, lists, out_control );
+ }
+ else
+ {
+ Init_Forces_Tab( system, control, data, workspace, lists, out_control );
+ }
t_elapsed = Get_Timing_Info( t_start );
data->timing.init_forces += t_elapsed;
-
#if defined(DEBUG_FOCUS)
- print_sparse_matrix (system, workspace);
fprintf( stderr, "init_forces - ");
#endif
-
- //analyze_hbonds (system, workspace, lists);
-
t_start = Get_Time( );
Compute_Bonded_Forces( system, control, data, workspace, lists, out_control );
t_elapsed = Get_Timing_Info( t_start );
data->timing.bonded += t_elapsed;
-
- //print_bond_list (system, workspace, lists);
- //exit (0);
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "bonded_forces - ");
#endif
t_start = Get_Time( );
- Compute_NonBonded_Forces( system, control, data, workspace,
- lists, out_control );
+ Compute_NonBonded_Forces( system, control, data, workspace,
+ lists, out_control );
t_elapsed = Get_Timing_Info( t_start );
data->timing.nonb += t_elapsed;
-
-#ifdef __DEBUG_CUDA__
- fprintf( stderr, "non_bonded_forces - %lf \n", t_elapsed);
-#endif
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "nonbondeds - ");
#endif
@@ -904,7 +1013,7 @@ void Compute_Forces( reax_system *system, control_params *control,
Print_Total_Force( system, control, data, workspace, lists, out_control );
Compare_Total_Forces( system, control, data, workspace, lists, out_control );
#endif
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "forces - ");
#endif
}
diff --git a/PuReMD-GPU/src/forces.h b/PuReMD-GPU/src/forces.h
index 73323f0419baf383d6bf671158ef85584a710728..0ef8b117c78de5e7b3cdc2311b4a492b4615a859 100644
--- a/PuReMD-GPU/src/forces.h
+++ b/PuReMD-GPU/src/forces.h
@@ -23,12 +23,14 @@
#include "mytypes.h"
+
void Init_Bonded_Force_Functions( control_params* );
void Compute_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Estimate_Storage_Sizes( reax_system*, control_params*, list**,
- int*, int*, int*, int* );
+ int*, int*, int*, int* );
+
#endif
diff --git a/PuReMD-GPU/src/four_body_interactions.c b/PuReMD-GPU/src/four_body_interactions.c
index c51601fa991203a77ec4840c10e74e15cfa42c87..25642871d23e389a96db943607f432aa68252a02 100644
--- a/PuReMD-GPU/src/four_body_interactions.c
+++ b/PuReMD-GPU/src/four_body_interactions.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -22,20 +23,21 @@
#include "bond_orders.h"
#include "box.h"
+#include "index_utils.h"
#include "list.h"
#include "lookup.h"
#include "vector.h"
#include "math.h"
-#include "index_utils.h"
+#define MIN_SINE 1e-10
real Calculate_Omega( rvec dvec_ij, real r_ij, rvec dvec_jk, real r_jk,
- rvec dvec_kl, real r_kl, rvec dvec_li, real r_li,
- three_body_interaction_data *p_ijk,
- three_body_interaction_data *p_jkl,
- rvec dcos_omega_di, rvec dcos_omega_dj,
- rvec dcos_omega_dk, rvec dcos_omega_dl,
- output_controls *out_control )
+ rvec dvec_kl, real r_kl, rvec dvec_li, real r_li,
+ three_body_interaction_data *p_ijk,
+ three_body_interaction_data *p_jkl,
+ rvec dcos_omega_di, rvec dcos_omega_dj,
+ rvec dcos_omega_dk, rvec dcos_omega_dl,
+ output_controls *out_control )
{
real unnorm_cos_omega, unnorm_sin_omega, omega;
real sin_ijk, cos_ijk, sin_jkl, cos_jkl;
@@ -49,11 +51,11 @@ real Calculate_Omega( rvec dvec_ij, real r_ij, rvec dvec_jk, real r_jk,
cos_jkl = COS( p_jkl->theta );
/* omega */
- unnorm_cos_omega = -rvec_Dot( dvec_ij,dvec_jk )*rvec_Dot( dvec_jk,dvec_kl ) +
- SQR( r_jk ) * rvec_Dot( dvec_ij,dvec_kl );
+ unnorm_cos_omega = -rvec_Dot( dvec_ij, dvec_jk ) * rvec_Dot( dvec_jk, dvec_kl ) +
+ SQR( r_jk ) * rvec_Dot( dvec_ij, dvec_kl );
rvec_Cross( cross_jk_kl, dvec_jk, dvec_kl );
unnorm_sin_omega = -r_jk * rvec_Dot( dvec_ij, cross_jk_kl );
- omega = atan2( unnorm_sin_omega, unnorm_cos_omega );
+ omega = atan2( unnorm_sin_omega, unnorm_cos_omega );
/* derivatives */
/* coef for adjusments to cos_theta's */
@@ -70,24 +72,25 @@ real Calculate_Omega( rvec dvec_ij, real r_ij, rvec dvec_jk, real r_jk,
hnhd = r_ij * r_kl * cos_ijk * sin_jkl;
hnhe = r_ij * r_kl * sin_ijk * cos_jkl;
+
poem = 2.0 * r_ij * r_kl * sin_ijk * sin_jkl;
- if( poem < 1e-20 ) poem = 1e-20;
+ if ( poem < 1e-20 ) poem = 1e-20;
- tel = (SQR(r_ij) + SQR(r_jk) + SQR(r_kl) - SQR(r_li)) -
- 2.0 * ( r_ij * r_jk * cos_ijk - r_ij * r_kl * cos_ijk * cos_jkl +
- r_jk * r_kl * cos_jkl );
+ tel = (SQR(r_ij) + SQR(r_jk) + SQR(r_kl) - SQR(r_li)) -
+ 2.0 * ( r_ij * r_jk * cos_ijk - r_ij * r_kl * cos_ijk * cos_jkl +
+ r_jk * r_kl * cos_jkl );
arg = tel / poem;
- if( arg > 1.0 )
+ if ( arg > 1.0 )
{
arg = 1.0;
}
- if( arg < -1.0 )
+ if ( arg < -1.0 )
{
arg = -1.0;
}
- /*fprintf( out_control->etor,
+ /*fprintf( out_control->etor,
"%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e\n",
htra, htrb, htrc, hthd, hthe, hnra, hnrc, hnhd, hnhe );
fprintf( out_control->etor, "%23.15e%23.15e%23.15e\n",
@@ -99,69 +102,72 @@ real Calculate_Omega( rvec dvec_ij, real r_ij, rvec dvec_jk, real r_jk,
fprintf( out_control->etor, "%23.15e%23.15e%23.15e%23.15e\n",
r_li, dvec_li[0], dvec_li[1], dvec_li[2] );
fprintf( out_control->etor, "%23.15e%23.15e%23.15e%23.15e\n",
- r_ij, r_jk, r_kl, r_li );
- fprintf( out_control->etor, "%23.15e%23.15e%23.15e%23.15e\n",
- cos_ijk, cos_jkl, sin_ijk, sin_jkl );
+ r_ij, r_jk, r_kl, r_li );
+ fprintf( out_control->etor, "%23.15e%23.15e%23.15e%23.15e\n",
+ cos_ijk, cos_jkl, sin_ijk, sin_jkl );
fprintf( out_control->etor, "%23.15e%23.15e%23.15e\n",
poem, tel, arg );*/
/* fprintf( out_control->etor, "%23.15e%23.15e%23.15e\n",
- -p_ijk->dcos_dk[0]/sin_ijk,
- -p_ijk->dcos_dk[1]/sin_ijk,
+ -p_ijk->dcos_dk[0]/sin_ijk,
+ -p_ijk->dcos_dk[1]/sin_ijk,
-p_ijk->dcos_dk[2]/sin_ijk );
fprintf( out_control->etor, "%23.15e%23.15e%23.15e\n",
- -p_jkl->dcos_dk[0]/sin_jkl,
- -p_jkl->dcos_dk[1]/sin_jkl,
+ -p_jkl->dcos_dk[0]/sin_jkl,
+ -p_jkl->dcos_dk[1]/sin_jkl,
-p_jkl->dcos_dk[2]/sin_jkl );*/
- if( sin_ijk >= 0 && sin_ijk <= MIN_SINE )
+ if ( sin_ijk >= 0 && sin_ijk <= MIN_SINE )
{
sin_ijk = MIN_SINE;
}
- else if( sin_ijk <= 0 && sin_ijk >= -MIN_SINE )
+ else if ( sin_ijk <= 0 && sin_ijk >= -MIN_SINE )
{
sin_ijk = -MIN_SINE;
}
- if( sin_jkl >= 0 && sin_jkl <= MIN_SINE )
+ if ( sin_jkl >= 0 && sin_jkl <= MIN_SINE )
{
sin_jkl = MIN_SINE;
}
- else if( sin_jkl <= 0 && sin_jkl >= -MIN_SINE )
+ else if ( sin_jkl <= 0 && sin_jkl >= -MIN_SINE )
{
sin_jkl = -MIN_SINE;
}
// dcos_omega_di
- rvec_ScaledSum( dcos_omega_di, (htra-arg*hnra)/r_ij, dvec_ij, -1., dvec_li );
- rvec_ScaledAdd( dcos_omega_di,-(hthd - arg*hnhd)/sin_ijk, p_ijk->dcos_dk );
+ rvec_ScaledSum( dcos_omega_di, (htra - arg * hnra) / r_ij, dvec_ij, -1., dvec_li );
+ rvec_ScaledAdd( dcos_omega_di, -(hthd - arg * hnhd) / sin_ijk, p_ijk->dcos_dk );
rvec_Scale( dcos_omega_di, 2.0 / poem, dcos_omega_di );
// dcos_omega_dj
- rvec_ScaledSum( dcos_omega_dj,-(htra-arg*hnra)/r_ij, dvec_ij,
- -htrb / r_jk, dvec_jk );
- rvec_ScaledAdd( dcos_omega_dj,-(hthd-arg*hnhd) / sin_ijk, p_ijk->dcos_dj );
- rvec_ScaledAdd( dcos_omega_dj,-(hthe-arg*hnhe) / sin_jkl, p_jkl->dcos_di );
+ rvec_ScaledSum( dcos_omega_dj, -(htra - arg * hnra) / r_ij, dvec_ij,
+ -htrb / r_jk, dvec_jk );
+ rvec_ScaledAdd( dcos_omega_dj, -(hthd - arg * hnhd) / sin_ijk, p_ijk->dcos_dj );
+ rvec_ScaledAdd( dcos_omega_dj, -(hthe - arg * hnhe) / sin_jkl, p_jkl->dcos_di );
rvec_Scale( dcos_omega_dj, 2.0 / poem, dcos_omega_dj );
// dcos_omega_dk
- rvec_ScaledSum( dcos_omega_dk,-(htrc-arg*hnrc) / r_kl, dvec_kl,
- htrb / r_jk, dvec_jk );
- rvec_ScaledAdd( dcos_omega_dk,-(hthd-arg*hnhd) / sin_ijk, p_ijk->dcos_di );
- rvec_ScaledAdd( dcos_omega_dk,-(hthe-arg*hnhe) / sin_jkl, p_jkl->dcos_dj );
+ rvec_ScaledSum( dcos_omega_dk, -(htrc - arg * hnrc) / r_kl, dvec_kl,
+ htrb / r_jk, dvec_jk );
+ rvec_ScaledAdd( dcos_omega_dk, -(hthd - arg * hnhd) / sin_ijk, p_ijk->dcos_di );
+ rvec_ScaledAdd( dcos_omega_dk, -(hthe - arg * hnhe) / sin_jkl, p_jkl->dcos_dj );
rvec_Scale( dcos_omega_dk, 2.0 / poem, dcos_omega_dk );
// dcos_omega_dl
- rvec_ScaledSum( dcos_omega_dl, (htrc-arg*hnrc) / r_kl, dvec_kl, 1., dvec_li );
- rvec_ScaledAdd( dcos_omega_dl,-(hthe-arg*hnhe) / sin_jkl, p_jkl->dcos_dk );
+ rvec_ScaledSum( dcos_omega_dl, (htrc - arg * hnrc) / r_kl, dvec_kl, 1., dvec_li );
+ rvec_ScaledAdd( dcos_omega_dl, -(hthe - arg * hnhe) / sin_jkl, p_jkl->dcos_dk );
rvec_Scale( dcos_omega_dl, 2.0 / poem, dcos_omega_dl );
- return omega;
+ return omega;
//return arg;
}
-void Four_Body_Interactions( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+
+
+
+void Four_Body_Interactions( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i, j, k, l, pi, pj, pk, pl, pij, plk;
int type_i, type_j, type_k, type_l;
@@ -212,31 +218,35 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
list *thb_intrs = (*lists) + THREE_BODIES;
- for( j = 0; j < system->N; ++j ) {
+ for ( j = 0; j < system->N; ++j )
+ {
type_j = system->atoms[j].type;
Delta_j = workspace->Delta_boc[j];
start_j = Start_Index(j, bonds);
end_j = End_Index(j, bonds);
- for( pk = start_j; pk < end_j; ++pk ) {
+ for ( pk = start_j; pk < end_j; ++pk )
+ {
pbond_jk = &( bonds->select.bond_list[pk] );
k = pbond_jk->nbr;
bo_jk = &( pbond_jk->bo_data );
BOA_jk = bo_jk->BO - control->thb_cut;
/* see if there are any 3-body interactions involving j&k
- where j is the central atom. Otherwise there is no point in
- trying to form a 4-body interaction out of this neighborhood */
- if( j < k && bo_jk->BO > control->thb_cut/*0*/ &&
- Num_Entries(pk, thb_intrs) ) {
+ where j is the central atom. Otherwise there is no point in
+ trying to form a 4-body interaction out of this neighborhood */
+ if ( j < k && bo_jk->BO > control->thb_cut/*0*/ &&
+ Num_Entries(pk, thb_intrs) )
+ {
start_k = Start_Index(k, bonds);
- end_k = End_Index(k, bonds);
+ end_k = End_Index(k, bonds);
pj = pbond_jk->sym_index; // pj points to j on k's list
- /* do the same check as above: are there any 3-body interactions
+ /* do the same check as above: are there any 3-body interactions
involving k&j where k is the central atom */
- if( Num_Entries(pj, thb_intrs) ) {
+ if ( Num_Entries(pj, thb_intrs) )
+ {
type_k = system->atoms[k].type;
Delta_k = workspace->Delta_boc[k];
r_jk = pbond_jk->d;
@@ -244,7 +254,7 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
start_pk = Start_Index(pk, thb_intrs );
end_pk = End_Index(pk, thb_intrs );
start_pj = Start_Index(pj, thb_intrs );
- end_pj = End_Index(pj, thb_intrs );
+ end_pj = End_Index(pj, thb_intrs );
exp_tor2_jk = EXP( -p_tor2 * BOA_jk );
exp_cot2_jk = EXP( -p_cot2 * SQR(BOA_jk - 1.5) );
@@ -255,14 +265,16 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
/* pick i up from j-k interaction where j is the centre atom */
- for( pi = start_pk; pi < end_pk; ++pi ) {
+ for ( pi = start_pk; pi < end_pk; ++pi )
+ {
p_ijk = &( thb_intrs->select.three_body_list[pi] );
pij = p_ijk->pthb; // pij is pointer to i on j's bond_list
pbond_ij = &( bonds->select.bond_list[pij] );
bo_ij = &( pbond_ij->bo_data );
- if( bo_ij->BO > control->thb_cut/*0*/ ) {
+ if ( bo_ij->BO > control->thb_cut/*0*/ )
+ {
i = p_ijk->thb;
type_i = system->atoms[i].type;
r_ij = pbond_ij->d;
@@ -272,17 +284,18 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
sin_ijk = SIN( theta_ijk );
cos_ijk = COS( theta_ijk );
//tan_ijk_i = 1. / TAN( theta_ijk );
- if( sin_ijk >= 0 && sin_ijk <= MIN_SINE )
+ if ( sin_ijk >= 0 && sin_ijk <= MIN_SINE )
tan_ijk_i = cos_ijk / MIN_SINE;
- else if( sin_ijk <= 0 && sin_ijk >= -MIN_SINE )
+ else if ( sin_ijk <= 0 && sin_ijk >= -MIN_SINE )
tan_ijk_i = cos_ijk / -MIN_SINE;
else tan_ijk_i = cos_ijk / sin_ijk;
exp_tor2_ij = EXP( -p_tor2 * BOA_ij );
- exp_cot2_ij = EXP( -p_cot2 * SQR(BOA_ij -1.5) );
+ exp_cot2_ij = EXP( -p_cot2 * SQR(BOA_ij - 1.5) );
/* pick l up from j-k intr. where k is the centre */
- for( pl = start_pj; pl < end_pj; ++pl ) {
+ for ( pl = start_pj; pl < end_pj; ++pl )
+ {
p_jkl = &( thb_intrs->select.three_body_list[pl] );
l = p_jkl->thb;
plk = p_jkl->pthb; //pointer to l on k's bond_list!
@@ -292,8 +305,9 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
fbh = &(system->reaxprm.fbp[ index_fbp(type_i,type_j,type_k,type_l,system->reaxprm.num_atom_types ) ]);
fbp = &(system->reaxprm.fbp[ index_fbp(type_i,type_j,type_k,type_l,system->reaxprm.num_atom_types )].prm[0]);
- if( i != l && fbh->cnt && bo_kl->BO > control->thb_cut/*0*/ &&
- bo_ij->BO * bo_jk->BO * bo_kl->BO > control->thb_cut/*0*/ ){
+ if ( i != l && fbh->cnt && bo_kl->BO > control->thb_cut/*0*/ &&
+ bo_ij->BO * bo_jk->BO * bo_kl->BO > control->thb_cut/*0*/ )
+ {
++num_frb_intrs;
r_kl = pbond_kl->d;
BOA_kl = bo_kl->BO - control->thb_cut;
@@ -302,77 +316,77 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
sin_jkl = SIN( theta_jkl );
cos_jkl = COS( theta_jkl );
//tan_jkl_i = 1. / TAN( theta_jkl );
- if( sin_jkl >= 0 && sin_jkl <= MIN_SINE )
+ if ( sin_jkl >= 0 && sin_jkl <= MIN_SINE )
tan_jkl_i = cos_jkl / MIN_SINE;
- else if( sin_jkl <= 0 && sin_jkl >= -MIN_SINE )
+ else if ( sin_jkl <= 0 && sin_jkl >= -MIN_SINE )
tan_jkl_i = cos_jkl / -MIN_SINE;
- else tan_jkl_i = cos_jkl /sin_jkl;
+ else tan_jkl_i = cos_jkl / sin_jkl;
- Sq_Distance_on_T3( system->atoms[l].x, system->atoms[i].x,
- &(system->box), dvec_li );
+ Sq_Distance_on_T3( system->atoms[l].x, system->atoms[i].x,
+ &(system->box), dvec_li );
r_li = rvec_Norm( dvec_li );
/* omega and its derivative */
- //cos_omega=Calculate_Omega(pbond_ij->dvec,r_ij,pbond_jk->dvec,
- omega = Calculate_Omega(pbond_ij->dvec, r_ij, pbond_jk->dvec,
- r_jk, pbond_kl->dvec, r_kl,
- dvec_li, r_li, p_ijk, p_jkl,
- dcos_omega_di, dcos_omega_dj,
- dcos_omega_dk, dcos_omega_dl,
- out_control);
+ //cos_omega=Calculate_Omega(pbond_ij->dvec,r_ij,pbond_jk->dvec,
+ omega = Calculate_Omega(pbond_ij->dvec, r_ij, pbond_jk->dvec,
+ r_jk, pbond_kl->dvec, r_kl,
+ dvec_li, r_li, p_ijk, p_jkl,
+ dcos_omega_di, dcos_omega_dj,
+ dcos_omega_dk, dcos_omega_dl,
+ out_control);
cos_omega = COS( omega );
cos2omega = COS( 2. * omega );
cos3omega = COS( 3. * omega );
/* end omega calculations */
/* torsion energy */
- exp_tor1 = EXP(fbp->p_tor1 * SQR(2.-bo_jk->BO_pi-f11_DjDk));
+ exp_tor1 = EXP(fbp->p_tor1 * SQR(2. - bo_jk->BO_pi - f11_DjDk));
exp_tor2_kl = EXP( -p_tor2 * BOA_kl );
- exp_cot2_kl = EXP( -p_cot2 * SQR(BOA_kl-1.5) );
- fn10 = (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_jk) *
- (1.0 - exp_tor2_kl);
-
- CV = 0.5 * ( fbp->V1 * (1.0 + cos_omega) +
- fbp->V2 * exp_tor1 * (1.0 - cos2omega) +
- fbp->V3 * (1.0 + cos3omega) );
- //CV = 0.5 * fbp->V1 * (1.0 + cos_omega) +
+ exp_cot2_kl = EXP( -p_cot2 * SQR(BOA_kl - 1.5) );
+ fn10 = (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_jk) *
+ (1.0 - exp_tor2_kl);
+
+ CV = 0.5 * ( fbp->V1 * (1.0 + cos_omega) +
+ fbp->V2 * exp_tor1 * (1.0 - cos2omega) +
+ fbp->V3 * (1.0 + cos3omega) );
+ //CV = 0.5 * fbp->V1 * (1.0 + cos_omega) +
// fbp->V2 * exp_tor1 * (1.0 - SQR(cos_omega)) +
// fbp->V3 * (0.5 + 2.0*CUBE(cos_omega) - 1.5 * cos_omega);
data->E_Tor += e_tor = fn10 * sin_ijk * sin_jkl * CV;
dfn11 = (-p_tor3 * exp_tor3_DjDk +
- (p_tor3 * exp_tor3_DjDk - p_tor4 * exp_tor4_DjDk) *
- (2.+exp_tor3_DjDk) * exp_tor34_inv) * exp_tor34_inv;
+ (p_tor3 * exp_tor3_DjDk - p_tor4 * exp_tor4_DjDk) *
+ (2. + exp_tor3_DjDk) * exp_tor34_inv) * exp_tor34_inv;
CEtors1 = sin_ijk * sin_jkl * CV;
- CEtors2 = -fn10 * 2.0 * fbp->p_tor1 * fbp->V2 * exp_tor1 *
- (2.0 - bo_jk->BO_pi - f11_DjDk) * (1.0 - SQR(cos_omega)) *
- sin_ijk * sin_jkl;
+ CEtors2 = -fn10 * 2.0 * fbp->p_tor1 * fbp->V2 * exp_tor1 *
+ (2.0 - bo_jk->BO_pi - f11_DjDk) * (1.0 - SQR(cos_omega)) *
+ sin_ijk * sin_jkl;
CEtors3 = CEtors2 * dfn11;
- CEtors4 = CEtors1 * p_tor2 * exp_tor2_ij *
- (1.0 - exp_tor2_jk) * (1.0 - exp_tor2_kl);
+ CEtors4 = CEtors1 * p_tor2 * exp_tor2_ij *
+ (1.0 - exp_tor2_jk) * (1.0 - exp_tor2_kl);
- CEtors5 = CEtors1 * p_tor2 * exp_tor2_jk *
- (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_kl);
+ CEtors5 = CEtors1 * p_tor2 * exp_tor2_jk *
+ (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_kl);
CEtors6 = CEtors1 * p_tor2 * exp_tor2_kl *
- (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_jk);
+ (1.0 - exp_tor2_ij) * (1.0 - exp_tor2_jk);
cmn = -fn10 * CV;
CEtors7 = cmn * sin_jkl * tan_ijk_i;
CEtors8 = cmn * sin_ijk * tan_jkl_i;
- CEtors9 = fn10 * sin_ijk * sin_jkl *
- (0.5 * fbp->V1 - 2.0 * fbp->V2 * exp_tor1 * cos_omega +
- 1.5 * fbp->V3 * (cos2omega + 2. * SQR(cos_omega)));
+ CEtors9 = fn10 * sin_ijk * sin_jkl *
+ (0.5 * fbp->V1 - 2.0 * fbp->V2 * exp_tor1 * cos_omega +
+ 1.5 * fbp->V3 * (cos2omega + 2. * SQR(cos_omega)));
//cmn = -fn10 * CV;
//CEtors7 = cmn * sin_jkl * cos_ijk;
//CEtors8 = cmn * sin_ijk * cos_jkl;
- //CEtors9 = fn10 * sin_ijk * sin_jkl *
+ //CEtors9 = fn10 * sin_ijk * sin_jkl *
// (0.5 * fbp->V1 - 2.0 * fbp->V2 * exp_tor1 * cos_omega +
// fbp->V3 * (6*SQR(cos_omega) - 1.50));
/* end of torsion energy */
@@ -380,38 +394,38 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
/* 4-body conjugation energy */
fn12 = exp_cot2_ij * exp_cot2_jk * exp_cot2_kl;
- data->E_Con += e_con = fbp->p_cot1 * fn12 *
- (1. + (SQR(cos_omega)-1.) * sin_ijk*sin_jkl);
+ data->E_Con += e_con = fbp->p_cot1 * fn12 *
+ (1. + (SQR(cos_omega) - 1.) * sin_ijk * sin_jkl);
- Cconj = -2.0 * fn12 * fbp->p_cot1 * p_cot2 *
- (1. + (SQR(cos_omega)-1.) * sin_ijk*sin_jkl);
+ Cconj = -2.0 * fn12 * fbp->p_cot1 * p_cot2 *
+ (1. + (SQR(cos_omega) - 1.) * sin_ijk * sin_jkl);
CEconj1 = Cconj * (BOA_ij - 1.5e0);
CEconj2 = Cconj * (BOA_jk - 1.5e0);
CEconj3 = Cconj * (BOA_kl - 1.5e0);
- CEconj4 = -fbp->p_cot1 * fn12 *
- (SQR(cos_omega) - 1.0) * sin_jkl * tan_ijk_i;
- CEconj5 = -fbp->p_cot1 * fn12 *
- (SQR(cos_omega) - 1.0) * sin_ijk * tan_jkl_i;
- //CEconj4 = -fbp->p_cot1 * fn12 *
+ CEconj4 = -fbp->p_cot1 * fn12 *
+ (SQR(cos_omega) - 1.0) * sin_jkl * tan_ijk_i;
+ CEconj5 = -fbp->p_cot1 * fn12 *
+ (SQR(cos_omega) - 1.0) * sin_ijk * tan_jkl_i;
+ //CEconj4 = -fbp->p_cot1 * fn12 *
// (SQR(cos_omega) - 1.0) * sin_jkl * cos_ijk;
- //CEconj5 = -fbp->p_cot1 * fn12 *
+ //CEconj5 = -fbp->p_cot1 * fn12 *
// (SQR(cos_omega) - 1.0) * sin_ijk * cos_jkl;
- CEconj6 = 2.0 * fbp->p_cot1 * fn12 *
- cos_omega * sin_ijk * sin_jkl;
+ CEconj6 = 2.0 * fbp->p_cot1 * fn12 *
+ cos_omega * sin_ijk * sin_jkl;
/* end 4-body conjugation energy */
//fprintf(stdout, "%6d %6d %6d %6d %7.3f %7.3f %7.3f %7.3f ",
// workspace->orig_id[i], workspace->orig_id[j],
- // workspace->orig_id[k], workspace->orig_id[l],
+ // workspace->orig_id[k], workspace->orig_id[l],
// omega, cos_omega, cos2omega, cos3omega );
- //fprintf(stdout,
+ //fprintf(stdout,
// "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
- // CEtors2, CEtors3, CEtors4, CEtors5,
+ // CEtors2, CEtors3, CEtors4, CEtors5,
// CEtors6, CEtors7, CEtors8, CEtors9 );
//fprintf(stdout, "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
- // theta_ijk, theta_jkl, sin_ijk,
+ // theta_ijk, theta_jkl, sin_ijk,
// sin_jkl, cos_jkl, tan_jkl_i );
/* forces */
@@ -420,37 +434,38 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
workspace->CdDelta[k] += CEtors3;
bo_ij->Cdbo += (CEtors4 + CEconj1);
bo_jk->Cdbo += (CEtors5 + CEconj2);
-
bo_kl->Cdbo += (CEtors6 + CEconj3);
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT) {
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT )
+ {
/* dcos_theta_ijk */
- rvec_ScaledAdd( system->atoms[i].f,
- CEtors7 + CEconj4, p_ijk->dcos_dk );
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors7 + CEconj4, p_ijk->dcos_dj );
- rvec_ScaledAdd( system->atoms[k].f,
- CEtors7 + CEconj4, p_ijk->dcos_di );
+ rvec_ScaledAdd( system->atoms[i].f,
+ CEtors7 + CEconj4, p_ijk->dcos_dk );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors7 + CEconj4, p_ijk->dcos_dj );
+ rvec_ScaledAdd( system->atoms[k].f,
+ CEtors7 + CEconj4, p_ijk->dcos_di );
/* dcos_theta_jkl */
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors8 + CEconj5, p_jkl->dcos_di );
- rvec_ScaledAdd( system->atoms[k].f,
- CEtors8 + CEconj5, p_jkl->dcos_dj );
- rvec_ScaledAdd( system->atoms[l].f,
- CEtors8 + CEconj5, p_jkl->dcos_dk );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors8 + CEconj5, p_jkl->dcos_di );
+ rvec_ScaledAdd( system->atoms[k].f,
+ CEtors8 + CEconj5, p_jkl->dcos_dj );
+ rvec_ScaledAdd( system->atoms[l].f,
+ CEtors8 + CEconj5, p_jkl->dcos_dk );
/* dcos_omega */
- rvec_ScaledAdd( system->atoms[i].f,
- CEtors9 + CEconj6, dcos_omega_di );
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors9 + CEconj6, dcos_omega_dj );
- rvec_ScaledAdd( system->atoms[k].f,
- CEtors9 + CEconj6, dcos_omega_dk );
- rvec_ScaledAdd( system->atoms[l].f,
- CEtors9 + CEconj6, dcos_omega_dl );
+ rvec_ScaledAdd( system->atoms[i].f,
+ CEtors9 + CEconj6, dcos_omega_di );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors9 + CEconj6, dcos_omega_dj );
+ rvec_ScaledAdd( system->atoms[k].f,
+ CEtors9 + CEconj6, dcos_omega_dk );
+ rvec_ScaledAdd( system->atoms[l].f,
+ CEtors9 + CEconj6, dcos_omega_dl );
}
- else {
+ else
+ {
ivec_Sum(rel_box_jl, pbond_jk->rel_box, pbond_kl->rel_box);
/* dcos_theta_ijk */
@@ -459,8 +474,8 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
rvec_iMultiply( ext_press, pbond_ij->rel_box, force );
rvec_Add( data->ext_press, ext_press );
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors7 + CEconj4, p_ijk->dcos_dj );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors7 + CEconj4, p_ijk->dcos_dj );
rvec_Scale( force, CEtors7 + CEconj4, p_ijk->dcos_di );
rvec_Add( system->atoms[k].f, force );
@@ -469,8 +484,8 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
/* dcos_theta_jkl */
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors8 + CEconj5, p_jkl->dcos_di );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors8 + CEconj5, p_jkl->dcos_di );
rvec_Scale( force, CEtors8 + CEconj5, p_jkl->dcos_dj );
rvec_Add( system->atoms[k].f, force );
@@ -483,14 +498,14 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
rvec_Add( data->ext_press, ext_press );
- /* dcos_omega */
+ /* dcos_omega */
rvec_Scale( force, CEtors9 + CEconj6, dcos_omega_di );
rvec_Add( system->atoms[i].f, force );
rvec_iMultiply( ext_press, pbond_ij->rel_box, force );
rvec_Add( data->ext_press, ext_press );
- rvec_ScaledAdd( system->atoms[j].f,
- CEtors9 + CEconj6, dcos_omega_dj );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEtors9 + CEconj6, dcos_omega_dj );
rvec_Scale( force, CEtors9 + CEconj6, dcos_omega_dk );
rvec_Add( system->atoms[k].f, force );
@@ -504,39 +519,39 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
/* This part is intended for a fully-flexible box */
- /* rvec_ScaledSum( temp_rvec,
- CEtors7 + CEconj4, p_ijk->dcos_dk, // i
+ /* rvec_ScaledSum( temp_rvec,
+ CEtors7 + CEconj4, p_ijk->dcos_dk, // i
CEtors9 + CEconj6, dcos_omega_di );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[i].x );
rtensor_Copy( total_rtensor, temp_rtensor );
- rvec_ScaledSum( temp_rvec,
+ rvec_ScaledSum( temp_rvec,
CEtors7 + CEconj4, p_ijk->dcos_dj, // j
CEtors8 + CEconj5, p_jkl->dcos_di );
- rvec_ScaledAdd( temp_rvec,
+ rvec_ScaledAdd( temp_rvec,
CEtors9 + CEconj6, dcos_omega_dj );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[j].x );
rtensor_Add( total_rtensor, temp_rtensor );
- rvec_ScaledSum( temp_rvec,
+ rvec_ScaledSum( temp_rvec,
CEtors7 + CEconj4, p_ijk->dcos_di, // k
CEtors8 + CEconj5, p_jkl->dcos_dj );
- rvec_ScaledAdd( temp_rvec,
+ rvec_ScaledAdd( temp_rvec,
CEtors9 + CEconj6, dcos_omega_dk );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[k].x );
rtensor_Add( total_rtensor, temp_rtensor );
- rvec_ScaledSum( temp_rvec,
+ rvec_ScaledSum( temp_rvec,
CEtors8 + CEconj5, p_jkl->dcos_dk, // l
CEtors9 + CEconj6, dcos_omega_dl );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[l].x );
rtensor_Copy( total_rtensor, temp_rtensor );
- if( pbond_ij->imaginary || pbond_jk->imaginary ||
+ if( pbond_ij->imaginary || pbond_jk->imaginary ||
pbond_kl->imaginary )
rtensor_ScaledAdd( data->flex_bar.P, -1., total_rtensor );
else
@@ -544,82 +559,82 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
}
#ifdef TEST_ENERGY
- /*fprintf( out_control->etor,
- //"%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
- //r_ij, r_jk, r_kl,
- "%12.8f%12.8f%12.8f%12.8f\n",
- cos_ijk, cos_jkl, sin_ijk, sin_jkl );*/
+ /*fprintf( out_control->etor,
+ //"%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
+ //r_ij, r_jk, r_kl,
+ "%12.8f%12.8f%12.8f%12.8f\n",
+ cos_ijk, cos_jkl, sin_ijk, sin_jkl );*/
// fprintf( out_control->etor, "%12.8f\n", dfn11 );
- fprintf( out_control->etor, "%12.8f%12.8f%12.8f\n",
- fn10, cos_omega, CV );
+ fprintf( out_control->etor, "%12.8f%12.8f%12.8f\n",
+ fn10, cos_omega, CV );
- fprintf( out_control->etor,
- "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
- CEtors2, CEtors3, CEtors4, CEtors5,
- CEtors6, CEtors7, CEtors8, CEtors9 );
+ fprintf( out_control->etor,
+ "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
+ CEtors2, CEtors3, CEtors4, CEtors5,
+ CEtors6, CEtors7, CEtors8, CEtors9 );
- /* fprintf( out_control->etor,
+ /* fprintf( out_control->etor,
"%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
htra, htrb, htrc, hthd, hthe, hnra, hnrc, hnhd, hnhe ); */
- fprintf( out_control->etor,
- "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
- CEconj1, CEconj2, CEconj3, CEconj4, CEconj5, CEconj6 );
+ fprintf( out_control->etor,
+ "%12.8f%12.8f%12.8f%12.8f%12.8f%12.8f\n",
+ CEconj1, CEconj2, CEconj3, CEconj4, CEconj5, CEconj6 );
/* fprintf(out_control->etor,"%23.15e%23.15e%23.15e%23.15e\n",
fbp->V1, fbp->V2, fbp->V3, fbp->p_tor1 );*/
- fprintf( out_control->etor,
- //"%6d%6d%6d%6d%23.15e%23.15e%23.15e%23.15e\n",
- "%6d%6d%6d%6d%12.8f%12.8f\n",
- workspace->orig_id[i], workspace->orig_id[j],
- workspace->orig_id[k], workspace->orig_id[l],
- e_tor, e_con );
+ fprintf( out_control->etor,
+ //"%6d%6d%6d%6d%23.15e%23.15e%23.15e%23.15e\n",
+ "%6d%6d%6d%6d%12.8f%12.8f\n",
+ workspace->orig_id[i], workspace->orig_id[j],
+ workspace->orig_id[k], workspace->orig_id[l],
+ e_tor, e_con );
//RAD2DEG(omega), BOA_jk, e_tor, data->E_Tor );
- fprintf( out_control->econ,
- "%6d%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e\n",
- workspace->orig_id[i], workspace->orig_id[j],
- workspace->orig_id[k], workspace->orig_id[l],
- RAD2DEG(omega), BOA_ij, BOA_jk, BOA_kl,
- e_con,data->E_Con );
-
- /* fprintf( out_control->etor,
- "%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n",
- (CEtors7 + CEconj4)*p_ijk->dcos_dk[0],
- (CEtors7 + CEconj4)*p_ijk->dcos_dk[1],
+ fprintf( out_control->econ,
+ "%6d%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e\n",
+ workspace->orig_id[i], workspace->orig_id[j],
+ workspace->orig_id[k], workspace->orig_id[l],
+ RAD2DEG(omega), BOA_ij, BOA_jk, BOA_kl,
+ e_con, data->E_Con );
+
+ /* fprintf( out_control->etor,
+ "%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n",
+ (CEtors7 + CEconj4)*p_ijk->dcos_dk[0],
+ (CEtors7 + CEconj4)*p_ijk->dcos_dk[1],
(CEtors7 + CEconj4)*p_ijk->dcos_dk[2],
- (CEtors7 + CEconj4)*p_ijk->dcos_dj[0],
- (CEtors7 + CEconj4)*p_ijk->dcos_dj[1],
+ (CEtors7 + CEconj4)*p_ijk->dcos_dj[0],
+ (CEtors7 + CEconj4)*p_ijk->dcos_dj[1],
(CEtors7 + CEconj4)*p_ijk->dcos_dj[2],
- (CEtors7 + CEconj4)*p_ijk->dcos_di[0],
- (CEtors7 + CEconj4)*p_ijk->dcos_di[1],
+ (CEtors7 + CEconj4)*p_ijk->dcos_di[0],
+ (CEtors7 + CEconj4)*p_ijk->dcos_di[1],
(CEtors7 + CEconj4)*p_ijk->dcos_di[2] ); */
- /* fprintf( out_control->etor,
+ /* fprintf( out_control->etor,
"%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n",
- (CEtors8 + CEconj5)*p_jkl->dcos_di[0],
- (CEtors8 + CEconj5)*p_jkl->dcos_di[1],
- (CEtors8 + CEconj5)*p_jkl->dcos_di[2],
- (CEtors8 + CEconj5)*p_jkl->dcos_dj[0],
- (CEtors8 + CEconj5)*p_jkl->dcos_dj[1],
- (CEtors8 + CEconj5)*p_jkl->dcos_dj[2],
- (CEtors8 + CEconj5)*p_jkl->dcos_dk[0],
- (CEtors8 + CEconj5)*p_jkl->dcos_dk[1],
+ (CEtors8 + CEconj5)*p_jkl->dcos_di[0],
+ (CEtors8 + CEconj5)*p_jkl->dcos_di[1],
+ (CEtors8 + CEconj5)*p_jkl->dcos_di[2],
+ (CEtors8 + CEconj5)*p_jkl->dcos_dj[0],
+ (CEtors8 + CEconj5)*p_jkl->dcos_dj[1],
+ (CEtors8 + CEconj5)*p_jkl->dcos_dj[2],
+ (CEtors8 + CEconj5)*p_jkl->dcos_dk[0],
+ (CEtors8 + CEconj5)*p_jkl->dcos_dk[1],
(CEtors8 + CEconj5)*p_jkl->dcos_dk[2] ); */
- fprintf( out_control->etor,
- "%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n",
- dcos_omega_di[0], dcos_omega_di[1], dcos_omega_di[2],
- dcos_omega_dj[0], dcos_omega_dj[1], dcos_omega_dj[2],
- dcos_omega_dk[0], dcos_omega_dk[1], dcos_omega_dk[2],
- dcos_omega_dl[0], dcos_omega_dl[1], dcos_omega_dl[2] );
+ fprintf( out_control->etor,
+ "%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n%12.8f%12.8f%12.8f\n",
+ dcos_omega_di[0], dcos_omega_di[1], dcos_omega_di[2],
+ dcos_omega_dj[0], dcos_omega_dj[1], dcos_omega_dj[2],
+ dcos_omega_dk[0], dcos_omega_dk[1], dcos_omega_dk[2],
+ dcos_omega_dl[0], dcos_omega_dl[1], dcos_omega_dl[2] );
#endif
#ifdef TEST_FORCES
- // Torsion Forces
- Add_dBOpinpi2(system, lists, j, pk, CEtors2, 0.,
- workspace->f_tor, workspace->f_tor);
+ /* Torsion Forces */
+ Add_dBOpinpi2(system, lists, j, pk, CEtors2, 0.,
+ workspace->f_tor, workspace->f_tor);
Add_dDelta( system, lists, j, CEtors3, workspace->f_tor );
Add_dDelta( system, lists, k, CEtors3, workspace->f_tor );
Add_dBO( system, lists, j, pij, CEtors4, workspace->f_tor );
@@ -639,7 +654,7 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
rvec_ScaledAdd( workspace->f_tor[k], CEtors9, dcos_omega_dk );
rvec_ScaledAdd( workspace->f_tor[l], CEtors9, dcos_omega_dl );
- // Conjugation Forces
+ /* Conjugation Forces */
Add_dBO( system, lists, j, pij, CEconj1, workspace->f_con );
Add_dBO( system, lists, j, pk, CEconj2, workspace->f_con );
Add_dBO( system, lists, k, plk, CEconj3, workspace->f_con );
@@ -666,12 +681,12 @@ void Four_Body_Interactions( reax_system *system, control_params *control,
} // pk loop ends
} // j loop
- /* fprintf( stderr, "4body: ext_press (%23.15e %23.15e %23.15e)\n",
+ /* fprintf( stderr, "4body: ext_press (%23.15e %23.15e %23.15e)\n",
data->ext_press[0], data->ext_press[1], data->ext_press[2] );*/
#ifdef TEST_FORCES
fprintf( stderr, "Number of torsion angles: %d\n", num_frb_intrs );
- fprintf( stderr, "Torsion Energy: %g\t Conjugation Energy: %g\n",
- data->E_Tor, data->E_Con );
+ fprintf( stderr, "Torsion Energy: %g\t Conjugation Energy: %g\n",
+ data->E_Tor, data->E_Con );
#endif
}
diff --git a/PuReMD-GPU/src/four_body_interactions.h b/PuReMD-GPU/src/four_body_interactions.h
index 8e8dd7c0991a747000e77b2d460711e433db52ef..65e315a94f95239ad8c7081a00c32cc0d3264cd3 100644
--- a/PuReMD-GPU/src/four_body_interactions.h
+++ b/PuReMD-GPU/src/four_body_interactions.h
@@ -23,10 +23,9 @@
#include "mytypes.h"
-#define MIN_SINE 1e-10
-
void Four_Body_Interactions( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
+
#endif
diff --git a/PuReMD-GPU/src/geo_tools.c b/PuReMD-GPU/src/geo_tools.c
new file mode 100644
index 0000000000000000000000000000000000000000..f3c3bd48140f9fd46ac3e460a0ab4bd33ade85b9
--- /dev/null
+++ b/PuReMD-GPU/src/geo_tools.c
@@ -0,0 +1,797 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#include <ctype.h>
+
+#include "geo_tools.h"
+#include "allocate.h"
+#include "box.h"
+#include "list.h"
+#include "restart.h"
+#include "tool_box.h"
+#include "vector.h"
+
+
+/********************* geo format routines ******************/
+void Count_Geo_Atoms( FILE *geo, reax_system *system )
+{
+ int i, serial;
+ rvec x;
+ char element[3], name[9], line[MAX_LINE + 1];
+
+ /* total number of atoms */
+ fscanf( geo, " %d", &(system->N) );
+
+ /* count atoms */
+ for ( i = 0; i < system->N; ++i )
+ {
+ fscanf( geo, CUSTOM_ATOM_FORMAT,
+ &serial, element, name, &x[0], &x[1], &x[2] );
+ Fit_to_Periodic_Box( &(system->box), &x );
+ }
+
+ fseek( geo, 0, SEEK_SET ); // set the pointer to the beginning of the file
+ fgets( line, MAX_LINE, geo );
+ fgets( line, MAX_LINE, geo );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "N = %d\n\n", system->N );
+#endif
+}
+
+
+char Read_Geo( char* geo_file, reax_system* system, control_params *control,
+ simulation_data *data, static_storage *workspace )
+{
+
+ FILE *geo;
+ char descriptor[9];
+ int i, serial, top;
+ real box_x, box_y, box_z, alpha, beta, gamma;
+ rvec x;
+ char element[3], name[9];
+ reax_atom *atom;
+
+ /* open the geometry file */
+ if ( (geo = fopen(geo_file, "r")) == NULL )
+ {
+ fprintf( stderr, "Error opening the geo file! terminating...\n" );
+ exit( FILE_NOT_FOUND );
+ }
+
+ /* read box information */
+ fscanf( geo, CUSTOM_BOXGEO_FORMAT,
+ descriptor, &box_x, &box_y, &box_z, &alpha, &beta, &gamma );
+ /* initialize the box */
+ Setup_Box( box_x, box_y, box_z, alpha, beta, gamma, &(system->box) );
+
+ /* count my atoms & allocate storage */
+ Count_Geo_Atoms( geo, system );
+ if ( PreAllocate_Space( system, control, workspace ) == FAILURE )
+ {
+ fprintf( stderr, "PreAllocate_Space: not enough memory!" );
+ fprintf( stderr, "terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* read in my atom info */
+ top = 0;
+ for ( i = 0; i < system->N; ++i )
+ {
+ fscanf( geo, CUSTOM_ATOM_FORMAT,
+ &serial, element, name, &x[0], &x[1], &x[2] );
+ Fit_to_Periodic_Box( &(system->box), &x );
+#if defined(DEBUG)
+ fprintf( stderr, "atom%d: %s %s %f %f %f\n",
+ serial, element, name, x[0], x[1], x[2] );
+#endif
+
+ atom = &(system->atoms[top]);
+ workspace->orig_id[i] = serial;
+ atom->type = Get_Atom_Type( &(system->reaxprm), element );
+ strcpy( atom->name, name );
+ rvec_Copy( atom->x, x );
+ rvec_MakeZero( atom->v );
+ rvec_MakeZero( atom->f );
+ atom->q = 0.;
+
+ top++;
+ }
+
+ fclose( geo );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "finished reading the geo file\n" );
+#endif
+
+ return SUCCESS;
+}
+
+
+int Read_Box_Info( reax_system *system, FILE *geo, int geo_format )
+{
+ char *cryst;
+ char line[MAX_LINE + 1];
+ char descriptor[9];
+ char s_a[12], s_b[12], s_c[12], s_alpha[12], s_beta[12], s_gamma[12];
+ char s_group[12], s_zValue[12];
+
+ /* initialize variables */
+ fseek( geo, 0, SEEK_SET ); // set the pointer to the beginning of the file
+
+ switch ( geo_format )
+ {
+ case PDB:
+ cryst = "CRYST1";
+ break;
+ default:
+ cryst = "BOX";
+ }
+
+ /* locate the cryst line in the geo file, read it and
+ initialize the big box */
+ while ( fgets( line, MAX_LINE, geo ) )
+ {
+ if ( strncmp( line, cryst, 6 ) == 0 )
+ {
+ if ( geo_format == PDB )
+ sscanf( line, PDB_CRYST1_FORMAT,
+ &descriptor[0],
+ &s_a[0], &s_b[0], &s_c[0],
+ &s_alpha[0], &s_beta[0], &s_gamma[0],
+ &s_group[0], &s_zValue[0] );
+
+ /* compute full volume tensor from the angles */
+ Setup_Box( atof(s_a), atof(s_b), atof(s_c),
+ atof(s_alpha), atof(s_beta), atof(s_gamma),
+ &(system->box) );
+ return SUCCESS;
+ }
+ }
+ if ( ferror( geo ) )
+ {
+ return FAILURE;
+ }
+
+ return FAILURE;
+}
+
+
+void Count_PDB_Atoms( FILE *geo, reax_system *system )
+{
+ char *endptr = NULL;
+ char line[MAX_LINE + 1];
+ char s_x[9], s_y[9], s_z[9];
+ rvec x;
+
+ /* initialize variables */
+ fseek( geo, 0, SEEK_SET ); /* set the pointer to the beginning of the file */
+ system->N = 0;
+
+ /* increment number of atoms for each line denoting an atom desc */
+ while ( fgets( line, MAX_LINE, geo ) )
+ {
+ if ( strncmp( line, "ATOM", 4 ) == 0 ||
+ strncmp( line, "HETATM", 6 ) == 0 )
+ {
+ system->N++;
+
+ strncpy( s_x, line + 30, 8 );
+ s_x[8] = 0;
+ strncpy( s_y, line + 38, 8 );
+ s_y[8] = 0;
+ strncpy( s_z, line + 46, 8 );
+ s_z[8] = 0;
+ Make_Point( strtod( s_x, &endptr ), strtod( s_y, &endptr ),
+ strtod( s_z, &endptr ), &x );
+ Fit_to_Periodic_Box( &(system->box), &x );
+ }
+ }
+
+#if defined(DEBUG)
+ fprintf( stderr, "count atoms:\n" );
+ fprintf( stderr, "N = %d\n\n", system->N );
+#endif
+}
+
+
+char Read_PDB( char* pdb_file, reax_system* system, control_params *control,
+ simulation_data *data, static_storage *workspace )
+{
+
+ FILE *pdb;
+ char **tmp;
+ char *s, *s1;
+ char descriptor[9], serial[9];
+ char atom_name[9], res_name[9], res_seq[9];
+ char s_x[9], s_y[9], s_z[9];
+ char occupancy[9], temp_factor[9];
+ char seg_id[9], element[9], charge[9];
+ char alt_loc, chain_id, icode;
+ char *endptr = NULL;
+ int i, c, c1, pdb_serial, top;
+ rvec x;
+ reax_atom *atom;
+
+ /* open pdb file */
+ if ( (pdb = fopen(pdb_file, "r")) == NULL )
+ {
+ fprintf( stderr, "fopen: error opening the pdb file! terminating...\n" );
+ exit( FILE_NOT_FOUND );
+ }
+
+ /* allocate memory for tokenizing pdb lines */
+ if ( Allocate_Tokenizer_Space( &s, &s1, &tmp ) == FAILURE )
+ {
+ fprintf( stderr, "Allocate_Tokenizer_Space: not enough memory!" );
+ fprintf( stderr, "terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* read box information */
+ if ( Read_Box_Info( system, pdb, PDB ) == FAILURE )
+ {
+ fprintf( stderr, "Read_Box_Info: no CRYST line in the pdb file!" );
+ fprintf( stderr, "terminating...\n" );
+ exit( INVALID_GEO );
+ }
+
+ Count_PDB_Atoms( pdb, system );
+ if ( PreAllocate_Space( system, control, workspace ) == FAILURE )
+ {
+ fprintf( stderr, "PreAllocate_Space: not enough memory!" );
+ fprintf( stderr, "terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* start reading and processing the pdb file */
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "starting to read the pdb file\n" );
+#endif
+ fseek( pdb, 0, SEEK_SET );
+ c = 0;
+ c1 = 0;
+ top = 0;
+ s[0] = 0;
+
+ while ( fgets( s, MAX_LINE, pdb ) )
+ {
+ /* read new line and tokenize it */
+ strncpy( s1, s, MAX_LINE - 1 );
+ c1 = Tokenize( s, &tmp );
+
+ /* process new line */
+ if ( strncmp(tmp[0], "ATOM", 4) == 0 || strncmp(tmp[0], "HETATM", 6) == 0 )
+ {
+ if ( strncmp(tmp[0], "ATOM", 4) == 0 )
+ {
+ strncpy( &descriptor[0], s1, 6 );
+ descriptor[6] = 0;
+ strncpy( &serial[0], s1 + 6, 5 );
+ serial[5] = 0;
+ strncpy( &atom_name[0], s1 + 12, 4 );
+ atom_name[4] = 0;
+ //strncpy( &serial[0], s1+6, 7 ); serial[7] = 0;
+ //strncpy( &atom_name[0], s1+13, 3 ); atom_name[3] = 0;
+ alt_loc = s1[16];
+ strncpy( &res_name[0], s1 + 17, 3 );
+ res_name[3] = 0;
+ chain_id = s1[21];
+ strncpy( &res_seq[0], s1 + 22, 4 );
+ res_seq[4] = 0;
+ icode = s1[26];
+ strncpy( &s_x[0], s1 + 30, 8 );
+ s_x[8] = 0;
+ strncpy( &s_y[0], s1 + 38, 8 );
+ s_y[8] = 0;
+ strncpy( &s_z[0], s1 + 46, 8 );
+ s_z[8] = 0;
+ strncpy( &occupancy[0], s1 + 54, 6 );
+ occupancy[6] = 0;
+ strncpy( &temp_factor[0], s1 + 60, 6 );
+ temp_factor[6] = 0;
+ strncpy( &seg_id[0], s1 + 72, 4 );
+ seg_id[4] = 0;
+ strncpy( &element[0], s1 + 76, 2 );
+ element[2] = 0;
+ strncpy( &charge[0], s1 + 78, 2 );
+ charge[2] = 0;
+ }
+ else if (strncmp(tmp[0], "HETATM", 6) == 0)
+ {
+ strncpy( &descriptor[0], s1, 6 );
+ descriptor[6] = 0;
+ strncpy( &serial[0], s1 + 6, 5 );
+ serial[5] = 0;
+ strncpy( &atom_name[0], s1 + 12, 4 );
+ atom_name[4] = 0;
+ //strncpy( &serial[0], s1+6, 7 ); serial[7] = 0;
+ //strncpy( &atom_name[0], s1+13, 3 ); atom_name[3] = 0;
+ alt_loc = s1[16];
+ strncpy( &res_name[0], s1 + 17, 3 );
+ res_name[3] = 0;
+ chain_id = s1[21];
+ strncpy( &res_seq[0], s1 + 22, 4 );
+ res_seq[4] = 0;
+ icode = s1[26];
+ strncpy( &s_x[0], s1 + 30, 8 );
+ s_x[8] = 0;
+ strncpy( &s_y[0], s1 + 38, 8 );
+ s_y[8] = 0;
+ strncpy( &s_z[0], s1 + 46, 8 );
+ s_z[8] = 0;
+ strncpy( &occupancy[0], s1 + 54, 6 );
+ occupancy[6] = 0;
+ strncpy( &temp_factor[0], s1 + 60, 6 );
+ temp_factor[6] = 0;
+ //strncpy( &seg_id[0], s1+72, 4 ); seg_id[4] = 0;
+ strncpy( &element[0], s1 + 76, 2 );
+ element[2] = 0;
+ strncpy( &charge[0], s1 + 78, 2 );
+ charge[2] = 0;
+ }
+
+ /* if the point is inside my_box, add it to my lists */
+ Make_Point( strtod( &s_x[0], &endptr ),
+ strtod( &s_y[0], &endptr ),
+ strtod( &s_z[0], &endptr ), &x );
+
+ Fit_to_Periodic_Box( &(system->box), &x );
+
+ /* store orig_id, type, name and coord info of the new atom */
+ atom = &(system->atoms[top]);
+ pdb_serial = (int) strtod( &serial[0], &endptr );
+ workspace->orig_id[top] = pdb_serial;
+
+ Trim_Spaces( element );
+ atom->type = Get_Atom_Type( &(system->reaxprm), element );
+ strcpy( atom->name, atom_name );
+
+ rvec_Copy( atom->x, x );
+ rvec_MakeZero( atom->v );
+ rvec_MakeZero( atom->f );
+ atom->q = 0;
+
+ top++;
+ // fprintf( stderr, "p%d: %6d%2d x:%8.3f%8.3f%8.3f"
+ // "q:%8.3f occ:%s temp:%s seg:%s elmnt:%s\n",
+ // system->my_rank,
+ // c, system->my_atoms[top].type,
+ // system->my_atoms[top].x[0],
+ // system->my_atoms[top].x[1],
+ // system->my_atoms[top].x[2],
+ // system->my_atoms[top].q, occupancy, temp_factor,
+ // seg_id, element );
+
+ //fprintf( stderr, "atom( %8.3f %8.3f %8.3f ) --> p%d\n",
+ // system->my_atoms[top].x[0], system->my_atoms[top].x[1],
+ // system->my_atoms[top].x[2], system->my_rank );
+
+ c++;
+ }
+
+ /* IMPORTANT: We do not check for the soundness of restrictions here.
+ When atom2 is on atom1's restricted list, and there is a restriction
+ on atom2, then atom1 has to be on atom2's restricted list, too.
+ However, we do not check if this is the case in the input file,
+ this is upto the user. */
+ else if (!strncmp( tmp[0], "CONECT", 6 ))
+ {
+ if ( control->restrict_bonds )
+ {
+ /* error check */
+ // Check_Input_Range( c1 - 2, 0, MAX_RESTRICT,
+ // "CONECT line exceeds max num restrictions allowed.\n" );
+
+ /* read bond restrictions */
+ // if( is_Valid_Serial( workspace, pdb_serial = atoi(tmp[1]) ) )
+ // ratom = workspace->map_serials[ pdb_serial ];
+
+ // workspace->restricted[ ratom ] = c1 - 2;
+ // for( i = 2; i < c1; ++i )
+ // {
+ // if( is_Valid_Serial(workspace, pdb_serial = atoi(tmp[i])) )
+ // workspace->restricted_list[ ratom ][ i-2 ] =
+ // workspace->map_serials[ pdb_serial ];
+ // }
+
+ // fprintf( stderr, "restriction on %d:", ratom );
+ // for( i = 0; i < workspace->restricted[ ratom ]; ++i )
+ // fprintf( stderr, " %d",
+ // workspace->restricted_list[ratom][i] );
+ // fprintf( stderr, "\n" );
+ }
+ }
+
+ /* clear previous input line */
+ s[0] = 0;
+ for ( i = 0; i < c1; ++i )
+ tmp[i][0] = 0;
+ }
+ if ( ferror( pdb ) )
+ {
+ return FAILURE;
+ }
+
+ fclose( pdb );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "finished reading the pdb file\n" );
+#endif
+
+ return SUCCESS;
+}
+
+
+/* PDB serials are written without regard to the order, we'll see if this
+ cause trouble, if so we'll have to rethink this approach
+ Also, we do not write connect lines yet.
+*/
+char Write_PDB( reax_system* system, list* bonds, simulation_data *data,
+ control_params *control, static_storage *workspace, output_controls *out_control )
+{
+ int i, buffer_req, buffer_len;
+ //int j, connect[4];
+ char name[8];
+ //real bo;
+ real alpha, beta, gamma;
+ reax_atom *p_atom;
+ char fname[MAX_STR];
+ char *line;
+ char *buffer;
+ FILE *pdb;
+
+ /* Allocation */
+ line = (char*) smalloc( sizeof(char) * PDB_ATOM_FORMAT_O_LENGTH, "geo:line" );
+ buffer_req = system->N * PDB_ATOM_FORMAT_O_LENGTH;
+
+ buffer = (char*) smalloc( sizeof(char) * buffer_req, "geo:buffer" );
+
+ pdb = NULL;
+ line[0] = 0;
+ buffer[0] = 0;
+ /* Writing Box information */
+ gamma = ACOS( (system->box.box[0][0] * system->box.box[1][0] +
+ system->box.box[0][1] * system->box.box[1][1] +
+ system->box.box[0][2] * system->box.box[1][2]) /
+ (system->box.box_norms[0] * system->box.box_norms[1]) );
+ beta = ACOS( (system->box.box[0][0] * system->box.box[2][0] +
+ system->box.box[0][1] * system->box.box[2][1] +
+ system->box.box[0][2] * system->box.box[2][2]) /
+ (system->box.box_norms[0] * system->box.box_norms[2]) );
+ alpha = ACOS( (system->box.box[2][0] * system->box.box[1][0] +
+ system->box.box[2][1] * system->box.box[1][1] +
+ system->box.box[2][2] * system->box.box[1][2]) /
+ (system->box.box_norms[2] * system->box.box_norms[1]) );
+
+ /*open pdb and write header*/
+ sprintf(fname, "%s-%d.pdb", control->sim_name, data->step);
+ pdb = fopen(fname, "w");
+ fprintf( pdb, PDB_CRYST1_FORMAT_O,
+ "CRYST1",
+ system->box.box_norms[0], system->box.box_norms[1],
+ system->box.box_norms[2],
+ RAD2DEG(alpha), RAD2DEG(beta), RAD2DEG(gamma), " ", 0 );
+ fprintf( out_control->log, "Box written\n" );
+ fflush( out_control->log );
+
+ /*write atom lines to buffer*/
+ for ( i = 0; i < system->N; i++)
+ {
+ p_atom = &(system->atoms[i]);
+ strncpy(name, p_atom->name, 8);
+ Trim_Spaces(name);
+ sprintf( line, PDB_ATOM_FORMAT_O,
+ "ATOM ", workspace->orig_id[i], p_atom->name, ' ', "REX", ' ', 1, ' ',
+ p_atom->x[0], p_atom->x[1], p_atom->x[2],
+ 1.0, 0.0, "0", name, " " );
+ fprintf( stderr, "PDB NAME <%s>\n", p_atom->name );
+ strncpy( buffer + i * PDB_ATOM_FORMAT_O_LENGTH, line,
+ PDB_ATOM_FORMAT_O_LENGTH );
+ }
+
+ buffer_len = system->N * PDB_ATOM_FORMAT_O_LENGTH;
+ buffer[buffer_len] = 0;
+
+ fprintf( pdb, "%s", buffer );
+ fclose( pdb );
+
+ /* Writing connect information */
+ /*
+ for(i=0; i < system->N; i++) {
+ count = 0;
+ for(j = Start_Index(i, bonds); j < End_Index(i, bonds); ++j) {
+ bo = bonds->bond_list[j].bo_data.BO;
+ if (bo > 0.3) {
+ connect[count] = bonds->bond_list[j].nbr+1;
+ count++;
+ }
+ }
+
+ fprintf( out_control->pdb, "%6s%5d", "CONECT", i+1 );
+ for( k=0; k < count; k++ )
+ fprintf( out_control->pdb, "%5d", connect[k] );
+ fprintf( out_control->pdb, "\n" );
+ }
+ */
+
+ free(buffer);
+ free(line);
+
+ return SUCCESS;
+}
+
+
+char Read_BGF( char* bgf_file, reax_system* system, control_params *control,
+ simulation_data *data, static_storage *workspace )
+{
+ FILE *bgf;
+ char **tokens;
+ char *line, *backup;
+ char descriptor[10], serial[10];
+ char atom_name[10], res_name[10], res_seq[10];
+ char s_x[12], s_y[12], s_z[12];
+ char occupancy[10], temp_factor[10];
+ char element[10], charge[10];
+ char chain_id;
+ char s_a[12], s_b[12], s_c[12], s_alpha[12], s_beta[12], s_gamma[12];
+ char *endptr = NULL;
+ int i, atom_cnt, token_cnt, bgf_serial, ratom = 0;
+
+ /* open biograf file */
+ if ( (bgf = fopen( bgf_file, "r" )) == NULL )
+ {
+ fprintf( stderr, "Error opening the bgf file!\n" );
+ exit( FILE_NOT_FOUND );
+ }
+
+ /* allocate memory for tokenizing biograf file lines */
+ line = (char*) malloc( sizeof(char) * MAX_LINE );
+ backup = (char*) malloc( sizeof(char) * MAX_LINE );
+ tokens = (char**) malloc( sizeof(char*) * MAX_TOKENS );
+ for ( i = 0; i < MAX_TOKENS; i++ )
+ {
+ tokens[i] = (char*) malloc( sizeof(char) * MAX_TOKEN_LEN );
+ }
+
+ /* count number of atoms in the pdb file */
+ system->N = 0;
+ line[0] = 0;
+
+ while ( fgets( line, MAX_LINE, bgf ) )
+ {
+ tokens[0][0] = 0;
+ token_cnt = Tokenize( line, &tokens );
+
+ if ( !strcmp( tokens[0], "ATOM" ) || !strcmp( tokens[0], "HETATM" ) )
+ {
+ (system->N)++;
+ }
+
+ line[0] = 0;
+ }
+ if ( ferror ( bgf ) )
+ {
+ return FAILURE;
+ }
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "system->N: %d\n", system->N );
+#endif
+
+ fclose( bgf );
+
+ /* memory allocations for atoms, atom maps, bond restrictions */
+// system->atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
+//
+// workspace->map_serials = (int*) calloc( MAX_ATOM_ID, sizeof(int) );
+// for ( i = 0; i < MAX_ATOM_ID; ++i )
+// {
+// workspace->map_serials[i] = -1;
+// }
+//
+// workspace->orig_id = (int*) calloc( system->N, sizeof(int) );
+// workspace->restricted = (int*) calloc( system->N, sizeof(int) );
+// workspace->restricted_list = (int**) calloc( system->N, sizeof(int*) );
+// for ( i = 0; i < system->N; ++i )
+// {
+// workspace->restricted_list[i] = (int*) calloc( MAX_RESTRICT, sizeof(int) );
+// }
+
+ //TODO: setup similar for BGF
+// Count_PDB_Atoms( pdb, system );
+ if ( PreAllocate_Space( system, control, workspace ) == FAILURE )
+ {
+ fprintf( stderr, "PreAllocate_Space: not enough memory!" );
+ fprintf( stderr, "terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* start reading and processing bgf file */
+ if ( (bgf = fopen( bgf_file, "r" )) == NULL )
+ {
+ fprintf( stderr, "Error opening the bgf file!\n" );
+ exit( FILE_NOT_FOUND );
+ }
+ atom_cnt = 0;
+ token_cnt = 0;
+
+ while ( fgets( line, MAX_LINE, bgf ) )
+ {
+ /* read new line and tokenize it */
+ strncpy( backup, line, MAX_LINE - 1 );
+ token_cnt = Tokenize( line, &tokens );
+
+ /* process new line */
+ if ( !strncmp(tokens[0], "ATOM", 4) || !strncmp(tokens[0], "HETATM", 6) )
+ {
+ if ( !strncmp(tokens[0], "ATOM", 4) )
+ {
+ strncpy( &descriptor[0], backup, 6 );
+ descriptor[6] = 0;
+ strncpy( &serial[0], backup + 7, 5 );
+ serial[5] = 0;
+ strncpy( &atom_name[0], backup + 13, 5 );
+ atom_name[5] = 0;
+ strncpy( &res_name[0], backup + 19, 3 );
+ res_name[3] = 0;
+ chain_id = backup[23];
+ strncpy( &res_seq[0], backup + 25, 5 );
+ res_seq[5] = 0;
+ strncpy( &s_x[0], backup + 30, 10 );
+ s_x[10] = 0;
+ strncpy( &s_y[0], backup + 40, 10 );
+ s_y[10] = 0;
+ strncpy( &s_z[0], backup + 50, 10 );
+ s_z[10] = 0;
+ strncpy( &element[0], backup + 61, 5 );
+ element[5] = 0;
+ strncpy( &occupancy[0], backup + 66, 3 );
+ occupancy[3] = 0;
+ strncpy( &temp_factor[0], backup + 69, 2 );
+ temp_factor[2] = 0;
+ strncpy( &charge[0], backup + 72, 8 );
+ charge[8] = 0;
+ }
+ else if ( !strncmp(tokens[0], "HETATM", 6) )
+ {
+ /* bgf hetatm:
+ (7x,i5,1x,a5,1x,a3,1x,a1,1x,a5,3f10.5,1x,a5,i3,i2,1x,f8.5) */
+ strncpy( &descriptor[0], backup, 6 );
+ descriptor[6] = 0;
+ strncpy( &serial[0], backup + 7, 5 );
+ serial[5] = 0;
+ strncpy( &atom_name[0], backup + 13, 5 );
+ atom_name[5] = 0;
+ strncpy( &res_name[0], backup + 19, 3 );
+ res_name[3] = 0;
+ chain_id = backup[23];
+ strncpy( &res_seq[0], backup + 25, 5 );
+ res_seq[5] = 0;
+ strncpy( &s_x[0], backup + 30, 10 );
+ s_x[10] = 0;
+ strncpy( &s_y[0], backup + 40, 10 );
+ s_y[10] = 0;
+ strncpy( &s_z[0], backup + 50, 10 );
+ s_z[10] = 0;
+ strncpy( &element[0], backup + 61, 5 );
+ element[5] = 0;
+ strncpy( &occupancy[0], backup + 66, 3 );
+ occupancy[3] = 0;
+ strncpy( &temp_factor[0], backup + 69, 2 );
+ temp_factor[2] = 0;
+ strncpy( &charge[0], backup + 72, 8 );
+ charge[8] = 0;
+ }
+
+ /* add to mapping */
+ bgf_serial = strtod( &serial[0], &endptr );
+ Check_Input_Range( bgf_serial, 0, MAX_ATOM_ID, "Invalid bgf serial" );
+ workspace->map_serials[ bgf_serial ] = atom_cnt;
+ workspace->orig_id[ atom_cnt ] = bgf_serial;
+ // fprintf( stderr, "map %d --> %d\n", bgf_serial, atom_cnt );
+
+ /* copy atomic positions */
+ system->atoms[atom_cnt].x[0] = strtod( &s_x[0], &endptr );
+ system->atoms[atom_cnt].x[1] = strtod( &s_y[0], &endptr );
+ system->atoms[atom_cnt].x[2] = strtod( &s_z[0], &endptr );
+
+ /* atom name and type */
+ strcpy( system->atoms[atom_cnt].name, atom_name );
+ Trim_Spaces( element );
+ system->atoms[atom_cnt].type =
+ Get_Atom_Type( &(system->reaxprm), element );
+
+ /* fprintf( stderr,
+ "a:%3d(%1d) c:%10.5f%10.5f%10.5f q:%10.5f occ:%s temp:%s seg_id:%s element:%s\n",
+ atom_cnt, system->atoms[ atom_cnt ].type,
+ system->atoms[ atom_cnt ].x[0],
+ system->atoms[ atom_cnt ].x[1], system->atoms[ atom_cnt ].x[2],
+ system->atoms[ atom_cnt ].q, occupancy, temp_factor,
+ seg_id, element ); */
+
+ atom_cnt++;
+ }
+ else if (!strncmp( tokens[0], "CRYSTX", 6 ))
+ {
+ sscanf( backup, BGF_CRYSTX_FORMAT,
+ &descriptor[0],
+ &s_a[0],
+ &s_b[0],
+ &s_c[0],
+ &s_alpha[0],
+ &s_beta[0],
+ &s_gamma[0] );
+
+ /* Compute full volume tensor from the angles */
+ Setup_Box( atof(s_a), atof(s_b), atof(s_c),
+ atof(s_alpha), atof(s_beta), atof(s_gamma),
+ &(system->box) );
+ }
+ else if (!strncmp( tokens[0], "CONECT", 6 ))
+ {
+ /* check number of restrictions */
+ Check_Input_Range( token_cnt - 2, 0, MAX_RESTRICT,
+ "CONECT line exceeds max restrictions allowed.\n" );
+
+ /* read bond restrictions */
+ if ( is_Valid_Serial( workspace, bgf_serial = atoi(tokens[1]) ) )
+ {
+ ratom = workspace->map_serials[ bgf_serial ];
+ }
+
+ workspace->restricted[ ratom ] = token_cnt - 2;
+ for ( i = 2; i < token_cnt; ++i )
+ {
+ if ( is_Valid_Serial( workspace, bgf_serial = atoi(tokens[i]) ) )
+ {
+ workspace->restricted_list[ ratom * system->N + (i - 2) ] =
+ workspace->map_serials[ bgf_serial ];
+ }
+ }
+
+ /* fprintf( stderr, "restriction on %d:", ratom );
+ for( i = 0; i < workspace->restricted[ ratom ]; ++i )
+ fprintf( stderr, " %d", workspace->restricted_list[ratom][i] );
+ fprintf( stderr, "\n" ); */
+ }
+
+ /* clear previous input line */
+ line[0] = 0;
+
+ for ( i = 0; i < token_cnt; ++i )
+ {
+ tokens[i][0] = 0;
+ }
+ }
+ if ( ferror ( bgf ) )
+ {
+ return FAILURE;
+ }
+
+ fclose( bgf );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "bgf file read\n" );
+#endif
+
+ return SUCCESS;
+}
diff --git a/PuReMD-GPU/src/pdb_tools.h b/PuReMD-GPU/src/geo_tools.h
similarity index 84%
rename from PuReMD-GPU/src/pdb_tools.h
rename to PuReMD-GPU/src/geo_tools.h
index 12518fc2daaecd735f5cb0f781a0f1c72e504aa7..4c44e3081e6105947b610916e95210e123b4a7d9 100644
--- a/PuReMD-GPU/src/pdb_tools.h
+++ b/PuReMD-GPU/src/geo_tools.h
@@ -1,9 +1,10 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
@@ -18,13 +19,20 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#ifndef __PDB_TOOLS_H_
-#define __PDB_TOOLS_H_
+#ifndef __GEO_TOOLS_H_
+#define __GEO_TOOLS_H_
#include "mytypes.h"
-/*
-PDB format :
+// CUSTOM_BOXGEO: BOXGEO box_x box_y box_z angle1 angle2 angle3
+#define CUSTOM_BOXGEO_FORMAT " %s %lf %lf %lf %lf %lf %lf"
+// CUSTOM ATOM: serial element name x y z
+#define CUSTOM_ATOM_FORMAT " %d %s %s %lf %lf %lf"
+
+char Read_Geo( char*, reax_system*, control_params*,
+ simulation_data*, static_storage* );
+
+/* PDB format :
http://www.rcsb.org/pdb/file_formats/pdb/pdbguide2.2/guide2.2_frame.html
#define PDB_ATOM_FORMAT "%6s%5d%4s%c%4s%c%4d%c%8s%8s%8s%6s%6s%4s%2s%2s\n"
@@ -94,24 +102,28 @@ COLUMNS DATA TYPE FIELD DEFINITION
67 - 70 Integer z Z value
*/
-//#define PDB_ATOM_FORMAT "ATOM %4d%4s%c%3s%c%4d%c%8.3f%8.3f%8.3f%6.2f%6.2f%-4s%2s%2s\n"
+//#define PDB_ATOM_FORMAT
+//"ATOM %4d%4s%c%3s%c%4d%c%8.3f%8.3f%8.3f%6.2f%6.2f%-4s%2s%2s\n"
#define PDB_ATOM_FORMAT "%6s%5d%4s%c%4s%c%4d%c%8s%8s%8s%6s%6s%4s%2s%2s\n"
+#define PDB_ATOM_FORMAT_LENGTH 71
#define PDB_HETATM_FORMAT "%6s%5d%4s%c%4s%c%4d%c%8s%8s%8s%6s%6s%2s%2s\n"
#define PDB_CONECT_FORMAT "%6s%5d%5d%5d%5d%5d\n"
#define PDB_CRYST1_FORMAT "%6s%9s%9s%9s%7s%7s%7s%11s%4s\n"
#define PDB_ATOM_FORMAT_O "%6s%5d %4s%c%3s %c%4d%c %8.3f%8.3f%8.3f%6.2f%6.2f %-4s%2s%2s\n"
+#define PDB_ATOM_FORMAT_O_LENGTH 81
#define PDB_CRYST1_FORMAT_O "%6s%9.3f%9.3f%9.3f%7.2f%7.2f%7.2f%11s%4d\n"
#define BGF_CRYSTX_FORMAT "%8s%11s%11s%11s%11s%11s%11s"
char Read_PDB( char*, reax_system*, control_params*,
- simulation_data*, static_storage* );
+ simulation_data*, static_storage* );
+
char Read_BGF( char*, reax_system*, control_params*,
- simulation_data*, static_storage* );
+ simulation_data*, static_storage* );
-char Write_PDB( reax_system*, control_params*, simulation_data*,
- static_storage*, list*, output_controls* );
+char Write_PDB( reax_system*, list*, simulation_data*,
+ control_params*, static_storage*, output_controls* );
#endif
diff --git a/PuReMD-GPU/src/grid.c b/PuReMD-GPU/src/grid.c
index fb09b409194a84b1646da3b779aad8b547ff9db3..2077d56080799f3674eecb3038a0f30bd318bd67 100644
--- a/PuReMD-GPU/src/grid.c
+++ b/PuReMD-GPU/src/grid.c
@@ -1,28 +1,29 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "grid.h"
+#include "index_utils.h"
#include "reset_utils.h"
#include "vector.h"
-#include "index_utils.h"
int Estimate_GCell_Population( reax_system* system )
@@ -34,36 +35,45 @@ int Estimate_GCell_Population( reax_system* system )
g = &( system->g );
Reset_Grid( g );
- for( l = 0; l < system->N; l++ ) {
+ for ( l = 0; l < system->N; l++ )
+ {
i = (int)(system->atoms[l].x[0] * g->inv_len[0]);
j = (int)(system->atoms[l].x[1] * g->inv_len[1]);
k = (int)(system->atoms[l].x[2] * g->inv_len[2]);
g->top[index_grid_3d (i, j, k, g)]++;
- // fprintf( stderr, "\tatom%-6d (%8.3f%8.3f%8.3f) --> (%3d%3d%3d)\n",
+ // fprintf( stderr, "\tatom%-6d (%8.3f%8.3f%8.3f) --> (%3d%3d%3d)\n",
// l, system->atoms[l].x[0], system->atoms[l].x[1], system->atoms[l].x[2],
// i, j, k );
}
max_atoms = 0;
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ )
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
if( max_atoms < g->top[index_grid_3d (i, j, k, g)] )
+ {
max_atoms = g->top[index_grid_3d (i, j, k, g)];
+ }
+ }
+ }
+ }
- return MAX(max_atoms*SAFE_ZONE, MIN_GCELL_POPL);
+ return MAX(max_atoms * SAFE_ZONE, MIN_GCELL_POPL);
}
void Allocate_Space_for_Grid( reax_system *system )
{
int i, j, k, l;
- grid *g = &(system->g);
-
- int total = g->ncell[0] * g->ncell[1] * g->ncell[2];
+ grid *g;
+ int total;
g = &(system->g);
- g->max_nbrs = (2*g->spread[0]+1) * (2*g->spread[1]+1) * (2*g->spread[2]+1)+3;
+ g->max_nbrs = (2 * g->spread[0] + 1) * (2 * g->spread[1] + 1) * (2 * g->spread[2] + 1) + 3;
+ total = g->ncell[0] * g->ncell[1] * g->ncell[2];
/* allocate space for the new grid */
g->top = (int*) calloc( total, sizeof( int ));
@@ -73,10 +83,14 @@ void Allocate_Space_for_Grid( reax_system *system )
g->nbrs = (ivec*) calloc( total * g->max_nbrs, sizeof( ivec ));
g->nbrs_cp = (rvec*) calloc( total * g->max_nbrs, sizeof( rvec ));
- for( i = 0; i < g->ncell[0]; i++ ) {
- for( j = 0; j < g->ncell[1]; j++ ) {
- for( k = 0; k < g->ncell[2]; k++ ) {
- for( l = 0; l < g->max_nbrs; ++l ){
+ for( i = 0; i < g->ncell[0]; i++ )
+ {
+ for( j = 0; j < g->ncell[1]; j++ )
+ {
+ for( k = 0; k < g->ncell[2]; k++ )
+ {
+ for( l = 0; l < g->max_nbrs; ++l )
+ {
g->nbrs[ index_grid_nbrs (i, j, k, l, g) ][0] = -1;
g->nbrs[ index_grid_nbrs (i, j, k, l, g) ][1] = -1;
g->nbrs[ index_grid_nbrs (i, j, k, l, g) ][2] = -1;
@@ -110,49 +124,74 @@ int Shift(int p, int dp, int dim, grid *g )
int dim_len = 0;
int newp = p + dp;
- switch( dim ) {
- case 0: dim_len = g->ncell[0];
- break;
- case 1: dim_len = g->ncell[1];
- break;
- case 2: dim_len = g->ncell[2];
+ switch ( dim )
+ {
+ case 0:
+ dim_len = g->ncell[0];
+ break;
+ case 1:
+ dim_len = g->ncell[1];
+ break;
+ case 2:
+ dim_len = g->ncell[2];
+ }
+
+ while ( newp < 0 )
+ {
+ newp = newp + dim_len;
+ }
+ while ( newp >= dim_len )
+ {
+ newp = newp - dim_len;
}
- while( newp < 0 ) newp = newp + dim_len;
- while( newp >= dim_len ) newp = newp - dim_len;
return newp;
}
/* finds the closest point between two grid cells denoted by c1 and c2.
periodic boundary conditions are taken into consideration as well. */
-void Find_Closest_Point( grid *g, int c1x, int c1y, int c1z,
- int c2x, int c2y, int c2z, rvec closest_point )
+void Find_Closest_Point( grid *g, int c1x, int c1y, int c1z,
+ int c2x, int c2y, int c2z, rvec closest_point )
{
int i, d;
ivec c1 = { c1x, c1y, c1z };
ivec c2 = { c2x, c2y, c2z };
- for( i = 0; i < 3; i++ ) {
- if( g->ncell[i] < 5 ) {
+ for ( i = 0; i < 3; i++ )
+ {
+ if ( g->ncell[i] < 5 )
+ {
closest_point[i] = NEG_INF - 1.;
continue;
}
d = c2[i] - c1[i];
- if( abs(d) <= g->ncell[i] / 2 ) {
- if( d > 0 )
+ if ( abs(d) <= g->ncell[i] / 2 )
+ {
+ if ( d > 0 )
+ {
closest_point[i] = c2[i] * g->len[i];
+ }
else if ( d == 0 )
+ {
closest_point[i] = NEG_INF - 1.;
+ }
else
+ {
closest_point[i] = ( c2[i] + 1 ) * g->len[i];
+ }
}
- else {
- if( d > 0 )
+ else
+ {
+ if ( d > 0 )
+ {
closest_point[i] = ( c2[i] - g->ncell[i] + 1 ) * g->len[i];
- else
+ }
+ else
+ {
closest_point[i] = ( c2[i] + g->ncell[i] ) * g->len[i];
+ }
}
}
}
@@ -168,29 +207,36 @@ void Find_Neighbor_GridCells( grid *g )
rvec *cp_stack;
/* pick up a cell in the grid */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
nbrs_stack = &( g->nbrs[ index_grid_nbrs (i, j, k, 0, g) ] );
cp_stack = &( g->nbrs_cp[ index_grid_nbrs (i, j, k, 0, g) ] );
stack_top = 0;
//fprintf( stderr, "grid1: %d %d %d\n", i, j, k );
/* choose an unmarked neighbor cell*/
- for( di = -g->spread[0]; di <= g->spread[0]; di++ ) {
+ for ( di = -g->spread[0]; di <= g->spread[0]; di++ )
+ {
x = Shift( i, di, 0, g );
- for( dj = -g->spread[1]; dj <= g->spread[1]; dj++ ) {
+ for ( dj = -g->spread[1]; dj <= g->spread[1]; dj++ )
+ {
y = Shift( j, dj, 1, g );
- for( dk = -g->spread[2]; dk <= g->spread[2]; dk++ ) {
+ for ( dk = -g->spread[2]; dk <= g->spread[2]; dk++ )
+ {
z = Shift( k, dk, 2, g );
//fprintf( stderr, "\tgrid2: %d %d %d\n", x, y, z );
- if( !g->mark[ index_grid_3d (x, y, z, g) ] ) {
+ if( !g->mark[ index_grid_3d (x, y, z, g) ] )
+ {
/*(di < 0 || // 9 combinations
- (di == 0 && dj < 0) || // 3 combinations
- (di == 0 && dj == 0 && dk < 0) ) )*/
+ (di == 0 && dj < 0) || // 3 combinations
+ (di == 0 && dj == 0 && dk < 0) ) )*/
/* put the neighbor cell into the stack and mark it */
nbrs_stack[stack_top][0] = x;
nbrs_stack[stack_top][1] = y;
@@ -198,8 +244,8 @@ void Find_Neighbor_GridCells( grid *g )
g->mark[ index_grid_3d(x,y,z,g) ] = 1;
Find_Closest_Point( g, i, j, k, x, y, z, cp_stack[stack_top] );
- //fprintf( stderr, "\tcp: %lf %lf %lf\n",
- // cp_stack[stack_top][0], cp_stack[stack_top][1],
+ //fprintf( stderr, "\tcp: %lf %lf %lf\n",
+ // cp_stack[stack_top][0], cp_stack[stack_top][1],
// cp_stack[stack_top][2]);
stack_top++;
}
@@ -220,6 +266,8 @@ void Find_Neighbor_GridCells( grid *g )
nbrs_stack[stack_top][2] = -1;
Reset_Marks( g, nbrs_stack, stack_top );
}
+ }
+ }
}
@@ -234,9 +282,13 @@ void Setup_Grid( reax_system* system )
/* determine number of grid cells in each direction */
ivec_rScale( ncell, 1. / g->cell_size, my_box->box_norms );
- for( d = 0; d < 3; ++d )
- if( ncell[d] <= 0 )
+ for ( d = 0; d < 3; ++d )
+ {
+ if ( ncell[d] <= 0 )
+ {
ncell[d] = 1;
+ }
+ }
/* find the number of grid cells */
g->total = ncell[0] * ncell[1] * ncell[2];
@@ -270,25 +322,34 @@ void Update_Grid( reax_system* system )
/* determine number of grid cells in each direction */
ivec_rScale( ncell, 1. / g->cell_size, my_box->box_norms );
- for( d = 0; d < 3; ++d )
- if( ncell[d] == 0 )
+ for ( d = 0; d < 3; ++d )
+ {
+ if ( ncell[d] == 0 )
+ {
ncell[d] = 1;
+ }
+ }
- if( ivec_isEqual( ncell, g->ncell ) ) {/* ncell are unchanged */
+ if ( ivec_isEqual( ncell, g->ncell ) ) /* ncell are unchanged */
+ {
/* update cell lengths */
rvec_iDivide( g->len, my_box->box_norms, g->ncell );
rvec_Invert( g->inv_len, g->len );
/* update closest point distances between gcells */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
nbrs = &( g->nbrs[ index_grid_nbrs (i, j, k, 0, g) ] );
nbrs_cp = &( g->nbrs_cp[ index_grid_nbrs (i, j, k, 0, g) ] );
//fprintf( stderr, "gridcell %d %d %d\n", i, j, k );
itr = 0;
- while( nbrs[itr][0] >= 0 ){
+ while ( nbrs[itr][0] >= 0 )
+ {
x = nbrs[itr][0];
y = nbrs[itr][1];
z = nbrs[itr][2];
@@ -297,9 +358,12 @@ void Update_Grid( reax_system* system )
++itr;
}
}
+ }
+ }
}
- else{ /* at least one of ncell has changed */
- Deallocate_Grid_Space( g );
+ else /* at least one of ncell has changed */
+ {
+ Deallocate_Grid_Space( g );
/* update number of grid cells */
g->total = ncell[0] * ncell[1] * ncell[2];
ivec_Copy( g->ncell, ncell );
@@ -311,10 +375,10 @@ void Update_Grid( reax_system* system )
Find_Neighbor_GridCells( g );
#if defined(DEBUG_FOCUS)
fprintf( stderr, "updated grid: " );
- fprintf( stderr, "ncell[%d %d %d] ",
- g->ncell[0], g->ncell[1], g->ncell[2] );
- fprintf( stderr, "len[%5.2f %5.2f %5.2f] ",
- g->len[0], g->len[1], g->len[2] );
+ fprintf( stderr, "ncell[%d %d %d] ",
+ g->ncell[0], g->ncell[1], g->ncell[2] );
+ fprintf( stderr, "len[%5.2f %5.2f %5.2f] ",
+ g->len[0], g->len[1], g->len[2] );
fprintf( stderr, "g->max_atoms = %d\n", g->max_atoms );
#endif
}
@@ -328,40 +392,59 @@ void Bin_Atoms( reax_system* system, static_storage *workspace )
grid *g = &( system->g );
Reset_Grid( g );
-
- for( l = 0; l < system->N; l++ ) {
+ for ( l = 0; l < system->N; l++ )
+ {
i = (int)(system->atoms[l].x[0] * g->inv_len[0]);
j = (int)(system->atoms[l].x[1] * g->inv_len[1]);
k = (int)(system->atoms[l].x[2] * g->inv_len[2]);
#ifdef __BNVT_FIX__
- if (i >= g->ncell[0]) i = g->ncell[0]-1;
- if (j >= g->ncell[1]) j = g->ncell[1]-1;
- if (k >= g->ncell[2]) k = g->ncell[2]-1;
+ if (i >= g->ncell[0])
+ {
+ i = g->ncell[0]-1;
+ }
+ if (j >= g->ncell[1])
+ {
+ j = g->ncell[1]-1;
+ }
+ if (k >= g->ncell[2])
+ {
+ k = g->ncell[2]-1;
+ }
#endif
g->atoms[ index_grid_atoms (i,j,k,g->top[ index_grid_3d (i,j,k,g) ], g) ] = l;
g->top[index_grid_3d (i,j,k,g) ]++;
- //fprintf( stderr, "\tatom%-6d (%8.3f%8.3f%8.3f) --> (%3d%3d%3d)\n",
- //l, system->atoms[l].x[0], system->atoms[l].x[1], system->atoms[l].x[2],
- //i, j, k );
+ // fprintf( stderr, "\tatom%-6d (%8.3f%8.3f%8.3f) --> (%3d%3d%3d)\n",
+ // l, system->atoms[l].x[0], system->atoms[l].x[1], system->atoms[l].x[2],
+ // i, j, k );
}
max_atoms = 0;
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ )
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
if( max_atoms < g->top[ index_grid_3d (i, j, k, g) ] )
+ {
max_atoms = g->top[ index_grid_3d (i, j, k, g) ];
+ }
+ }
+ }
+ }
/* check if current gcell->max_atoms is safe */
- if( max_atoms >= g->max_atoms * SAFE_ZONE )
- workspace->realloc.gcell_atoms = MAX(max_atoms*SAFE_ZONE,MIN_GCELL_POPL);
+ if ( max_atoms >= g->max_atoms * SAFE_ZONE )
+ {
+ workspace->realloc.gcell_atoms = MAX(max_atoms * SAFE_ZONE, MIN_GCELL_POPL);
+ }
}
-inline void reax_atom_Copy( reax_atom *dest, reax_atom *src )
+static inline void reax_atom_Copy( reax_atom *dest, reax_atom *src )
{
dest->type = src->type;
rvec_Copy( dest->x, src->x );
@@ -370,30 +453,37 @@ inline void reax_atom_Copy( reax_atom *dest, reax_atom *src )
}
-void Copy_Storage( reax_system *system, static_storage *workspace,
- int top, int old_id, int old_type,
- int *num_H, real *v, real *s, real *t,
- int *orig_id, rvec *f_old )
+void Copy_Storage( reax_system *system, static_storage *workspace,
+ int top, int old_id, int old_type,
+ int *num_H, real *v, real *s, real *t,
+ int *orig_id, rvec *f_old )
{
int i;
- for( i = 0; i < RESTART+1; ++i )
+ for ( i = 0; i < RESTART + 1; ++i )
+ {
v[ index_wkspace_sys (i,top, system->N) ] = workspace->v[ index_wkspace_sys (i,old_id, system->N) ];
+ }
- for( i = 0; i < 3; ++i ) {
- s[ index_wkspace_sys (i,top, system->N) ] = workspace->s[ index_wkspace_sys (i,old_id, system->N) ];
- t[ index_wkspace_sys (i,top, system->N) ] = workspace->t[ index_wkspace_sys (i,old_id, system->N) ];
+ for ( i = 0; i < 3; ++i )
+ {
+ s[ index_wkspace_sys(i,top, system->N) ] = workspace->s[ index_wkspace_sys(i,old_id, system->N) ];
+ t[ index_wkspace_sys(i,top, system->N) ] = workspace->t[ index_wkspace_sys(i,old_id, system->N) ];
}
orig_id[top] = workspace->orig_id[old_id];
- workspace->Hdia_inv[top] = 1. / system->reaxprm.sbp[ old_type ].eta;
workspace->b_s[top] = -system->reaxprm.sbp[ old_type ].chi;
- workspace->b_t[top] = -1.0;
+ workspace->b_t[top] = -1.0;
- if( system->reaxprm.sbp[ old_type ].p_hbond == 1 ) // H atom
+ if ( system->reaxprm.sbp[ old_type ].p_hbond == 1 ) // H atom
+ {
workspace->hbond_index[top] = (*num_H)++;
- else workspace->hbond_index[top] = -1;
+ }
+ else
+ {
+ workspace->hbond_index[top] = -1;
+ }
rvec_Copy( f_old[top], workspace->f_old[old_id] );
}
@@ -404,12 +494,12 @@ void Free_Storage( static_storage *workspace )
free( workspace->v );
free( workspace->s );
free( workspace->t );
- free( workspace->orig_id );
+ free( workspace->orig_id );
}
-void Assign_New_Storage( static_storage *workspace,
- real *v, real *s, real *t,
+void Assign_New_Storage( static_storage *workspace,
+ real *v, real *s, real *t,
int *orig_id, rvec *f_old )
{
workspace->v = v;
@@ -425,14 +515,20 @@ void Assign_New_Storage( static_storage *workspace,
void Cluster_Atoms( reax_system *system, static_storage *workspace )
{
- int i, j, k, l, top, old_id, num_H = 0;
+ int i, j, k, l, top, old_id, num_H;
reax_atom *old_atom;
- grid *g = &( system->g );
- reax_atom *new_atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
- int *orig_id = (int *) calloc( system->N, sizeof( int ) );
+ grid *g;
+ reax_atom *new_atoms;
+ int *orig_id ;
real *v;
real *s, *t;
- rvec *f_old = (rvec*) calloc( system->N, sizeof(rvec) );
+ rvec *f_old;
+
+ num_H = 0;
+ g = &( system->g );
+ new_atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
+ orig_id = (int *) calloc( system->N, sizeof( int ) );
+ f_old = (rvec*) calloc( system->N, sizeof(rvec) );
s = (real*) calloc( 3, sizeof( real ) * system->N );
t = (real*) calloc( 3, sizeof( real ) * system->N );
@@ -440,24 +536,30 @@ void Cluster_Atoms( reax_system *system, static_storage *workspace )
top = 0;
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
g->start[ index_grid_3d (i, j, k, g) ] = top;
- for( l = 0; l < g->top[ index_grid_3d (i, j, k, g) ]; ++l ) {
+ for( l = 0; l < g->top[ index_grid_3d (i, j, k, g) ]; ++l )
+ {
old_id = g->atoms[ index_grid_atoms (i, j, k, l, g) ];
old_atom = &( system->atoms[old_id] );
// fprintf( stderr, "%d <-- %d\n", top, old_id );
reax_atom_Copy( &(new_atoms[top]), old_atom );
- Copy_Storage( system, workspace, top, old_id, old_atom->type,
- &num_H, v, s, t, orig_id, f_old );
+ Copy_Storage( system, workspace, top, old_id, old_atom->type,
+ &num_H, v, s, t, orig_id, f_old );
++top;
}
g->end[ index_grid_3d (i, j, k, g) ] = top;
}
+ }
+ }
free( system->atoms );
diff --git a/PuReMD-GPU/src/init_md.c b/PuReMD-GPU/src/init_md.c
index 2a2ce1270e2c694722e489b9a3f38f8dd48177a1..d1b40c6a224208ba3b24e00682b6872be8c16752 100644
--- a/PuReMD-GPU/src/init_md.c
+++ b/PuReMD-GPU/src/init_md.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -25,8 +26,8 @@
#include "forces.h"
#include "grid.h"
#include "index_utils.h"
-#include "lin_alg.h"
#include "integrate.h"
+#include "lin_alg.h"
#include "neighbors.h"
#include "list.h"
#include "lookup.h"
@@ -34,21 +35,20 @@
#include "reset_utils.h"
#include "system_props.h"
#include "traj.h"
+#include "tool_box.h"
#include "vector.h"
-void Generate_Initial_Velocities(reax_system *system, real T )
+void Generate_Initial_Velocities( reax_system *system, real T )
{
int i;
real scale, norm;
- if( T <= 0.1 )
+ if ( T <= 0.1 )
{
- for ( i = 0; i < system->N; i++ )
- {
+ for (i = 0; i < system->N; i++)
rvec_MakeZero( system->atoms[i].v );
- }
#if defined(DEBUG)
fprintf( stderr, "no random velocities...\n" );
@@ -56,73 +56,74 @@ void Generate_Initial_Velocities(reax_system *system, real T )
}
else
{
- for( i = 0; i < system->N; i++ )
+ for ( i = 0; i < system->N; i++ )
{
rvec_Random( system->atoms[i].v );
norm = rvec_Norm_Sqr( system->atoms[i].v );
- scale = SQRT( system->reaxprm.sbp[ system->atoms[i].type ].mass *
- norm / (3.0 * K_B * T) );
+ scale = SQRT( system->reaxprm.sbp[ system->atoms[i].type ].mass *
+ norm / (3.0 * K_B * T) );
- rvec_Scale( system->atoms[i].v, 1.0/scale, system->atoms[i].v );
+ rvec_Scale( system->atoms[i].v, 1.0 / scale, system->atoms[i].v );
- /*
- fprintf( stderr, "v = %f %f %f\n",
- system->atoms[i].v[0],system->atoms[i].v[1],system->atoms[i].v[2]);
- fprintf( stderr, "scale = %f\n", scale );
- fprintf( stderr, "v = %f %f %f\n",
- system->atoms[i].v[0],system->atoms[i].v[1],system->atoms[i].v[2]);
- */
+ /*fprintf( stderr, "v = %f %f %f\n",
+ system->atoms[i].v[0],system->atoms[i].v[1],system->atoms[i].v[2]);
+ fprintf( stderr, "scale = %f\n", scale );
+ fprintf( stderr, "v = %f %f %f\n",
+ system->atoms[i].v[0],system->atoms[i].v[1],system->atoms[i].v[2]);*/
}
}
}
-void Init_System( reax_system *system, control_params *control,
+void Init_System( reax_system *system, control_params *control,
simulation_data *data )
{
int i;
rvec dx;
- if( !control->restart )
+ if ( !control->restart )
{
Reset_Atoms( system );
}
Compute_Total_Mass( system, data );
-
Compute_Center_of_Mass( system, data, stderr );
/* reposition atoms */
// just fit the atoms to the periodic box
- if( control->reposition_atoms == 0 )
+ if ( control->reposition_atoms == 0 )
{
rvec_MakeZero( dx );
}
// put the center of mass to the center of the box
- else if( control->reposition_atoms == 1 )
+ else if ( control->reposition_atoms == 1 )
{
rvec_Scale( dx, 0.5, system->box.box_norms );
rvec_ScaledAdd( dx, -1., data->xcm );
}
// put the center of mass to the origin
- else if( control->reposition_atoms == 2 ) {
+ else if ( control->reposition_atoms == 2 )
+ {
rvec_Scale( dx, -1., data->xcm );
}
- else {
+ else
+ {
fprintf( stderr, "UNKNOWN OPTION: reposition_atoms. Terminating...\n" );
exit( UNKNOWN_OPTION );
}
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
Inc_on_T3( system->atoms[i].x, dx, &(system->box) );
- /*fprintf( stderr, "%6d%2d%8.3f%8.3f%8.3f\n",
- i, system->atoms[i].type,
+ /*fprintf( stderr, "%6d%2d%8.3f%8.3f%8.3f\n",
+ i, system->atoms[i].type,
system->atoms[i].x[0], system->atoms[i].x[1], system->atoms[i].x[2] );*/
}
/* Initialize velocities so that desired init T can be attained */
- if( !control->restart || (control->restart && control->random_vel) ) {
+ if ( !control->restart || (control->restart && control->random_vel) )
+ {
Generate_Initial_Velocities( system, control->T_init );
}
@@ -130,96 +131,153 @@ void Init_System( reax_system *system, control_params *control,
}
-void Init_Simulation_Data( reax_system *system, control_params *control,
- simulation_data *data, output_controls *out_control,
- evolve_function *Evolve )
+void Init_Simulation_Data( reax_system *system, control_params *control,
+ simulation_data *data, output_controls *out_control, evolve_function *Evolve )
{
Reset_Simulation_Data( data );
- if( !control->restart )
+ if ( !control->restart )
+ {
data->step = data->prev_steps = 0;
+ }
- switch( control->ensemble ) {
- case NVE:
- data->N_f = 3 * system->N;
- *Evolve = Velocity_Verlet_NVE;
- break;
+ switch ( control->ensemble )
+ {
+ case NVE:
+ data->N_f = 3 * system->N;
+ *Evolve = Velocity_Verlet_NVE;
+ break;
- case NVT:
- data->N_f = 3 * system->N + 1;
- //control->Tau_T = 100 * data->N_f * K_B * control->T_final;
- if( !control->restart || (control->restart && control->random_vel) ) {
- data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
- data->N_f * K_B * control->T );
- data->therm.v_xi = data->therm.G_xi * control->dt;
- data->therm.v_xi_old = 0;
- data->therm.xi = 0;
+ case NVT:
+ data->N_f = 3 * system->N + 1;
+ //control->Tau_T = 100 * data->N_f * K_B * control->T_final;
+ if ( !control->restart || (control->restart && control->random_vel) )
+ {
+ data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
+ data->N_f * K_B * control->T );
+ data->therm.v_xi = data->therm.G_xi * control->dt;
+ data->therm.v_xi_old = 0;
+ data->therm.xi = 0;
#if defined(DEBUG_FOCUS)
- fprintf( stderr, "init_md: G_xi=%f Tau_T=%f E_kin=%f N_f=%f v_xi=%f\n",
- data->therm.G_xi, control->Tau_T, data->E_Kin,
- data->N_f, data->therm.v_xi );
+ fprintf( stderr, "init_md: G_xi=%f Tau_T=%f E_kin=%f N_f=%f v_xi=%f\n",
+ data->therm.G_xi, control->Tau_T, data->E_Kin,
+ data->N_f, data->therm.v_xi );
#endif
- }
+ }
- *Evolve = Velocity_Verlet_Nose_Hoover_NVT_Klein;
- break;
+ *Evolve = Velocity_Verlet_Nose_Hoover_NVT_Klein;
+ break;
- case NPT: // Anisotropic NPT
- fprintf( stderr, "THIS OPTION IS NOT YET IMPLEMENTED! TERMINATING...\n" );
- exit( UNKNOWN_OPTION );
- data->N_f = 3 * system->N + 9;
- if( !control->restart ) {
- data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
- data->N_f * K_B * control->T );
- data->therm.v_xi = data->therm.G_xi * control->dt;
- data->iso_bar.eps = 0.33333 * log(system->box.volume);
- //data->inv_W = 1. / (data->N_f*K_B*control->T*SQR(control->Tau_P));
- //Compute_Pressure( system, data, workspace );
- }
- *Evolve = Velocity_Verlet_Berendsen_Isotropic_NPT;
- break;
+ case NPT: // Anisotropic NPT
+ fprintf( stderr, "THIS OPTION IS NOT YET IMPLEMENTED! TERMINATING...\n" );
+ exit( UNKNOWN_OPTION );
+ data->N_f = 3 * system->N + 9;
+ if ( !control->restart )
+ {
+ data->therm.G_xi = control->Tau_T * (2.0 * data->E_Kin -
+ data->N_f * K_B * control->T );
+ data->therm.v_xi = data->therm.G_xi * control->dt;
+ data->iso_bar.eps = 0.33333 * log(system->box.volume);
+ //data->inv_W = 1. / (data->N_f*K_B*control->T*SQR(control->Tau_P));
+ //Compute_Pressure( system, data, workspace );
+ }
+ *Evolve = Velocity_Verlet_Berendsen_Isotropic_NPT;
+ break;
- case sNPT: // Semi-Isotropic NPT
- data->N_f = 3 * system->N + 4;
- *Evolve = Velocity_Verlet_Berendsen_SemiIsotropic_NPT;
- break;
+ case sNPT: // Semi-Isotropic NPT
+ data->N_f = 3 * system->N + 4;
+ *Evolve = Velocity_Verlet_Berendsen_SemiIsotropic_NPT;
+ break;
- case iNPT: // Isotropic NPT
- data->N_f = 3 * system->N + 2;
- *Evolve = Velocity_Verlet_Berendsen_Isotropic_NPT;
- break;
+ case iNPT: // Isotropic NPT
+ data->N_f = 3 * system->N + 2;
+ *Evolve = Velocity_Verlet_Berendsen_Isotropic_NPT;
+ break;
- case bNVT: //berendensen NVT
- data->N_f = 3 * system->N + 1;
- *Evolve = Velocity_Verlet_Berendsen_NVT;
- break;
+ case bNVT:
+ data->N_f = 3 * system->N + 1;
+ *Evolve = Velocity_Verlet_Berendsen_NVT;
+ fprintf (stderr, " Initializing Velocity_Verlet_Berendsen_NVT .... \n");
+ break;
- default:
- break;
+ default:
+ break;
}
Compute_Kinetic_Energy( system, data );
- /* init timing info for the host*/
+ /* init timing info */
data->timing.start = Get_Time( );
data->timing.total = data->timing.start;
data->timing.nbrs = 0;
data->timing.init_forces = 0;
data->timing.bonded = 0;
data->timing.nonb = 0;
- data->timing.QEq = 0;
- data->timing.matvecs = 0;
+ data->timing.QEq = ZERO;
+ data->timing.QEq_sort_mat_rows = ZERO;
+ data->timing.pre_comp = ZERO;
+ data->timing.pre_app = ZERO;
+ data->timing.solver_iters = 0;
+ data->timing.solver_spmv = ZERO;
+ data->timing.solver_vector_ops = ZERO;
+ data->timing.solver_orthog = ZERO;
+ data->timing.solver_tri_solve = ZERO;
}
-void Init_Workspace( reax_system *system, control_params *control,
+/* Initialize Taper params */
+void Init_Taper( control_params *control )
+{
+ real d1, d7;
+ real swa, swa2, swa3;
+ real swb, swb2, swb3;
+
+ swa = control->r_low;
+ swb = control->r_cut;
+
+ if ( fabs( swa ) > 0.01 )
+ {
+ fprintf( stderr, "Warning: non-zero value for lower Taper-radius cutoff\n" );
+ }
+
+ if ( swb < 0 )
+ {
+ fprintf( stderr, "Negative value for upper Taper-radius cutoff\n" );
+ exit( INVALID_INPUT );
+ }
+ else if ( swb < 5 )
+ {
+ fprintf( stderr, "Warning: low value for upper Taper-radius cutoff:%f\n",
+ swb );
+ }
+
+ d1 = swb - swa;
+ d7 = POW( d1, 7.0 );
+ swa2 = SQR( swa );
+ swa3 = CUBE( swa );
+ swb2 = SQR( swb );
+ swb3 = CUBE( swb );
+
+ control->Tap7 = 20.0 / d7;
+ control->Tap6 = -70.0 * (swa + swb) / d7;
+ control->Tap5 = 84.0 * (swa2 + 3.0 * swa * swb + swb2) / d7;
+ control->Tap4 = -35.0 * (swa3 + 9.0 * swa2 * swb + 9.0 * swa * swb2 + swb3 ) / d7;
+ control->Tap3 = 140.0 * (swa3 * swb + 3.0 * swa2 * swb2 + swa * swb3 ) / d7;
+ control->Tap2 = -210.0 * (swa3 * swb2 + swa2 * swb3) / d7;
+ control->Tap1 = 140.0 * swa3 * swb3 / d7;
+ control->Tap0 = (-35.0 * swa3 * swb2 * swb2 + 21.0 * swa2 * swb3 * swb2 +
+ 7.0 * swa * swb3 * swb3 + swb3 * swb3 * swb ) / d7;
+}
+
+
+void Init_Workspace( reax_system *system, control_params *control,
static_storage *workspace )
-{
+{
int i;
/* Allocate space for hydrogen bond list */
@@ -231,35 +289,27 @@ void Init_Workspace( reax_system *system, control_params *control,
workspace->Deltap_boc = (real *) malloc( system->N * sizeof( real ) );
workspace->dDeltap_self = (rvec *) malloc( system->N * sizeof( rvec ) );
- workspace->Delta = (real *) malloc( system->N * sizeof( real ) );
- workspace->Delta_lp = (real *) malloc( system->N * sizeof( real ) );
+ workspace->Delta = (real *) malloc( system->N * sizeof( real ) );
+ workspace->Delta_lp = (real *) malloc( system->N * sizeof( real ) );
workspace->Delta_lp_temp = (real *) malloc( system->N * sizeof( real ) );
- workspace->dDelta_lp = (real *) malloc( system->N * sizeof( real ) );
+ workspace->dDelta_lp = (real *) malloc( system->N * sizeof( real ) );
workspace->dDelta_lp_temp = (real *) malloc( system->N * sizeof( real ) );
workspace->Delta_e = (real *) malloc( system->N * sizeof( real ) );
workspace->Delta_boc = (real *) malloc( system->N * sizeof( real ) );
- workspace->nlp = (real *) malloc( system->N * sizeof( real ) );
- workspace->nlp_temp = (real *) malloc( system->N * sizeof( real ) );
- workspace->Clp = (real *) malloc( system->N * sizeof( real ) );
+ workspace->nlp = (real *) malloc( system->N * sizeof( real ) );
+ workspace->nlp_temp = (real *) malloc( system->N * sizeof( real ) );
+ workspace->Clp = (real *) malloc( system->N * sizeof( real ) );
workspace->CdDelta = (real *) malloc( system->N * sizeof( real ) );
- workspace->vlpex = (real *) malloc( system->N * sizeof( real ) );
+ workspace->vlpex = (real *) malloc( system->N * sizeof( real ) );
/* QEq storage */
- //workspace->H = NULL;
- //workspace->L = NULL;
- //workspace->U = NULL;
- //
- workspace->H.start = NULL;
- workspace->L.start = NULL;
- workspace->U.start = NULL;
-
- workspace->H.entries = NULL;
- workspace->L.entries = NULL;
- workspace->U.entries = NULL;
-
+ workspace->H = NULL;
+ workspace->H_sp = NULL;
+ workspace->L = NULL;
+ workspace->U = NULL;
+ workspace->Hdia_inv = NULL;
workspace->droptol = (real *) calloc( system->N, sizeof( real ) );
workspace->w = (real *) calloc( system->N, sizeof( real ) );
- workspace->Hdia_inv = (real *) calloc( system->N, sizeof( real ) );
workspace->b = (real *) calloc( system->N * 2, sizeof( real ) );
workspace->b_s = (real *) calloc( system->N, sizeof( real ) );
workspace->b_t = (real *) calloc( system->N, sizeof( real ) );
@@ -273,25 +323,27 @@ void Init_Workspace( reax_system *system, control_params *control,
// workspace->s_oldest = (real *) calloc( system->N, sizeof( real ) );
// workspace->t_oldest = (real *) calloc( system->N, sizeof( real ) );
- for( i = 0; i < system->N; ++i ) {
- workspace->Hdia_inv[i] = 1./system->reaxprm.sbp[system->atoms[i].type].eta;
+ for ( i = 0; i < system->N; ++i )
+ {
workspace->b_s[i] = -system->reaxprm.sbp[ system->atoms[i].type ].chi;
workspace->b_t[i] = -1.0;
workspace->b[i] = -system->reaxprm.sbp[ system->atoms[i].type ].chi;
- workspace->b[i+system->N] = -1.0;
+ workspace->b[i + system->N] = -1.0;
}
+ //TODO: conditionally allocate based on solver selection
/* GMRES storage */
- workspace->y = (real *) calloc( RESTART+1, sizeof( real ) );
- workspace->z = (real *) calloc( RESTART+1, sizeof( real ) );
- workspace->g = (real *) calloc( RESTART+1, sizeof( real ) );
- workspace->hs = (real *) calloc( RESTART+1, sizeof( real ) );
- workspace->hc = (real *) calloc( RESTART+1, sizeof( real ) );
-
- workspace->rn = (real *) calloc( (RESTART+1)*system->N*2, sizeof( real) );
- workspace->v = (real *) calloc( (RESTART+1)*system->N, sizeof( real) );
- workspace->h = (real *) calloc( (RESTART+1)*(RESTART+1), sizeof( real) );
+ workspace->y = (real *) calloc( RESTART + 1, sizeof( real ) );
+ //TODO: unused?
+ workspace->z = (real *) calloc( RESTART + 1, sizeof( real ) );
+ workspace->g = (real *) calloc( RESTART + 1, sizeof( real ) );
+ workspace->h = (real *) calloc( (RESTART + 1) * (RESTART + 1), sizeof( real ) );
+ workspace->hs = (real *) calloc( RESTART + 1, sizeof( real ) );
+ workspace->hc = (real *) calloc( RESTART + 1, sizeof( real ) );
+ //TODO: unused?
+ workspace->rn = (real *) calloc( (RESTART + 1) * system->N * 2, sizeof( real ) );
+ workspace->v = (real *) calloc( (RESTART + 1) * system->N, sizeof( real ) );
/* CG storage */
workspace->r = (real *) calloc( system->N, sizeof( real ) );
@@ -304,20 +356,25 @@ void Init_Workspace( reax_system *system, control_params *control,
workspace->f_old = (rvec *) malloc( system->N * sizeof( rvec ) );
workspace->v_const = (rvec *) malloc( system->N * sizeof( rvec ) );
-
/* storage for analysis */
- if( control->molec_anal || control->diffusion_coef )
+ if ( control->molec_anal || control->diffusion_coef )
{
workspace->mark = (int *) calloc( system->N, sizeof(int) );
workspace->old_mark = (int *) calloc( system->N, sizeof(int) );
}
- else
+ else
+ {
workspace->mark = workspace->old_mark = NULL;
+ }
- if( control->diffusion_coef )
+ if ( control->diffusion_coef )
+ {
workspace->x_old = (rvec *) calloc( system->N, sizeof( rvec ) );
- else workspace->x_old = NULL;
-
+ }
+ else
+ {
+ workspace->x_old = NULL;
+ }
#ifdef TEST_FORCES
workspace->dDelta = (rvec *) malloc( system->N * sizeof( rvec ) );
@@ -344,9 +401,14 @@ void Init_Workspace( reax_system *system, control_params *control,
workspace->realloc.gcell_atoms = -1;
Reset_Workspace( system, workspace );
+
+ /* Initialize Taper function */
+ Init_Taper( control );
}
-void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *data, list *slist, int N)
+
+void compare_far_neighbors( int *test, int *start, int *end,
+ far_neighbor_data *data, list *slist, int N )
{
int index = 0;
int count = 0;
@@ -369,16 +431,19 @@ void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *
}
*/
-
- for (i = 0; i < N; i++){
- index = Start_Index (i, slist);
+ for (i = 0; i < N; i++)
+ {
+ index = Start_Index( i, slist );
//fprintf (stderr, "GPU : Neighbors of atom --> %d (start: %d , end: %d )\n", i, start[i], end[i]);
-
- for (j = start[i]; j < end[i]; j++){
+ for (j = start[i]; j < end[i]; j++)
+ {
gpu = data[j];
- if (i < data[j].nbr) continue;
+ if (i < data[j].nbr)
+ {
+ continue;
+ }
/*
if (i < data[j].nbr) {
//fprintf (stderr, " atom %d and neighbor %d @ index %d\n", i, data[j].nbr, j);
@@ -386,7 +451,6 @@ void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *
int dest = i;
int x;
-
for (x = start[src]; x < end[src]; x++) {
if (data[x].nbr != dest) continue;
@@ -431,9 +495,11 @@ void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *
cpu = slist->select.far_nbr_list[index];
//if ( (gpu.nbr != cpu.nbr) || (gpu.d != cpu.d) ){
//if ( (gpu->d != cpu->d) ){
- if ( (gpu.nbr != cpu.nbr) || (gpu.d != cpu.d) ||
- (cpu.dvec[0] != gpu.dvec[0]) || (cpu.dvec[1] != gpu.dvec[1]) || (cpu.dvec[2] != gpu.dvec[2]) ||
- (cpu.rel_box[0] != gpu.rel_box[0]) || (cpu.rel_box[1] != gpu.rel_box[1]) || (cpu.rel_box[2] != gpu.rel_box[2])) {
+ if ( (gpu.nbr != cpu.nbr) || (gpu.d != cpu.d)
+ ||(cpu.dvec[0] != gpu.dvec[0]) || (cpu.dvec[1] != gpu.dvec[1])
+ || (cpu.dvec[2] != gpu.dvec[2]) || (cpu.rel_box[0] != gpu.rel_box[0])
+ || (cpu.rel_box[1] != gpu.rel_box[1]) || (cpu.rel_box[2] != gpu.rel_box[2]))
+ {
//if ( (gpu.dvec[0] != i) || (gpu.dvec[1] != i) ||(gpu.dvec[2] != i) ||
// (gpu.rel_box[0] != i) || (gpu.rel_box[1] != i) ||(gpu.rel_box[2] != i) ) {
//if (memcmp (&gpu, &cpu, FAR_NEIGHBOR_SIZE - RVEC_SIZE - INT_SIZE )){
@@ -457,16 +523,16 @@ void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *
count ++;
}
- //fprintf (stderr, "GPU (neighbor %d , d %d )\n", gpu->nbr, gpu->d);
- index ++;
+ //fprintf (stderr, "GPU (neighbor %d , d %d )\n", gpu->nbr, gpu->d);
+ index ++;
}
- if (index != End_Index (i, slist))
+ if (index != End_Index( i, slist ))
{
fprintf( stderr,
"End index does not match for atom --> %d end index (%d) Cpu (%d, %d ) gpu (%d, %d)\n",
i, index, Start_Index (i, slist), End_Index(i, slist),
- start[i], end[i]);
+ start[i], end[i] );
exit( 10 );
}
}
@@ -518,112 +584,118 @@ void compare_far_neighbors (int *test, int *start, int *end, far_neighbor_data *
}
-void Init_Lists( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+void Init_Lists( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i, num_nbrs, num_hbonds, num_bonds, num_3body, Htop;
int *hb_top, *bond_top;
- real t_start, t_elapsed;
-
num_nbrs = Estimate_NumNeighbors( system, control, workspace, lists );
-
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, "Serial NumNeighbors ---> %d \n", num_nbrs);
-#endif
-
- if( !Make_List(system->N, num_nbrs, TYP_FAR_NEIGHBOR, (*lists)+FAR_NBRS ) ) {
+ if ( !Make_List(system->N, num_nbrs, TYP_FAR_NEIGHBOR, (*lists) + FAR_NBRS) )
+ {
fprintf(stderr, "Problem in initializing far nbrs list. Terminating!\n");
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
#if defined(DEBUG_FOCUS)
- fprintf( stderr, "memory allocated: far_nbrs = %ldMB\n",
- num_nbrs * sizeof(far_neighbor_data) / (1024*1024) );
-#endif
-
- t_start = Get_Time ();
- Generate_Neighbor_Lists(system,control,data,workspace,lists,out_control);
- t_elapsed = Get_Timing_Info ( t_start );
-
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " Timing Generate Neighbors %lf \n", t_elapsed );
+ fprintf( stderr, "memory allocated: far_nbrs = %ldMB\n",
+ num_nbrs * sizeof(far_neighbor_data) / (1024 * 1024) );
#endif
+ Generate_Neighbor_Lists(system, control, data, workspace, lists, out_control);
Htop = 0;
hb_top = (int*) calloc( system->N, sizeof(int) );
bond_top = (int*) calloc( system->N, sizeof(int) );
num_3body = 0;
- Estimate_Storage_Sizes( system, control, lists,
+ Estimate_Storage_Sizes( system, control, lists,
&Htop, hb_top, bond_top, &num_3body );
- Allocate_Matrix( &(workspace->H), system->N, Htop );
-
+ if ( Allocate_Matrix( workspace->H, system->N, Htop ) == FAILURE )
+ {
+ fprintf( stderr, "Not enough space for init matrices. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ /* TODO: better estimate for H_sp?
+ * If so, need to refactor Estimate_Storage_Sizes
+ * to use various cut-off distances as parameters
+ * (non-bonded, hydrogen, 3body, etc.) */
+ if ( Allocate_Matrix( workspace->H_sp, system->N, Htop ) == FAILURE )
+ {
+ fprintf( stderr, "Not enough space for init matrices. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
#if defined(DEBUG_FOCUS)
fprintf( stderr, "estimated storage - Htop: %d\n", Htop );
- fprintf( stderr, "memory allocated: H = %ldMB\n",
- Htop * sizeof(sparse_matrix_entry) / (1024*1024) );
+ fprintf( stderr, "memory allocated: H = %ldMB\n",
+ Htop * sizeof(sparse_matrix_entry) / (1024 * 1024) );
#endif
workspace->num_H = 0;
- if( control->hb_cut > 0 ) {
+ if ( control->hb_cut > 0 )
+ {
/* init H indexes */
- for( i = 0; i < system->N; ++i )
- if( system->reaxprm.sbp[ system->atoms[i].type ].p_hbond == 1 ) // H atom
+ for ( i = 0; i < system->N; ++i )
+ {
+ // H atom
+ if ( system->reaxprm.sbp[ system->atoms[i].type ].p_hbond == 1 )
+ {
workspace->hbond_index[i] = workspace->num_H++;
- else workspace->hbond_index[i] = -1;
-
- Allocate_HBond_List( system->N, workspace->num_H, workspace->hbond_index,
- hb_top, (*lists)+HBONDS );
- num_hbonds = hb_top[system->N-1];
+ }
+ else
+ {
+ workspace->hbond_index[i] = -1;
+ }
+ }
-#ifdef __DEBUG_CUDA__
- fprintf( stderr, "Serial num_hbonds: %d\n", num_hbonds );
-#endif
+ Allocate_HBond_List( system->N, workspace->num_H, workspace->hbond_index,
+ hb_top, (*lists) + HBONDS );
+ num_hbonds = hb_top[system->N - 1];
#if defined(DEBUG_FOCUS)
fprintf( stderr, "estimated storage - num_hbonds: %d\n", num_hbonds );
- fprintf( stderr, "memory allocated: hbonds = %ldMB\n",
- num_hbonds * sizeof(hbond_data) / (1024*1024) );
+ fprintf( stderr, "memory allocated: hbonds = %ldMB\n",
+ num_hbonds * sizeof(hbond_data) / (1024 * 1024) );
#endif
}
/* bonds list */
- Allocate_Bond_List( system->N, bond_top, (*lists)+BONDS );
- num_bonds = bond_top[system->N-1];
+ Allocate_Bond_List( system->N, bond_top, (*lists) + BONDS );
+ num_bonds = bond_top[system->N - 1];
#if defined(DEBUG_FOCUS)
fprintf( stderr, "estimated storage - num_bonds: %d\n", num_bonds );
- fprintf( stderr, "memory allocated: bonds = %ldMB\n",
- num_bonds * sizeof(bond_data) / (1024*1024) );
+ fprintf( stderr, "memory allocated: bonds = %ldMB\n",
+ num_bonds * sizeof(bond_data) / (1024 * 1024) );
#endif
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " host num_3body : %d \n", num_3body);
- fprintf (stderr, " host num_bonds : %d \n", num_bonds);
-#endif
+//fprintf (stderr, " **** sizeof 3 body : %d \n", sizeof (three_body_interaction_data));
+//fprintf (stderr, " **** num_3body : %d \n", num_3body);
+//fprintf (stderr, " **** num_bonds : %d \n", num_bonds);
/* 3bodies list */
- if(!Make_List(num_bonds, num_3body, TYP_THREE_BODY, (*lists)+THREE_BODIES )) {
+ if (!Make_List(num_bonds, num_3body, TYP_THREE_BODY, (*lists) + THREE_BODIES))
+ {
fprintf( stderr, "Problem in initializing angles list. Terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
#if defined(DEBUG_FOCUS)
fprintf( stderr, "estimated storage - num_3body: %d\n", num_3body );
- fprintf( stderr, "memory allocated: 3-body = %ldMB\n",
- num_3body * sizeof(three_body_interaction_data) / (1024*1024) );
+ fprintf( stderr, "memory allocated: 3-body = %ldMB\n",
+ num_3body * sizeof(three_body_interaction_data) / (1024 * 1024) );
#endif
#ifdef TEST_FORCES
- if(!Make_List( system->N, num_bonds * 8, TYP_DDELTA, (*lists) + DDELTA )) {
+ if (!Make_List( system->N, num_bonds * 8, TYP_DDELTA, (*lists) + DDELTA ))
+ {
fprintf( stderr, "Problem in initializing dDelta list. Terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
- if( !Make_List( num_bonds, num_bonds*MAX_BONDS*3, TYP_DBO, (*lists)+DBO ) ) {
+ if ( !Make_List( num_bonds, num_bonds * MAX_BONDS * 3, TYP_DBO, (*lists) + DBO ) )
+ {
fprintf( stderr, "Problem in initializing dBO list. Terminating!\n" );
- exit( INIT_ERR );
+ exit( CANNOT_INITIALIZE );
}
#endif
@@ -632,83 +704,91 @@ void Init_Lists( reax_system *system, control_params *control,
}
-void Init_Out_Controls(reax_system *system, control_params *control,
+void Init_Out_Controls(reax_system *system, control_params *control,
static_storage *workspace, output_controls *out_control)
{
char temp[1000];
/* Init trajectory file */
- if( out_control->write_steps > 0 ) {
+ if ( out_control->write_steps > 0 )
+ {
strcpy( temp, control->sim_name );
strcat( temp, ".trj" );
out_control->trj = fopen( temp, "w" );
out_control->write_header( system, control, workspace, out_control );
}
- if( out_control->energy_update_freq > 0 ) {
+ if ( out_control->energy_update_freq > 0 )
+ {
/* Init out file */
strcpy( temp, control->sim_name );
strcat( temp, ".out" );
out_control->out = fopen( temp, "w" );
fprintf( out_control->out, "%-6s%16s%16s%16s%11s%11s%13s%13s%13s\n",
- "step", "total energy", "poten. energy", "kin. energy",
- "temp.", "target", "volume", "press.", "target" );
+ "step", "total energy", "poten. energy", "kin. energy",
+ "temp.", "target", "volume", "press.", "target" );
fflush( out_control->out );
/* Init potentials file */
strcpy( temp, control->sim_name );
strcat( temp, ".pot" );
out_control->pot = fopen( temp, "w" );
- fprintf( out_control->pot,
- "%-6s%13s%13s%13s%13s%13s%13s%13s%13s%13s%13s%13s\n",
- "step", "ebond", "eatom", "elp", "eang", "ecoa", "ehb",
- "etor", "econj", "evdw","ecoul", "epol" );
+ fprintf( out_control->pot,
+ "%-6s%13s%13s%13s%13s%13s%13s%13s%13s%13s%13s%13s\n",
+ "step", "ebond", "eatom", "elp", "eang", "ecoa", "ehb",
+ "etor", "econj", "evdw", "ecoul", "epol" );
fflush( out_control->pot );
/* Init log file */
strcpy( temp, control->sim_name );
strcat( temp, ".log" );
out_control->log = fopen( temp, "w" );
- fprintf( out_control->log, "%-6s%10s%10s%10s%10s%10s%10s%10s\n",
- "step", "total", "neighbors", "init", "bonded",
- "nonbonded", "QEq", "matvec" );
+ fprintf( out_control->log, "%-6s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s\n",
+ "step", "total", "neighbors", "init", "bonded",
+ "nonbonded", "QEq", "QEq Sort", "S iters", "Pre Comp", "Pre App",
+ "S spmv", "S vec ops", "S orthog", "S tsolve" );
}
/* Init pressure file */
- if( control->ensemble == NPT ||
- control->ensemble == iNPT ||
- control->ensemble == sNPT ) {
+ if ( control->ensemble == NPT ||
+ control->ensemble == iNPT ||
+ control->ensemble == sNPT )
+ {
strcpy( temp, control->sim_name );
strcat( temp, ".prs" );
out_control->prs = fopen( temp, "w" );
fprintf( out_control->prs, "%-6s%13s%13s%13s%13s%13s%13s%13s%13s\n",
- "step", "norm_x", "norm_y", "norm_z",
- "press_x", "press_y", "press_z", "target_p", "volume" );
+ "step", "norm_x", "norm_y", "norm_z",
+ "press_x", "press_y", "press_z", "target_p", "volume" );
fflush( out_control->prs );
}
/* Init molecular analysis file */
- if( control->molec_anal ) {
+ if ( control->molec_anal )
+ {
sprintf( temp, "%s.mol", control->sim_name );
out_control->mol = fopen( temp, "w" );
- if( control->num_ignored ) {
+ if ( control->num_ignored )
+ {
sprintf( temp, "%s.ign", control->sim_name );
out_control->ign = fopen( temp, "w" );
- }
+ }
}
/* Init electric dipole moment analysis file */
- if( control->dipole_anal ) {
+ if ( control->dipole_anal )
+ {
strcpy( temp, control->sim_name );
strcat( temp, ".dpl" );
out_control->dpl = fopen( temp, "w" );
- fprintf( out_control->dpl,
- "Step Molecule Count Avg. Dipole Moment Norm\n" );
+ fprintf( out_control->dpl,
+ "Step Molecule Count Avg. Dipole Moment Norm\n" );
fflush( out_control->dpl );
}
/* Init diffusion coef analysis file */
- if( control->diffusion_coef ) {
+ if ( control->diffusion_coef )
+ {
strcpy( temp, control->sim_name );
strcat( temp, ".drft" );
out_control->drft = fopen( temp, "w" );
@@ -836,21 +916,22 @@ void Init_Out_Controls(reax_system *system, control_params *control,
#endif
/* Error handling */
- /* if ( out_control->out == NULL || out_control->pot == NULL ||
- out_control->log == NULL || out_control->mol == NULL ||
- out_control->dpl == NULL || out_control->drft == NULL ||
+ /* if ( out_control->out == NULL || out_control->pot == NULL ||
+ out_control->log == NULL || out_control->mol == NULL ||
+ out_control->dpl == NULL || out_control->drft == NULL ||
out_control->pdb == NULL )
{
fprintf( stderr, "FILE OPEN ERROR. TERMINATING..." );
- exit( CANNOT_OPEN_OUTFILE );
+ exit( CANNOT_OPEN_FILE );
}*/
}
-void Initialize(reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace, list **lists,
+void Initialize(reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
output_controls *out_control, evolve_function *Evolve)
{
+ real start, end;
Randomize();
Init_System( system, control, data );
@@ -870,10 +951,16 @@ void Initialize(reax_system *system, control_params *control,
Init_Force_Test_Functions( );
#endif
- if( control->tabulate )
+ if ( control->tabulate )
+ {
+ start = Get_Time ();
Make_LR_Lookup_Table( system, control );
+ end = Get_Timing_Info (start);
+
+ //fprintf (stderr, "Time for LR Lookup Table calculation is %f \n", end );
+ }
#if defined(DEBUG_FOCUS)
- fprintf( stderr, "data structures have been initialized...\n" );
+ fprintf( stderr, "data structures have been initialized...\n" );
#endif
}
diff --git a/PuReMD-GPU/src/init_md.h b/PuReMD-GPU/src/init_md.h
index 8c23806594a8f2b107ddb884efbf68e7b5fe27ff..947d81e6e50e96f325742c6d024c6011a900152c 100644
--- a/PuReMD-GPU/src/init_md.h
+++ b/PuReMD-GPU/src/init_md.h
@@ -31,10 +31,10 @@ extern "C" {
void Initialize( reax_system*, control_params*, simulation_data*,
static_storage*, list**, output_controls*, evolve_function* );
-void Generate_Initial_Velocities(reax_system *, real );
+void Generate_Initial_Velocities( reax_system *, real );
-void Init_Out_Controls(reax_system *, control_params *, static_storage *,
- output_controls *);
+void Init_Out_Controls( reax_system *, control_params *, static_storage *,
+ output_controls * );
#ifdef __cplusplus
}
diff --git a/PuReMD-GPU/src/integrate.c b/PuReMD-GPU/src/integrate.c
index 482a9c89a302c052e9ac44ae2de446c61b1c6a3e..d65406f8824697a396c65337bd197ad771320e81 100644
--- a/PuReMD-GPU/src/integrate.c
+++ b/PuReMD-GPU/src/integrate.c
@@ -1,32 +1,32 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "integrate.h"
-
#include "allocate.h"
#include "box.h"
#include "forces.h"
#include "grid.h"
#include "neighbors.h"
#include "print_utils.h"
-#include "QEq.h"
+#include "qeq.h"
#include "reset_utils.h"
#include "restart.h"
#include "system_props.h"
@@ -34,9 +34,10 @@
#include "list.h"
-void Velocity_Verlet_NVE(reax_system* system, control_params* control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+
+void Velocity_Verlet_NVE(reax_system* system, control_params* control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i, steps, renbr;
real inv_m, dt, dt_sqr;
@@ -46,53 +47,50 @@ void Velocity_Verlet_NVE(reax_system* system, control_params* control,
dt_sqr = SQR(dt);
steps = data->step - data->prev_steps;
renbr = (steps % control->reneighbor == 0);
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "step%d: ", data->step );
#endif
- for( i = 0; i < system->N; i++ ) {
+ for ( i = 0; i < system->N; i++ )
+ {
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
- rvec_ScaledSum( dx, dt, system->atoms[i].v,
- 0.5 * dt_sqr * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledSum( dx, dt, system->atoms[i].v,
+ 0.5 * dt_sqr * -F_CONV * inv_m, system->atoms[i].f );
Inc_on_T3( system->atoms[i].x, dx, &( system->box ) );
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
}
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet1 - ");
#endif
Reallocate( system, workspace, lists, renbr );
Reset( system, control, data, workspace, lists );
- if( renbr )
- {
- Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
- }
+ if ( renbr )
+ Generate_Neighbor_Lists( system, control, data, workspace,
+ lists, out_control );
Compute_Forces( system, control, data, workspace, lists, out_control );
- for( i = 0; i < system->N; i++ )
+ for ( i = 0; i < system->N; i++ )
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
}
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet2\n");
#endif
}
-void Velocity_Verlet_Nose_Hoover_NVT_Klein(reax_system* system,
- control_params* control,
- simulation_data *data,
- static_storage *workspace,
- list **lists,
+
+void Velocity_Verlet_Nose_Hoover_NVT_Klein(reax_system* system,
+ control_params* control,
+ simulation_data *data,
+ static_storage *workspace,
+ list **lists,
output_controls *out_control )
{
int i, itr, steps, renbr;
@@ -106,22 +104,17 @@ void Velocity_Verlet_Nose_Hoover_NVT_Klein(reax_system* system,
therm = &( data->therm );
steps = data->step - data->prev_steps;
renbr = (steps % control->reneighbor == 0);
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "step%d: ", data->step );
#endif
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " Entering Velocity_Verlet_Nose_Hoover_NVT_Klein: coef to update velocity --> %6.10f\n", therm->v_xi_old);
-#endif
-
/* Compute x(t + dt) and copy old forces */
- for (i=0; i < system->N; i++)
+ for (i = 0; i < system->N; i++)
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
rvec_ScaledSum( dx, dt - 0.5 * dt_sqr * therm->v_xi, system->atoms[i].v,
- 0.5 * dt_sqr * inv_m * -F_CONV, system->atoms[i].f );
+ 0.5 * dt_sqr * inv_m * -F_CONV, system->atoms[i].f );
Inc_on_T3( system->atoms[i].x, dx, &(system->box) );
@@ -129,105 +122,88 @@ void Velocity_Verlet_Nose_Hoover_NVT_Klein(reax_system* system,
}
/* Compute xi(t + dt) */
therm->xi += ( therm->v_xi * dt + 0.5 * dt_sqr * therm->G_xi );
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet1 - " );
#endif
Reallocate( system, workspace, lists, renbr );
Reset( system, control, data, workspace, lists );
-
- if( renbr )
- {
- Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
- }
-
+ if ( renbr )
+ Generate_Neighbor_Lists( system, control, data, workspace,
+ lists, out_control );
/* Calculate Forces at time (t + dt) */
- Compute_Forces( system,control,data, workspace, lists, out_control );
+ Compute_Forces( system, control, data, workspace, lists, out_control );
/* Compute iteration constants for each atom's velocity */
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
- rvec_Scale( workspace->v_const[i],
- 1.0 - 0.5 * dt * therm->v_xi, system->atoms[i].v );
- rvec_ScaledAdd( workspace->v_const[i],
- 0.5 * dt * inv_m * -F_CONV, workspace->f_old[i] );
- rvec_ScaledAdd( workspace->v_const[i],
- 0.5 * dt * inv_m * -F_CONV, system->atoms[i].f );
+ rvec_Scale( workspace->v_const[i],
+ 1.0 - 0.5 * dt * therm->v_xi, system->atoms[i].v );
+ rvec_ScaledAdd( workspace->v_const[i],
+ 0.5 * dt * inv_m * -F_CONV, workspace->f_old[i] );
+ rvec_ScaledAdd( workspace->v_const[i],
+ 0.5 * dt * inv_m * -F_CONV, system->atoms[i].f );
#if defined(DEBUG)
- fprintf( stderr, "atom%d: inv_m=%f, C1=%f, C2=%f, v_const=%f %f %f\n",
- i, inv_m, 1.0 - 0.5 * dt * therm->v_xi,
- 0.5 * dt * inv_m * -F_CONV, workspace->v_const[i][0],
- workspace->v_const[i][1], workspace->v_const[i][2] );
+ fprintf( stderr, "atom%d: inv_m=%f, C1=%f, C2=%f, v_const=%f %f %f\n",
+ i, inv_m, 1.0 - 0.5 * dt * therm->v_xi,
+ 0.5 * dt * inv_m * -F_CONV, workspace->v_const[i][0],
+ workspace->v_const[i][1], workspace->v_const[i][2] );
#endif
}
v_xi_new = therm->v_xi_old + 2.0 * dt * therm->G_xi;
E_kin_new = G_xi_new = v_xi_old = 0;
itr = 0;
- do {
- itr++;
+ do
+ {
+ itr++;
/* new values become old in this iteration */
v_xi_old = v_xi_new;
coef_v = 1.0 / (1.0 + 0.5 * dt * v_xi_old);
E_kin_new = 0;
-
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " *********** coef to update velocity --> %6.10f, %6.10f, %6.10f\n", coef_v, dt, therm->v_xi_old);
- //print_sys_atoms (system);
-#endif
-
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
{
rvec_Scale( system->atoms[i].v, coef_v, workspace->v_const[i] );
- E_kin_new += ( 0.5*system->reaxprm.sbp[system->atoms[i].type].mass *
- rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
+ E_kin_new += ( 0.5 * system->reaxprm.sbp[system->atoms[i].type].mass *
+ rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
#if defined(DEBUG)
- fprintf( stderr, "itr%d-atom%d: coef_v = %f, v_xi_old = %f\n",
- itr, i, coef_v, v_xi_old );
+ fprintf( stderr, "itr%d-atom%d: coef_v = %f, v_xi_old = %f\n",
+ itr, i, coef_v, v_xi_old );
#endif
}
- G_xi_new = control->Tau_T * ( 2.0 * E_kin_new -
- data->N_f * K_B * control->T );
+ G_xi_new = control->Tau_T * ( 2.0 * E_kin_new -
+ data->N_f * K_B * control->T );
v_xi_new = therm->v_xi + 0.5 * dt * ( therm->G_xi + G_xi_new );
-
#if defined(DEBUG)
fprintf( stderr, "itr%d: G_xi_new = %f, v_xi_new = %f, v_xi_old = %f\n",
- itr, G_xi_new, v_xi_new, v_xi_old );
+ itr, G_xi_new, v_xi_new, v_xi_old );
#endif
}
- while( fabs(v_xi_new - v_xi_old ) > 1e-5 );
+ while ( fabs(v_xi_new - v_xi_old ) > 1e-5 );
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " Iteration Count in NVE --> %d \n", itr );
-#endif
-
-#ifndef __BUILD_DEBUG__
therm->v_xi_old = therm->v_xi;
therm->v_xi = v_xi_new;
- therm->G_xi = G_xi_new;
-#endif
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr,"vel scale\n" );
-#endif
+ therm->G_xi = G_xi_new;
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "vel scale\n" );
+#endif
}
-/* uses Berendsen-type coupling for both T and P.
- All box dimensions are scaled by the same amount,
+
+/* uses Berendsen-type coupling for both T and P.
+ All box dimensions are scaled by the same amount,
there is no change in the angles between axes. */
-void Velocity_Verlet_Berendsen_Isotropic_NPT( reax_system* system,
- control_params* control,
+void Velocity_Verlet_Berendsen_Isotropic_NPT( reax_system* system,
+ control_params* control,
simulation_data *data,
- static_storage *workspace,
- list **lists,
+ static_storage *workspace,
+ list **lists,
output_controls *out_control )
{
int i, steps, renbr;
@@ -237,94 +213,102 @@ void Velocity_Verlet_Berendsen_Isotropic_NPT( reax_system* system,
dt = control->dt;
steps = data->step - data->prev_steps;
renbr = (steps % control->reneighbor == 0);
-
#if defined(DEBUG_FOCUS)
- //fprintf( out_control->prs,
- // "tau_t: %g tau_p: %g dt/tau_t: %g dt/tau_p: %g\n",
+ //fprintf( out_control->prs,
+ // "tau_t: %g tau_p: %g dt/tau_t: %g dt/tau_p: %g\n",
//control->Tau_T, control->Tau_P, dt / control->Tau_T, dt / control->Tau_P );
fprintf( stderr, "step %d: ", data->step );
#endif
/* velocity verlet, 1st part */
- for( i = 0; i < system->N; i++ )
+ for ( i = 0; i < system->N; i++ )
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
/* Compute x(t + dt) */
- rvec_ScaledSum( dx, dt, system->atoms[i].v,
- 0.5 * -F_CONV * inv_m * SQR(dt), system->atoms[i].f );
+ rvec_ScaledSum( dx, dt, system->atoms[i].v,
+ 0.5 * -F_CONV * inv_m * SQR(dt), system->atoms[i].f );
Inc_on_T3( system->atoms[i].x, dx, &(system->box) );
/* Compute v(t + dt/2) */
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * -F_CONV * inv_m * dt, system->atoms[i].f );
- /*fprintf( stderr, "%6d %15.8f %15.8f %15.8f %15.8f %15.8f %15.8f\n",
- workspace->orig_id[i],
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * -F_CONV * inv_m * dt, system->atoms[i].f );
+ /*fprintf( stderr, "%6d %15.8f %15.8f %15.8f %15.8f %15.8f %15.8f\n",
+ workspace->orig_id[i],
system->atoms[i].x[0], system->atoms[i].x[1], system->atoms[i].x[2],
- 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[0],
- 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[1],
+ 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[0],
+ 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[1],
0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[2] ); */
}
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet1 - " );
#endif
- Reallocate( system, workspace, lists, renbr );
+ Reallocate( system, workspace, lists, renbr );
Reset( system, control, data, workspace, lists );
- if( renbr ) {
+ if ( renbr )
+ {
Update_Grid( system );
Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
+ lists, out_control );
}
Compute_Forces( system, control, data, workspace, lists, out_control );
/* velocity verlet, 2nd part */
- for( i = 0; i < system->N; i++ ) {
+ for ( i = 0; i < system->N; i++ )
+ {
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
/* Compute v(t + dt) */
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
- /* fprintf( stderr, "%6d %15f %15f %15f %15.8f %15.8f %15.8f\n",
- workspace->orig_id[i],
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ /* fprintf( stderr, "%6d %15f %15f %15f %15.8f %15.8f %15.8f\n",
+ workspace->orig_id[i],
system->atoms[i].v[0], system->atoms[i].v[1], system->atoms[i].v[2],
- 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[0],
- 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[1],
+ 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[0],
+ 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[1],
0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[2] );*/
}
- //Compute_Kinetic_Energy( system, data );
+ //TODO: commented out for GPU version, why?
+#ifndef HAVE_CUDA
+ Compute_Kinetic_Energy( system, data );
+#endif
Compute_Pressure_Isotropic( system, control, data, out_control );
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet2 - " );
#endif
/* pressure scaler */
mu = POW( 1.0 + (dt / control->Tau_P[0]) * (data->iso_bar.P - control->P[0]),
- 1.0 / 3 );
- if( mu < MIN_dV )
+ 1.0 / 3 );
+ if ( mu < MIN_dV )
mu = MIN_dV;
- else if( mu > MAX_dV )
+ else if ( mu > MAX_dV )
mu = MAX_dV;
/* temperature scaler */
lambda = 1.0 + (dt / control->Tau_T) * (control->T / data->therm.T - 1.0);
- if( lambda < MIN_dT )
+ if ( lambda < MIN_dT )
lambda = MIN_dT;
else if (lambda > MAX_dT )
lambda = MAX_dT;
lambda = SQRT( lambda );
/* Scale velocities and positions at t+dt */
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
rvec_Scale( system->atoms[i].v, lambda, system->atoms[i].v );
- /* IMPORTANT: What Adri does with scaling positions first to
- unit coordinates and then back to cartesian coordinates essentially
- is scaling the coordinates with mu^2. However, this causes unphysical
+ /* IMPORTANT: What Adri does with scaling positions first to
+ unit coordinates and then back to cartesian coordinates essentially
+ is scaling the coordinates with mu^2. However, this causes unphysical
modifications on the system because box dimensions
are being scaled with mu! We need to discuss this with Adri! */
rvec_Scale( system->atoms[i].x, mu, system->atoms[i].x );
}
- //Compute_Kinetic_Energy( system, data );
-#if defined(DEBUG_FOCUS)
+ //TODO: commented out for GPU version, why?
+#ifndef HAVE_CUDA
+ Compute_Kinetic_Energy( system, data );
+#endif
+
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "scaling - " );
#endif
@@ -335,14 +319,14 @@ void Velocity_Verlet_Berendsen_Isotropic_NPT( reax_system* system,
}
-/* uses Berendsen-type coupling for both T and P.
- All box dimensions are scaled by the same amount,
+/* uses Berendsen-type coupling for both T and P.
+ All box dimensions are scaled by the same amount,
there is no change in the angles between axes. */
-void Velocity_Verlet_Berendsen_SemiIsotropic_NPT( reax_system* system,
- control_params* control,
+void Velocity_Verlet_Berendsen_SemiIsotropic_NPT( reax_system* system,
+ control_params* control,
simulation_data *data,
- static_storage *workspace,
- list **lists,
+ static_storage *workspace,
+ list **lists,
output_controls *out_control )
{
int i, d, steps, renbr;
@@ -352,120 +336,139 @@ void Velocity_Verlet_Berendsen_SemiIsotropic_NPT( reax_system* system,
dt = control->dt;
steps = data->step - data->prev_steps;
renbr = (steps % control->reneighbor == 0);
-
#if defined(DEBUG_FOCUS)
- //fprintf( out_control->prs,
- // "tau_t: %g tau_p: %g dt/tau_t: %g dt/tau_p: %g\n",
+ //fprintf( out_control->prs,
+ // "tau_t: %g tau_p: %g dt/tau_t: %g dt/tau_p: %g\n",
//control->Tau_T, control->Tau_P, dt / control->Tau_T, dt / control->Tau_P );
fprintf( stderr, "step %d: ", data->step );
#endif
/* velocity verlet, 1st part */
- for( i = 0; i < system->N; i++ ) {
- inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
+ for ( i = 0; i < system->N; i++ )
+ {
+ inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
/* Compute x(t + dt) */
- rvec_ScaledSum( dx, dt, system->atoms[i].v,
- 0.5 * -F_CONV * inv_m * SQR(dt), system->atoms[i].f );
+ rvec_ScaledSum( dx, dt, system->atoms[i].v,
+ 0.5 * -F_CONV * inv_m * SQR(dt), system->atoms[i].f );
Inc_on_T3( system->atoms[i].x, dx, &(system->box) );
/* Compute v(t + dt/2) */
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * -F_CONV * inv_m * dt, system->atoms[i].f );
- /*fprintf( stderr, "%6d %15.8f %15.8f %15.8f %15.8f %15.8f %15.8f\n",
- workspace->orig_id[i],
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * -F_CONV * inv_m * dt, system->atoms[i].f );
+ /*fprintf( stderr, "%6d %15.8f %15.8f %15.8f %15.8f %15.8f %15.8f\n",
+ workspace->orig_id[i],
system->atoms[i].x[0], system->atoms[i].x[1], system->atoms[i].x[2],
- 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[0],
- 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[1],
+ 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[0],
+ 0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[1],
0.5 * SQR(dt) * -F_CONV * inv_m * system->atoms[i].f[2] ); */
}
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet1 - " );
#endif
- Reallocate( system, workspace, lists, renbr );
+ Reallocate( system, workspace, lists, renbr );
Reset( system, control, data, workspace, lists );
- if( renbr ) {
+ if ( renbr )
+ {
Update_Grid( system );
- Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
+ Generate_Neighbor_Lists( system, control, data, workspace,
+ lists, out_control );
}
Compute_Forces( system, control, data, workspace, lists, out_control );
/* velocity verlet, 2nd part */
- for( i = 0; i < system->N; i++ ) {
+ for ( i = 0; i < system->N; i++ )
+ {
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
/* Compute v(t + dt) */
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
- /* fprintf( stderr, "%6d %15f %15f %15f %15.8f %15.8f %15.8f\n",
- workspace->orig_id[i],
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ /* fprintf( stderr, "%6d %15f %15f %15f %15.8f %15.8f %15.8f\n",
+ workspace->orig_id[i],
system->atoms[i].v[0], system->atoms[i].v[1], system->atoms[i].v[2],
- 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[0],
- 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[1],
+ 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[0],
+ 0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[1],
0.5 * dt * -F_CONV * inv_m * system->atoms[i].f[2] );*/
}
- //Compute_Kinetic_Energy( system, data );
+ //TODO: commented out for GPU version, why?
+#ifndef HAVE_CUDA
+ Compute_Kinetic_Energy( system, data );
+#endif
Compute_Pressure_Isotropic( system, control, data, out_control );
-
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "verlet2 - " );
#endif
/* pressure scaler */
- for( d = 0; d < 3; ++d ){
- mu[d] = POW( 1.0+(dt/control->Tau_P[d])*(data->tot_press[d]-control->P[d]),
- 1.0 / 3 );
- if( mu[d] < MIN_dV )
+ for ( d = 0; d < 3; ++d )
+ {
+ mu[d] = POW( 1.0 + (dt / control->Tau_P[d]) * (data->tot_press[d] - control->P[d]),
+ 1.0 / 3 );
+ if ( mu[d] < MIN_dV )
+ {
mu[d] = MIN_dV;
- else if( mu[d] > MAX_dV )
+ }
+ else if ( mu[d] > MAX_dV )
+ {
mu[d] = MAX_dV;
+ }
}
/* temperature scaler */
lambda = 1.0 + (dt / control->Tau_T) * (control->T / data->therm.T - 1.0);
- if( lambda < MIN_dT )
+ if ( lambda < MIN_dT )
+ {
lambda = MIN_dT;
+ }
else if (lambda > MAX_dT )
+ {
lambda = MAX_dT;
+ }
lambda = SQRT( lambda );
/* Scale velocities and positions at t+dt */
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
rvec_Scale( system->atoms[i].v, lambda, system->atoms[i].v );
- /* IMPORTANT: What Adri does with scaling positions first to
- unit coordinates and then back to cartesian coordinates essentially
- is scaling the coordinates with mu^2. However, this causes unphysical
+ /* IMPORTANT: What Adri does with scaling positions first to
+ unit coordinates and then back to cartesian coordinates essentially
+ is scaling the coordinates with mu^2. However, this causes unphysical
modifications on the system because box dimensions
are being scaled with mu! We need to discuss this with Adri! */
- for( d = 0; d < 3; ++d )
+ for ( d = 0; d < 3; ++d )
system->atoms[i].x[d] = system->atoms[i].x[d] * mu[d];
}
- //Compute_Kinetic_Energy( system, data );
-#if defined(DEBUG_FOCUS)
+ //TODO: commented out for GPU version, why?
+#ifndef HAVE_CUDA
+ Compute_Kinetic_Energy( system, data );
+#endif
+
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "scaling - " );
#endif
Update_Box_SemiIsotropic( &(system->box), mu );
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "updated box & grid\n" );
#endif
}
+
/************************************************/
/* BELOW FUNCTIONS ARE NOT BEING USED ANYMORE! */
/* */
/*!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!*/
/*!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!*/
/************************************************/
+
#ifdef ANISOTROPIC
-void Velocity_Verlet_Nose_Hoover_NVT(reax_system* system,
- control_params* control,
- simulation_data *data,
- static_storage *workspace,
- list **lists,
- output_controls *out_control )
+void Velocity_Verlet_Nose_Hoover_NVT(reax_system* system,
+ control_params* control,
+ simulation_data *data,
+ static_storage *workspace,
+ list **lists,
+ output_controls *out_control )
{
int i;
real inv_m;
@@ -473,73 +476,77 @@ void Velocity_Verlet_Nose_Hoover_NVT(reax_system* system,
real dt_sqr = SQR(dt);
rvec dx;
- for (i=0; i < system->N; i++)
+ for (i = 0; i < system->N; i++)
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
// Compute x(t + dt)
- rvec_ScaledSum( dx, dt, system->atoms[i].v,
- 0.5 * dt_sqr * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledSum( dx, dt, system->atoms[i].v,
+ 0.5 * dt_sqr * -F_CONV * inv_m, system->atoms[i].f );
Inc_on_T3_Gen( system->atoms[i].x, dx, &(system->box) );
// Compute v(t + dt/2)
- rvec_ScaledAdd( system->atoms[i].v,
- -0.5 * dt * data->therm.xi, system->atoms[i].v );
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledAdd( system->atoms[i].v,
+ -0.5 * dt * data->therm.xi, system->atoms[i].v );
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
}
// Compute zeta(t + dt/2), E_Kininetic(t + dt/2)
// IMPORTANT: What will be the initial value of zeta? and what is g?
- data->therm.xi += 0.5 * dt * control->Tau_T *
- ( 2.0 * data->E_Kin - data->N_f * K_B * control->T );
+ data->therm.xi += 0.5 * dt * control->Tau_T *
+ ( 2.0 * data->E_Kin - data->N_f * K_B * control->T );
Reset( system, control, data, workspace );
- fprintf(out_control->log,"reset-"); fflush( out_control->log );
+ fprintf(out_control->log, "reset-");
+ fflush( out_control->log );
- Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
- fprintf(out_control->log,"nbrs-"); fflush( out_control->log );
+ Generate_Neighbor_Lists( system, control, data, workspace,
+ lists, out_control );
+ fprintf(out_control->log, "nbrs-");
+ fflush( out_control->log );
/* QEq( system, control, workspace, lists[FAR_NBRS], out_control );
fprintf(out_control->log,"qeq-"); fflush( out_control->log ); */
Compute_Forces( system, control, data, workspace, lists, out_control );
- fprintf(out_control->log,"forces\n"); fflush( out_control->log );
+ fprintf(out_control->log, "forces\n");
+ fflush( out_control->log );
- //Compute_Kinetic_Energy( system, data );
+ //TODO: commented out for GPU version, why?
+#ifndef HAVE_CUDA
+ Compute_Kinetic_Energy( system, data );
+#endif
- for( i = 0; i < system->N; i++ )
+ for ( i = 0; i < system->N; i++ )
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
// compute v(t + dt)
- rvec_ScaledAdd( system->atoms[i].v,
- -0.5 * dt * data->therm.xi, system->atoms[i].v );
- rvec_ScaledAdd( system->atoms[i].v,
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledAdd( system->atoms[i].v,
+ -0.5 * dt * data->therm.xi, system->atoms[i].v );
+ rvec_ScaledAdd( system->atoms[i].v,
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
}
// Compute zeta(t + dt)
- data->therm.xi += 0.5*dt * control->Tau_T * ( 2.0 * data->E_Kin -
- data->N_f * K_B * control->T );
+ data->therm.xi += 0.5 * dt * control->Tau_T * ( 2.0 * data->E_Kin -
+ data->N_f * K_B * control->T );
- fprintf( out_control->log,"Xi: %8.3f %8.3f %8.3f\n",
- data->therm.xi, data->E_Kin, data->N_f * K_B * control->T );
+ fprintf( out_control->log, "Xi: %8.3f %8.3f %8.3f\n",
+ data->therm.xi, data->E_Kin, data->N_f * K_B * control->T );
fflush( out_control->log );
}
-void Velocity_Verlet_Isotropic_NPT( reax_system* system,
- control_params* control,
- simulation_data *data,
- static_storage *workspace,
- list **lists,
+
+void Velocity_Verlet_Isotropic_NPT( reax_system* system, control_params* control,
+ simulation_data *data, static_storage *workspace, list **lists,
output_controls *out_control )
{
int i, itr;
- real deps, v_eps_new=0, v_eps_old=0, G_xi_new;
- real dxi, v_xi_new=0, v_xi_old=0, a_eps_new;
+ real deps, v_eps_new = 0, v_eps_old = 0, G_xi_new;
+ real dxi, v_xi_new = 0, v_xi_old = 0, a_eps_new;
real inv_m, exp_deps, inv_3V;
real E_kin, P_int, P_int_const;
real coef_v, coef_v_eps;
@@ -552,37 +559,37 @@ void Velocity_Verlet_Isotropic_NPT( reax_system* system,
// Here we just calculate how much to increment eps, xi, v_eps, v_xi.
// Commits are done after positions and velocities of atoms are updated
- // because position, velocity updates uses v_eps, v_xi terms;
- // yet we need EXP( deps ) to be able to calculate
- // positions and velocities accurately.
- iso_bar->a_eps = control->Tau_P *
- ( 3.0 * box->volume * (iso_bar->P - control->P) +
- 6.0 * data->E_Kin / data->N_f ) - iso_bar->v_eps * therm->v_xi;
+ // because position, velocity updates uses v_eps, v_xi terms;
+ // yet we need EXP( deps ) to be able to calculate
+ // positions and velocities accurately.
+ iso_bar->a_eps = control->Tau_P *
+ ( 3.0 * box->volume * (iso_bar->P - control->P) +
+ 6.0 * data->E_Kin / data->N_f ) - iso_bar->v_eps * therm->v_xi;
deps = dt * iso_bar->v_eps + 0.5 * dt_sqr * iso_bar->a_eps;
exp_deps = EXP( deps );
- therm->G_xi = control->Tau_T * ( 2.0 * data->E_Kin +
- SQR( iso_bar->v_eps ) / control->Tau_P -
- (data->N_f +1) * K_B * control->T );
+ therm->G_xi = control->Tau_T * ( 2.0 * data->E_Kin +
+ SQR( iso_bar->v_eps ) / control->Tau_P -
+ (data->N_f + 1) * K_B * control->T );
dxi = therm->v_xi * dt + 0.5 * therm->G_xi * dt_sqr;
- fprintf(out_control->log, "a: %12.6f eps: %12.6f deps: %12.6f\n",
+ fprintf(out_control->log, "a: %12.6f eps: %12.6f deps: %12.6f\n",
iso_bar->a_eps, iso_bar->v_eps, iso_bar->eps);
- fprintf(out_control->log, "G: %12.6f xi : %12.6f dxi : %12.6f\n",
+ fprintf(out_control->log, "G: %12.6f xi : %12.6f dxi : %12.6f\n",
therm->G_xi, therm->v_xi, therm->xi );
// Update positions and velocities
- // NOTE: v_old, v_xi_old, v_eps_old are meant to be the old values
+ // NOTE: v_old, v_xi_old, v_eps_old are meant to be the old values
// in the iteration not the old values at time t or before!
- for (i=0; i < system->N; i++)
+ for (i = 0; i < system->N; i++)
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
// Compute x(t + dt)
- rvec_ScaledSum( workspace->a[i], -F_CONV * inv_m, system->atoms[i].f,
- -( (2.0 + 3.0/data->N_f) * iso_bar->v_eps + therm->v_xi ),
- system->atoms[i].v );
- rvec_ScaledSum( dx, dt, system->atoms[i].v,
+ rvec_ScaledSum( workspace->a[i], -F_CONV * inv_m, system->atoms[i].f,
+ -( (2.0 + 3.0 / data->N_f) * iso_bar->v_eps + therm->v_xi ),
+ system->atoms[i].v );
+ rvec_ScaledSum( dx, dt, system->atoms[i].v,
0.5 * dt_sqr, workspace->a[i] );
Inc_on_T3( system->atoms[i].x, dx, &(system->box) );
rvec_Scale( system->atoms[i].x, exp_deps, system->atoms[i].x );
@@ -597,39 +604,40 @@ void Velocity_Verlet_Isotropic_NPT( reax_system* system,
// Calculate new forces, f(t + dt)
Reset( system, control, data, workspace );
- fprintf(out_control->log,"reset-"); fflush( out_control->log );
+ fprintf(out_control->log, "reset-");
+ fflush( out_control->log );
- Generate_Neighbor_Lists( system, control, data, workspace,
- lists, out_control );
- fprintf(out_control->log,"nbrs-"); fflush( out_control->log );
+ Generate_Neighbor_Lists( system, control, data, workspace,
+ lists, out_control );
+ fprintf(out_control->log, "nbrs-");
+ fflush( out_control->log );
/* QEq( system, control, workspace, lists[FAR_NBRS], out_control );
fprintf(out_control->log,"qeq-"); fflush( out_control->log ); */
Compute_Forces( system, control, data, workspace, lists, out_control );
- fprintf(out_control->log,"forces\n"); fflush( out_control->log );
-
+ fprintf(out_control->log, "forces\n");
+ fflush( out_control->log );
// Compute iteration constants for each atom's velocity and for P_internal
// Compute kinetic energy for initial velocities of the iteration
P_int_const = E_kin = 0;
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
{
inv_m = 1.0 / system->reaxprm.sbp[system->atoms[i].type].mass;
- rvec_ScaledSum( dv, 0.5 * dt, workspace->a[i],
- 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
+ rvec_ScaledSum( dv, 0.5 * dt, workspace->a[i],
+ 0.5 * dt * -F_CONV * inv_m, system->atoms[i].f );
rvec_Add( dv, system->atoms[i].v );
rvec_Scale( workspace->v_const[i], exp_deps, dv );
- P_int_const += ( -F_CONV *
- rvec_Dot( system->atoms[i].f, system->atoms[i].x ) );
+ P_int_const += ( -F_CONV *
+ rvec_Dot( system->atoms[i].f, system->atoms[i].x ) );
- E_kin += (0.5 * system->reaxprm.sbp[system->atoms[i].type].mass *
- rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
+ E_kin += (0.5 * system->reaxprm.sbp[system->atoms[i].type].mass *
+ rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
}
-
// Compute initial p_int
inv_3V = 1.0 / (3.0 * system->box.volume);
P_int = inv_3V * ( 2.0 * E_kin + P_int_const );
@@ -645,42 +653,38 @@ void Velocity_Verlet_Isotropic_NPT( reax_system* system,
v_xi_old = v_xi_new;
v_eps_old = v_eps_new;
-
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
{
- coef_v = 1.0 / (1.0 + 0.5 * dt * exp_deps *
- ( (2.0 + 3.0/data->N_f) * v_eps_old + v_xi_old ) );
+ coef_v = 1.0 / (1.0 + 0.5 * dt * exp_deps *
+ ( (2.0 + 3.0 / data->N_f) * v_eps_old + v_xi_old ) );
rvec_Scale( system->atoms[i].v, coef_v, workspace->v_const[i] );
}
-
coef_v_eps = 1.0 / (1.0 + 0.5 * dt * v_xi_old);
- a_eps_new = 3.0 * control->Tau_P *
- ( system->box.volume * (P_int - control->P) + 2.0 * E_kin / data->N_f );
- v_eps_new = coef_v_eps * ( iso_bar->v_eps +
+ a_eps_new = 3.0 * control->Tau_P *
+ ( system->box.volume * (P_int - control->P) + 2.0 * E_kin / data->N_f );
+ v_eps_new = coef_v_eps * ( iso_bar->v_eps +
0.5 * dt * ( iso_bar->a_eps + a_eps_new ) );
-
- G_xi_new = control->Tau_T * ( 2.0 * E_kin +
- SQR( v_eps_old ) / control->Tau_P -
+ G_xi_new = control->Tau_T * ( 2.0 * E_kin +
+ SQR( v_eps_old ) / control->Tau_P -
(data->N_f + 1) * K_B * control->T );
v_xi_new = therm->v_xi + 0.5 * dt * ( therm->G_xi + G_xi_new );
-
E_kin = 0;
- for( i = 0; i < system->N; ++i )
- E_kin += (0.5 * system->reaxprm.sbp[system->atoms[i].type].mass *
- rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
-
- P_int = inv_3V * ( 2.0*E_kin + P_int_const );
+ for ( i = 0; i < system->N; ++i )
+ {
+ E_kin += (0.5 * system->reaxprm.sbp[system->atoms[i].type].mass *
+ rvec_Dot( system->atoms[i].v, system->atoms[i].v ) );
+ }
+ P_int = inv_3V * ( 2.0 * E_kin + P_int_const );
- fprintf( out_control->log,
- "itr %d E_kin: %8.3f veps_n:%8.3f veps_o:%8.3f vxi_n:%8.3f vxi_o: %8.3f\n",
- itr, E_kin, v_eps_new, v_eps_old, v_xi_new, v_xi_old );
+ fprintf( out_control->log,
+ "itr %d E_kin: %8.3f veps_n:%8.3f veps_o:%8.3f vxi_n:%8.3f vxi_o: %8.3f\n",
+ itr, E_kin, v_eps_new, v_eps_old, v_xi_new, v_xi_old );
}
- while( fabs(v_eps_new - v_eps_old) + fabs(v_xi_new - v_xi_old) > 2e-3 );
-
+ while ( FABS(v_eps_new - v_eps_old) + fabs(v_xi_new - v_xi_old) > 2e-3 );
therm->v_xi_old = therm->v_xi;
therm->v_xi = v_xi_new;
@@ -690,36 +694,30 @@ void Velocity_Verlet_Isotropic_NPT( reax_system* system,
iso_bar->v_eps = v_eps_new;
iso_bar->a_eps = a_eps_new;
- fprintf( out_control->log, "V: %8.3ff\tsides{%8.3f, %8.3f, %8.3f}\n",
- system->box.volume,
- system->box.box[0][0],system->box.box[1][1],system->box.box[2][2] );
- fprintf(out_control->log,"eps:\ta- %8.3f v- %8.3f eps- %8.3f\n",
+ fprintf( out_control->log, "V: %8.3ff\tsides{%8.3f, %8.3f, %8.3f}\n",
+ system->box.volume,
+ system->box.box[0][0], system->box.box[1][1], system->box.box[2][2] );
+ fprintf(out_control->log, "eps:\ta- %8.3f v- %8.3f eps- %8.3f\n",
iso_bar->a_eps, iso_bar->v_eps, iso_bar->eps);
- fprintf(out_control->log,"xi: \tG- %8.3f v- %8.3f xi - %8.3f\n",
+ fprintf(out_control->log, "xi: \tG- %8.3f v- %8.3f xi - %8.3f\n",
therm->G_xi, therm->v_xi, therm->xi);
}
#endif
-/* uses Berendsen-type coupling for both T and P.
- All box dimensions are scaled by the same amount,
+/* uses Berendsen-type coupling for both T and P.
+ All box dimensions are scaled by the same amount,
there is no change in the angles between axes. */
-void Velocity_Verlet_Berendsen_NVT( reax_system* system,
- control_params* control,
- simulation_data *data,
- static_storage *workspace,
- list **lists,
- output_controls *out_control
- )
+void Velocity_Verlet_Berendsen_NVT( reax_system* system, control_params* control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
int i, steps, renbr;
real inv_m, dt, lambda;
rvec dx;
reax_atom *atom;
- fprintf (stderr, " Velocity_Verlet_Berendsen_NVT: step :%d \n", data->step);
-
#if defined(DEBUG_FOCUS)
fprintf( stderr, "step%d\n", data->step );
#endif
@@ -729,12 +727,19 @@ void Velocity_Verlet_Berendsen_NVT( reax_system* system,
renbr = (steps % control->reneighbor == 0);
/* velocity verlet, 1st part */
- for( i = 0; i < system->N; i++ ) {
+ for ( i = 0; i < system->N; i++ )
+ {
atom = &(system->atoms[i]);
inv_m = 1.0 / system->reaxprm.sbp[atom->type].mass;
/* Compute x(t + dt) */
rvec_ScaledSum( dx, dt, atom->v, 0.5 * -F_CONV * inv_m * SQR(dt), atom->f );
- rvec_Add( atom->x, dx );
+
+ //TODO: used rvec_Add in GPU version -- which is correct?
+ /* bNVT fix - Metin's suggestion */
+ /* ORIGINAL CHANGE -- CHECK THE branch serial-bnvt for the fix */
+ //rvec_Add( atom->x, dx );
+ Inc_on_T3( atom->x, dx, &( system->box ) );
+
/* Compute v(t + dt/2) */
rvec_ScaledAdd( atom->v, 0.5 * -F_CONV * inv_m * dt, atom->f );
}
@@ -746,42 +751,50 @@ void Velocity_Verlet_Berendsen_NVT( reax_system* system,
Reallocate( system, workspace, lists, renbr );
Reset( system, control, data, workspace, lists );
- if( renbr )
+ if ( renbr )
+ {
Generate_Neighbor_Lists( system, control, data, workspace, lists, out_control );
+ }
- Compute_Forces( system, control, data, workspace,
- lists, out_control );
+ Compute_Forces( system, control, data, workspace, lists, out_control );
/* velocity verlet, 2nd part */
- for( i = 0; i < system->N; i++ ) {
+ for ( i = 0; i < system->N; i++ )
+ {
atom = &(system->atoms[i]);
inv_m = 1.0 / system->reaxprm.sbp[atom->type].mass;
/* Compute v(t + dt) */
rvec_ScaledAdd( atom->v, 0.5 * dt * -F_CONV * inv_m, atom->f );
}
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf(stderr, "step%d: verlet2 done\n", data->step);
#endif
/* temperature scaler */
Compute_Kinetic_Energy( system, data );
lambda = 1.0 + (dt / control->Tau_T) * (control->T / data->therm.T - 1.0);
- if( lambda < MIN_dT )
+ if ( lambda < MIN_dT )
+ {
lambda = MIN_dT;
+ }
else if (lambda > MAX_dT )
+ {
lambda = MAX_dT;
+ }
lambda = SQRT( lambda );
+ fprintf( stderr, "step:%d lambda -> %f \n", data->step, lambda );
+
/* Scale velocities and positions at t+dt */
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
atom = &(system->atoms[i]);
rvec_Scale( atom->v, lambda, atom->v );
}
Compute_Kinetic_Energy( system, data );
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "step%d: scaled velocities\n",
- data->step );
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "step%d: scaled velocities\n", data->step );
#endif
}
diff --git a/PuReMD-GPU/src/integrate.h b/PuReMD-GPU/src/integrate.h
index 6f5848f0de84e8a50ef2c5090194618b61f185fc..55b36c559f96d5e3d45ab69177e26f0d4136d8e1 100644
--- a/PuReMD-GPU/src/integrate.h
+++ b/PuReMD-GPU/src/integrate.h
@@ -23,29 +23,30 @@
#include "mytypes.h"
+
void Velocity_Verlet_NVE( reax_system*, control_params*, simulation_data*,
static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Nose_Hoover_NVT( reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Nose_Hoover_NVT_Klein( reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Flexible_NPT( reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Isotropic_NPT( reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Berendsen_Isotropic_NPT( reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Berendsen_SemiIsotropic_NPT( reax_system*, control_params*,
- simulation_data*,
- static_storage*, list**,
- output_controls* );
+ simulation_data*, static_storage*, list**, output_controls* );
+
void Velocity_Verlet_Berendsen_NVT( reax_system* , control_params* ,
- simulation_data *, static_storage *,
- list **, output_controls * );
+ simulation_data *, static_storage *, list **, output_controls * );
+
#endif
diff --git a/PuReMD-GPU/src/lin_alg.c b/PuReMD-GPU/src/lin_alg.c
index cb141d475b0e2cf702901ed551287e0e238cdcd6..fb1e25bb83273730fcd74f82acad76ec7e5336e1 100644
--- a/PuReMD-GPU/src/lin_alg.c
+++ b/PuReMD-GPU/src/lin_alg.c
@@ -1,319 +1,1654 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "lin_alg.h"
+#include "allocate.h"
#include "list.h"
+#include "print_utils.h"
+#include "tool_box.h"
#include "vector.h"
-#include "index_utils.h"
-void Sparse_MatVec( sparse_matrix *A, real *x, real *b )
+typedef enum
+{
+ LOWER = 0,
+ UPPER = 1,
+} TRIANGULARITY;
+
+
+/* global to make OpenMP shared (Sparse_MatVec) */
+#ifdef _OPENMP
+real *b_local = NULL;
+#endif
+/* global to make OpenMP shared (apply_preconditioner) */
+real *Dinv_L = NULL, *Dinv_U = NULL;
+/* global to make OpenMP shared (tri_solve_level_sched) */
+int levels = 1;
+int levels_L = 1, levels_U = 1;
+unsigned int *row_levels_L = NULL, *level_rows_L = NULL, *level_rows_cnt_L = NULL;
+unsigned int *row_levels_U = NULL, *level_rows_U = NULL, *level_rows_cnt_U = NULL;
+unsigned int *row_levels, *level_rows, *level_rows_cnt;
+unsigned int *top = NULL;
+/* global to make OpenMP shared (graph_coloring) */
+unsigned int *color = NULL;
+unsigned int *to_color = NULL;
+unsigned int *conflict = NULL;
+unsigned int *temp_ptr;
+unsigned int *recolor = NULL;
+unsigned int recolor_cnt;
+unsigned int *color_top = NULL;
+/* global to make OpenMP shared (sort_colors) */
+unsigned int *permuted_row_col = NULL;
+unsigned int *permuted_row_col_inv = NULL;
+real *y_p = NULL;
+/* global to make OpenMP shared (permute_vector) */
+real *x_p = NULL;
+unsigned int *mapping = NULL;
+sparse_matrix *H_full;
+sparse_matrix *H_p;
+/* global to make OpenMP shared (jacobi_iter) */
+real *Dinv_b = NULL, *rp = NULL, *rp2 = NULL, *rp3 = NULL;
+
+
+/* sparse matrix-vector product Ax=b
+ * where:
+ * A: lower triangular matrix, stored in CSR format
+ * x: vector
+ * b: vector (result) */
+static void Sparse_MatVec( const sparse_matrix * const A,
+ const real * const x, real * const b )
{
int i, j, k, n, si, ei;
real H;
+#ifdef _OPENMP
+ unsigned int tid;
+#endif
n = A->n;
- for( i = 0; i < n; ++i )
- b[i] = 0;
+ Vector_MakeZero( b, n );
+
+#ifdef _OPENMP
+ tid = omp_get_thread_num();
+
+ #pragma omp master
+ {
+
+ /* keep b_local for program duration to avoid allocate/free
+ * overhead per Sparse_MatVec call*/
+ if ( b_local == NULL )
+ {
+ if ( (b_local = (real*) malloc( omp_get_num_threads() * n * sizeof(real))) == NULL )
+ {
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ }
+
+ #pragma omp barrier
+
+ Vector_MakeZero( (real * const)b_local, omp_get_num_threads() * n );
+
+#endif
+ #pragma omp for schedule(static)
+ for ( i = 0; i < n; ++i )
+ {
+ si = A->start[i];
+ ei = A->start[i + 1] - 1;
+
+ for ( k = si; k < ei; ++k )
+ {
+ j = A->j[k];
+ H = A->val[k];
+#ifdef _OPENMP
+ b_local[tid * n + j] += H * x[i];
+ b_local[tid * n + i] += H * x[j];
+#else
+ b[j] += H * x[i];
+ b[i] += H * x[j];
+#endif
+ }
+
+ // the diagonal entry is the last one in
+#ifdef _OPENMP
+ b_local[tid * n + i] += A->val[k] * x[i];
+#else
+ b[i] += A->val[k] * x[i];
+#endif
+ }
+#ifdef _OPENMP
+ #pragma omp for schedule(static)
+ for ( i = 0; i < n; ++i )
+ {
+ for ( j = 0; j < omp_get_num_threads(); ++j )
+ {
+ b[i] += b_local[j * n + i];
+ }
+ }
+#endif
+
+}
+
+
+/* Transpose A and copy into A^T
+ *
+ * A: stored in CSR
+ * A_t: stored in CSR
+ */
+void Transpose( const sparse_matrix const *A, sparse_matrix const *A_t )
+{
+ unsigned int i, j, pj, *A_t_top;
+
+ if ( (A_t_top = (unsigned int*) calloc( A->n + 1, sizeof(unsigned int))) == NULL )
+ {
+ fprintf( stderr, "Not enough space for matrix tranpose. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ memset( A_t->start, 0, (A->n + 1) * sizeof(unsigned int) );
+
+ /* count nonzeros in each column of A^T, store one row greater (see next loop) */
+ for ( i = 0; i < A->n; ++i )
+ {
+ for ( pj = A->start[i]; pj < A->start[i + 1]; ++pj )
+ {
+ ++A_t->start[A->j[pj] + 1];
+ }
+ }
+
+ /* setup the row pointers for A^T */
+ for ( i = 1; i <= A->n; ++i )
+ {
+ A_t_top[i] = A_t->start[i] = A_t->start[i] + A_t->start[i - 1];
+ }
+
+ /* fill in A^T */
+ for ( i = 0; i < A->n; ++i )
+ {
+ for ( pj = A->start[i]; pj < A->start[i + 1]; ++pj )
+ {
+ j = A->j[pj];
+ A_t->j[A_t_top[j]] = i;
+ A_t->val[A_t_top[j]] = A->val[pj];
+ ++A_t_top[j];
+ }
+ }
+
+ free( A_t_top );
+}
+
+
+/* Transpose A in-place
+ *
+ * A: stored in CSR
+ */
+void Transpose_I( sparse_matrix * const A )
+{
+ sparse_matrix * A_t;
+
+ if ( Allocate_Matrix( A_t, A->n, A->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for transposing matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ Transpose( A, A_t );
+
+ memcpy( A->start, A_t->start, sizeof(int) * (A_t->n + 1) );
+ memcpy( A->j, A_t->j, sizeof(int) * (A_t->start[A_t->n]) );
+ memcpy( A->val, A_t->val, sizeof(real) * (A_t->start[A_t->n]) );
+
+ Deallocate_Matrix( A_t );
+}
+
+
+/* Apply diagonal inverse (Jacobi) preconditioner to system residual
+ *
+ * Hdia_inv: diagonal inverse preconditioner (constructed using H)
+ * y: current residual
+ * x: preconditioned residual
+ * N: length of preconditioner and vectors (# rows in H)
+ */
+static void diag_pre_app( const real * const Hdia_inv, const real * const y,
+ real * const x, const int N )
+{
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < N; ++i )
+ {
+ x[i] = y[i] * Hdia_inv[i];
+ }
+}
+
+
+/* Solve triangular system LU*x = y using level scheduling
+ *
+ * LU: lower/upper triangular, stored in CSR
+ * y: constants in linear system (RHS)
+ * x: solution
+ * tri: triangularity of LU (lower/upper)
+ *
+ * Assumptions:
+ * LU has non-zero diagonals
+ * Each row of LU has at least one non-zero (i.e., no rows with all zeros) */
+static void tri_solve( const sparse_matrix * const LU, const real * const y,
+ real * const x, const TRIANGULARITY tri )
+{
+ int i, pj, j, si, ei;
+ real val;
+
+ #pragma omp master
+ {
+ if ( tri == LOWER )
+ {
+ for ( i = 0; i < LU->n; ++i )
+ {
+ x[i] = y[i];
+ si = LU->start[i];
+ ei = LU->start[i + 1];
+ for ( pj = si; pj < ei - 1; ++pj )
+ {
+ j = LU->j[pj];
+ val = LU->val[pj];
+ x[i] -= val * x[j];
+ }
+ x[i] /= LU->val[pj];
+ }
+ }
+ else
+ {
+ for ( i = LU->n - 1; i >= 0; --i )
+ {
+ x[i] = y[i];
+ si = LU->start[i];
+ ei = LU->start[i + 1];
+ for ( pj = si + 1; pj < ei; ++pj )
+ {
+ j = LU->j[pj];
+ val = LU->val[pj];
+ x[i] -= val * x[j];
+ }
+ x[i] /= LU->val[si];
+ }
+ }
+ }
+}
+
+
+/* Solve triangular system LU*x = y using level scheduling
+ *
+ * LU: lower/upper triangular, stored in CSR
+ * y: constants in linear system (RHS)
+ * x: solution
+ * tri: triangularity of LU (lower/upper)
+ * find_levels: perform level search if positive, otherwise reuse existing levels
+ *
+ * Assumptions:
+ * LU has non-zero diagonals
+ * Each row of LU has at least one non-zero (i.e., no rows with all zeros) */
+static void tri_solve_level_sched( const sparse_matrix * const LU, const real * const y,
+ real * const x, const TRIANGULARITY tri, int find_levels )
+{
+ int i, j, pj, local_row, local_level;
+
+ #pragma omp master
+ {
+ if ( tri == LOWER )
+ {
+ row_levels = row_levels_L;
+ level_rows = level_rows_L;
+ level_rows_cnt = level_rows_cnt_L;
+ levels = levels_L;
+ }
+ else
+ {
+ row_levels = row_levels_U;
+ level_rows = level_rows_U;
+ level_rows_cnt = level_rows_cnt_U;
+ levels = levels_U;
+ }
+
+ if ( row_levels == NULL || level_rows == NULL || level_rows_cnt == NULL )
+ {
+ if ( (row_levels = (unsigned int*) malloc((size_t)LU->n * sizeof(unsigned int))) == NULL
+ || (level_rows = (unsigned int*) malloc((size_t)LU->n * sizeof(unsigned int))) == NULL
+ || (level_rows_cnt = (unsigned int*) malloc((size_t)(LU->n + 1) * sizeof(unsigned int))) == NULL )
+ {
+ fprintf( stderr, "Not enough space for triangular solve via level scheduling. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ if ( top == NULL )
+ {
+ if ( (top = (unsigned int*) malloc((size_t)(LU->n + 1) * sizeof(unsigned int))) == NULL )
+ {
+ fprintf( stderr, "Not enough space for triangular solve via level scheduling. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ /* find levels (row dependencies in substitutions) */
+ if ( find_levels == TRUE )
+ {
+ memset( row_levels, 0, LU->n * sizeof(unsigned int) );
+ memset( level_rows_cnt, 0, LU->n * sizeof(unsigned int) );
+ memset( top, 0, LU->n * sizeof(unsigned int) );
+ levels = 1;
+
+ if ( tri == LOWER )
+ {
+ for ( i = 0; i < LU->n; ++i )
+ {
+ local_level = 1;
+ for ( pj = LU->start[i]; pj < LU->start[i + 1] - 1; ++pj )
+ {
+ local_level = MAX( local_level, row_levels[LU->j[pj]] + 1 );
+ }
+
+ levels = MAX( levels, local_level );
+ row_levels[i] = local_level;
+ ++level_rows_cnt[local_level];
+ }
+
+//#if defined(DEBUG)
+ fprintf(stderr, "levels(L): %d\n", levels);
+ fprintf(stderr, "NNZ(L): %d\n", LU->start[LU->n]);
+//#endif
+ }
+ else
+ {
+ for ( i = LU->n - 1; i >= 0; --i )
+ {
+ local_level = 1;
+ for ( pj = LU->start[i] + 1; pj < LU->start[i + 1]; ++pj )
+ {
+ local_level = MAX( local_level, row_levels[LU->j[pj]] + 1 );
+ }
+
+ levels = MAX( levels, local_level );
+ row_levels[i] = local_level;
+ ++level_rows_cnt[local_level];
+ }
+
+//#if defined(DEBUG)
+ fprintf(stderr, "levels(U): %d\n", levels);
+ fprintf(stderr, "NNZ(U): %d\n", LU->start[LU->n]);
+//#endif
+ }
+
+ for ( i = 1; i < levels + 1; ++i )
+ {
+ level_rows_cnt[i] += level_rows_cnt[i - 1];
+ top[i] = level_rows_cnt[i];
+ }
+
+ for ( i = 0; i < LU->n; ++i )
+ {
+ level_rows[top[row_levels[i] - 1]] = i;
+ ++top[row_levels[i] - 1];
+ }
+ }
+ }
+
+ #pragma omp barrier
+
+ /* perform substitutions by level */
+ if ( tri == LOWER )
+ {
+ for ( i = 0; i < levels; ++i )
+ {
+ #pragma omp for schedule(static)
+ for ( j = level_rows_cnt[i]; j < level_rows_cnt[i + 1]; ++j )
+ {
+ local_row = level_rows[j];
+ x[local_row] = y[local_row];
+ for ( pj = LU->start[local_row]; pj < LU->start[local_row + 1] - 1; ++pj )
+ {
+ x[local_row] -= LU->val[pj] * x[LU->j[pj]];
+
+ }
+ x[local_row] /= LU->val[pj];
+ }
+ }
+ }
+ else
+ {
+ for ( i = 0; i < levels; ++i )
+ {
+ #pragma omp for schedule(static)
+ for ( j = level_rows_cnt[i]; j < level_rows_cnt[i + 1]; ++j )
+ {
+ local_row = level_rows[j];
+ x[local_row] = y[local_row];
+ for ( pj = LU->start[local_row] + 1; pj < LU->start[local_row + 1]; ++pj )
+ {
+ x[local_row] -= LU->val[pj] * x[LU->j[pj]];
+
+ }
+ x[local_row] /= LU->val[LU->start[local_row]];
+ }
+ }
+ }
+
+ #pragma omp master
+ {
+ /* save level info for re-use if performing repeated triangular solves via preconditioning */
+ if ( tri == LOWER )
+ {
+ row_levels_L = row_levels;
+ level_rows_L = level_rows;
+ level_rows_cnt_L = level_rows_cnt;
+ levels_L = levels;
+ }
+ else
+ {
+ row_levels_U = row_levels;
+ level_rows_U = level_rows;
+ level_rows_cnt_U = level_rows_cnt;
+ levels_U = levels;
+ }
+ }
+
+ #pragma omp barrier
+}
+
+
+static void compute_H_full( const sparse_matrix * const H )
+{
+ int count, i, pj;
+ sparse_matrix *H_t;
+
+ if ( Allocate_Matrix( H_t, H->n, H->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for full H. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* Set up the sparse matrix data structure for A. */
+ Transpose( H, H_t );
+
+ count = 0;
+ for ( i = 0; i < H->n; ++i )
+ {
+ H_full->start[i] = count;
+
+ /* H: symmetric, lower triangular portion only stored */
+ for ( pj = H->start[i]; pj < H->start[i + 1]; ++pj )
+ {
+ H_full->val[count] = H->val[pj];
+ H_full->j[count] = H->j[pj];
+ ++count;
+ }
+ /* H^T: symmetric, upper triangular portion only stored;
+ * skip diagonal from H^T, as included from H above */
+ for ( pj = H_t->start[i] + 1; pj < H_t->start[i + 1]; ++pj )
+ {
+ H_full->val[count] = H_t->val[pj];
+ H_full->j[count] = H_t->j[pj];
+ ++count;
+ }
+ }
+ H_full->start[i] = count;
+
+ Deallocate_Matrix( H_t );
+}
+
+
+/* Iterative greedy shared-memory parallel graph coloring
+ *
+ * A: matrix to use for coloring, stored in CSR format;
+ * rows represent vertices, columns of entries within a row represent adjacent vertices
+ * (i.e., dependent rows for elimination during LU factorization)
+ * tri: triangularity of LU (lower/upper)
+ * color: vertex color (1-based)
+ *
+ * Reference:
+ * Umit V. Catalyurek et al.
+ * Graph Coloring Algorithms for Multi-core
+ * and Massively Threaded Architectures
+ * Parallel Computing, 2012
+ */
+void graph_coloring( const sparse_matrix * const A, const TRIANGULARITY tri )
+{
+ #pragma omp parallel
+ {
+#define MAX_COLOR (500)
+ int i, pj, v;
+ unsigned int temp;
+ int *fb_color;
+
+ #pragma omp master
+ {
+ memset( color, 0, sizeof(unsigned int) * A->n );
+ recolor_cnt = A->n;
+ }
+
+ /* ordering of vertices to color depends on triangularity of factor
+ * for which coloring is to be used for */
+ if ( tri == LOWER )
+ {
+ #pragma omp for schedule(static)
+ for ( i = 0; i < A->n; ++i )
+ {
+ to_color[i] = i;
+ }
+ }
+ else
+ {
+ #pragma omp for schedule(static)
+ for ( i = 0; i < A->n; ++i )
+ {
+ to_color[i] = A->n - 1 - i;
+ }
+ }
+
+ if ( (fb_color = (int*) malloc(sizeof(int) * MAX_COLOR)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for graph coloring. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ #pragma omp barrier
+
+ while ( recolor_cnt > 0 )
+ {
+ memset( fb_color, -1, sizeof(int) * MAX_COLOR );
+
+ /* color vertices */
+ #pragma omp for schedule(static)
+ for ( i = 0; i < recolor_cnt; ++i )
+ {
+ v = to_color[i];
+
+ /* colors of adjacent vertices are forbidden */
+ for ( pj = A->start[v]; pj < A->start[v + 1]; ++pj )
+ {
+ if ( v != A->j[pj] )
+ {
+ fb_color[color[A->j[pj]]] = v;
+ }
+ }
+
+ /* search for min. color which is not in conflict with adjacent vertices;
+ * start at 1 since 0 is default (invalid) color for all vertices */
+ for ( pj = 1; fb_color[pj] == v; ++pj );
+
+ /* assign discovered color (no conflict in neighborhood of adjacent vertices) */
+ color[v] = pj;
+ }
+
+ /* determine if recoloring required */
+ //TODO: switch to reduction on recolor_cnt (+) via parallel scan through recolor
+ #pragma omp master
+ {
+ temp = recolor_cnt;
+ recolor_cnt = 0;
+
+ for ( i = 0; i < temp; ++i )
+ {
+ v = to_color[i];
+
+ /* search for color conflicts with adjacent vertices */
+ for ( pj = A->start[v]; pj < A->start[v + 1]; ++pj )
+ {
+ if ( color[v] == color[A->j[pj]] && v > A->j[pj] )
+ {
+ conflict[recolor_cnt] = v;
+ color[v] = 0;
+ ++recolor_cnt;
+ break;
+ }
+ }
+ }
+
+ temp_ptr = to_color;
+ to_color = conflict;
+ conflict = temp_ptr;
+ }
+
+ #pragma omp barrier
+ }
+
+ free( fb_color );
+
+//#if defined(DEBUG)
+// #pragma omp master
+// {
+// for ( i = 0; i < A->n; ++i )
+// printf("Vertex: %5d, Color: %5d\n", i, color[i] );
+// }
+//#endif
+
+ #pragma omp barrier
+ }
+}
+
+
+/* Sort coloring
+ *
+ * n: number of entries in coloring
+ * tri: coloring to triangular factor to use (lower/upper)
+ */
+void sort_colors( const unsigned int n, const TRIANGULARITY tri )
+{
+ unsigned int i;
+
+ memset( color_top, 0, sizeof(unsigned int) * (n + 1) );
+
+ /* sort vertices by color (ascending within a color)
+ * 1) count colors
+ * 2) determine offsets of color ranges
+ * 3) sort by color
+ *
+ * note: color is 1-based */
+ for ( i = 0; i < n; ++i )
+ {
+ ++color_top[color[i]];
+ }
+ for ( i = 1; i < n + 1; ++i )
+ {
+ color_top[i] += color_top[i - 1];
+ }
+ for ( i = 0; i < n; ++i )
+ {
+ permuted_row_col[color_top[color[i] - 1]] = i;
+ ++color_top[color[i] - 1];
+ }
+
+ /* invert mapping to get map from current row/column to permuted (new) row/column */
+ for ( i = 0; i < n; ++i )
+ {
+ permuted_row_col_inv[permuted_row_col[i]] = i;
+ }
+}
+
+
+/* Apply permutation Q^T*x or Q*x based on graph coloring
+ *
+ * color: vertex color (1-based); vertices represent matrix rows/columns
+ * x: vector to permute (in-place)
+ * n: number of entries in x
+ * invert_map: if TRUE, use Q^T, otherwise use Q
+ * tri: coloring to triangular factor to use (lower/upper)
+ */
+static void permute_vector( real * const x, const unsigned int n, const int invert_map,
+ const TRIANGULARITY tri )
+{
+ unsigned int i;
+
+ #pragma omp master
+ {
+ if ( x_p == NULL )
+ {
+ if ( (x_p = (real*) malloc(sizeof(real) * n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for permuting vector. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ if ( invert_map == TRUE )
+ {
+ mapping = permuted_row_col_inv;
+ }
+ else
+ {
+ mapping = permuted_row_col;
+ }
+ }
+
+ #pragma omp barrier
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < n; ++i )
+ {
+ x_p[i] = x[mapping[i]];
+ }
+
+ #pragma omp master
+ {
+ memcpy( x, x_p, sizeof(real) * n );
+ }
+
+ #pragma omp barrier
+}
+
+
+/* Apply permutation Q^T*(LU)*Q based on graph coloring
+ *
+ * color: vertex color (1-based); vertices represent matrix rows/columns
+ * LU: matrix to permute, stored in CSR format
+ * tri: triangularity of LU (lower/upper)
+ */
+void permute_matrix( sparse_matrix * const LU, const TRIANGULARITY tri )
+{
+ int i, pj, nr, nc;
+ sparse_matrix *LUtemp;
+
+ if ( Allocate_Matrix( LUtemp, LU->n, LU->m ) == FAILURE )
+ {
+ fprintf( stderr, "Not enough space for graph coloring (factor permutation). Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* count nonzeros in each row of permuted factor (re-use color_top for counting) */
+ memset( color_top, 0, sizeof(unsigned int) * (LU->n + 1) );
+
+ if ( tri == LOWER )
+ {
+ for ( i = 0; i < LU->n; ++i )
+ {
+ nr = permuted_row_col_inv[i];
+
+ for ( pj = LU->start[i]; pj < LU->start[i + 1]; ++pj )
+ {
+ nc = permuted_row_col_inv[LU->j[pj]];
+
+ if ( nc <= nr )
+ {
+ ++color_top[nr + 1];
+ }
+ /* correct entries to maintain triangularity (lower) */
+ else
+ {
+ ++color_top[nc + 1];
+ }
+ }
+ }
+ }
+ else
+ {
+ for ( i = LU->n - 1; i >= 0; --i )
+ {
+ nr = permuted_row_col_inv[i];
+
+ for ( pj = LU->start[i]; pj < LU->start[i + 1]; ++pj )
+ {
+ nc = permuted_row_col_inv[LU->j[pj]];
+
+ if ( nc >= nr )
+ {
+ ++color_top[nr + 1];
+ }
+ /* correct entries to maintain triangularity (upper) */
+ else
+ {
+ ++color_top[nc + 1];
+ }
+ }
+ }
+ }
+
+ for ( i = 1; i < LU->n + 1; ++i )
+ {
+ color_top[i] += color_top[i - 1];
+ }
+
+ memcpy( LUtemp->start, color_top, sizeof(unsigned int) * (LU->n + 1) );
+
+ /* permute factor */
+ if ( tri == LOWER )
+ {
+ for ( i = 0; i < LU->n; ++i )
+ {
+ nr = permuted_row_col_inv[i];
+
+ for ( pj = LU->start[i]; pj < LU->start[i + 1]; ++pj )
+ {
+ nc = permuted_row_col_inv[LU->j[pj]];
+
+ if ( nc <= nr )
+ {
+ LUtemp->j[color_top[nr]] = nc;
+ LUtemp->val[color_top[nr]] = LU->val[pj];
+ ++color_top[nr];
+ }
+ /* correct entries to maintain triangularity (lower) */
+ else
+ {
+ LUtemp->j[color_top[nc]] = nr;
+ LUtemp->val[color_top[nc]] = LU->val[pj];
+ ++color_top[nc];
+ }
+ }
+ }
+ }
+ else
+ {
+ for ( i = LU->n - 1; i >= 0; --i )
+ {
+ nr = permuted_row_col_inv[i];
+
+ for ( pj = LU->start[i]; pj < LU->start[i + 1]; ++pj )
+ {
+ nc = permuted_row_col_inv[LU->j[pj]];
+
+ if ( nc >= nr )
+ {
+ LUtemp->j[color_top[nr]] = nc;
+ LUtemp->val[color_top[nr]] = LU->val[pj];
+ ++color_top[nr];
+ }
+ /* correct entries to maintain triangularity (upper) */
+ else
+ {
+ LUtemp->j[color_top[nc]] = nr;
+ LUtemp->val[color_top[nc]] = LU->val[pj];
+ ++color_top[nc];
+ }
+ }
+ }
+ }
+
+ memcpy( LU->start, LUtemp->start, sizeof(unsigned int) * (LU->n + 1) );
+ memcpy( LU->j, LUtemp->j, sizeof(unsigned int) * LU->start[LU->n] );
+ memcpy( LU->val, LUtemp->val, sizeof(real) * LU->start[LU->n] );
+
+ Deallocate_Matrix( LUtemp );
+}
+
+
+/* Setup routines to build permuted QEq matrix H (via graph coloring),
+ * used for preconditioning (incomplete factorizations computed based on
+ * permuted H)
+ *
+ * H: symmetric, lower triangular portion only, stored in CSR format;
+ * H is permuted in-place
+ */
+sparse_matrix * setup_graph_coloring( sparse_matrix * const H )
+{
+ if ( color == NULL )
+ {
+ /* internal storage for graph coloring (global to facilitate simultaneous access to OpenMP threads) */
+ if ( (color = (unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (to_color =(unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (conflict = (unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (recolor = (unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (color_top = (unsigned int*) malloc(sizeof(unsigned int) * (H->n + 1))) == NULL ||
+ (permuted_row_col = (unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (permuted_row_col_inv = (unsigned int*) malloc(sizeof(unsigned int) * H->n)) == NULL ||
+ (y_p = (real*) malloc(sizeof(real) * H->n)) == NULL ||
+ (Allocate_Matrix( H_p, H->n, H->m ) == FAILURE ) ||
+ (Allocate_Matrix( H_full, H->n, 2 * H->m - H->n ) == FAILURE ) )
+ {
+ fprintf( stderr, "not enough memory for graph coloring. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ compute_H_full( H );
+
+ graph_coloring( H_full, LOWER );
+ sort_colors( H_full->n, LOWER );
+
+ memcpy( H_p->start, H->start, sizeof(int) * (H->n + 1) );
+ memcpy( H_p->j, H->j, sizeof(int) * (H->start[H->n]) );
+ memcpy( H_p->val, H->val, sizeof(real) * (H->start[H->n]) );
+ permute_matrix( H_p, LOWER );
+
+ return H_p;
+}
+
+
+/* Jacobi iteration using truncated Neumann series: x_{k+1} = Gx_k + D^{-1}b
+ * where:
+ * G = I - D^{-1}R
+ * R = triangular matrix
+ * D = diagonal matrix, diagonals from R
+ *
+ * Note: used during the backsolves when applying preconditioners with
+ * triangular factors in iterative linear solvers
+ *
+ * Note: Newmann series arises from series expansion of the inverse of
+ * the coefficient matrix in the triangular system */
+static void jacobi_iter( const sparse_matrix * const R, const real * const Dinv,
+ const real * const b, real * const x, const TRIANGULARITY tri, const
+ unsigned int maxiter )
+{
+ unsigned int i, k, si = 0, ei = 0, iter;
+
+ iter = 0;
+
+ #pragma omp master
+ {
+ if ( Dinv_b == NULL )
+ {
+ if ( (Dinv_b = (real*) malloc(sizeof(real) * R->n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for Jacobi iteration matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ if ( rp == NULL )
+ {
+ if ( (rp = (real*) malloc(sizeof(real) * R->n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for Jacobi iteration matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ if ( rp2 == NULL )
+ {
+ if ( (rp2 = (real*) malloc(sizeof(real) * R->n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for Jacobi iteration matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ }
+
+ #pragma omp barrier
+
+ Vector_MakeZero( rp, R->n );
+
+ /* precompute and cache, as invariant in loop below */
+ #pragma omp for schedule(static)
+ for ( i = 0; i < R->n; ++i )
+ {
+ Dinv_b[i] = Dinv[i] * b[i];
+ }
+
+ do
+ {
+ // x_{k+1} = G*x_{k} + Dinv*b;
+ #pragma omp for schedule(guided)
+ for ( i = 0; i < R->n; ++i )
+ {
+ if (tri == LOWER)
+ {
+ si = R->start[i];
+ ei = R->start[i + 1] - 1;
+ }
+ else
+ {
+
+ si = R->start[i] + 1;
+ ei = R->start[i + 1];
+ }
+
+ rp2[i] = 0.;
+
+ for ( k = si; k < ei; ++k )
+ {
+ rp2[i] += R->val[k] * rp[R->j[k]];
+ }
+
+ rp2[i] *= -Dinv[i];
+ rp2[i] += Dinv_b[i];
+ }
+
+ #pragma omp master
+ {
+ rp3 = rp;
+ rp = rp2;
+ rp2 = rp3;
+ }
+
+ #pragma omp barrier
+
+ ++iter;
+ }
+ while ( iter < maxiter );
+
+ Vector_Copy( x, rp, R->n );
+}
+
+
+/* Solve triangular system LU*x = y using level scheduling
+ *
+ * workspace: data struct containing matrices, lower/upper triangular, stored in CSR
+ * control: data struct containing parameters
+ * y: constants in linear system (RHS)
+ * x: solution
+ * fresh_pre: parameter indicating if this is a newly computed (fresh) preconditioner
+ *
+ * Assumptions:
+ * Matrices have non-zero diagonals
+ * Each row of a matrix has at least one non-zero (i.e., no rows with all zeros) */
+static void apply_preconditioner( const static_storage * const workspace,
+ const control_params * const control, const real * const y,
+ real * const x, const int fresh_pre )
+{
+ int i, si;
+
+ switch ( control->pre_app_type )
+ {
+ case NONE_PA:
+ break;
+ case TRI_SOLVE_PA:
+ switch ( control->pre_comp_type )
+ {
+ case DIAG_PC:
+ diag_pre_app( workspace->Hdia_inv, y, x, workspace->H->n );
+ break;
+ case ICHOLT_PC:
+ case ILU_PAR_PC:
+ case ILUT_PAR_PC:
+ tri_solve( workspace->L, y, x, LOWER );
+ tri_solve( workspace->U, x, x, UPPER );
+ break;
+ default:
+ fprintf( stderr, "Unrecognized preconditioner application method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ }
+ break;
+ case TRI_SOLVE_LEVEL_SCHED_PA:
+ switch ( control->pre_comp_type )
+ {
+ case DIAG_PC:
+ diag_pre_app( workspace->Hdia_inv, y, x, workspace->H->n );
+ break;
+ case ICHOLT_PC:
+ case ILU_PAR_PC:
+ case ILUT_PAR_PC:
+ tri_solve_level_sched( workspace->L, y, x, LOWER, fresh_pre );
+ tri_solve_level_sched( workspace->U, x, x, UPPER, fresh_pre );
+ break;
+ default:
+ fprintf( stderr, "Unrecognized preconditioner application method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ }
+ break;
+ case TRI_SOLVE_GC_PA:
+ switch ( control->pre_comp_type )
+ {
+ case DIAG_PC:
+ fprintf( stderr, "Unsupported preconditioner computation/application method combination. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ case ICHOLT_PC:
+ case ILU_PAR_PC:
+ case ILUT_PAR_PC:
+ #pragma omp master
+ {
+ memcpy( y_p, y, sizeof(real) * workspace->H->n );
+ }
- for( i = 0; i < n; ++i ) {
- si = A->start[i];
- ei = A->start[i+1]-1;
+ #pragma omp barrier
- for( k = si; k < ei; ++k ) {
- j = A->entries[k].j;
- H = A->entries[k].val;
- b[j] += H * x[i];
- b[i] += H * x[j];
+ permute_vector( y_p, workspace->H->n, FALSE, LOWER );
+ tri_solve_level_sched( workspace->L, y_p, x, LOWER, fresh_pre );
+ tri_solve_level_sched( workspace->U, x, x, UPPER, fresh_pre );
+ permute_vector( x, workspace->H->n, TRUE, UPPER );
+ break;
+ default:
+ fprintf( stderr, "Unrecognized preconditioner application method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
}
+ break;
+ case JACOBI_ITER_PA:
+ switch ( control->pre_comp_type )
+ {
+ case DIAG_PC:
+ fprintf( stderr, "Unsupported preconditioner computation/application method combination. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ case ICHOLT_PC:
+ case ILU_PAR_PC:
+ case ILUT_PAR_PC:
+ #pragma omp master
+ {
+ if ( Dinv_L == NULL )
+ {
+ if ( (Dinv_L = (real*) malloc(sizeof(real) * workspace->L->n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for Jacobi iteration matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ }
- // the diagonal entry is the last one in
- b[i] += A->entries[k].val * x[i];
- }
-}
+ #pragma omp barrier
+ /* construct D^{-1}_L */
+ if ( fresh_pre == TRUE )
+ {
+ #pragma omp for schedule(static)
+ for ( i = 0; i < workspace->L->n; ++i )
+ {
+ si = workspace->L->start[i + 1] - 1;
+ Dinv_L[i] = 1. / workspace->L->val[si];
+ }
+ }
-void Forward_Subs( sparse_matrix *L, real *b, real *y )
-{
- int i, pj, j, si, ei;
- real val;
+ jacobi_iter( workspace->L, Dinv_L, y, x, LOWER, control->pre_app_jacobi_iters );
- for( i = 0; i < L->n; ++i ) {
- y[i] = b[i];
- si = L->start[i];
- ei = L->start[i+1];
- for( pj = si; pj < ei-1; ++pj ){
- j = L->entries[pj].j;
- val = L->entries[pj].val;
- y[i] -= val * y[j];
- }
- y[i] /= L->entries[pj].val;
- }
-}
+ #pragma omp master
+ {
+ if ( Dinv_U == NULL )
+ {
+ if ( (Dinv_U = (real*) malloc(sizeof(real) * workspace->U->n)) == NULL )
+ {
+ fprintf( stderr, "not enough memory for Jacobi iteration matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ }
+ #pragma omp barrier
-void Backward_Subs( sparse_matrix *U, real *y, real *x )
-{
- int i, pj, j, si, ei;
- real val;
+ /* construct D^{-1}_U */
+ if ( fresh_pre == TRUE )
+ {
+ #pragma omp for schedule(static)
+ for ( i = 0; i < workspace->U->n; ++i )
+ {
+ si = workspace->U->start[i];
+ Dinv_U[i] = 1. / workspace->U->val[si];
+ }
+ }
- for( i = U->n-1; i >= 0; --i ) {
- x[i] = y[i];
- si = U->start[i];
- ei = U->start[i+1];
- for( pj = si+1; pj < ei; ++pj ){
- j = U->entries[pj].j;
- val = U->entries[pj].val;
- x[i] -= val * x[j];
+ jacobi_iter( workspace->U, Dinv_U, y, x, UPPER, control->pre_app_jacobi_iters );
+ break;
+ default:
+ fprintf( stderr, "Unrecognized preconditioner application method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
}
- x[i] /= U->entries[si].val;
+ break;
+ default:
+ fprintf( stderr, "Unrecognized preconditioner application method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+
}
+
+ return;
}
-int GMRES( static_storage *workspace, sparse_matrix *H,
- real *b, real tol, real *x, FILE *fout, reax_system* system)
+/* generalized minimual residual iterative solver for sparse linear systems */
+int GMRES( const static_storage * const workspace, const control_params * const control,
+ simulation_data * const data, const sparse_matrix * const H,
+ const real * const b, const real tol, real * const x,
+ const FILE * const fout, const int fresh_pre )
{
- int i, j, k, itr, N;
- real cc, tmp1, tmp2, temp, bnorm;
+ int i, j, k, itr, N, g_j, g_itr;
+ real cc, tmp1, tmp2, temp, ret_temp, bnorm, time_start;
N = H->n;
- bnorm = Norm( b, N );
-
- /* apply the diagonal pre-conditioner to rhs */
- for( i = 0; i < N; ++i )
- workspace->b_prc[i] = b[i] * workspace->Hdia_inv[i];
-
- /* GMRES outer-loop */
- for( itr = 0; itr < MAX_ITR; ++itr ) {
- /* calculate r0 */
- Sparse_MatVec( H, x, workspace->b_prm );
- for( i = 0; i < N; ++i )
- workspace->b_prm[i] *= workspace->Hdia_inv[i]; /* pre-conditioner */
+ #pragma omp parallel default(none) private(i, j, k, itr, bnorm, ret_temp) \
+ shared(N, cc, tmp1, tmp2, temp, time_start, g_itr, g_j, stderr)
+ {
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ bnorm = Norm( b, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
+ if ( control->pre_comp_type == DIAG_PC )
+ {
+ /* apply preconditioner to RHS */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ apply_preconditioner( workspace, control, b, workspace->b_prc, fresh_pre );
+ #pragma omp master
+ {
+ data->timing.pre_app += Get_Timing_Info( time_start );
+ }
+ }
- Vector_Sum(&workspace->v[ index_wkspace_sys (0,0,system->N) ], 1.,workspace->b_prc, -1., workspace->b_prm, N);
- workspace->g[0] = Norm( &workspace->v[index_wkspace_sys (0,0,system->N)], N );
- Vector_Scale( &workspace->v[ index_wkspace_sys (0,0,system->N) ], 1.0/workspace->g[0], &workspace->v[index_wkspace_sys(0,0,system->N)], N );
+ /* GMRES outer-loop */
+ for ( itr = 0; itr < MAX_ITR; ++itr )
+ {
+ /* calculate r0 */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ Sparse_MatVec( H, x, workspace->b_prm );
+ #pragma omp master
+ {
+ data->timing.solver_spmv += Get_Timing_Info( time_start );
+ }
- /* GMRES inner-loop */
- for( j = 0; j < RESTART && fabs(workspace->g[j]) / bnorm > tol; j++ ) {
- /* matvec */
- Sparse_MatVec( H, &workspace->v[index_wkspace_sys(j,0,system->N)], &workspace->v[index_wkspace_sys(j+1,0,system->N)] );
+ if ( control->pre_comp_type == DIAG_PC )
+ {
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ apply_preconditioner( workspace, control, workspace->b_prm, workspace->b_prm, FALSE );
+ #pragma omp master
+ {
+ data->timing.pre_app += Get_Timing_Info( time_start );
+ }
+ }
- for( k = 0; k < N; ++k )
- workspace->v[ index_wkspace_sys (j+1,k,system->N)] *= workspace->Hdia_inv[k]; /*pre-conditioner*/
+ if ( control->pre_comp_type == DIAG_PC )
+ {
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ Vector_Sum( workspace->v, 1., workspace->b_prc, -1., workspace->b_prm, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
+ }
+ else
+ {
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ Vector_Sum( workspace->v, 1., b, -1., workspace->b_prm, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
+ }
- /* apply modified Gram-Schmidt to orthogonalize the new residual */
- for( i = 0; i <= j; i++ ) {
- workspace->h[ index_wkspace_res (i,j) ] = Dot( &workspace->v[index_wkspace_sys(i,0,system->N)], &workspace->v[index_wkspace_sys(j+1,0,system->N)], N );
- Vector_Add( &workspace->v[index_wkspace_sys(j+1,0,system->N)],
- -workspace->h[index_wkspace_res (i,j) ], &workspace->v[index_wkspace_sys(i,0,system->N)], N );
+ if ( control->pre_comp_type != DIAG_PC )
+ {
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ apply_preconditioner( workspace, control, workspace->v, workspace->v,
+ itr == 0 ? fresh_pre : FALSE );
+ #pragma omp master
+ {
+ data->timing.pre_app += Get_Timing_Info( time_start );
+ }
}
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ ret_temp = Norm( workspace->v, N );
+ #pragma omp single
+ {
+ workspace->g[0] = ret_temp;
+ }
+ Vector_Scale( workspace->v, 1. / workspace->g[0], workspace->v, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
- workspace->h[ index_wkspace_res (j+1,j) ] = Norm( &workspace->v[index_wkspace_sys(j+1,0,system->N)], N );
- Vector_Scale( &workspace->v[index_wkspace_sys(j+1,0,system->N)],
- 1. / workspace->h[ index_wkspace_res (j+1,j) ], &workspace->v[index_wkspace_sys(j+1,0,system->N)], N );
- // fprintf( stderr, "%d-%d: orthogonalization completed.\n", itr, j );
+ /* GMRES inner-loop */
+ for ( j = 0; j < RESTART && FABS(workspace->g[j]) / bnorm > tol; j++ )
+ {
+ /* matvec */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ Sparse_MatVec( H, workspace->v + j * N, workspace->v + (j + 1) * N );
+ #pragma omp master
+ {
+ data->timing.solver_spmv += Get_Timing_Info( time_start );
+ }
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ apply_preconditioner( workspace, control,
+ workspace->v + (j + 1) * N, workspace->v + (j + 1) * N, FALSE );
+ #pragma omp master
+ {
+ data->timing.pre_app += Get_Timing_Info( time_start );
+ }
- /* Givens rotations on the upper-Hessenberg matrix to make it U */
- for( i = 0; i <= j; i++ ) {
- if( i == j ) {
- cc = SQRT( SQR(workspace->h[ index_wkspace_res (j,j) ])+SQR(workspace->h[ index_wkspace_res (j+1,j) ]) );
- workspace->hc[j] = workspace->h[ index_wkspace_res (j,j) ] / cc;
- workspace->hs[j] = workspace->h[ index_wkspace_res (j+1,j) ] / cc;
+ if ( control->pre_comp_type == DIAG_PC )
+ {
+ /* apply modified Gram-Schmidt to orthogonalize the new residual */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ for ( i = 0; i <= j; i++ )
+ {
+ workspace->h[(RESTART + 1) * i + j] =
+ Dot( workspace->v + i * N, workspace->v + (j + 1) * N, N );
+ Vector_Add( workspace->v + (j + 1) * N, -workspace->h[(RESTART + 1) * i + j],
+ workspace->v + i * N, N );
+ }
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
+ }
+ else
+ {
+ //TODO: investigate correctness of not explicitly orthogonalizing first few vectors
+ /* apply modified Gram-Schmidt to orthogonalize the new residual */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ for ( i = 0; i < j - 1; i++ )
+ {
+ workspace->h[(RESTART + 1) * i + j] = 0;
+ }
+ }
+
+ for ( i = MAX(j - 1, 0); i <= j; i++ )
+ {
+ ret_temp = Dot( workspace->v + i * N, workspace->v + (j + 1) * N, N );
+ #pragma omp single
+ {
+ workspace->h[(RESTART + 1) * i + j] = ret_temp;
+ }
+ Vector_Add( workspace->v + (j + 1) * N,
+ -workspace->h[(RESTART + 1) * i + j], workspace->v + i * N, N );
+ }
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
}
- tmp1 = workspace->hc[i] * workspace->h[ index_wkspace_res (i,j) ] +
- workspace->hs[i] * workspace->h[ index_wkspace_res (i+1,j) ];
- tmp2 = -workspace->hs[i] * workspace->h[ index_wkspace_res (i,j) ] +
- workspace->hc[i] * workspace->h[ index_wkspace_res (i+1,j) ];
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ }
+ ret_temp = Norm( workspace->v + (j + 1) * N, N );
+ #pragma omp single
+ {
+ workspace->h[(RESTART + 1) * (j + 1) + j] = ret_temp;
+ }
+ Vector_Scale( workspace->v + (j + 1) * N,
+ 1. / workspace->h[(RESTART + 1) * (j + 1) + j],
+ workspace->v + (j + 1) * N, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
+#if defined(DEBUG)
+ fprintf( stderr, "%d-%d: orthogonalization completed.\n", itr, j );
+#endif
- workspace->h[ index_wkspace_res (i,j) ] = tmp1;
- workspace->h[ index_wkspace_res (i+1,j) ] = tmp2;
- }
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ if ( control->pre_comp_type == DIAG_PC )
+ {
+ /* Givens rotations on the upper-Hessenberg matrix to make it U */
+ for ( i = 0; i <= j; i++ )
+ {
+ if ( i == j )
+ {
+ cc = SQRT( SQR(workspace->h[(RESTART + 1) * j + j])
+ + SQR(workspace->h[(RESTART + 1) * (j + 1) + j]) );
+ workspace->hc[j] = workspace->h[(RESTART + 1) * j + j] / cc;
+ workspace->hs[j] = workspace->h[(RESTART + 1) * (j + 1) + j] / cc;
+ }
+
+ tmp1 = workspace->hc[i] * workspace->h[(RESTART + 1) * i + j] +
+ workspace->hs[i] * workspace->h[(RESTART + 1) * (i + 1) + j];
+ tmp2 = -workspace->hs[i] * workspace->h[(RESTART + 1) * i + j] +
+ workspace->hc[i] * workspace->h[(RESTART + 1) * (i + 1) + j];
+
+ workspace->h[(RESTART + 1) * i + j] = tmp1;
+ workspace->h[(RESTART + 1) * (i + 1) + j] = tmp2;
+ }
+ }
+ else
+ {
+ //TODO: investigate correctness of not explicitly orthogonalizing first few vectors
+ /* Givens rotations on the upper-Hessenberg matrix to make it U */
+ for ( i = MAX(j - 1, 0); i <= j; i++ )
+ {
+ if ( i == j )
+ {
+ cc = SQRT( SQR(workspace->h[(RESTART + 1) * j + j])
+ + SQR(workspace->h[(RESTART + 1) * (j + 1) + j]) );
+ workspace->hc[j] = workspace->h[(RESTART + 1) * j + j] / cc;
+ workspace->hs[j] = workspace->h[(RESTART + 1) * (j + 1) + j] / cc;
+ }
+
+ tmp1 = workspace->hc[i] * workspace->h[(RESTART + 1) * i + j] +
+ workspace->hs[i] * workspace->h[(RESTART + 1) * (i + 1) + j];
+ tmp2 = -workspace->hs[i] * workspace->h[(RESTART + 1) * i + j] +
+ workspace->hc[i] * workspace->h[(RESTART + 1) * (i + 1) + j];
+
+ workspace->h[(RESTART + 1) * i + j] = tmp1;
+ workspace->h[(RESTART + 1) * (i + 1) + j] = tmp2;
+ }
+ }
+
+ /* apply Givens rotations to the rhs as well */
+ tmp1 = workspace->hc[j] * workspace->g[j];
+ tmp2 = -workspace->hs[j] * workspace->g[j];
+ workspace->g[j] = tmp1;
+ workspace->g[j + 1] = tmp2;
+ data->timing.solver_orthog += Get_Timing_Info( time_start );
+ }
- /* apply Givens rotations to the rhs as well */
- tmp1 = workspace->hc[j] * workspace->g[j];
- tmp2 = -workspace->hs[j] * workspace->g[j];
- workspace->g[j] = tmp1;
- workspace->g[j+1] = tmp2;
+ #pragma omp barrier
- // fprintf( stderr, "h: " );
- // for( i = 0; i <= j+1; ++i )
- // fprintf( stderr, "%.6f ", workspace->h[i][j] );
- // fprintf( stderr, "\n" );
- //fprintf( stderr, "res: %.15e\n", workspace->g[j+1] );
- }
+ //fprintf( stderr, "h: " );
+ //for( i = 0; i <= j+1; ++i )
+ //fprintf( stderr, "%.6f ", workspace->h[i][j] );
+ //fprintf( stderr, "\n" );
+ //fprintf( stderr, "res: %.15e\n", workspace->g[j+1] );
+ }
+ /* solve Hy = g: H is now upper-triangular, do back-substitution */
+ #pragma omp master
+ {
+ time_start = Get_Time( );
+ for ( i = j - 1; i >= 0; i-- )
+ {
+ temp = workspace->g[i];
+ for ( k = j - 1; k > i; k-- )
+ {
+ temp -= workspace->h[(RESTART + 1) * i + k] * workspace->y[k];
+ }
- /* solve Hy = g.
- H is now upper-triangular, do back-substitution */
- for( i = j-1; i >= 0; i-- ) {
- temp = workspace->g[i];
- for( k = j-1; k > i; k-- )
- temp -= workspace->h[ index_wkspace_res (i,k) ] * workspace->y[k];
+ workspace->y[i] = temp / workspace->h[(RESTART + 1) * i + i];
+ }
+ data->timing.solver_tri_solve += Get_Timing_Info( time_start );
- workspace->y[i] = temp / workspace->h[ index_wkspace_res (i,i) ];
- }
+ /* update x = x_0 + Vy */
+ time_start = Get_Time( );
+ }
+ Vector_MakeZero( workspace->p, N );
+ for ( i = 0; i < j; i++ )
+ {
+ Vector_Add( workspace->p, workspace->y[i], workspace->v + i * N, N );
+ }
+ Vector_Add( x, 1., workspace->p, N );
+ #pragma omp master
+ {
+ data->timing.solver_vector_ops += Get_Timing_Info( time_start );
+ }
- /* update x = x_0 + Vy */
- for( i = 0; i < j; i++ )
- Vector_Add( x, workspace->y[i], &workspace->v[index_wkspace_sys(i,0,system->N)], N );
+ /* stopping condition */
+ if ( FABS(workspace->g[j]) / bnorm <= tol )
+ {
+ break;
+ }
+ }
- /* stopping condition */
- if( fabs(workspace->g[j]) / bnorm <= tol )
- break;
+ #pragma omp master
+ {
+ g_itr = itr;
+ g_j = j;
+ }
}
// Sparse_MatVec( H, x, workspace->b_prm );
// for( i = 0; i < N; ++i )
- // workspace->b_prm[i] *= workspace->Hdia_inv[i];
+ // workspace->b_prm[i] *= workspace->Hdia_inv[i];
// fprintf( fout, "\n%10s%15s%15s\n", "b_prc", "b_prm", "x" );
// for( i = 0; i < N; ++i )
- // fprintf( fout, "%10.5f%15.12f%15.12f\n",
+ // fprintf( fout, "%10.5f%15.12f%15.12f\n",
// workspace->b_prc[i], workspace->b_prm[i], x[i] );*/
- // fprintf(fout,"GMRES outer:%d, inner:%d iters - residual norm: %25.20f\n",
+ // fprintf(fout,"GMRES outer:%d, inner:%d iters - residual norm: %25.20f\n",
// itr, j, fabs( workspace->g[j] ) / bnorm );
- // data->timing.matvec += itr * RESTART + j;
+ // data->timing.solver_iters += itr * RESTART + j;
- if( itr >= MAX_ITR ) {
+ if ( g_itr >= MAX_ITR )
+ {
fprintf( stderr, "GMRES convergence failed\n" );
// return -1;
- return itr * (RESTART+1) + j + 1;
+ return g_itr * (RESTART + 1) + g_j + 1;
}
- return itr * (RESTART+1) + j + 1;
+ return g_itr * (RESTART + 1) + g_j + 1;
}
-int GMRES_HouseHolder( static_storage *workspace, sparse_matrix *H,
- real *b, real tol, real *x, FILE *fout, reax_system *system)
+int GMRES_HouseHolder( const static_storage * const workspace, const control_params * const control,
+ simulation_data * const data, const sparse_matrix * const H,
+ const real * const b, real tol, real * const x,
+ const FILE * const fout, const int fresh_pre )
{
int i, j, k, itr, N;
real cc, tmp1, tmp2, temp, bnorm;
- real v[10000], z[RESTART+2][10000], w[RESTART+2];
- real u[RESTART+2][10000];
+ real v[10000], z[RESTART + 2][10000], w[RESTART + 2];
+ real u[RESTART + 2][10000];
N = H->n;
bnorm = Norm( b, N );
/* apply the diagonal pre-conditioner to rhs */
- for( i = 0; i < N; ++i )
- workspace->b_prc[i] = b[i] * workspace->Hdia_inv[i];
+ for ( i = 0; i < N; ++i )
+ {
+ workspace->b_prc[i] = b[i] * workspace->Hdia_inv[i];
+ }
// memset( x, 0, sizeof(real) * N );
/* GMRES outer-loop */
- for( itr = 0; itr < MAX_ITR; ++itr ) {
+ for ( itr = 0; itr < MAX_ITR; ++itr )
+ {
/* compute z = r0 */
- Sparse_MatVec( H, x, workspace->b_prm );
- for( i = 0; i < N; ++i )
+ Sparse_MatVec( H, x, workspace->b_prm );
+ for ( i = 0; i < N; ++i )
+ {
workspace->b_prm[i] *= workspace->Hdia_inv[i]; /* pre-conditioner */
+ }
Vector_Sum( z[0], 1., workspace->b_prc, -1., workspace->b_prm, N );
- Vector_MakeZero( w, RESTART+1 );
+ Vector_MakeZero( w, RESTART + 1 );
w[0] = Norm( z[0], N );
Vector_Copy( u[0], z[0], N );
u[0][0] += ( u[0][0] < 0.0 ? -1 : 1 ) * w[0];
Vector_Scale( u[0], 1 / Norm( u[0], N ), u[0], N );
- w[0] *= ( u[0][0] < 0.0 ? 1 :-1 );
+ w[0] *= ( u[0][0] < 0.0 ? 1 : -1 );
// fprintf( stderr, "\n\n%12.6f\n", w[0] );
/* GMRES inner-loop */
- for( j = 0; j < RESTART && fabs( w[j] ) / bnorm > tol; j++ ) {
+ for ( j = 0; j < RESTART && fabs( w[j] ) / bnorm > tol; j++ )
+ {
/* compute v_j */
Vector_Scale( z[j], -2 * u[j][j], u[j], N );
z[j][j] += 1.; /* due to e_j */
- for( i = j-1; i >= 0; --i )
- Vector_Add( z[j]+i, -2 * Dot( u[i]+i, z[j]+i, N-i ), u[i]+i, N-i );
-
+ for ( i = j - 1; i >= 0; --i )
+ {
+ Vector_Add( z[j] + i, -2 * Dot( u[i] + i, z[j] + i, N - i ), u[i] + i, N - i );
+ }
/* matvec */
Sparse_MatVec( H, z[j], v );
- for( k = 0; k < N; ++k )
+ for ( k = 0; k < N; ++k )
+ {
v[k] *= workspace->Hdia_inv[k]; /* pre-conditioner */
+ }
- for( i = 0; i <= j; ++i )
- Vector_Add( v+i, -2 * Dot( u[i]+i, v+i, N-i ), u[i]+i, N-i );
-
+ for ( i = 0; i <= j; ++i )
+ {
+ Vector_Add( v + i, -2 * Dot( u[i] + i, v + i, N - i ), u[i] + i, N - i );
+ }
- if( !Vector_isZero( v + (j+1), N - (j+1) ) ) {
+ if ( !Vector_isZero( v + (j + 1), N - (j + 1) ) )
+ {
/* compute the HouseHolder unit vector u_j+1 */
- for( i = 0; i <= j; ++i )
- u[j+1][i] = 0;
+ for ( i = 0; i <= j; ++i )
+ {
+ u[j + 1][i] = 0;
+ }
- Vector_Copy( u[j+1] + (j+1), v + (j+1), N - (j+1) );
+ Vector_Copy( u[j + 1] + (j + 1), v + (j + 1), N - (j + 1) );
- u[j+1][j+1] += ( v[j+1]<0.0 ? -1:1 ) * Norm( v+(j+1), N-(j+1) );
+ u[j + 1][j + 1] += ( v[j + 1] < 0.0 ? -1 : 1 ) * Norm( v + (j + 1), N - (j + 1) );
- Vector_Scale( u[j+1], 1 / Norm( u[j+1], N ), u[j+1], N );
+ Vector_Scale( u[j + 1], 1 / Norm( u[j + 1], N ), u[j + 1], N );
/* overwrite v with P_m+1 * v */
- v[j+1] -= 2 * Dot( u[j+1]+(j+1), v+(j+1), N-(j+1) ) * u[j+1][j+1];
- Vector_MakeZero( v + (j+2), N - (j+2) );
+ v[j + 1] -= 2 * Dot( u[j + 1] + (j + 1), v + (j + 1), N - (j + 1) ) * u[j + 1][j + 1];
+ Vector_MakeZero( v + (j + 2), N - (j + 2) );
// Vector_Add( v, -2 * Dot( u[j+1], v, N ), u[j+1], N );
}
/* prev Givens rots on the upper-Hessenberg matrix to make it U */
- for( i = 0; i < j; i++ ) {
- tmp1 = workspace->hc[i] * v[i] + workspace->hs[i] * v[i+1];
- tmp2 = -workspace->hs[i] * v[i] + workspace->hc[i] * v[i+1];
+ for ( i = 0; i < j; i++ )
+ {
+ tmp1 = workspace->hc[i] * v[i] + workspace->hs[i] * v[i + 1];
+ tmp2 = -workspace->hs[i] * v[i] + workspace->hc[i] * v[i + 1];
v[i] = tmp1;
- v[i+1] = tmp2;
+ v[i + 1] = tmp2;
}
/* apply the new Givens rotation to H and right-hand side */
- if( fabs(v[j+1]) >= ALMOST_ZERO ) {
- cc = SQRT( SQR( v[j] ) + SQR( v[j+1] ) );
+ if ( fabs(v[j + 1]) >= ALMOST_ZERO )
+ {
+ cc = SQRT( SQR( v[j] ) + SQR( v[j + 1] ) );
workspace->hc[j] = v[j] / cc;
- workspace->hs[j] = v[j+1] / cc;
+ workspace->hs[j] = v[j + 1] / cc;
- tmp1 = workspace->hc[j] * v[j] + workspace->hs[j] * v[j+1];
- tmp2 = -workspace->hs[j] * v[j] + workspace->hc[j] * v[j+1];
+ tmp1 = workspace->hc[j] * v[j] + workspace->hs[j] * v[j + 1];
+ tmp2 = -workspace->hs[j] * v[j] + workspace->hc[j] * v[j + 1];
v[j] = tmp1;
- v[j+1] = tmp2;
+ v[j + 1] = tmp2;
/* Givens rotations to rhs */
tmp1 = workspace->hc[j] * w[j];
tmp2 = -workspace->hs[j] * w[j];
w[j] = tmp1;
- w[j+1] = tmp2;
+ w[j + 1] = tmp2;
}
/* extend R */
- for( i = 0; i <= j; ++i )
- workspace->h[ index_wkspace_res (i,j) ] = v[i];
+ for ( i = 0; i <= j; ++i )
+ {
+ workspace->h[(RESTART + 1) * i + j] = v[i];
+ }
// fprintf( stderr, "h:" );
@@ -326,12 +1661,15 @@ int GMRES_HouseHolder( static_storage *workspace, sparse_matrix *H,
/* solve Hy = w.
H is now upper-triangular, do back-substitution */
- for( i = j-1; i >= 0; i-- ) {
- temp = w[i];
- for( k = j-1; k > i; k-- )
- temp -= workspace->h[ index_wkspace_res (i,k) ] * workspace->y[k];
+ for ( i = j - 1; i >= 0; i-- )
+ {
+ temp = w[i];
+ for ( k = j - 1; k > i; k-- )
+ {
+ temp -= workspace->h[(RESTART + 1) * i + k] * workspace->y[k];
+ }
- workspace->y[i] = temp / workspace->h[ index_wkspace_res (i,i) ];
+ workspace->y[i] = temp / workspace->h[(RESTART + 1) * i + i];
}
// fprintf( stderr, "y: " );
@@ -345,9 +1683,9 @@ int GMRES_HouseHolder( static_storage *workspace, sparse_matrix *H,
// {
// Vector_Copy( v, z, N );
// v[i] += workspace->y[i];
- //
+ //
// Vector_Sum( z, 1., v, -2 * Dot( u[i], v, N ), u[i], N );
- // }
+ // }
//
// fprintf( stderr, "\nz: " );
// for( k = 0; k < N; ++k )
@@ -358,16 +1696,20 @@ int GMRES_HouseHolder( static_storage *workspace, sparse_matrix *H,
// fprintf( stderr, "%6.2f ", x[i] );
// Vector_Add( x, 1, z, N );
- for( i = j-1; i >= 0; i-- )
+ for ( i = j - 1; i >= 0; i-- )
+ {
Vector_Add( x, workspace->y[i], z[i], N );
+ }
// fprintf( stderr, "\nx_aft: " );
// for( i = 0; i < N; ++i )
// fprintf( stderr, "%6.2f ", x[i] );
/* stopping condition */
- if( fabs( w[j] ) / bnorm <= tol )
+ if ( fabs( w[j] ) / bnorm <= tol )
+ {
break;
+ }
}
// Sparse_MatVec( H, x, workspace->b_prm );
@@ -376,152 +1718,26 @@ int GMRES_HouseHolder( static_storage *workspace, sparse_matrix *H,
// fprintf( fout, "\n%10s%15s%15s\n", "b_prc", "b_prm", "x" );
// for( i = 0; i < N; ++i )
- // fprintf( fout, "%10.5f%15.12f%15.12f\n",
+ // fprintf( fout, "%10.5f%15.12f%15.12f\n",
// workspace->b_prc[i], workspace->b_prm[i], x[i] );
- //fprintf( fout,"GMRES outer:%d, inner:%d iters - residual norm: %15.10f\n",
+ //fprintf( fout,"GMRES outer:%d, inner:%d iters - residual norm: %15.10f\n",
// itr, j, fabs( workspace->g[j] ) / bnorm );
- if( itr >= MAX_ITR ) {
- fprintf( stderr, "GMRES convergence failed\n" );
- // return -1;
- return itr * (RESTART+1) + j + 1;
- }
-
- return itr * (RESTART+1) + j + 1;
-}
-
-
-int PGMRES( static_storage *workspace, sparse_matrix *H, real *b, real tol,
- sparse_matrix *L, sparse_matrix *U, real *x, FILE *fout, reax_system *system )
-{
- int i, j, k, itr, N;
- real cc, tmp1, tmp2, temp, bnorm;
-
- N = H->n;
- bnorm = Norm( b, N );
-
- /* GMRES outer-loop */
- for( itr = 0; itr < MAX_ITR; ++itr )
+ if ( itr >= MAX_ITR )
{
- /* calculate r0 */
- Sparse_MatVec( H, x, workspace->b_prm );
- Vector_Sum( &workspace->v[index_wkspace_sys(0,0,system->N)], 1., b, -1., workspace->b_prm, N );
- Forward_Subs( L, &workspace->v[index_wkspace_sys(0,0,system->N)], &workspace->v[index_wkspace_sys(0,0,system->N)] );
- Backward_Subs( U, &workspace->v[index_wkspace_sys(0,0,system->N)], &workspace->v[index_wkspace_sys(0,0,system->N)] );
- workspace->g[0] = Norm( &workspace->v[index_wkspace_sys(0,0,system->N)], N );
- Vector_Scale( &workspace->v[index_wkspace_sys(0,0,system->N)], 1. / workspace->g[0], &workspace->v[index_wkspace_sys (0,0,system->N)], N );
- //fprintf( stderr, "res: %.15e\n", workspace->g[0] );
-
- /* GMRES inner-loop */
- for( j = 0; j < RESTART && fabs(workspace->g[j]) / bnorm > tol; j++ )
- {
- /* matvec */
- Sparse_MatVec( H, &workspace->v[index_wkspace_sys (j,0,system->N)], &workspace->v[index_wkspace_sys (j+1,0,system->N)] );
- Forward_Subs( L, &workspace->v[index_wkspace_sys(j+1,0,system->N)], &workspace->v[index_wkspace_sys(j+1,0,system->N)] );
- Backward_Subs( U, &workspace->v[index_wkspace_sys(j+1,0,system->N)], &workspace->v[index_wkspace_sys(j+1,0,system->N)] );
-
- /* apply modified Gram-Schmidt to orthogonalize the new residual */
- for( i = 0; i < j-1; i++ )
- {
- workspace->h[ index_wkspace_res (i,j)] = 0;
- }
-
- //for( i = 0; i <= j; i++ ) {
- for( i = MAX(j-1,0); i <= j; i++ ) {
- workspace->h[index_wkspace_res (i,j)] = Dot( &workspace->v[index_wkspace_sys (i,0,system->N)], &workspace->v[index_wkspace_sys(j+1,0,system->N)], N );
- Vector_Add( &workspace->v[index_wkspace_sys(j+1,0,system->N)],-workspace->h[ index_wkspace_res (i,j) ], &workspace->v[index_wkspace_sys(i,0,system->N)], N );
- }
-
- workspace->h[index_wkspace_res (j+1,j) ] = Norm( &workspace->v[index_wkspace_sys (j+1,0,system->N)], N );
- Vector_Scale( &workspace->v[index_wkspace_sys(j+1,0,system->N)],
- 1. / workspace->h[ index_wkspace_res (j+1,j)], &workspace->v[index_wkspace_sys(j+1,0,system->N)], N );
- // fprintf( stderr, "%d-%d: orthogonalization completed.\n", itr, j );
-
- /* Givens rotations on the upper-Hessenberg matrix to make it U */
- for( i = MAX(j-1,0); i <= j; i++ )
- {
- if( i == j )
- {
- cc = SQRT( SQR(workspace->h[ index_wkspace_res (j,j) ])+SQR(workspace->h[ index_wkspace_res (j+1,j) ]) );
- workspace->hc[j] = workspace->h[ index_wkspace_res (j,j) ] / cc;
- workspace->hs[j] = workspace->h[ index_wkspace_res (j+1,j) ] / cc;
- }
-
- tmp1 = workspace->hc[i] * workspace->h[ index_wkspace_res (i,j) ] +
- workspace->hs[i] * workspace->h[index_wkspace_res (i+1,j) ];
- tmp2 = -workspace->hs[i] * workspace->h[index_wkspace_res (i,j)] +
- workspace->hc[i] * workspace->h[index_wkspace_res (i+1,j) ];
-
- workspace->h[ index_wkspace_res (i,j) ] = tmp1;
- workspace->h[ index_wkspace_res (i+1,j) ] = tmp2;
- }
-
- /* apply Givens rotations to the rhs as well */
- tmp1 = workspace->hc[j] * workspace->g[j];
- tmp2 = -workspace->hs[j] * workspace->g[j];
- workspace->g[j] = tmp1;
- workspace->g[j+1] = tmp2;
-
- //fprintf( stderr, "h: " );
- //for( i = 0; i <= j+1; ++i )
- //fprintf( stderr, "%.6f ", workspace->h[i][j] );
- //fprintf( stderr, "\n" );
- //fprintf( stderr, "res: %.15e\n", workspace->g[j+1] );
- }
-
-
- /* solve Hy = g: H is now upper-triangular, do back-substitution */
- for( i = j-1; i >= 0; i-- )
- {
- temp = workspace->g[i];
- for( k = j-1; k > i; k-- )
- {
- temp -= workspace->h[ index_wkspace_res (i,k) ] * workspace->y[k];
- }
-
- workspace->y[i] = temp / workspace->h[index_wkspace_res (i,i)];
- }
-
- /* update x = x_0 + Vy */
- Vector_MakeZero( workspace->p, N );
- for( i = 0; i < j; i++ )
- Vector_Add( workspace->p, workspace->y[i], &workspace->v[index_wkspace_sys(i,0,system->N)], N );
- //Backward_Subs( U, workspace->p, workspace->p );
- //Forward_Subs( L, workspace->p, workspace->p );
- Vector_Add( x, 1., workspace->p, N );
-
- /* stopping condition */
- if( fabs(workspace->g[j]) / bnorm <= tol )
- {
- break;
- }
- }
-
- // Sparse_MatVec( H, x, workspace->b_prm );
- // for( i = 0; i < N; ++i )
- // workspace->b_prm[i] *= workspace->Hdia_inv[i];
- // fprintf( fout, "\n%10s%15s%15s\n", "b_prc", "b_prm", "x" );
- // for( i = 0; i < N; ++i )
- // fprintf( fout, "%10.5f%15.12f%15.12f\n",
- // workspace->b_prc[i], workspace->b_prm[i], x[i] );*/
-
- // fprintf(fout,"GMRES outer:%d, inner:%d iters - residual norm: %25.20f\n",
- // itr, j, fabs( workspace->g[j] ) / bnorm );
- // data->timing.matvec += itr * RESTART + j;
-
- if( itr >= MAX_ITR ) {
fprintf( stderr, "GMRES convergence failed\n" );
// return -1;
- return itr * (RESTART+1) + j + 1;
+ return itr * (RESTART + 1) + j + 1;
}
- return itr * (RESTART+1) + j + 1;
+ return itr * (RESTART + 1) + j + 1;
}
-int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
- sparse_matrix *L, sparse_matrix *U, real *x, FILE *fout, reax_system* system )
+/* Preconditioned Conjugate Gradient */
+int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
+ sparse_matrix *L, sparse_matrix *U, real *x, FILE *fout )
{
int i, N;
real tmp, alpha, beta, b_norm, r_norm;
@@ -537,12 +1753,12 @@ int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
//Print_Soln( workspace, x, q, b, N );
//fprintf( stderr, "res: %.15e\n", r_norm );
- Forward_Subs( L, workspace->r, workspace->d );
- Backward_Subs( U, workspace->d, workspace->p );
+ tri_solve( L, workspace->r, workspace->d, LOWER );
+ tri_solve( U, workspace->d, workspace->p, UPPER );
sig_new = Dot( workspace->r, workspace->p, N );
sig0 = sig_new;
- for( i = 0; i < 200 && r_norm/b_norm > tol; ++i )
+ for ( i = 0; i < 200 && r_norm / b_norm > tol; ++i )
{
//for( i = 0; i < 200 && sig_new > SQR(tol) * sig0; ++i ) {
Sparse_MatVec( A, workspace->p, workspace->q );
@@ -556,8 +1772,8 @@ int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
r_norm = Norm(workspace->r, N);
//fprintf( stderr, "res: %.15e\n", r_norm );
- Forward_Subs( L, workspace->r, workspace->d );
- Backward_Subs( U, workspace->d, workspace->d );
+ tri_solve( L, workspace->r, workspace->d, LOWER );
+ tri_solve( U, workspace->d, workspace->d, UPPER );
sig_old = sig_new;
sig_new = Dot( workspace->r, workspace->d, N );
beta = sig_new / sig_old;
@@ -565,7 +1781,8 @@ int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
}
//fprintf( fout, "CG took %d iterations\n", i );
- if( i >= 200 ) {
+ if ( i >= 200 )
+ {
fprintf( stderr, "CG convergence failed!\n" );
return i;
}
@@ -574,8 +1791,9 @@ int PCG( static_storage *workspace, sparse_matrix *A, real *b, real tol,
}
-int CG( static_storage *workspace, sparse_matrix *H,
- real *b, real tol, real *x, FILE *fout, reax_system *system)
+/* Conjugate Gradient */
+int CG( static_storage *workspace, sparse_matrix *H,
+ real *b, real tol, real *x, FILE *fout )
{
int i, j, N;
real tmp, alpha, beta, b_norm;
@@ -587,29 +1805,34 @@ int CG( static_storage *workspace, sparse_matrix *H,
Sparse_MatVec( H, x, workspace->q );
Vector_Sum( workspace->r , 1., b, -1., workspace->q, N );
- for( j = 0; j < N; ++j )
+ for ( j = 0; j < N; ++j )
+ {
workspace->d[j] = workspace->r[j] * workspace->Hdia_inv[j];
+ }
sig_new = Dot( workspace->r, workspace->d, N );
sig0 = sig_new;
//Print_Soln( workspace, x, q, b, N );
- //fprintf( stderr, "sig_new: %24.15e, d_norm:%24.15e, q_norm:%24.15e\n",
+ //fprintf( stderr, "sig_new: %24.15e, d_norm:%24.15e, q_norm:%24.15e\n",
// sqrt(sig_new), Norm(workspace->d,N), Norm(workspace->q,N) );
//fprintf( stderr, "sig_new: %f\n", sig_new );
- for( i = 0; i < 300 && SQRT(sig_new) / b_norm > tol; ++i ) {
+ for ( i = 0; i < 300 && SQRT(sig_new) / b_norm > tol; ++i )
+ {
//for( i = 0; i < 300 && sig_new > SQR(tol)*sig0; ++i ) {
Sparse_MatVec( H, workspace->d, workspace->q );
tmp = Dot( workspace->d, workspace->q, N );
//fprintf( stderr, "tmp: %f\n", tmp );
- alpha = sig_new / tmp;
+ alpha = sig_new / tmp;
Vector_Add( x, alpha, workspace->d, N );
//fprintf( stderr, "d_norm:%24.15e, q_norm:%24.15e, tmp:%24.15e\n",
// Norm(workspace->d,N), Norm(workspace->q,N), tmp );
- Vector_Add( workspace->r, -alpha, workspace->q, N );
- for( j = 0; j < N; ++j )
+ Vector_Add( workspace->r, -alpha, workspace->q, N );
+ for ( j = 0; j < N; ++j )
+ {
workspace->p[j] = workspace->r[j] * workspace->Hdia_inv[j];
+ }
sig_old = sig_new;
sig_new = Dot( workspace->r, workspace->p, N );
@@ -620,7 +1843,8 @@ int CG( static_storage *workspace, sparse_matrix *H,
fprintf( stderr, "CG took %d iterations\n", i );
- if( i >= 300 ) {
+ if ( i >= 300 )
+ {
fprintf( stderr, "CG convergence failed!\n" );
return i;
}
@@ -630,8 +1854,8 @@ int CG( static_storage *workspace, sparse_matrix *H,
/* Steepest Descent */
-int SDM( static_storage *workspace, sparse_matrix *H,
- real *b, real tol, real *x, FILE *fout )
+int SDM( static_storage *workspace, sparse_matrix *H,
+ real *b, real tol, real *x, FILE *fout )
{
int i, j, N;
real tmp, alpha, beta, b_norm;
@@ -643,23 +1867,28 @@ int SDM( static_storage *workspace, sparse_matrix *H,
Sparse_MatVec( H, x, workspace->q );
Vector_Sum( workspace->r , 1., b, -1., workspace->q, N );
- for( j = 0; j < N; ++j )
+ for ( j = 0; j < N; ++j )
+ {
workspace->d[j] = workspace->r[j] * workspace->Hdia_inv[j];
+ }
sig = Dot( workspace->r, workspace->d, N );
sig0 = sig;
- for( i = 0; i < 300 && SQRT(sig) / b_norm > tol; ++i ) {
+ for ( i = 0; i < 300 && SQRT(sig) / b_norm > tol; ++i )
+ {
Sparse_MatVec( H, workspace->d, workspace->q );
sig = Dot( workspace->r, workspace->d, N );
tmp = Dot( workspace->d, workspace->q, N );
- alpha = sig / tmp;
+ alpha = sig / tmp;
Vector_Add( x, alpha, workspace->d, N );
Vector_Add( workspace->r, -alpha, workspace->q, N );
- for( j = 0; j < N; ++j )
+ for ( j = 0; j < N; ++j )
+ {
workspace->d[j] = workspace->r[j] * workspace->Hdia_inv[j];
+ }
//fprintf( stderr, "d_norm:%24.15e, q_norm:%24.15e, tmp:%24.15e\n",
// Norm(workspace->d,N), Norm(workspace->q,N), tmp );
@@ -667,10 +1896,55 @@ int SDM( static_storage *workspace, sparse_matrix *H,
fprintf( stderr, "SDM took %d iterations\n", i );
- if( i >= 300 ) {
+ if ( i >= 300 )
+ {
fprintf( stderr, "SDM convergence failed!\n" );
return i;
}
return i;
}
+
+
+/* Estimate the stability of a 2-side preconditioning scheme
+ * using the factorization A \approx LU. Specifically, estimate the 1-norm of A^{-1}
+ * using the 1-norm of (LU)^{-1}e, with e = [1 1 ... 1]^T through 2 triangular solves:
+ * 1) Ly = e
+ * 2) Ux = y where y = Ux
+ * That is, we seek to solve e = LUx for unknown x
+ *
+ * Reference: Incomplete LU Preconditioning with the Multilevel Fast Multipole Algorithm
+ * for Electromagnetic Scattering, SIAM J. Sci. Computing, 2007 */
+real condest( const sparse_matrix * const L, const sparse_matrix * const U )
+{
+ unsigned int i, N;
+ real *e, c;
+
+ N = L->n;
+
+ if ( (e = (real*) malloc(sizeof(real) * N)) == NULL )
+ {
+ fprintf( stderr, "Not enough memory for condest. Terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ memset( e, 1., N * sizeof(real) );
+
+ tri_solve( L, e, e, LOWER );
+ tri_solve( U, e, e, UPPER );
+
+ /* compute 1-norm of vector e */
+ c = FABS(e[0]);
+ for ( i = 1; i < N; ++i)
+ {
+ if ( FABS(e[i]) > c )
+ {
+ c = FABS(e[i]);
+ }
+
+ }
+
+ free( e );
+
+ return c;
+}
diff --git a/PuReMD-GPU/src/lin_alg.h b/PuReMD-GPU/src/lin_alg.h
index a515a959494a6eca40fe9f338d2a08118ff3e39a..317afbf94cf2a26a4f48be4a0ad9c66bcef42085 100644
--- a/PuReMD-GPU/src/lin_alg.h
+++ b/PuReMD-GPU/src/lin_alg.h
@@ -21,28 +21,31 @@
#ifndef __LIN_ALG_H_
#define __LIN_ALG_H_
-#define SIGN(x) (x < 0.0 ? -1 : 1);
-
#include "mytypes.h"
-int GMRES( static_storage*, sparse_matrix*,
- real*, real, real*, FILE* , reax_system* );
+void Transpose( const sparse_matrix const *, sparse_matrix const * );
+void Transpose_I( sparse_matrix * const );
-int GMRES_HouseHolder( static_storage*, sparse_matrix*,
- real*, real, real*, FILE* , reax_system* );
+sparse_matrix * setup_graph_coloring( sparse_matrix * const );
-int PGMRES( static_storage*, sparse_matrix*, real*, real,
- sparse_matrix*, sparse_matrix*, real*, FILE*, reax_system* );
+int GMRES( const static_storage * const, const control_params * const,
+ simulation_data * const, const sparse_matrix * const,
+ const real * const, const real, real * const,
+ const FILE * const, const int );
-int PCG( static_storage*, sparse_matrix*, real*, real,
- sparse_matrix*, sparse_matrix*, real*, FILE*, reax_system* );
+int GMRES_HouseHolder( const static_storage * const, const control_params * const,
+ simulation_data * const, const sparse_matrix * const,
+ const real * const, const real, real * const,
+ const FILE * const, const int );
int CG( static_storage*, sparse_matrix*,
- real*, real, real*, FILE*, reax_system* );
+ real*, real, real*, FILE* );
+
+int SDM( static_storage*, sparse_matrix*,
+ real*, real, real*, FILE* );
-int uyduruk_GMRES( static_storage*, sparse_matrix*,
- real*, real, real*, int, FILE*, reax_system* );
+real condest( const sparse_matrix * const, const sparse_matrix * const );
#endif
diff --git a/PuReMD-GPU/src/list.c b/PuReMD-GPU/src/list.c
index c6f0e55ebad4fc59c07f253a1d216d3242115aff..c52a4cc1cf2b2a8c1d32fdda71c8b0aa7808992a 100644
--- a/PuReMD-GPU/src/list.c
+++ b/PuReMD-GPU/src/list.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -21,9 +22,9 @@
#include "list.h"
-char Make_List(int n, int num_intrs, int type, list* l)
+int Make_List( int n, int num_intrs, int type, list* l )
{
- char success=1;
+ int ret = SUCCESS;
l->n = n;
l->num_intrs = num_intrs;
@@ -31,116 +32,170 @@ char Make_List(int n, int num_intrs, int type, list* l)
l->index = (int*) malloc( n * sizeof(int) );
l->end_index = (int*) malloc( n * sizeof(int) );
- if (l->index == NULL) success = 0;
- if (l->end_index == NULL) success = 0;
+ if (l->index == NULL)
+ {
+ ret = FAILURE;
+ }
+ if (l->end_index == NULL)
+ {
+ ret = FAILURE;
+ }
l->type = type;
- switch(type)
+ switch (type)
{
- case TYP_VOID:
- l->select.v = (void *) malloc(l->num_intrs*sizeof(void));
- if (l->select.v == NULL) success = 0;
- break;
-
- case TYP_THREE_BODY:
- l->select.three_body_list = (three_body_interaction_data*)
- malloc(l->num_intrs*sizeof(three_body_interaction_data));
- if (l->select.three_body_list == NULL) success = 0;
- break;
-
- case TYP_BOND:
- l->select.bond_list = (bond_data*)
+ case TYP_VOID:
+ l->select.v = (void *) malloc(l->num_intrs * sizeof(void));
+ if (l->select.v == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_THREE_BODY:
+ l->select.three_body_list = (three_body_interaction_data*)
+ malloc(l->num_intrs * sizeof(three_body_interaction_data));
+ if (l->select.three_body_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_BOND:
+ l->select.bond_list = (bond_data*)
malloc(l->num_intrs * sizeof(bond_data));
- if (l->select.bond_list == NULL) success = 0;
- break;
-
- case TYP_DBO:
- l->select.dbo_list = (dbond_data*)
+ if (l->select.bond_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_DBO:
+ l->select.dbo_list = (dbond_data*)
malloc(l->num_intrs * sizeof(dbond_data));
- if (l->select.dbo_list == NULL) success = 0;
- break;
-
- case TYP_DDELTA:
- l->select.dDelta_list = (dDelta_data*)
- malloc(l->num_intrs*sizeof(dDelta_data));
- if (l->select.dDelta_list == NULL) success = 0;
- break;
-
- case TYP_FAR_NEIGHBOR:
- l->select.far_nbr_list = (far_neighbor_data*)
- malloc(l->num_intrs*sizeof(far_neighbor_data));
- if (l->select.far_nbr_list == NULL) success = 0;
- break;
-
- case TYP_NEAR_NEIGHBOR:
- l->select.near_nbr_list = (near_neighbor_data*)
- malloc(l->num_intrs*sizeof(near_neighbor_data));
- if (l->select.near_nbr_list == NULL) success = 0;
- break;
-
- case TYP_HBOND:
- l->select.hbond_list = (hbond_data*)
+ if (l->select.dbo_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_DDELTA:
+ l->select.dDelta_list = (dDelta_data*)
+ malloc(l->num_intrs * sizeof(dDelta_data));
+ if (l->select.dDelta_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_FAR_NEIGHBOR:
+ l->select.far_nbr_list = (far_neighbor_data*)
+ malloc(l->num_intrs * sizeof(far_neighbor_data));
+ if (l->select.far_nbr_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_NEAR_NEIGHBOR:
+ l->select.near_nbr_list = (near_neighbor_data*)
+ malloc(l->num_intrs * sizeof(near_neighbor_data));
+ if (l->select.near_nbr_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ case TYP_HBOND:
+ l->select.hbond_list = (hbond_data*)
malloc( l->num_intrs * sizeof(hbond_data) );
- if (l->select.hbond_list == NULL) success = 0;
- break;
-
- default:
- l->select.v = (void *) malloc(l->num_intrs*sizeof(void));
- if (l->select.v == NULL) success = 0;
- l->type = TYP_VOID;
- break;
+ if (l->select.hbond_list == NULL)
+ {
+ ret = FAILURE;
+ }
+ break;
+
+ default:
+ l->select.v = (void *) malloc(l->num_intrs * sizeof(void));
+ if (l->select.v == NULL)
+ {
+ ret = FAILURE;
+ }
+ l->type = TYP_VOID;
+ break;
}
- return success;
+ return ret;
}
-void Delete_List(list* l)
+void Delete_List( list* l )
{
- if( l->index != NULL )
+ if ( l->index != NULL )
+ {
free(l->index);
- if( l->end_index != NULL )
+ }
+ if ( l->end_index != NULL )
+ {
free(l->end_index);
+ }
- switch(l->type)
+ switch (l->type)
{
- case TYP_VOID:
- if( l->select.v != NULL )
- free(l->select.v);
- break;
- case TYP_THREE_BODY:
- if( l->select.three_body_list != NULL )
- free(l->select.three_body_list);
- break;
- case TYP_BOND:
- if( l->select.bond_list != NULL )
- free(l->select.bond_list);
- break;
- case TYP_DBO:
- if( l->select.dbo_list != NULL )
- free(l->select.dbo_list);
- break;
- case TYP_DDELTA:
- if( l->select.dDelta_list != NULL )
- free(l->select.dDelta_list);
- break;
- case TYP_FAR_NEIGHBOR:
- if( l->select.far_nbr_list != NULL )
- free(l->select.far_nbr_list);
- break;
- case TYP_NEAR_NEIGHBOR:
- if( l->select.near_nbr_list != NULL )
- free(l->select.near_nbr_list);
- break;
- case TYP_HBOND:
- if( l->select.hbond_list != NULL )
- free(l->select.hbond_list);
- break;
-
- default:
- // Report fatal error
- break;
+ case TYP_VOID:
+ if ( l->select.v != NULL )
+ {
+ free(l->select.v);
+ }
+ break;
+ case TYP_THREE_BODY:
+ if ( l->select.three_body_list != NULL )
+ {
+ free(l->select.three_body_list);
+ }
+ break;
+ case TYP_BOND:
+ if ( l->select.bond_list != NULL )
+ {
+ free(l->select.bond_list);
+ }
+ break;
+ case TYP_DBO:
+ if ( l->select.dbo_list != NULL )
+ {
+ free(l->select.dbo_list);
+ }
+ break;
+ case TYP_DDELTA:
+ if ( l->select.dDelta_list != NULL )
+ {
+ free(l->select.dDelta_list);
+ }
+ break;
+ case TYP_FAR_NEIGHBOR:
+ if ( l->select.far_nbr_list != NULL )
+ {
+ free(l->select.far_nbr_list);
+ }
+ break;
+ case TYP_NEAR_NEIGHBOR:
+ if ( l->select.near_nbr_list != NULL )
+ {
+ free(l->select.near_nbr_list);
+ }
+ break;
+ case TYP_HBOND:
+ if ( l->select.hbond_list != NULL )
+ {
+ free(l->select.hbond_list);
+ }
+ break;
+
+ default:
+ fprintf( stderr, "Unrecognized list type. Terminating...\n" );
+ exit( UNKNOWN_OPTION );
+ break;
}
-}
+}
diff --git a/PuReMD-GPU/src/list.h b/PuReMD-GPU/src/list.h
index b90c41419271ca6b859be08ea4005fbe9107c029..5ee4544212218488e6fa84477f8a446f66e73544 100644
--- a/PuReMD-GPU/src/list.h
+++ b/PuReMD-GPU/src/list.h
@@ -24,7 +24,7 @@
#include "mytypes.h"
-char Make_List( int, int, int, list* );
+int Make_List( int, int, int, list* );
void Delete_List( list* );
diff --git a/PuReMD-GPU/src/lookup.c b/PuReMD-GPU/src/lookup.c
index c439709dc09c77775ed716a39db797fa8c831585..b67bf5b7b96e91562a34ab2af3bbe421a1f5c19c 100644
--- a/PuReMD-GPU/src/lookup.c
+++ b/PuReMD-GPU/src/lookup.c
@@ -1,28 +1,28 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "lookup.h"
-#include "two_body_interactions.h"
-
#include "index_utils.h"
+#include "two_body_interactions.h"
void Make_Lookup_Table(real xmin, real xmax, int n,
@@ -33,44 +33,48 @@ void Make_Lookup_Table(real xmin, real xmax, int n,
t->xmin = xmin;
t->xmax = xmax;
t->n = n;
- t->dx = (xmax - xmin)/(n-1);
+ t->dx = (xmax - xmin) / (n - 1);
t->inv_dx = 1.0 / t->dx;
- t->a = (n-1)/(xmax-xmin);
- t->y = (real*) malloc(n*sizeof(real));
+ t->a = (n - 1) / (xmax - xmin);
+ t->y = (real*) malloc(n * sizeof(real));
- for(i=0; i < n; i++)
- t->y[i] = f(i*t->dx + t->xmin);
+ for (i = 0; i < n; i++)
+ t->y[i] = f(i * t->dx + t->xmin);
- // //fprintf(stdout,"dx = %lf\n",t->dx);
+ // fprintf(stdout,"dx = %lf\n",t->dx);
// for(i=0; i < n; i++)
- // //fprintf( stdout,"%d %lf %lf %lf\n",
+ // fprintf( stdout,"%d %lf %lf %lf\n",
// i, i/t->a+t->xmin, t->y[i], exp(i/t->a+t->xmin) );
}
/* Fills solution into x. Warning: will modify c and d! */
void Tridiagonal_Solve( const real *a, const real *b,
- real *c, real *d, real *x, unsigned int n){
+ real *c, real *d, real *x, unsigned int n)
+{
int i;
real id;
/* Modify the coefficients. */
- c[0] /= b[0]; /* Division by zero risk. */
- d[0] /= b[0]; /* Division by zero would imply a singular matrix. */
- for(i = 1; i < n; i++){
- id = (b[i] - c[i-1] * a[i]); /* Division by zero risk. */
- c[i] /= id; /* Last value calculated is redundant. */
- d[i] = (d[i] - d[i-1] * a[i])/id;
+ c[0] /= b[0]; /* Division by zero risk. */
+ d[0] /= b[0]; /* Division by zero would imply a singular matrix. */
+ for (i = 1; i < n; i++)
+ {
+ id = (b[i] - c[i - 1] * a[i]); /* Division by zero risk. */
+ c[i] /= id; /* Last value calculated is redundant. */
+ d[i] = (d[i] - d[i - 1] * a[i]) / id;
}
/* Now back substitute. */
x[n - 1] = d[n - 1];
- for(i = n - 2; i >= 0; i--)
+ for (i = n - 2; i >= 0; i--)
+ {
x[i] = d[i] - c[i] * x[i + 1];
+ }
}
-void Natural_Cubic_Spline( const real *h, const real *f,
+void Natural_Cubic_Spline( const real *h, const real *f,
cubic_spline_coef *coef, unsigned int n )
{
int i;
@@ -84,43 +88,53 @@ void Natural_Cubic_Spline( const real *h, const real *f,
v = (real*) malloc( n * sizeof(real) );
/* build the linear system */
- a[0] = a[1] = a[n-1] = 0;
- for( i = 2; i < n-1; ++i )
- a[i] = h[i-1];
+ a[0] = a[1] = a[n - 1] = 0;
+ for ( i = 2; i < n - 1; ++i )
+ {
+ a[i] = h[i - 1];
+ }
- b[0] = b[n-1] = 0;
- for( i = 1; i < n-1; ++i )
- b[i] = 2 * (h[i-1] + h[i]);
+ b[0] = b[n - 1] = 0;
+ for ( i = 1; i < n - 1; ++i )
+ {
+ b[i] = 2 * (h[i - 1] + h[i]);
+ }
- c[0] = c[n-2] = c[n-1] = 0;
- for( i = 1; i < n-2; ++i )
+ c[0] = c[n - 2] = c[n - 1] = 0;
+ for ( i = 1; i < n - 2; ++i )
+ {
c[i] = h[i];
+ }
- d[0] = d[n-1] = 0;
- for( i = 1; i < n-1; ++i )
- d[i] = 6 * ((f[i+1]-f[i])/h[i] - (f[i]-f[i-1])/h[i-1]);
+ d[0] = d[n - 1] = 0;
+ for ( i = 1; i < n - 1; ++i )
+ {
+ d[i] = 6 * ((f[i + 1] - f[i]) / h[i] - (f[i] - f[i - 1]) / h[i - 1]);
+ }
- /*//fprintf( stderr, "i a b c d\n" );
+ /*fprintf( stderr, "i a b c d\n" );
for( i = 0; i < n; ++i )
- //fprintf( stderr, "%d %f %f %f %f\n", i, a[i], b[i], c[i], d[i] );*/
+ fprintf( stderr, "%d %f %f %f %f\n", i, a[i], b[i], c[i], d[i] );*/
v[0] = 0;
- v[n-1] = 0;
- Tridiagonal_Solve( &(a[1]), &(b[1]), &(c[1]), &(d[1]), &(v[1]), n-2 );
-
- for( i = 1; i < n; ++i ){
- coef[i-1].d = (v[i] - v[i-1]) / (6*h[i-1]);
- coef[i-1].c = v[i]/2;
- coef[i-1].b = (f[i]-f[i-1])/h[i-1] + h[i-1]*(2*v[i] + v[i-1])/6;
- coef[i-1].a = f[i];
+ v[n - 1] = 0;
+ Tridiagonal_Solve( &(a[1]), &(b[1]), &(c[1]), &(d[1]), &(v[1]), n - 2 );
+
+ for ( i = 1; i < n; ++i )
+ {
+ coef[i - 1].d = (v[i] - v[i - 1]) / (6 * h[i - 1]);
+ coef[i - 1].c = v[i] / 2;
+ coef[i - 1].b = (f[i] - f[i - 1]) / h[i - 1] + h[i - 1] * (2 * v[i] + v[i - 1]) / 6;
+ coef[i - 1].a = f[i];
}
- /*//fprintf( stderr, "i v coef\n" );
+ /*fprintf( stderr, "i v coef\n" );
for( i = 0; i < n; ++i )
- //fprintf( stderr, "%d %f %f %f %f %f\n",
- i, v[i], coef[i].a, coef[i].b, coef[i].c, coef[i].d ); */
+ fprintf( stderr, "%d %f %f %f %f %f\n",
+ i, v[i], coef[i].a, coef[i].b, coef[i].c, coef[i].d ); */
}
+
void Complete_Cubic_Spline( const real *h, const real *f, real v0, real vlast,
cubic_spline_coef *coef, unsigned int n )
{
@@ -136,39 +150,48 @@ void Complete_Cubic_Spline( const real *h, const real *f, real v0, real vlast,
/* build the linear system */
a[0] = 0;
- for( i = 1; i < n; ++i )
- a[i] = h[i-1];
+ for ( i = 1; i < n; ++i )
+ {
+ a[i] = h[i - 1];
+ }
- b[0] = 2*h[0];
- for( i = 1; i < n; ++i )
- b[i] = 2 * (h[i-1] + h[i]);
+ b[0] = 2 * h[0];
+ for ( i = 1; i < n; ++i )
+ {
+ b[i] = 2 * (h[i - 1] + h[i]);
+ }
- c[n-1] = 0;
- for( i = 0; i < n-1; ++i )
+ c[n - 1] = 0;
+ for ( i = 0; i < n - 1; ++i )
+ {
c[i] = h[i];
+ }
- d[0] = 6 * (f[1]-f[0])/h[0] - 6 * v0;
- d[n-1] = 6 * vlast - 6 * (f[n-1]-f[n-2]/h[n-2]);
- for( i = 1; i < n-1; ++i )
- d[i] = 6 * ((f[i+1]-f[i])/h[i] - (f[i]-f[i-1])/h[i-1]);
+ d[0] = 6 * (f[1] - f[0]) / h[0] - 6 * v0;
+ d[n - 1] = 6 * vlast - 6 * (f[n - 1] - f[n - 2] / h[n - 2]);
+ for ( i = 1; i < n - 1; ++i )
+ {
+ d[i] = 6 * ((f[i + 1] - f[i]) / h[i] - (f[i] - f[i - 1]) / h[i - 1]);
+ }
- /*//fprintf( stderr, "i a b c d\n" );
+ /*fprintf( stderr, "i a b c d\n" );
for( i = 0; i < n; ++i )
- //fprintf( stderr, "%d %f %f %f %f\n", i, a[i], b[i], c[i], d[i] );*/
+ fprintf( stderr, "%d %f %f %f %f\n", i, a[i], b[i], c[i], d[i] );*/
Tridiagonal_Solve( &(a[0]), &(b[0]), &(c[0]), &(d[0]), &(v[0]), n );
// Tridiagonal_Solve( &(a[1]), &(b[1]), &(c[1]), &(d[1]), &(v[1]), n-2 );
- for( i = 1; i < n; ++i ){
- coef[i-1].d = (v[i] - v[i-1]) / (6*h[i-1]);
- coef[i-1].c = v[i]/2;
- coef[i-1].b = (f[i]-f[i-1])/h[i-1] + h[i-1]*(2*v[i] + v[i-1])/6;
- coef[i-1].a = f[i];
+ for ( i = 1; i < n; ++i )
+ {
+ coef[i - 1].d = (v[i] - v[i - 1]) / (6 * h[i - 1]);
+ coef[i - 1].c = v[i] / 2;
+ coef[i - 1].b = (f[i] - f[i - 1]) / h[i - 1] + h[i - 1] * (2 * v[i] + v[i - 1]) / 6;
+ coef[i - 1].a = f[i];
}
- /*//fprintf( stderr, "i v coef\n" );
+ /*fprintf( stderr, "i v coef\n" );
for( i = 0; i < n; ++i )
- //fprintf( stderr, "%d %f %f %f %f %f\n",
- i, v[i], coef[i].a, coef[i].b, coef[i].c, coef[i].d ); */
+ fprintf( stderr, "%d %f %f %f %f %f\n",
+ i, v[i], coef[i].a, coef[i].b, coef[i].c, coef[i].d ); */
}
@@ -178,21 +201,24 @@ void LR_Lookup( LR_lookup_table *t, real r, LR_data *y )
real base, dif;
i = (int)(r * t->inv_dx);
- if( i == 0 ) ++i;
- base = (real)(i+1) * t->dx;
+ if ( i == 0 )
+ {
+ ++i;
+ }
+ base = (real)(i + 1) * t->dx;
dif = r - base;
- ////fprintf( stderr, "r: %f, i: %d, base: %f, dif: %f\n", r, i, base, dif );
+ //fprintf( stderr, "r: %f, i: %d, base: %f, dif: %f\n", r, i, base, dif );
- y->e_vdW = ((t->vdW[i].d*dif + t->vdW[i].c)*dif + t->vdW[i].b)*dif +
- t->vdW[i].a;
- y->CEvd = ((t->CEvd[i].d*dif + t->CEvd[i].c)*dif +
- t->CEvd[i].b)*dif + t->CEvd[i].a;
+ y->e_vdW = ((t->vdW[i].d * dif + t->vdW[i].c) * dif + t->vdW[i].b) * dif +
+ t->vdW[i].a;
+ y->CEvd = ((t->CEvd[i].d * dif + t->CEvd[i].c) * dif +
+ t->CEvd[i].b) * dif + t->CEvd[i].a;
//y->CEvd = (3*t->vdW[i].d*dif + 2*t->vdW[i].c)*dif + t->vdW[i].b;
- y->e_ele = ((t->ele[i].d*dif + t->ele[i].c)*dif + t->ele[i].b)*dif +
- t->ele[i].a;
- y->CEclmb = ((t->CEclmb[i].d*dif + t->CEclmb[i].c)*dif + t->CEclmb[i].b)*dif +
- t->CEclmb[i].a;
+ y->e_ele = ((t->ele[i].d * dif + t->ele[i].c) * dif + t->ele[i].b) * dif +
+ t->ele[i].a;
+ y->CEclmb = ((t->CEclmb[i].d * dif + t->CEclmb[i].c) * dif + t->CEclmb[i].b) * dif +
+ t->CEclmb[i].a;
y->H = y->e_ele * EV_to_KCALpMOL / C_ele;
//y->H = ((t->H[i].d*dif + t->H[i].c)*dif + t->H[i].b)*dif + t->H[i].a;
@@ -221,147 +247,158 @@ void Make_LR_Lookup_Table( reax_system *system, control_params *control )
num_atom_types = system->reaxprm.num_atom_types;
dr = control->r_cut / control->tabulate;
- h = (real*) malloc( (control->tabulate+1) * sizeof(real) );
- fh = (real*) malloc( (control->tabulate+1) * sizeof(real) );
- fvdw = (real*) malloc( (control->tabulate+1) * sizeof(real) );
- fCEvd = (real*) malloc( (control->tabulate+1) * sizeof(real) );
- fele = (real*) malloc( (control->tabulate+1) * sizeof(real) );
- fCEclmb = (real*) malloc( (control->tabulate+1) * sizeof(real) );
-
- /* allocate Long-Range LookUp Table space based on
+ h = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+ fh = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+ fvdw = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+ fCEvd = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+ fele = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+ fCEclmb = (real*) malloc( (control->tabulate + 1) * sizeof(real) );
+
+ /* allocate Long-Range LookUp Table space based on
number of atom types in the ffield file */
- //LR = (LR_lookup_table**) malloc( num_atom_types * sizeof(LR_lookup_table*) );
- //for( i = 0; i < num_atom_types; ++i )
- // LR[i] = (LR_lookup_table*) malloc(num_atom_types * sizeof(LR_lookup_table));
-
- LR = (LR_lookup_table*) malloc(num_atom_types * num_atom_types * sizeof(LR_lookup_table));
+ LR = (LR_lookup_table*) malloc( num_atom_types * num_atom_types * sizeof(LR_lookup_table) );
/* most atom types in ffield file will not exist in the current
simulation. to avoid unnecessary lookup table space, determine
the atom types that exist in the current simulation */
- for( i = 0; i < MAX_ATOM_TYPES; ++i )
+ for ( i = 0; i < MAX_ATOM_TYPES; ++i )
+ {
existing_types[i] = 0;
- for( i = 0; i < system->N; ++i )
+ }
+ for ( i = 0; i < system->N; ++i )
+ {
existing_types[ system->atoms[i].type ] = 1;
+ }
/* fill in the lookup table entries for existing atom types.
only lower half should be enough. */
- for( i = 0; i < num_atom_types; ++i )
- if( existing_types[i] )
- for( j = i; j < num_atom_types; ++j )
- if( existing_types[j] ) {
- LR[ index_lr (i,j,num_atom_types) ].xmin = 0;
- LR[ index_lr (i,j,num_atom_types) ].xmax = control->r_cut;
- LR[ index_lr (i,j,num_atom_types) ].n = control->tabulate + 1;
- LR[ index_lr (i,j,num_atom_types) ].dx = dr;
- LR[ index_lr (i,j,num_atom_types) ].inv_dx = control->tabulate / control->r_cut;
- LR[ index_lr (i,j,num_atom_types) ].y = (LR_data*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(LR_data));
- LR[ index_lr (i,j,num_atom_types) ].H = (cubic_spline_coef*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(cubic_spline_coef));
- LR[ index_lr (i,j,num_atom_types) ].vdW = (cubic_spline_coef*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(cubic_spline_coef));
- LR[ index_lr (i,j,num_atom_types) ].CEvd = (cubic_spline_coef*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(cubic_spline_coef));
- LR[ index_lr (i,j,num_atom_types) ].ele = (cubic_spline_coef*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(cubic_spline_coef));
- LR[ index_lr (i,j,num_atom_types) ].CEclmb = (cubic_spline_coef*)
- malloc(LR[ index_lr (i,j,num_atom_types) ].n * sizeof(cubic_spline_coef));
-
- for( r = 1; r <= control->tabulate; ++r ) {
- LR_vdW_Coulomb( system, control, i, j, r * dr, &(LR[ index_lr (i,j,num_atom_types) ].y[r]) );
- h[r] = LR[ index_lr (i,j,num_atom_types) ].dx;
- fh[r] = LR[ index_lr (i,j,num_atom_types) ].y[r].H;
- fvdw[r] = LR[ index_lr (i,j,num_atom_types) ].y[r].e_vdW;
- fCEvd[r] = LR[ index_lr (i,j,num_atom_types) ].y[r].CEvd;
- fele[r] = LR[ index_lr (i,j,num_atom_types) ].y[r].e_ele;
- fCEclmb[r] = LR[ index_lr (i,j,num_atom_types) ].y[r].CEclmb;
-
- if( r == 1 ){
- v0_vdw = LR[ index_lr (i,j,num_atom_types) ].y[r].CEvd;
- v0_ele = LR[ index_lr (i,j,num_atom_types) ].y[r].CEclmb;
+ for ( i = 0; i < num_atom_types; ++i )
+ {
+ if ( existing_types[i] )
+ {
+ for ( j = i; j < num_atom_types; ++j )
+ {
+ if ( existing_types[j] )
+ {
+ LR[ index_lr(i,j,num_atom_types) ].xmin = 0;
+ LR[ index_lr(i,j,num_atom_types) ].xmax = control->r_cut;
+ LR[ index_lr(i,j,num_atom_types) ].n = control->tabulate + 1;
+ LR[ index_lr(i,j,num_atom_types) ].dx = dr;
+ LR[ index_lr(i,j,num_atom_types) ].inv_dx = control->tabulate / control->r_cut;
+ LR[ index_lr(i,j,num_atom_types) ].y = (LR_data*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(LR_data) );
+ LR[ index_lr(i,j,num_atom_types) ].H = (cubic_spline_coef*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(cubic_spline_coef) );
+ LR[ index_lr(i,j,num_atom_types) ].vdW = (cubic_spline_coef*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(cubic_spline_coef) );
+ LR[ index_lr(i,j,num_atom_types) ].CEvd = (cubic_spline_coef*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(cubic_spline_coef) );
+ LR[ index_lr(i,j,num_atom_types) ].ele = (cubic_spline_coef*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(cubic_spline_coef) );
+ LR[ index_lr(i,j,num_atom_types) ].CEclmb = (cubic_spline_coef*)
+ malloc( LR[index_lr(i,j,num_atom_types)].n * sizeof(cubic_spline_coef) );
+
+ for ( r = 1; r <= control->tabulate; ++r )
+ {
+ LR_vdW_Coulomb( system, control, i, j, r * dr,
+ &(LR[ index_lr(i,j,num_atom_types) ].y[r]) );
+ h[r] = LR[ index_lr(i,j,num_atom_types) ].dx;
+ fh[r] = LR[ index_lr(i,j,num_atom_types) ].y[r].H;
+ fvdw[r] = LR[ index_lr(i,j,num_atom_types) ].y[r].e_vdW;
+ fCEvd[r] = LR[ index_lr(i,j,num_atom_types) ].y[r].CEvd;
+ fele[r] = LR[ index_lr(i,j,num_atom_types) ].y[r].e_ele;
+ fCEclmb[r] = LR[ index_lr(i,j,num_atom_types) ].y[r].CEclmb;
+
+ if ( r == 1 )
+ {
+ v0_vdw = LR[ index_lr(i,j,num_atom_types) ].y[r].CEvd;
+ v0_ele = LR[ index_lr(i,j,num_atom_types) ].y[r].CEclmb;
}
- else if( r == control->tabulate ){
- vlast_vdw = LR[ index_lr (i,j,num_atom_types) ].y[r].CEvd;
- vlast_ele = LR[ index_lr (i,j,num_atom_types) ].y[r].CEclmb;
+ else if ( r == control->tabulate )
+ {
+ vlast_vdw = LR[ index_lr(i,j,num_atom_types) ].y[r].CEvd;
+ vlast_ele = LR[ index_lr(i,j,num_atom_types) ].y[r].CEclmb;
}
}
- /*//fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fh" );
+ /*fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fh" );
for( r = 1; r <= control->tabulate; ++r )
- //fprintf( stderr, "%f %f %f\n", r * dr, h[r], fh[r] ); */
- Natural_Cubic_Spline( &h[1], &fh[1],
- &(LR[ index_lr (i,j,num_atom_types) ].H[1]), control->tabulate+1 );
+ fprintf( stderr, "%f %f %f\n", r * dr, h[r], fh[r] ); */
+ Natural_Cubic_Spline( &h[1], &fh[1],
+ &(LR[ index_lr(i,j,num_atom_types) ].H[1]), control->tabulate + 1 );
- /*//fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fvdw" );
+ /*fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fvdw" );
for( r = 1; r <= control->tabulate; ++r )
- //fprintf( stderr, "%f %f %f\n", r * dr, h[r], fvdw[r] );
- //fprintf( stderr, "v0_vdw: %f, vlast_vdw: %f\n", v0_vdw, vlast_vdw );
- */
- Complete_Cubic_Spline( &h[1], &fvdw[1], v0_vdw, vlast_vdw,
- &(LR[ index_lr (i,j,num_atom_types) ].vdW[1]), control->tabulate+1 );
- Natural_Cubic_Spline( &h[1], &fCEvd[1],
- &(LR[ index_lr (i,j,num_atom_types) ].CEvd[1]), control->tabulate+1 );
-
- /*//fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fele" );
+ fprintf( stderr, "%f %f %f\n", r * dr, h[r], fvdw[r] );
+ fprintf( stderr, "v0_vdw: %f, vlast_vdw: %f\n", v0_vdw, vlast_vdw );
+ */
+ Complete_Cubic_Spline( &h[1], &fvdw[1], v0_vdw, vlast_vdw,
+ &(LR[ index_lr(i,j,num_atom_types) ].vdW[1]), control->tabulate + 1 );
+ Natural_Cubic_Spline( &h[1], &fCEvd[1],
+ &(LR[ index_lr(i,j,num_atom_types) ].CEvd[1]), control->tabulate + 1 );
+
+ /*fprintf( stderr, "%-6s %-6s %-6s\n", "r", "h", "fele" );
for( r = 1; r <= control->tabulate; ++r )
- //fprintf( stderr, "%f %f %f\n", r * dr, h[r], fele[r] );
- //fprintf( stderr, "v0_ele: %f, vlast_ele: %f\n", v0_ele, vlast_ele );
- */
- Complete_Cubic_Spline( &h[1], &fele[1], v0_ele, vlast_ele,
- &(LR[ index_lr (i,j,num_atom_types) ].ele[1]), control->tabulate+1 );
- Natural_Cubic_Spline( &h[1], &fCEclmb[1],
- &(LR[ index_lr (i,j,num_atom_types) ].CEclmb[1]), control->tabulate+1 );
+ fprintf( stderr, "%f %f %f\n", r * dr, h[r], fele[r] );
+ fprintf( stderr, "v0_ele: %f, vlast_ele: %f\n", v0_ele, vlast_ele );
+ */
+ Complete_Cubic_Spline( &h[1], &fele[1], v0_ele, vlast_ele,
+ &(LR[ index_lr(i,j,num_atom_types) ].ele[1]), control->tabulate + 1 );
+ Natural_Cubic_Spline( &h[1], &fCEclmb[1],
+ &(LR[ index_lr(i,j,num_atom_types) ].CEclmb[1]), control->tabulate + 1 );
}
+ }
+ }
+ }
/***** //test LR-Lookup table
- evdw_maxerr = 0;
- eele_maxerr = 0;
- for( i = 0; i < num_atom_types; ++i )
- if( existing_types[i] )
- for( j = i; j < num_atom_types; ++j )
- if( existing_types[j] ) {
- for( r = 1; r <= 100; ++r ) {
- rand_dist = (real)rand()/RAND_MAX * control->r_cut;
- LR_vdW_Coulomb( system, control, i, j, rand_dist, &y );
- LR_Lookup( &(LR[i][j]), rand_dist, &y_spline );
-
- evdw_abserr = fabs(y.e_vdW - y_spline.e_vdW);
- evdw_relerr = fabs(evdw_abserr / y.e_vdW);
- fvdw_abserr = fabs(y.CEvd - y_spline.CEvd);
- fvdw_relerr = fabs(fvdw_abserr / y.CEvd);
- eele_abserr = fabs(y.e_ele - y_spline.e_ele);
- eele_relerr = fabs(eele_abserr / y.e_ele);
- fele_abserr = fabs(y.CEclmb - y_spline.CEclmb);
- fele_relerr = fabs(fele_abserr / y.CEclmb);
-
- if( evdw_relerr > 1e-10 || eele_relerr > 1e-10 ){
- //fprintf( stderr, "rand_dist = %24.15e\n", rand_dist );
- //fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
- y.H, y_spline.H,
- fabs(y.H-y_spline.H), fabs((y.H-y_spline.H)/y.H) );
-
- //fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
- y.e_vdW, y_spline.e_vdW, evdw_abserr, evdw_relerr );
- //fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
- y.CEvd, y_spline.CEvd, fvdw_abserr, fvdw_relerr );
-
- //fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
- y.e_ele, y_spline.e_ele, eele_abserr, eele_relerr );
- //fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
- y.CEclmb, y_spline.CEclmb, fele_abserr, fele_relerr );
- }
-
- if( evdw_relerr > evdw_maxerr )
- evdw_maxerr = evdw_relerr;
- if( eele_relerr > eele_maxerr )
- eele_maxerr = eele_relerr;
- }
- }
- //fprintf( stderr, "evdw_maxerr: %24.15e\n", evdw_maxerr );
- //fprintf( stderr, "eele_maxerr: %24.15e\n", eele_maxerr );
- *******/
-
+ evdw_maxerr = 0;
+ eele_maxerr = 0;
+ for( i = 0; i < num_atom_types; ++i )
+ if( existing_types[i] )
+ for( j = i; j < num_atom_types; ++j )
+ if( existing_types[j] ) {
+ for( r = 1; r <= 100; ++r ) {
+ rand_dist = (real)rand()/RAND_MAX * control->r_cut;
+ LR_vdW_Coulomb( system, control, i, j, rand_dist, &y );
+ LR_Lookup( &(LR[i][j]), rand_dist, &y_spline );
+
+ evdw_abserr = fabs(y.e_vdW - y_spline.e_vdW);
+ evdw_relerr = fabs(evdw_abserr / y.e_vdW);
+ fvdw_abserr = fabs(y.CEvd - y_spline.CEvd);
+ fvdw_relerr = fabs(fvdw_abserr / y.CEvd);
+ eele_abserr = fabs(y.e_ele - y_spline.e_ele);
+ eele_relerr = fabs(eele_abserr / y.e_ele);
+ fele_abserr = fabs(y.CEclmb - y_spline.CEclmb);
+ fele_relerr = fabs(fele_abserr / y.CEclmb);
+
+ if( evdw_relerr > 1e-10 || eele_relerr > 1e-10 ){
+ fprintf( stderr, "rand_dist = %24.15e\n", rand_dist );
+ fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
+ y.H, y_spline.H,
+ fabs(y.H-y_spline.H), fabs((y.H-y_spline.H)/y.H) );
+
+ fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
+ y.e_vdW, y_spline.e_vdW, evdw_abserr, evdw_relerr );
+ fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
+ y.CEvd, y_spline.CEvd, fvdw_abserr, fvdw_relerr );
+
+ fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
+ y.e_ele, y_spline.e_ele, eele_abserr, eele_relerr );
+ fprintf( stderr, "%24.15e %24.15e %24.15e %24.15e\n",
+ y.CEclmb, y_spline.CEclmb, fele_abserr, fele_relerr );
+ }
+
+ if( evdw_relerr > evdw_maxerr )
+ evdw_maxerr = evdw_relerr;
+ if( eele_relerr > eele_maxerr )
+ eele_maxerr = eele_relerr;
+ }
+ }
+ fprintf( stderr, "evdw_maxerr: %24.15e\n", evdw_maxerr );
+ fprintf( stderr, "eele_maxerr: %24.15e\n", eele_maxerr );
+ *******/
+
free(h);
free(fh);
free(fvdw);
@@ -383,24 +420,26 @@ real Lookup( real x, lookup_table* t )
real b;
int i;
- /* if ( x < t->xmin)
- {
- //fprintf(stderr,"Domain check %lf > %lf\n",t->xmin,x);
- exit(0);
+ /*
+ if ( x < t->xmin)
+ {
+ fprintf(stderr,"Domain check %lf > %lf\n",t->xmin,x);
+ exit(0);
}
- if ( x > t->xmax)
+ if ( x > t->xmax)
{
- //fprintf(stderr,"Domain check %lf < %lf\n",t->xmax,x);
- exit(0);
- } */
+ fprintf(stderr,"Domain check %lf < %lf\n",t->xmax,x);
+ exit(0);
+ }
+ */
i = Lookup_Index_Of( x, t );
x1 = i * t->dx + t->xmin;
- x2 = (i+1) * t->dx + t->xmin;
+ x2 = (i + 1) * t->dx + t->xmin;
- b = ( x2 * t->y[i] - x1 * t->y[i+1] ) * t->inv_dx;
- // //fprintf( stdout,"SLookup_Entry: %d, %lf, %lf, %lf, %lf: %lf, %lf\n",
+ b = ( x2 * t->y[i] - x1 * t->y[i + 1] ) * t->inv_dx;
+ // fprintf( stdout,"SLookup_Entry: %d, %lf, %lf, %lf, %lf: %lf, %lf\n",
// i,x1,x2,x,b,t->one_over_dx*(t->y[i+1]-t->y[i])*x+b,exp(x));
- return t->inv_dx * ( t->y[i+1] - t->y[i] ) * x + b;
+ return t->inv_dx * ( t->y[i + 1] - t->y[i] ) * x + b;
}
diff --git a/PuReMD-GPU/src/mytypes.h b/PuReMD-GPU/src/mytypes.h
index 0eb1856a578df369aba28d8dbfb55dcf57348e9d..b04a9de39c8dbf487cac6d02a5e098a220862d5a 100644
--- a/PuReMD-GPU/src/mytypes.h
+++ b/PuReMD-GPU/src/mytypes.h
@@ -28,12 +28,6 @@
#define GLOBAL __global__
#define HOST_DEVICE __host__ __device__
- #include <cuda_runtime.h>
- #include <cuda.h>
- #include <cuda_runtime_api.h>
-
- #include <cublas_v2.h>
- #include <cusparse_v2.h>
#if __CUDA_ARCH__ < 600
#define MYATOMICADD myAtomicAdd
#else
@@ -55,14 +49,25 @@
#include "config.h"
#endif
-#include "math.h"
-//#include "random.h"
-#include "stdio.h"
-#include "stdlib.h"
-#include "string.h"
-#include "sys/time.h"
-#include "time.h"
-#include "zlib.h"
+#include <math.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/time.h>
+#include <time.h>
+
+#ifdef _OPENMP
+ #include <omp.h>
+#endif
+
+#ifdef HAVE_CUDA
+ #include <cuda_runtime.h>
+ #include <cuda.h>
+ #include <cuda_runtime_api.h>
+
+ #include <cublas_v2.h>
+ #include <cusparse_v2.h>
+#endif
//#define DEBUG_FOCUS
//#define TEST_FORCES
@@ -75,6 +80,7 @@
#define TRUE 1
#define FALSE 0
+#define LOG log
#define EXP exp
#define SQRT sqrt
#define POW pow
@@ -82,6 +88,8 @@
#define COS cos
#define SIN sin
#define TAN tan
+#define FABS fabs
+#define FMOD fmod
#define SQR(x) ((x)*(x))
#define CUBE(x) ((x)*(x)*(x))
@@ -90,6 +98,15 @@
#define MAX( x, y ) (((x) > (y)) ? (x) : (y))
#define MIN( x, y ) (((x) < (y)) ? (x) : (y))
+/* NaN IEEE 754 representation for C99 in math.h
+ * Note: function choice must match REAL typedef below */
+#ifdef NAN
+ #define IS_NAN_REAL(a) (isnan(a))
+#else
+ #warn "No support for NaN"
+ #define NAN_REAL(a) (0)
+#endif
+
#define PI 3.14159265
#define C_ele 332.06371
//#define K_B 503.398008 // kcal/mol/K
@@ -106,7 +123,11 @@
#define AVOGNR 6.0221367e23
#define P_CONV 1.0e-24 * AVOGNR * JOULES_to_CAL
-#define MAX_STR 100 // MAX STRing length (used for naming)
+#define MAX_STR 1024
+#define MAX_LINE 1024
+#define MAX_TOKENS 1024
+#define MAX_TOKEN_LEN 1024
+
#define MAX_ATOM_ID 100000
#define MAX_RESTRICT 15
#define MAX_MOLECULE_SIZE 20
@@ -124,23 +145,7 @@
#define MAX_ITR 10
#define RESTART 50
-#define FILE_NOT_FOUND_ERR 10
-#define UNKNOWN_ATOM_TYPE_ERR 11
-#define CANNOT_OPEN_OUTFILE 12
-#define INIT_ERR 13
-#define INSUFFICIENT_SPACE 14
-#define UNKNOWN_OPTION 15
-#define INVALID_INPUT 16
-
-#define C_ATOM 0
-#define H_ATOM 1
-#define O_ATOM 2
-#define N_ATOM 3
-#define S_ATOM 4
-#define SI_ATOM 5
-#define GE_ATOM 6
-#define X_ATOM 7
-
+/* tolerance used for validating GPU results against host */
#define GPU_TOLERANCE 1e-5
#define ZERO 0.000000000000000e+00
@@ -157,6 +162,7 @@
#define DANGER_ZONE 0.95
#define LOOSE_ZONE 0.75
+//TODO: make enum
#define RES_GRID_ATOMS 0x01
#define RES_GRID_TOP 0x02
#define RES_GRID_MARK 0x03
@@ -165,17 +171,21 @@
#define RES_GRID_NBRS 0x06
#define RES_GRID_NBRS_CP 0x07
+//TODO: make enum
#define RES_SYSTEM_ATOMS 0x10
#define RES_SYSTEM_SIMULATION_BOX 0x11
+//TODO: make enum
#define RES_REAX_INT_SBP 0x20
#define RES_REAX_INT_TBP 0x21
#define RES_REAX_INT_THBP 0x22
#define RES_REAX_INT_HBP 0x23
#define RES_REAX_INT_FBP 0x24
+//TODO: make enum
#define RES_SIMULATION_DATA 0x30
+//TODO: make enum
#define RES_STORAGE 0x401
#define RES_STORAGE_HBOND_INDEX 0x402
#define RES_STORAGE_TOTAL_BOND_ORDER 0x403
@@ -229,13 +239,17 @@
#define RES_STORAGE_RESTRICTED_LIST 0x432
#define RES_STORAGE_ORIG_ID 0x433
+//TODO: make enum
#define RES_CONTROL_PARAMS 0x50
+//TODO: make enum
#define RES_GLOBAL_PARAMS 0x60
+//TODO: make enum
#define RES_SPARSE_MATRIX_INDEX 0x70
#define RES_SPARSE_MATRIX_ENTRY 0x71
+//TODO: make enum
#define RES_LR_LOOKUP_Y 0x80
#define RES_LR_LOOKUP_H 0x81
#define RES_LR_LOOKUP_VDW 0x82
@@ -244,6 +258,7 @@
#define RES_LR_LOOKUP_CECLMB 0x85
#define RES_LR_LOOKUP_TABLE 0x86
+//TODO: make enum
#define RES_SCRATCH 0x90
#define LIST_INDEX 0x00
@@ -314,17 +329,78 @@ typedef real rvec[3];
typedef int ivec[3];
typedef real rtensor[3][3];
-enum {NVE, NVT, NPT, sNPT, iNPT, ensNR, bNVT};
-enum {FAR_NBRS, NEAR_NBRS, THREE_BODIES, BONDS, OLD_BONDS,
- HBONDS, DBO, DDELTA, LIST_N
- };
-enum {TYP_VOID, TYP_THREE_BODY, TYP_BOND, TYP_HBOND, TYP_DBO,
- TYP_DDELTA, TYP_FAR_NEIGHBOR, TYP_NEAR_NEIGHBOR, TYP_N
- };
-enum {UNKNOWN, WATER};
-enum {NO_ANALYSIS, FRAGMENTS, REACTIONS, NUM_ANALYSIS};
-enum {WRITE_ASCII, WRITE_BINARY, RF_N};
-enum {XYZ, PDB, BGF, ASCII_RESTART, BINARY_RESTART, GF_N};
+/* config params */
+enum ensemble
+{
+ NVE = 0, NVT = 1, NPT = 2, sNPT = 3, iNPT = 4, ensNR = 5, bNVT = 6,
+};
+
+enum interaction_list_offets
+{
+ FAR_NBRS = 0, NEAR_NBRS = 1, THREE_BODIES = 2, BONDS = 3, OLD_BONDS = 4,
+ HBONDS = 5, DBO = 6, DDELTA = 7, LIST_N = 8,
+};
+
+enum interaction_type
+{
+ TYP_VOID = 0, TYP_THREE_BODY = 1, TYP_BOND = 2, TYP_HBOND = 3, TYP_DBO = 4,
+ TYP_DDELTA = 5, TYP_FAR_NEIGHBOR = 6, TYP_NEAR_NEIGHBOR = 7, TYP_N = 8,
+};
+
+enum errors
+{
+ FILE_NOT_FOUND = -10,
+ UNKNOWN_ATOM_TYPE = -11,
+ CANNOT_OPEN_FILE = -12,
+ CANNOT_INITIALIZE = -13,
+ INSUFFICIENT_MEMORY = -14,
+ UNKNOWN_OPTION = -15,
+ INVALID_INPUT = -16,
+ INVALID_GEO = -17,
+ NUMERIC_BREAKDOWN = -18,
+ RUNTIME_ERROR = -19,
+};
+
+enum atoms
+{
+ C_ATOM = 0, H_ATOM = 1, O_ATOM = 2, N_ATOM = 3,
+ S_ATOM = 4, SI_ATOM = 5, GE_ATOM = 6, X_ATOM = 7,
+};
+
+enum molecule_type
+{
+ UNKNOWN = 0, WATER = 1,
+};
+
+enum molecular_analysis_type
+{
+ NO_ANALYSIS = 0, FRAGMENTS = 1, REACTIONS = 2, NUM_ANALYSIS = 3,
+};
+
+enum restart_format
+{
+ WRITE_ASCII = 0, WRITE_BINARY = 1, RF_N = 2,
+};
+
+enum geo_formats
+{
+ CUSTOM = 0, PDB = 1, BGF = 2, ASCII_RESTART = 3, BINARY_RESTART = 4, GF_N = 5,
+};
+
+enum solver
+{
+ GMRES_S = 0, GMRES_H_S = 1, CG_S = 2, SDM_S = 3,
+};
+
+enum pre_comp
+{
+ DIAG_PC = 0, ICHOLT_PC = 1, ILU_PAR_PC = 2, ILUT_PAR_PC = 3, ILU_SUPERLU_MT_PC = 4,
+};
+
+enum pre_app
+{
+ NONE_PA = 0, TRI_SOLVE_PA = 1, TRI_SOLVE_LEVEL_SCHED_PA = 2, TRI_SOLVE_GC_PA = 3, JACOBI_ITER_PA = 4,
+};
/* Global params mapping */
@@ -502,33 +578,36 @@ typedef struct
{
int num_atom_types;
global_parameters gp;
- global_parameters d_gp;
-
single_body_parameters *sbp;
- single_body_parameters *d_sbp;
-
two_body_parameters *tbp;
- two_body_parameters *d_tbp;
-
three_body_header *thbp;
- three_body_header *d_thbp;
-
hbond_parameters *hbp;
- hbond_parameters *d_hbp;
-
four_body_header *fbp;
- four_body_header *d_fbp;
+#ifdef HAVE_CUDA
+ global_parameters d_gp;
+ single_body_parameters *d_sbp;
+ two_body_parameters *d_tbp;
+ three_body_header *d_thbp;
+ hbond_parameters *d_hbp;
+ four_body_header *d_fbp;
+#endif
} reax_interaction;
typedef struct
{
- rvec x; /* Position, velocity, force on atom */
+ /* Position, velocity, force on atom */
+ rvec x;
rvec v;
rvec f;
- real q; /* Charge on the atom */
- int type; /* Type of this atom */
+
+ /* Charge on the atom */
+ real q;
+
+ /* Type of this atom */
+ int type;
+
char name[5];
char spare[7];
} reax_atom;
@@ -561,9 +640,6 @@ typedef struct
rvec len;
rvec inv_len;
- //CUDA
- int max_cuda_nbrs; //TODO remove this not used anymore
-
int *atoms;
int *top;
int *mark;
@@ -578,7 +654,16 @@ typedef struct
{
int N;
- //CUDA
+ reax_atom *atoms;
+ reax_interaction reaxprm;
+ simulation_box box;
+ grid g;
+
+#ifdef HAVE_CUDA
+ reax_atom *d_atoms;
+ simulation_box *d_box;
+ grid d_g;
+
//int max_thb_intrs;
int max_sparse_matrix_entries;
int num_nbrs;
@@ -586,17 +671,7 @@ typedef struct
int num_hbonds;
int num_thbodies;
int init_thblist;
-
- reax_atom *atoms;
- reax_atom *d_atoms;
-
- reax_interaction reaxprm;
-
- simulation_box box;
- simulation_box *d_box;
-
- grid g;
- grid d_g;
+#endif
} reax_system;
@@ -616,23 +691,22 @@ typedef struct
2 : NPT (Parrinello-Rehman-Nose-Hoover) Anisotropic
3 : sNPT (Parrinello-Rehman-Nose-Hoover) semiisotropic
4 : iNPT (Parrinello-Rehman-Nose-Hoover) isotropic */
- int ensemble;
- int nsteps;
- int periodic_boundaries;
- int restrict_bonds;
- int tabulate;
+ int ensemble;
+ int nsteps;
+ int periodic_boundaries;
+ int restrict_bonds;
+ int tabulate;
ivec periodic_images;
real dt;
int reneighbor;
real vlist_cut;
real nbr_cut;
- real r_cut, r_low; // upper and lower taper
+ real r_cut, r_sp_cut, r_low; // upper, reduced upper, and lower taper
real bo_cut;
real thb_cut;
real hb_cut;
real Tap7, Tap6, Tap5, Tap4, Tap3, Tap2, Tap1, Tap0;
- real q_err;
int max_far_nbrs;
real T_init, T_final, T;
@@ -656,16 +730,26 @@ typedef struct
int freq_diffusion_coef;
int restrict_type;
- int refactor;
- real droptol;
+ unsigned int qeq_solver_type;
+ real qeq_solver_q_err;
+ real qeq_domain_sparsity;
+ unsigned int qeq_domain_sparsify_enabled;
+ unsigned int pre_comp_type;
+ unsigned int pre_comp_refactor;
+ real pre_comp_droptol;
+ unsigned int pre_comp_sweeps;
+ unsigned int pre_app_type;
+ unsigned int pre_app_jacobi_iters;
int molec_anal;
int freq_molec_anal;
real bg_cut;
int num_ignored;
- int ignore[MAX_ATOM_TYPES];
+ int ignore[MAX_ATOM_TYPES];
+#ifdef HAVE_CUDA
void *d_control;
+#endif
} control_params;
@@ -720,7 +804,14 @@ typedef struct
real bonded;
real nonb;
real QEq;
- int matvecs;
+ real QEq_sort_mat_rows;
+ real pre_comp;
+ real pre_app;
+ int solver_iters;
+ real solver_spmv;
+ real solver_vector_ops;
+ real solver_orthog;
+ real solver_tri_solve;
} reax_timing;
@@ -776,9 +867,11 @@ typedef struct
rvec tot_press;
reax_timing timing;
- //CUDA
+
+#ifdef HAVE_CUDA
reax_timing d_timing;
void *d_simulation_data;
+#endif
} simulation_data;
@@ -789,8 +882,9 @@ typedef struct
real theta, cos_theta;
rvec dcos_di, dcos_dj, dcos_dk;
- //CUDA
+#ifdef HAVE_CUDA
int i, j, k;
+#endif
} three_body_interaction_data;
@@ -813,9 +907,11 @@ typedef struct
rvec dvec;
// real H; //, Tap, inv_dr3gamij_1, inv_dr3gamij_3;
- //CUDA
+#ifdef HAVE_CUDA
//int sym_index;
//rvec h_f;
+#endif
+
char spare[16];
} far_neighbor_data;
@@ -868,6 +964,7 @@ typedef struct
rvec dvec;
bond_order_data bo_data;
+#ifdef HAVE_CUDA
//single body -- lone pair
real scratch;
@@ -887,42 +984,47 @@ typedef struct
//compute_total_forces
rvec t_f;
+#endif
} bond_data;
+/* compressed row storage (crs) format
+ * See, e.g.,
+ * http://netlib.org/linalg/html_templates/node91.html#SECTION00931100000000000000
+ *
+ * m: number of nonzeros (NNZ) ALLOCATED
+ * n: number of rows
+ * start: row pointer (last element contains ACTUAL NNZ)
+ * j: column index for corresponding matrix entry
+ * val: matrix entry
+ * */
typedef struct
{
- int j;
- real val;
-} sparse_matrix_entry;
-
-
-typedef struct
-{
- int n, m;
- int *start;
- //CUDA
- int *end;
- sparse_matrix_entry *entries;
-
- int *j;
+ unsigned int n, m;
+ unsigned int *start;
+#ifdef HAVE_CUDA
+ unsigned int *end;
+#endif
+ unsigned int *j;
real *val;
-
} sparse_matrix;
typedef struct
{
- int estimate_nbrs;
int num_far;
int Htop;
int hbonds;
int num_hbonds;
int bonds;
int num_bonds;
- int thbody;
int num_3body;
int gcell_atoms;
+
+#ifdef HAVE_CUDA
+ int estimate_nbrs;
+ int thbody;
+#endif
} reallocate_data;
@@ -937,7 +1039,7 @@ typedef struct
rvec *dDeltap_self;
/* QEq storage */
- sparse_matrix H, L, U;
+ sparse_matrix *H, *H_sp, *L, *U;
real *droptol;
real *w;
real *Hdia_inv;
@@ -990,6 +1092,7 @@ typedef struct
} static_storage;
+/* interaction lists */
typedef struct
{
int n;
@@ -1127,25 +1230,25 @@ typedef void (*evolve_function)(reax_system*, control_params*,
list**, output_controls*);
typedef real (*lookup_function)(real);
-extern lookup_table Exp, Sqrt, Cube_Root, Four_Third_Root, Cos, Sin, ACos;
+extern lookup_table Exp, Sqrt, Cube_Root, Four_Third_Root, Cos, Sin, ACos;
extern LR_lookup_table *LR;
-
typedef void (*get_far_neighbors_function)(rvec, rvec, simulation_box*,
- control_params*, far_neighbor_data*,
- int*);
+ control_params*, far_neighbor_data*, int*);
+
+extern reax_timing d_timing;
-/* CUDA structures */
+#ifdef HAVE_CUDA
extern list *dev_lists;
extern static_storage *dev_workspace;
extern LR_lookup_table *d_LR;
-extern reax_timing d_timing;
-//Scratch Pad usage.
+/* scratch Pad usage */
extern void *scratch;
extern int BLOCKS, BLOCKS_POW_2, BLOCK_SIZE;
extern int MATVEC_BLOCKS;
+#endif
#endif
diff --git a/PuReMD-GPU/src/neighbors.c b/PuReMD-GPU/src/neighbors.c
index 5f425e672080d2d4a272f7aca1859c45d8dde17d..7a005f081d57e26cc86bde6501e72bd64d6bc2cc 100644
--- a/PuReMD-GPU/src/neighbors.c
+++ b/PuReMD-GPU/src/neighbors.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -26,51 +27,11 @@
#include "list.h"
#include "reset_utils.h"
#include "system_props.h"
+#include "tool_box.h"
#include "vector.h"
-int Are_Far_Neighbors( rvec x1, rvec x2, simulation_box *box,
- real cutoff, far_neighbor_data *data )
-{
- real norm_sqr, d, tmp;
- int i;
-
- norm_sqr = 0;
-
- for( i = 0; i < 3; i++ ) {
- d = x2[i] - x1[i];
- tmp = SQR(d);
-
- if( tmp >= SQR( box->box_norms[i] / 2.0 ) ) {
- if( x2[i] > x1[i] ) {
- d -= box->box_norms[i];
- data->rel_box[i] = -1;
- }
- else {
- d += box->box_norms[i];
- data->rel_box[i] = +1;
- }
-
- data->dvec[i] = d;
- norm_sqr += SQR(d);
- }
- else {
- data->dvec[i] = d;
- norm_sqr += tmp;
- data->rel_box[i] = 0;
- }
- }
-
- if( norm_sqr <= SQR(cutoff) ){
- data->d = sqrt(norm_sqr);
- return 1;
- }
-
- return 0;
-}
-
-
-void Generate_Neighbor_Lists( reax_system *system, control_params *control,
+void Generate_Neighbor_Lists( reax_system *system, control_params *control,
simulation_data *data, static_storage *workspace,
list **lists, output_controls *out_control )
{
@@ -86,54 +47,61 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
far_neighbor_data *nbr_data;
real t_start, t_elapsed;
+ t_start = Get_Time( );
// fprintf( stderr, "\n\tentered nbrs - " );
g = &( system->g );
far_nbrs = (*lists) + FAR_NBRS;
Bin_Atoms( system, workspace );
-
- t_start = Get_Time( );
-
// fprintf( stderr, "atoms sorted - " );
num_far = 0;
/* first pick up a cell in the grid */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
- nbrs = &g->nbrs[ index_grid_nbrs (i,j,k,0,g) ];
- nbrs_cp = &g->nbrs_cp[ index_grid_nbrs (i,j,k,0,g) ];
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
+ nbrs = &g->nbrs[ index_grid_nbrs(i,j,k,0,g) ];
+ nbrs_cp = &g->nbrs_cp[ index_grid_nbrs(i,j,k,0,g) ];
//fprintf( stderr, "gridcell %d %d %d\n", i, j, k );
/* pick up an atom from the current cell */
- for(l = 0; l < g->top[ index_grid_3d (i,j,k,g) ]; ++l ){
- atom1 = g->atoms[ index_grid_atoms (i,j,k,l,g) ];
+ for(l = 0; l < g->top[ index_grid_3d(i,j,k,g) ]; ++l )
+ {
+ atom1 = g->atoms[ index_grid_atoms(i,j,k,l,g) ];
Set_Start_Index( atom1, num_far, far_nbrs );
//fprintf( stderr, "\tatom %d\n", atom1 );
itr = 0;
- while( nbrs[itr][0] >= 0 ){
+ while ( nbrs[itr][0] >= 0 )
+ {
x = nbrs[itr][0];
y = nbrs[itr][1];
z = nbrs[itr][2];
//fprintf( stderr, "\t\tgridcell %d %d %d\n", x, y, z );
- if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
- SQR(control->vlist_cut) ) {
- nbr_atoms = &g->atoms[ index_grid_atoms (x,y,z,0,g) ];
- max = g->top[ index_grid_3d (x,y,z,g) ];
+ if ( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
+ SQR(control->vlist_cut) )
+ {
+ nbr_atoms = &g->atoms[ index_grid_atoms(x,y,z,0,g) ];
+ max = g->top[ index_grid_3d(x,y,z,g) ];
//fprintf( stderr, "\t\tmax: %d\n", max );
/* pick up another atom from the neighbor cell */
- for( m = 0; m < max; ++m ) {
+ for ( m = 0; m < max; ++m )
+ {
atom2 = nbr_atoms[m];
- if( atom1 > atom2 ) {
+ if ( atom1 > atom2 )
+ {
nbr_data = &(far_nbrs->select.far_nbr_list[num_far]);
- if(Are_Far_Neighbors(system->atoms[atom1].x,
- system->atoms[atom2].x,
- &(system->box), control->vlist_cut,
- nbr_data)) {
+ //fprintf (stderr, " %f %f %f \n", nbr_data->dvec[0], nbr_data->dvec[1], nbr_data->dvec[2]);
+ if (Are_Far_Neighbors(system->atoms[atom1].x,
+ system->atoms[atom2].x,
+ &(system->box), control->vlist_cut,
+ nbr_data))
+ {
nbr_data->nbr = atom2;
-
++num_far;
}
}
@@ -144,20 +112,22 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
}
Set_End_Index( atom1, num_far, far_nbrs );
- //fprintf(stderr, "i:%d, start: %d, end: %d - itr: %d\n",
+ //fprintf(stderr, "i:%d, start: %d, end: %d - itr: %d\n",
// atom1,Start_Index(atom1,far_nbrs),End_Index(atom1,far_nbrs),
- // itr);
+ // itr);
}
}
+ }
+ }
- fprintf (stderr, " TOTAL HOST NEIGHBORS : %d \n", num_far);
-
- if( num_far > far_nbrs->num_intrs * DANGER_ZONE ) {
+ if ( num_far > far_nbrs->num_intrs * DANGER_ZONE )
+ {
workspace->realloc.num_far = num_far;
- if( num_far > far_nbrs->num_intrs ){
+ if ( num_far > far_nbrs->num_intrs )
+ {
fprintf( stderr, "step%d-ran out of space on far_nbrs: top=%d, max=%d",
- data->step, num_far, far_nbrs->num_intrs );
- exit( INSUFFICIENT_SPACE );
+ data->step, num_far, far_nbrs->num_intrs );
+ exit( INSUFFICIENT_MEMORY );
}
}
@@ -165,25 +135,24 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
data->timing.nbrs += t_elapsed;
#if defined(DEBUG)
- for( i = 0; i < system->N; ++i ) {
- qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
- Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
- compare_far_nbrs );
+ for ( i = 0; i < system->N; ++i )
+ {
+ qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
+ Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
+ compare_far_nbrs );
}
#endif
-
-#if defined(DEBUG_FOCUS)
- //fprintf( stderr, "nbrs - ");
- //fprintf( stderr, "nbrs done, num_far: %d\n", num_far );
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "nbrs - ");
+ fprintf( stderr, "nbrs done, num_far: %d\n", num_far );
#endif
-
#if defined(TEST_ENERGY)
//Print_Far_Neighbors( system, control, workspace, lists );
#endif
}
-int Estimate_NumNeighbors( reax_system *system, control_params *control,
+int Estimate_NumNeighbors( reax_system *system, control_params *control,
static_storage *workspace, list **lists )
{
int i, j, k, l, m, itr;
@@ -195,53 +164,63 @@ int Estimate_NumNeighbors( reax_system *system, control_params *control,
rvec *nbrs_cp;
grid *g;
far_neighbor_data nbr_data;
-
+#ifdef HAVE_CUDA
int start = 0, finish = 0;
+#endif
// fprintf( stderr, "\n\tentered nbrs - " );
g = &( system->g );
Bin_Atoms( system, workspace );
// fprintf( stderr, "atoms sorted - " );
num_far = 0;
+#ifdef HAVE_CUDA
g->max_cuda_nbrs = 0;
+#endif
/* first pick up a cell in the grid */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
- nbrs = &g->nbrs[index_grid_nbrs (i,j,k,0,g) ];
- nbrs_cp = &g->nbrs_cp[index_grid_nbrs (i,j,k,0,g) ];
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
+ nbrs = &g->nbrs[ index_grid_nbrs(i,j,k,0,g) ];
+ nbrs_cp = &g->nbrs_cp[ index_grid_nbrs(i,j,k,0,g) ];
//fprintf( stderr, "gridcell %d %d %d\n", i, j, k );
/* pick up an atom from the current cell */
- for(l = 0; l < g->top[index_grid_3d (i,j,k,g) ]; ++l ){
- atom1 = g->atoms[index_grid_atoms (i,j,k,l,g) ];
- start = num_far;
+ for(l = 0; l < g->top[ index_grid_3d(i,j,k,g) ]; ++l )
+ {
+ atom1 = g->atoms[ index_grid_atoms(i,j,k,l,g) ];
itr = 0;
- while( nbrs[itr][0] >= 0 ){
+ while ( nbrs[itr][0] >= 0 )
+ {
x = nbrs[itr][0];
y = nbrs[itr][1];
z = nbrs[itr][2];
//fprintf( stderr, "\t\tgridcell %d %d %d\n", x, y, z );
- if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
- SQR(control->vlist_cut) ) {
- nbr_atoms = &g->atoms[index_grid_atoms (x,y,z,0,g) ];
- max = g->top[index_grid_3d (x,y,z,g) ];
+ if ( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
+ SQR(control->vlist_cut) )
+ {
+ nbr_atoms = &g->atoms[ index_grid_atoms(x,y,z,0,g) ];
+ max = g->top[ index_grid_3d(x,y,z,g) ];
//fprintf( stderr, "\t\tmax: %d\n", max );
/* pick up another atom from the neighbor cell -
- we have to compare atom1 with its own periodic images as well,
- that's why there is also equality in the if stmt below */
- for( m = 0; m < max; ++m ) {
+ we have to compare atom1 with its own periodic images as well,
+ that's why there is also equality in the if stmt below */
+ for ( m = 0; m < max; ++m )
+ {
atom2 = nbr_atoms[m];
//if( nbrs[itr+1][0] >= 0 || atom1 > atom2 ) {
- if( atom1 > atom2 ) {
- if(Are_Far_Neighbors(system->atoms[atom1].x,
- system->atoms[atom2].x,
- &(system->box), control->vlist_cut,
- &nbr_data))
+ if ( atom1 > atom2 )
+ {
+ if (Are_Far_Neighbors(system->atoms[atom1].x,
+ system->atoms[atom2].x,
+ &(system->box), control->vlist_cut,
+ &nbr_data))
++num_far;
}
}
@@ -250,38 +229,46 @@ int Estimate_NumNeighbors( reax_system *system, control_params *control,
++itr;
}
- // finish note
+#ifdef HAVE_CUDA
finish = num_far;
- if (g->max_cuda_nbrs <= (finish - start)){
+ if (g->max_cuda_nbrs <= (finish - start))
+ {
g->max_cuda_nbrs = finish - start;
}
+#endif
}
}
+ }
+ }
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "estimate nbrs done, num_far: %d\n", num_far );
#endif
-
return num_far * SAFE_ZONE;
}
-//Code not used anymore
#if defined DONE
-
-void Choose_Neighbor_Finder( reax_system *system, control_params *control,
+void Choose_Neighbor_Finder( reax_system *system, control_params *control,
get_far_neighbors_function *Get_Far_Neighbors )
{
- if( control->periodic_boundaries )
+ if ( control->periodic_boundaries )
{
- if( system->box.box_norms[0] > 2.0 * control->vlist_cut &&
+ if ( system->box.box_norms[0] > 2.0 * control->vlist_cut &&
system->box.box_norms[1] > 2.0 * control->vlist_cut &&
system->box.box_norms[2] > 2.0 * control->vlist_cut )
+ {
(*Get_Far_Neighbors) = Get_Periodic_Far_Neighbors_Big_Box;
- else (*Get_Far_Neighbors) = Get_Periodic_Far_Neighbors_Small_Box;
+ }
+ else
+ {
+ (*Get_Far_Neighbors) = Get_Periodic_Far_Neighbors_Small_Box;
+ }
}
else
+ {
(*Get_Far_Neighbors) = Get_NonPeriodic_Far_Neighbors;
+ }
}
@@ -327,18 +314,28 @@ inline int can_Bond( static_storage *workspace, int atom1, int atom2 )
// fprintf( stderr, "can bond %6d %6d?\n", atom1, atom2 );
- if( !workspace->restricted[ atom1 ] && !workspace->restricted[ atom2 ] )
- return 1;
+ if ( !workspace->restricted[ atom1 ] && !workspace->restricted[ atom2 ] )
+ {
+ return FALSE;
+ }
- for( i = 0; i < workspace->restricted[ atom1 ]; ++i )
- if( workspace->restricted_list[ atom1 ][i] == atom2 )
- return 1;
+ for ( i = 0; i < workspace->restricted[ atom1 ]; ++i )
+ {
+ if ( workspace->restricted_list[ atom1 ][i] == atom2 )
+ {
+ return FALSE;
+ }
+ }
- for( i = 0; i < workspace->restricted[ atom2 ]; ++i )
- if( workspace->restricted_list[ atom2 ][i] == atom1 )
- return 1;
+ for ( i = 0; i < workspace->restricted[ atom2 ]; ++i )
+ {
+ if ( workspace->restricted_list[ atom2 ][i] == atom1 )
+ {
+ return FALSE;
+ }
+ }
- return 0;
+ return TRUE;
}
@@ -347,17 +344,20 @@ inline int is_Near_Neighbor( list *near_nbrs, int atom1, int atom2 )
{
int i;
- for( i=Start_Index(atom1,near_nbrs); i<End_Index(atom1,near_nbrs); ++i )
- if( near_nbrs->select.near_nbr_list[i].nbr == atom2 )
+ for ( i = Start_Index(atom1, near_nbrs); i < End_Index(atom1, near_nbrs); ++i )
+ {
+ if ( near_nbrs->select.near_nbr_list[i].nbr == atom2 )
{
// fprintf( stderr, "near neighbors %6d %6d\n", atom1, atom2 );
- return 1;
+ return FALSE;
}
+ }
- return 0;
+ return TRUE;
}
-void Generate_Neighbor_Lists( reax_system *system, control_params *control,
+
+void Generate_Neighbor_Lists( reax_system *system, control_params *control,
simulation_data *data, static_storage *workspace,
list **lists, output_controls *out_control )
{
@@ -368,21 +368,20 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
int num_far;
int c, count;
int grid_top;
- grid *g = &( system->g );
+ grid *g = &( system->g );
list *far_nbrs = (*lists) + FAR_NBRS;
//int hb_type1, hb_type2;
//list *hbonds = (*lists) + HBOND;
//int top_hbond1, top_hbond2;
get_far_neighbors_function Get_Far_Neighbors;
far_neighbor_data new_nbrs[125];
-#ifndef REORDER_ATOMS
- int l, m;
-#endif
// fprintf( stderr, "\n\tentered nbrs - " );
- if( control->ensemble == iNPT || control->ensemble == sNPT ||
+ if ( control->ensemble == iNPT || control->ensemble == sNPT ||
control->ensemble == NPT )
+ {
Update_Grid( system );
+ }
// fprintf( stderr, "grid updated - " );
Bin_Atoms( system, out_control );
@@ -394,9 +393,9 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
#endif
Choose_Neighbor_Finder( system, control, &Get_Far_Neighbors );
- // fprintf( stderr, "function chosen - " );
+ // fprintf( stderr, "function chosen - " );
- Reset_Neighbor_Lists( system, workspace, lists );
+ Reset_Neighbor_Lists( system, workspace, lists );
// fprintf( stderr, "lists cleared - " );
num_far = 0;
@@ -404,9 +403,12 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
c = 0;
/* first pick up a cell in the grid */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
nbrs = g->nbrs[i][j][k];
nbrs_cp = g->nbrs_cp[i][j][k];
@@ -414,119 +416,137 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
//#ifdef REORDER_ATOMS
// for(atom1 = g->start[i][j][k]; atom1 < g->end[i][j][k]; atom1++)
//#else
- for(l = 0; l < g->top[i][j][k]; ++l ){
+ for (l = 0; l < g->top[i][j][k]; ++l )
+ {
atom1 = g->atoms[i][j][k][l];
Set_End_Index( atom1, num_far, far_nbrs );
// fprintf( stderr, "atom %d:\n", atom1 );
itr = 0;
- while( nbrs[itr][0] > 0 ){
+ while ( nbrs[itr][0] > 0 )
+ {
x = nbrs[itr][0];
y = nbrs[itr][1];
z = nbrs[itr][2];
- // if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
- // SQR(control->r_cut))
+ // if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
+ // SQR(control->r_cut))
nbr_atoms = g->atoms[x][y][z];
max_atoms = g->top[x][y][z];
/* pick up another atom from the neighbor cell -
- we have to compare atom1 with its own periodic images as well,
+ we have to compare atom1 with its own periodic images as well,
that's why there is also equality in the if stmt below */
//#ifdef REORDER_ATOMS
//for(atom2=g->start[x][y][z]; atom2<g->end[x][y][z]; atom2++)
//#else
- for( m = 0, atom2=nbr_atoms[m]; m < max; ++m, atom2=nbr_atoms[m] )
- if( atom1 >= atom2 ) {
+ for ( m = 0, atom2 = nbr_atoms[m]; m < max; ++m, atom2 = nbr_atoms[m] )
+ {
+ if ( atom1 >= atom2 )
+ {
//fprintf( stderr, "\tatom2 %d", atom2 );
//top_near1 = End_Index( atom1, near_nbrs );
//Set_Start_Index( atom1, num_far, far_nbrs );
//hb_type1=system->reaxprm.sbp[system->atoms[atom1].type].p_hbond;
Get_Far_Neighbors( system->atoms[atom1].x,
- system->atoms[atom2].x,
- &(system->box), control, new_nbrs, &count );
+ system->atoms[atom2].x,
+ &(system->box), control, new_nbrs, &count );
fprintf( stderr, "\t%d count:%d\n", atom2, count );
- for( c = 0; c < count; ++c )
- if(atom1 != atom2 || (atom1 == atom2 && new_nbrs[c].d>=0.1)){
+ for ( c = 0; c < count; ++c )
+ {
+ if (atom1 != atom2 || (atom1 == atom2 && new_nbrs[c].d >= 0.1))
+ {
Set_Far_Neighbor(&(far_nbrs->select.far_nbr_list[num_far]),
- atom2, new_nbrs[c].d, 1.0,
- new_nbrs[c].dvec, new_nbrs[c].rel_box );
+ atom2, new_nbrs[c].d, 1.0,
+ new_nbrs[c].dvec, new_nbrs[c].rel_box );
++num_far;
/*fprintf(stderr,"FARNBR:%6d%6d%8.3f[%8.3f%8.3f%8.3f]\n",
- atom1, atom2, new_nbrs[c].d,
- new_nbrs[c].dvec[0], new_nbrs[c].dvec[1],
+ atom1, atom2, new_nbrs[c].d,
+ new_nbrs[c].dvec[0], new_nbrs[c].dvec[1],
new_nbrs[c].dvec[2] ); */
- /* hydrogen bond lists */
- /*if( control->hb_cut > 0.1 &&
+ /* hydrogen bond lists */
+ /*if( control->hb_cut > 0.1 &&
new_nbrs[c].d <= control->hb_cut ) {
- // fprintf( stderr, "%d %d\n", atom1, atom2 );
- hb_type2=system->reaxprm.sbp[system->atoms[atom2].type].p_hbond;
- if( hb_type1 == 1 && hb_type2 == 2 ) {
- top_hbond1=End_Index(workspace->hbond_index[atom1],hbonds);
- Set_Near_Neighbor(&(hbonds->select.hbond_list[top_hbond1]),
- atom2, new_nbrs[c].d, 1.0, new_nbrs[c].dvec,
- new_nbrs[c].rel_box );
- Set_End_Index( workspace->hbond_index[atom1],
- top_hbond1 + 1, hbonds );
- }
- else if( hb_type1 == 2 && hb_type2 == 1 ) {
- top_hbond2 = End_Index( workspace->hbond_index[atom2], hbonds );
- Set_Near_Neighbor(&(hbonds->select.hbond_list[top_hbond2]),
- atom1, new_nbrs[c].d, -1.0, new_nbrs[c].dvec,
- new_nbrs[c].rel_box );
- Set_End_Index( workspace->hbond_index[atom2],
- top_hbond2 + 1, hbonds );
- }*/
+ // fprintf( stderr, "%d %d\n", atom1, atom2 );
+ hb_type2=system->reaxprm.sbp[system->atoms[atom2].type].p_hbond;
+ if( hb_type1 == 1 && hb_type2 == 2 ) {
+ top_hbond1=End_Index(workspace->hbond_index[atom1],hbonds);
+ Set_Near_Neighbor(&(hbonds->select.hbond_list[top_hbond1]),
+ atom2, new_nbrs[c].d, 1.0, new_nbrs[c].dvec,
+ new_nbrs[c].rel_box );
+ Set_End_Index( workspace->hbond_index[atom1],
+ top_hbond1 + 1, hbonds );
+ }
+ else if( hb_type1 == 2 && hb_type2 == 1 ) {
+ top_hbond2 = End_Index( workspace->hbond_index[atom2], hbonds );
+ Set_Near_Neighbor(&(hbonds->select.hbond_list[top_hbond2]),
+ atom1, new_nbrs[c].d, -1.0, new_nbrs[c].dvec,
+ new_nbrs[c].rel_box );
+ Set_End_Index( workspace->hbond_index[atom2],
+ top_hbond2 + 1, hbonds );
+ }*/
}
}
+ }
}
+ }
Set_End_Index( atom1, top_far1, far_nbrs );
}
}
+ }
+ }
fprintf( stderr, "nbrs done-" );
+
/* apply restrictions on near neighbors only */
- if( (data->step - data->prev_steps) < control->restrict_bonds ) {
- for( atom1 = 0; atom1 < system->N; ++atom1 )
- if( workspace->restricted[ atom1 ] ) {
+ if ( (data->step - data->prev_steps) < control->restrict_bonds )
+ {
+ for ( atom1 = 0; atom1 < system->N; ++atom1 )
+ {
+ if ( workspace->restricted[ atom1 ] )
+ {
// fprintf( stderr, "atom1: %d\n", atom1 );
top_near1 = End_Index( atom1, near_nbrs );
- for( j = 0; j < workspace->restricted[ atom1 ]; ++j )
- if(!is_Near_Neighbor(near_nbrs, atom1,
- atom2 = workspace->restricted_list[atom1][j])) {
+ for ( j = 0; j < workspace->restricted[ atom1 ]; ++j )
+ {
+ if (is_Near_Neighbor(near_nbrs, atom1,
+ atom2 = workspace->restricted_list[atom1][j]) == FALSE)
+ {
fprintf( stderr, "%3d-%3d: added bond by applying restrictions!\n",
- atom1, atom2 );
+ atom1, atom2 );
- top_near2 = End_Index( atom2, near_nbrs );
+ top_near2 = End_Index( atom2, near_nbrs );
- /* we just would like to get the nearest image, so a call to
+ /* we just would like to get the nearest image, so a call to
Get_Periodic_Far_Neighbors_Big_Box is good enough. */
- Get_Periodic_Far_Neighbors_Big_Box( system->atoms[ atom1 ].x,
- system->atoms[ atom2 ].x,
- &(system->box), control,
- new_nbrs, &count );
+ Get_Periodic_Far_Neighbors_Big_Box( system->atoms[ atom1 ].x,
+ system->atoms[ atom2 ].x,
+ &(system->box), control,
+ new_nbrs, &count );
Set_Near_Neighbor( &(near_nbrs->select.near_nbr_list[ top_near1 ]),
- atom2, new_nbrs[c].d, 1.0,
- new_nbrs[c].dvec, new_nbrs[c].rel_box );
+ atom2, new_nbrs[c].d, 1.0,
+ new_nbrs[c].dvec, new_nbrs[c].rel_box );
++top_near1;
Set_Near_Neighbor( &(near_nbrs->select.near_nbr_list[ top_near2 ]),
- atom1, new_nbrs[c].d, -1.0,
- new_nbrs[c].dvec, new_nbrs[c].rel_box );
- Set_End_Index( atom2, top_near2+1, near_nbrs );
+ atom1, new_nbrs[c].d, -1.0,
+ new_nbrs[c].dvec, new_nbrs[c].rel_box );
+ Set_End_Index( atom2, top_near2 + 1, near_nbrs );
}
+ }
Set_End_Index( atom1, top_near1, near_nbrs );
}
+ }
}
// fprintf( stderr, "restrictions applied-" );
@@ -534,56 +554,61 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
/* verify nbrlists, count num_intrs, sort nearnbrs */
near_nbrs->num_intrs = 0;
far_nbrs->num_intrs = 0;
- for( i = 0; i < system->N-1; ++i ) {
- if( End_Index(i, near_nbrs) > Start_Index(i+1, near_nbrs) ) {
- fprintf( stderr,
- "step%3d: nearnbr list of atom%d is overwritten by atom%d\n",
- data->step, i+1, i );
- exit( 1 );
+ for ( i = 0; i < system->N - 1; ++i )
+ {
+ if ( End_Index(i, near_nbrs) > Start_Index(i + 1, near_nbrs) )
+ {
+ fprintf( stderr,
+ "step%3d: nearnbr list of atom%d is overwritten by atom%d\n",
+ data->step, i + 1, i );
+ exit( RUNTIME_ERROR );
}
near_nbrs->num_intrs += Num_Entries(i, near_nbrs);
- if( End_Index(i, far_nbrs) > Start_Index(i+1, far_nbrs) ) {
- fprintf( stderr,
- "step%3d: farnbr list of atom%d is overwritten by atom%d\n",
- data->step, i+1, i );
- exit( 1 );
+ if ( End_Index(i, far_nbrs) > Start_Index(i + 1, far_nbrs) )
+ {
+ fprintf( stderr,
+ "step%3d: farnbr list of atom%d is overwritten by atom%d\n",
+ data->step, i + 1, i );
+ exit( RUNTIME_ERROR );
}
far_nbrs->num_intrs += Num_Entries(i, far_nbrs);
}
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
qsort( &(near_nbrs->select.near_nbr_list[ Start_Index(i, near_nbrs) ]),
- Num_Entries(i, near_nbrs), sizeof(near_neighbor_data),
- compare_near_nbrs );
+ Num_Entries(i, near_nbrs), sizeof(near_neighbor_data),
+ compare_near_nbrs );
}
// fprintf( stderr, "near nbrs sorted\n" );
+
#ifdef TEST_ENERGY
/* for( i = 0; i < system->N; ++i ) {
- qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
- Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
- compare_far_nbrs );
+ qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
+ Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
+ compare_far_nbrs );
} */
- fprintf( stderr, "Near neighbors/atom: %d (compare to 150)\n",
- num_near / system->N );
- fprintf( stderr, "Far neighbors per atom: %d (compare to %d)\n",
- num_far / system->N, control->max_far_nbrs );
+ fprintf( stderr, "Near neighbors/atom: %d (compare to 150)\n",
+ num_near / system->N );
+ fprintf( stderr, "Far neighbors per atom: %d (compare to %d)\n",
+ num_far / system->N, control->max_far_nbrs );
#endif
//fprintf( stderr, "step%d: num of nearnbrs = %6d num of farnbrs: %6d\n",
// data->step, num_near, num_far );
- //fprintf( stderr, "\talloc nearnbrs = %6d alloc farnbrs: %6d\n",
- // system->N * near_nbrs->intrs_per_unit,
+ //fprintf( stderr, "\talloc nearnbrs = %6d alloc farnbrs: %6d\n",
+ // system->N * near_nbrs->intrs_per_unit,
// system->N * far_nbrs->intrs_per_unit );
}
-void Generate_Neighbor_Lists( reax_system *system, control_params *control,
+void Generate_Neighbor_Lists( reax_system *system, control_params *control,
simulation_data *data, static_storage *workspace,
list **lists, output_controls *out_control )
{
@@ -603,73 +628,84 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
far_nbrs = (*lists) + FAR_NBRS;
// fprintf( stderr, "\n\tentered nbrs - " );
- if( control->ensemble == iNPT ||
- control->ensemble == sNPT ||
+ if ( control->ensemble == iNPT ||
+ control->ensemble == sNPT ||
control->ensemble == NPT )
+ {
Update_Grid( system );
+ }
// fprintf( stderr, "grid updated - " );
Bin_Atoms( system, out_control );
// fprintf( stderr, "atoms sorted - " );
Choose_Neighbor_Finder( system, control, &Get_Far_Neighbors );
- // fprintf( stderr, "function chosen - " );
- Reset_Neighbor_Lists( system, workspace, lists );
+ // fprintf( stderr, "function chosen - " );
+ Reset_Neighbor_Lists( system, workspace, lists );
// fprintf( stderr, "lists cleared - " );
num_far = 0;
c = 0;
/* first pick up a cell in the grid */
- for( i = 0; i < g->ncell[0]; i++ )
- for( j = 0; j < g->ncell[1]; j++ )
- for( k = 0; k < g->ncell[2]; k++ ) {
+ for ( i = 0; i < g->ncell[0]; i++ )
+ {
+ for ( j = 0; j < g->ncell[1]; j++ )
+ {
+ for ( k = 0; k < g->ncell[2]; k++ )
+ {
nbrs = g->nbrs[i][j][k];
nbrs_cp = g->nbrs_cp[i][j][k];
fprintf( stderr, "gridcell %d %d %d\n", i, j, k );
/* pick up an atom from the current cell */
- for(l = 0; l < g->top[i][j][k]; ++l ){
+ for (l = 0; l < g->top[i][j][k]; ++l )
+ {
atom1 = g->atoms[i][j][k][l];
Set_Start_Index( atom1, num_far, far_nbrs );
fprintf( stderr, "\tatom %d\n", atom1 );
itr = 0;
- while( nbrs[itr][0] > 0 ){
+ while ( nbrs[itr][0] > 0 )
+ {
x = nbrs[itr][0];
y = nbrs[itr][1];
z = nbrs[itr][2];
fprintf( stderr, "\t\tgridcell %d %d %d\n", x, y, z );
- // if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
- // SQR(control->r_cut))
+ // if( DistSqr_to_CP(nbrs_cp[itr], system->atoms[atom1].x ) <=
+ // SQR(control->r_cut))
nbr_atoms = g->atoms[x][y][z];
max = g->top[x][y][z];
fprintf( stderr, "\t\tmax: %d\n", max );
/* pick up another atom from the neighbor cell -
- we have to compare atom1 with its own periodic images as well,
+ we have to compare atom1 with its own periodic images as well,
that's why there is also equality in the if stmt below */
- for( m = 0, atom2=nbr_atoms[m]; m < max; ++m, atom2=nbr_atoms[m] )
- if( atom1 >= atom2 ) {
+ for ( m = 0, atom2 = nbr_atoms[m]; m < max; ++m, atom2 = nbr_atoms[m] )
+ {
+ if ( atom1 >= atom2 )
+ {
Get_Far_Neighbors( system->atoms[atom1].x,
- system->atoms[atom2].x,
- &(system->box), control, new_nbrs, &count );
+ system->atoms[atom2].x,
+ &(system->box), control, new_nbrs, &count );
fprintf( stderr, "\t\t\t%d count:%d\n", atom2, count );
- for( c = 0; c < count; ++c )
- if(atom1 != atom2 || (atom1 == atom2 && new_nbrs[c].d>=0.1)){
+ for ( c = 0; c < count; ++c )
+ if (atom1 != atom2 || (atom1 == atom2 && new_nbrs[c].d >= 0.1))
+ {
Set_Far_Neighbor(&(far_nbrs->select.far_nbr_list[num_far]),
- atom2, new_nbrs[c].d, 1.0,
- new_nbrs[c].dvec, new_nbrs[c].rel_box );
+ atom2, new_nbrs[c].d, 1.0,
+ new_nbrs[c].dvec, new_nbrs[c].rel_box );
++num_far;
/*fprintf(stderr,"FARNBR:%6d%6d%8.3f[%8.3f%8.3f%8.3f]\n",
- atom1, atom2, new_nbrs[c].d,
- new_nbrs[c].dvec[0], new_nbrs[c].dvec[1],
+ atom1, atom2, new_nbrs[c].d,
+ new_nbrs[c].dvec[0], new_nbrs[c].dvec[1],
new_nbrs[c].dvec[2] ); */
}
}
+ }
++itr;
}
@@ -677,22 +713,26 @@ void Generate_Neighbor_Lists( reax_system *system, control_params *control,
Set_End_Index( atom1, num_far, far_nbrs );
}
}
+ }
+ }
- far_nbrs->num_intrs = num_far;
+ far_nbrs->num_intrs = num_far;
fprintf( stderr, "nbrs done, num_far: %d\n", num_far );
#if defined(DEBUG)
- for( i = 0; i < system->N; ++i ) {
- qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
- Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
- compare_far_nbrs );
+ for ( i = 0; i < system->N; ++i )
+ {
+ qsort( &(far_nbrs->select.far_nbr_list[ Start_Index(i, far_nbrs) ]),
+ Num_Entries(i, far_nbrs), sizeof(far_neighbor_data),
+ compare_far_nbrs );
}
fprintf( stderr, "step%d: num of farnbrs=%6d\n", data->step, num_far );
- fprintf( stderr, "\tallocated farnbrs: %6d\n",
- system->N * far_nbrs->intrs_per_unit );
+ fprintf( stderr, "\tallocated farnbrs: %6d\n",
+ system->N * far_nbrs->intrs_per_unit );
#endif
}
+
#endif
diff --git a/PuReMD-GPU/src/neighbors.h b/PuReMD-GPU/src/neighbors.h
index 64c14ad29d5194006aacb057a7d80ef54aeee8e4..8eb5cfc2696f4d354edcf3751dedfd315c6762a3 100644
--- a/PuReMD-GPU/src/neighbors.h
+++ b/PuReMD-GPU/src/neighbors.h
@@ -30,10 +30,8 @@ void Generate_Neighbor_Lists( reax_system*, control_params*, simulation_data*,
int Estimate_NumNeighbors( reax_system*, control_params*,
static_storage*, list** );
-int Are_Far_Neighbors( rvec, rvec, simulation_box*, real, far_neighbor_data* );
-
-static inline HOST_DEVICE int index_grid_debug (int x, int y, int z, int blocksize)
+static inline HOST_DEVICE int index_grid_debug( int x, int y, int z, int blocksize )
{
return x * 8 * 8 * blocksize +
y * 8 * blocksize +
diff --git a/PuReMD-GPU/src/pdb_tools.c b/PuReMD-GPU/src/pdb_tools.c
deleted file mode 100644
index a7102da2cf8d3023956539960f93f4e61c116a81..0000000000000000000000000000000000000000
--- a/PuReMD-GPU/src/pdb_tools.c
+++ /dev/null
@@ -1,628 +0,0 @@
-/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
-
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
- Hasan Metin Aktulga, haktulga@cs.purdue.edu
- Ananth Y Grama, ayg@cs.purdue.edu
-
- This program is free software; you can redistribute it and/or
- modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
- the License, or (at your option) any later version.
-
- This program is distributed in the hope that it will be useful,
- but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
- See the GNU General Public License for more details:
- <http://www.gnu.org/licenses/>.
- ----------------------------------------------------------------------*/
-
-#include "pdb_tools.h"
-#include "box.h"
-#include "list.h"
-#include "param.h"
-#include "restart.h"
-#include "ctype.h"
-
-
-int is_Valid_Serial( static_storage *workspace, int serial )
-{
- if ( workspace->map_serials[ serial ] < 0 )
- {
- fprintf( stderr, "CONECT line includes invalid pdb serial number %d.\n",
- serial );
- fprintf( stderr, "Please correct the input file.Terminating...\n" );
- exit( INVALID_INPUT );
- }
-
- return 1;
-}
-
-
-int Check_Input_Range( int val, int lo, int hi, char *message )
-{
- if ( val < lo || val > hi )
- {
- fprintf( stderr, "%s\nInput %d - Out of range %d-%d. Terminating...\n",
- message, val, lo, hi );
- exit( INVALID_INPUT );
- }
-
- return 1;
-}
-
-
-void Trim_Spaces( char *element )
-{
- int i, j;
-
- for ( i = 0; element[i] == ' '; ++i ); // skip initial space chars
-
- for ( j = i; j < strlen(element) && element[j] != ' '; ++j )
- element[j - i] = toupper( element[j] ); // make uppercase, move to beginning
- element[j - i] = 0; // finalize the string
-}
-
-
-char Read_PDB( char* pdb_file, reax_system* system, control_params *control,
- simulation_data *data, static_storage *workspace )
-{
-
- FILE *pdb;
- char **tmp;
- char *s, *s1;
- char descriptor[9], serial[9];
- char atom_name[9], res_name[9], res_seq[9];
- char s_x[9], s_y[9], s_z[9];
- char occupancy[9], temp_factor[9];
- char seg_id[9], element[9], charge[9];
- char alt_loc, chain_id, icode;
- char s_a[10], s_b[10], s_c[10], s_alpha[9], s_beta[9], s_gamma[9];
- char s_group[12], s_zValue[9];
- char *endptr = NULL;
- int i, c, c1, pdb_serial, ratom = 0;
- /* open pdb file */
- if ( (pdb = fopen(pdb_file, "r")) == NULL )
- {
- fprintf( stderr, "Error opening the pdb file!\n" );
- exit( FILE_NOT_FOUND_ERR );
- }
-
-
- /* allocate memory for tokenizing pdb lines */
- s = (char*) malloc( sizeof(char) * MAX_LINE );
- s1 = (char*) malloc( sizeof(char) * MAX_LINE );
- tmp = (char**) malloc( sizeof(char*) * MAX_TOKENS );
- for ( i = 0; i < MAX_TOKENS; i++ )
- tmp[i] = (char*) malloc( sizeof(char) * MAX_TOKEN_LEN );
-
-
- /* count number of atoms in the pdb file */
- system->N = 0;
- while (!feof(pdb))
- {
- s[0] = 0;
- fgets( s, MAX_LINE, pdb );
-
- tmp[0][0] = 0;
- c = Tokenize( s, &tmp );
-
- if ( strncmp( tmp[0], "ATOM", 4 ) == 0 ||
- strncmp( tmp[0], "HETATM", 6 ) == 0 )
- (system->N)++;
- }
- fclose(pdb);
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "system->N: %d\n", system->N );
-#endif
-
- /* memory allocations for atoms, atom maps, bond restrictions */
- system->atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
-
- workspace->map_serials = (int*) calloc( MAX_ATOM_ID, sizeof(int) );
- for ( i = 0; i < MAX_ATOM_ID; ++i )
- workspace->map_serials[i] = -1;
-
- workspace->orig_id = (int*) calloc( system->N, sizeof(int) );
- workspace->restricted = (int*) calloc( system->N, sizeof(int) );
- workspace->restricted_list = (int*) calloc( system->N * MAX_RESTRICT, sizeof(int) );
-
- //for( i = 0; i < system->N; ++i )
- // workspace->restricted_list[i] = (int*) calloc( MAX_RESTRICT, sizeof(int) );
-
-
- /* start reading and processing pdb file */
- pdb = fopen(pdb_file, "r");
- c = 0;
- c1 = 0;
-
- while (!feof(pdb))
- {
- /* clear previous input line */
- s[0] = 0;
- for ( i = 0; i < c1; ++i )
- tmp[i][0] = 0;
-
- /* read new line and tokenize it */
- fgets( s, MAX_LINE, pdb );
- strncpy( s1, s, MAX_LINE - 1 );
- c1 = Tokenize( s, &tmp );
-
- /* process new line */
- if ( strncmp(tmp[0], "ATOM", 4) == 0 || strncmp(tmp[0], "HETATM", 6) == 0 )
- {
- if ( strncmp(tmp[0], "ATOM", 4) == 0 )
- {
- strncpy( &descriptor[0], s1, 6 );
- descriptor[6] = 0;
- strncpy( &serial[0], s1 + 6, 5 );
- serial[5] = 0;
- strncpy( &atom_name[0], s1 + 12, 4 );
- atom_name[4] = 0;
- alt_loc = s1[16];
- strncpy( &res_name[0], s1 + 17, 3 );
- res_name[3] = 0;
- chain_id = s1[21];
- strncpy( &res_seq[0], s1 + 22, 4 );
- res_seq[4] = 0;
- icode = s1[26];
- strncpy( &s_x[0], s1 + 30, 8 );
- s_x[8] = 0;
- strncpy( &s_y[0], s1 + 38, 8 );
- s_y[8] = 0;
- strncpy( &s_z[0], s1 + 46, 8 );
- s_z[8] = 0;
- strncpy( &occupancy[0], s1 + 54, 6 );
- occupancy[6] = 0;
- strncpy( &temp_factor[0], s1 + 60, 6 );
- temp_factor[6] = 0;
- strncpy( &seg_id[0], s1 + 72, 4 );
- seg_id[4] = 0;
- strncpy( &element[0], s1 + 76, 2 );
- element[2] = 0;
- strncpy( &charge[0], s1 + 78, 2 );
- charge[2] = 0;
- }
- else if (strncmp(tmp[0], "HETATM", 6) == 0)
- {
- strncpy( &descriptor[0], s1, 6 );
- descriptor[6] = 0;
- strncpy( &serial[0], s1 + 6, 5 );
- serial[5] = 0;
- strncpy( &atom_name[0], s1 + 12, 4 );
- atom_name[4] = 0;
- alt_loc = s1[16];
- strncpy( &res_name[0], s1 + 17, 3 );
- res_name[3] = 0;
- chain_id = s1[21];
- strncpy( &res_seq[0], s1 + 22, 4 );
- res_seq[4] = 0;
- icode = s1[26];
- strncpy( &s_x[0], s1 + 30, 8 );
- s_x[8] = 0;
- strncpy( &s_y[0], s1 + 38, 8 );
- s_y[8] = 0;
- strncpy( &s_z[0], s1 + 46, 8 );
- s_z[8] = 0;
- strncpy( &occupancy[0], s1 + 54, 6 );
- occupancy[6] = 0;
- strncpy( &temp_factor[0], s1 + 60, 6 );
- temp_factor[6] = 0;
- //strncpy( &seg_id[0], s1+72, 4 ); seg_id[4] = 0;
- strncpy( &element[0], s1 + 76, 2 );
- element[2] = 0;
- strncpy( &charge[0], s1 + 78, 2 );
- charge[2] = 0;
- }
-
-
- /* add to mapping */
- pdb_serial = strtod( &serial[0], &endptr );
- Check_Input_Range( pdb_serial, 0, MAX_ATOM_ID, "Invalid pdb_serial" );
- workspace->map_serials[ pdb_serial ] = c;
- workspace->orig_id[ c ] = pdb_serial;
- // fprintf( stderr, "map %d --> %d\n", pdb_serial, c );
-
-
- /* copy atomic positions */
- system->atoms[c].x[0] = strtod( &s_x[0], &endptr );
- system->atoms[c].x[1] = strtod( &s_y[0], &endptr );
- system->atoms[c].x[2] = strtod( &s_z[0], &endptr );
-
- /* atom name and type */
- strcpy( system->atoms[c].name, atom_name );
- Trim_Spaces( element );
- system->atoms[c].type = Get_Atom_Type( &(system->reaxprm), element );
-
- /* fprintf( stderr,
- "%d%8.3f%8.3f%8.3fq:%8.3f occ:%s temp:%s seg_id:%s element:%s\n",
- system->atoms[c].type,
- system->atoms[c].x[0], system->atoms[c].x[1], system->atoms[c].x[2],
- system->atoms[c].q, occupancy, temp_factor, seg_id, element ); */
- c++;
- }
- else if (!strncmp( tmp[0], "CRYST1", 6 ))
- {
- sscanf( s1, PDB_CRYST1_FORMAT,
- &descriptor[0],
- &s_a[0],
- &s_b[0],
- &s_c[0],
- &s_alpha[0],
- &s_beta[0],
- &s_gamma[0],
- &s_group[0],
- &s_zValue[0] );
-
- /* Compute full volume tensor from the angles */
- Init_Box_From_CRYST( atof(s_a), atof(s_b), atof(s_c),
- atof(s_alpha), atof(s_beta), atof(s_gamma),
- &(system->box) );
- }
-
- /* IMPORTANT: We do not check for the soundness of restrictions here.
- When atom2 is on atom1's restricted list, and there is a restriction on
- atom2, then atom1 has to be on atom2's restricted list, too. However,
- we do not check if this is the case in the input file,
- this is upto the user. */
- else if (!strncmp( tmp[0], "CONECT", 6 ))
- {
- /* error check */
- //fprintf(stderr, "CONECT: %d\n", c1 );
- Check_Input_Range( c1 - 2, 0, MAX_RESTRICT,
- "CONECT line exceeds max restrictions allowed.\n" );
-
- /* read bond restrictions */
- if ( is_Valid_Serial( workspace, pdb_serial = atoi(tmp[1]) ) )
- ratom = workspace->map_serials[ pdb_serial ];
-
- workspace->restricted[ ratom ] = c1 - 2;
- for ( i = 2; i < c1; ++i )
- {
- if ( is_Valid_Serial( workspace, pdb_serial = atoi(tmp[i]) ) )
- workspace->restricted_list[ (ratom * MAX_RESTRICT) + (i - 2) ] =
- workspace->map_serials[ pdb_serial ];
- }
-
- /* fprintf( stderr, "restriction on %d:", ratom );
- for( i = 0; i < workspace->restricted[ ratom ]; ++i )
- fprintf( stderr, " %d", workspace->restricted_list[ratom][i] );
- fprintf( stderr, "\n" ); */
- }
- }
-
- fclose(pdb);
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "pdb file read\n" );
-#endif
-
- return 1;
-}
-
-
-char Write_PDB( reax_system* system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list* bonds, output_controls *out_control )
-{
- int i, j, k, count;
- int connect[4];
- char temp[MAX_STR], name[10];
- real bo;
- real alpha, beta, gamma;
-
-
- /* open output pdb file */
- sprintf( temp, "%s%d.pdb", control->sim_name, data->step );
- out_control->pdb = fopen( temp, "w" );
-
-
- /* Writing Box information */
- /* Write full volume tensor from the angles (as soon as possible) TODO_SOON */
- gamma = acos( (system->box.box[0][0] * system->box.box[1][0] +
- system->box.box[0][1] * system->box.box[1][1] +
- system->box.box[0][2] * system->box.box[1][2]) /
- (system->box.box_norms[0] * system->box.box_norms[1]));
- beta = acos( (system->box.box[0][0] * system->box.box[2][0] +
- system->box.box[0][1] * system->box.box[2][1] +
- system->box.box[0][2] * system->box.box[2][2]) /
- (system->box.box_norms[0] * system->box.box_norms[2]));
- alpha = acos( (system->box.box[2][0] * system->box.box[1][0] +
- system->box.box[2][1] * system->box.box[1][1] +
- system->box.box[2][2] * system->box.box[1][2]) /
- (system->box.box_norms[2] * system->box.box_norms[1]));
-
- fprintf(out_control->pdb, PDB_CRYST1_FORMAT_O,
- "CRYST1",
- system->box.box_norms[0],
- system->box.box_norms[1],
- system->box.box_norms[2],
- RAD2DEG(alpha),
- RAD2DEG(beta),
- RAD2DEG(gamma),
- " ",
- 0);
- fprintf( out_control->log, "Box written\n" );
- fflush( out_control->log );
-
- /* Writing atom information */
- for (i = 0; i < system->N; i++)
- {
- strncpy( name, system->reaxprm.sbp[system->atoms[i].type].name, 2 );
- name[2] = '\0';
- fprintf( out_control->pdb, PDB_ATOM_FORMAT_O,
- "ATOM ",
- workspace->orig_id[i],
- name,
- ' ',
- "REX",
- ' ',
- 1,
- ' ',
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2],
- 1.0,
- 0.0,
- "0",
- name,
- " " );
- }
-
- fprintf( out_control->log, "ATOM written\n" );
- fflush( out_control->log );
-
- /* Writing connect information */
- for (i = 0; i < system->N; i++)
- {
- count = 0;
-
- for (j = Start_Index(i, bonds); j < End_Index(i, bonds); ++j)
- {
- bo = bonds->select.bond_list[j].bo_data.BO;
- if (bo > 0.3)
- {
- connect[count] = workspace->orig_id[bonds->select.bond_list[j].nbr];
- count++;
- }
- }
-
- fprintf( out_control->pdb, "%6s%6d", "CONECT", workspace->orig_id[i] );
- for ( k = 0; k < count; k++ )
- fprintf( out_control->pdb, "%6d", connect[k] );
- fprintf( out_control->pdb, "\n" );
- }
-
- fprintf( out_control->pdb, "END\n" );
-
- fclose( out_control->pdb );
-
- return 1;
-}
-
-
-char Read_BGF( char* bgf_file, reax_system* system, control_params *control,
- simulation_data *data, static_storage *workspace )
-{
- FILE *bgf;
- char **tokens;
- char *line, *backup;
- char descriptor[10], serial[10];
- char atom_name[10], res_name[10], res_seq[10];
- char s_x[12], s_y[12], s_z[12];
- char occupancy[10], temp_factor[10];
- char element[10], charge[10];
- char chain_id;
- char s_a[12], s_b[12], s_c[12], s_alpha[12], s_beta[12], s_gamma[12];
- char *endptr = NULL;
- int i, atom_cnt, token_cnt, bgf_serial, ratom = 0;
-
- /* open biograf file */
- if ( (bgf = fopen( bgf_file, "r" )) == NULL )
- {
- fprintf( stderr, "Error opening the bgf file!\n" );
- exit( FILE_NOT_FOUND_ERR );
- }
-
-
- /* allocate memory for tokenizing biograf file lines */
- line = (char*) malloc( sizeof(char) * MAX_LINE );
- backup = (char*) malloc( sizeof(char) * MAX_LINE );
- tokens = (char**) malloc( sizeof(char*) * MAX_TOKENS );
- for ( i = 0; i < MAX_TOKENS; i++ )
- tokens[i] = (char*) malloc( sizeof(char) * MAX_TOKEN_LEN );
-
-
- /* count number of atoms in the pdb file */
- system->N = 0;
- while ( !feof( bgf ) )
- {
- line[0] = 0;
- fgets( line, MAX_LINE, bgf );
-
- tokens[0][0] = 0;
- token_cnt = Tokenize( line, &tokens );
-
- if ( !strcmp( tokens[0], "ATOM" ) || !strcmp( tokens[0], "HETATM" ) )
- (system->N)++;
- }
- //fprintf( stderr, "system->N: %d\n", system->N );
- fclose( bgf );
-
-
- /* memory allocations for atoms, atom maps, bond restrictions */
- system->atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
-
- workspace->map_serials = (int*) calloc( MAX_ATOM_ID, sizeof(int) );
- for ( i = 0; i < MAX_ATOM_ID; ++i )
- workspace->map_serials[i] = -1;
-
- workspace->orig_id = (int*) calloc( system->N, sizeof(int) );
- workspace->restricted = (int*) calloc( system->N, sizeof(int) );
- workspace->restricted_list = (int*) calloc( system->N * MAX_RESTRICT, sizeof(int) );
- //for( i = 0; i < system->N; ++i )
- // workspace->restricted_list[i] = (int*) calloc( MAX_RESTRICT, sizeof(int) );
-
-
- /* start reading and processing pdb file */
- bgf = fopen( bgf_file, "r" );
- atom_cnt = 0;
- token_cnt = 0;
-
- while ( !feof( bgf ) )
- {
- /* clear previous input line */
- line[0] = 0;
- for ( i = 0; i < token_cnt; ++i )
- tokens[i][0] = 0;
-
- /* read new line and tokenize it */
- fgets( line, MAX_LINE, bgf );
- strncpy( backup, line, MAX_LINE - 1 );
- token_cnt = Tokenize( line, &tokens );
-
- /* process new line */
- if ( !strncmp(tokens[0], "ATOM", 4) || !strncmp(tokens[0], "HETATM", 6) )
- {
- if ( !strncmp(tokens[0], "ATOM", 4) )
- {
- strncpy( &descriptor[0], backup, 6 );
- descriptor[6] = 0;
- strncpy( &serial[0], backup + 7, 5 );
- serial[5] = 0;
- strncpy( &atom_name[0], backup + 13, 5 );
- atom_name[5] = 0;
- strncpy( &res_name[0], backup + 19, 3 );
- res_name[3] = 0;
- chain_id = backup[23];
- strncpy( &res_seq[0], backup + 25, 5 );
- res_seq[5] = 0;
- strncpy( &s_x[0], backup + 30, 10 );
- s_x[10] = 0;
- strncpy( &s_y[0], backup + 40, 10 );
- s_y[10] = 0;
- strncpy( &s_z[0], backup + 50, 10 );
- s_z[10] = 0;
- strncpy( &element[0], backup + 61, 5 );
- element[5] = 0;
- strncpy( &occupancy[0], backup + 66, 3 );
- occupancy[3] = 0;
- strncpy( &temp_factor[0], backup + 69, 2 );
- temp_factor[2] = 0;
- strncpy( &charge[0], backup + 72, 8 );
- charge[8] = 0;
- }
- else if ( !strncmp(tokens[0], "HETATM", 6) )
- {
- /* bgf hetatm:
- (7x,i5,1x,a5,1x,a3,1x,a1,1x,a5,3f10.5,1x,a5,i3,i2,1x,f8.5) */
- strncpy( &descriptor[0], backup, 6 );
- descriptor[6] = 0;
- strncpy( &serial[0], backup + 7, 5 );
- serial[5] = 0;
- strncpy( &atom_name[0], backup + 13, 5 );
- atom_name[5] = 0;
- strncpy( &res_name[0], backup + 19, 3 );
- res_name[3] = 0;
- chain_id = backup[23];
- strncpy( &res_seq[0], backup + 25, 5 );
- res_seq[5] = 0;
- strncpy( &s_x[0], backup + 30, 10 );
- s_x[10] = 0;
- strncpy( &s_y[0], backup + 40, 10 );
- s_y[10] = 0;
- strncpy( &s_z[0], backup + 50, 10 );
- s_z[10] = 0;
- strncpy( &element[0], backup + 61, 5 );
- element[5] = 0;
- strncpy( &occupancy[0], backup + 66, 3 );
- occupancy[3] = 0;
- strncpy( &temp_factor[0], backup + 69, 2 );
- temp_factor[2] = 0;
- strncpy( &charge[0], backup + 72, 8 );
- charge[8] = 0;
- }
-
-
- /* add to mapping */
- bgf_serial = strtod( &serial[0], &endptr );
- Check_Input_Range( bgf_serial, 0, MAX_ATOM_ID, "Invalid bgf serial" );
- workspace->map_serials[ bgf_serial ] = atom_cnt;
- workspace->orig_id[ atom_cnt ] = bgf_serial;
- // fprintf( stderr, "map %d --> %d\n", bgf_serial, atom_cnt );
-
-
- /* copy atomic positions */
- system->atoms[atom_cnt].x[0] = strtod( &s_x[0], &endptr );
- system->atoms[atom_cnt].x[1] = strtod( &s_y[0], &endptr );
- system->atoms[atom_cnt].x[2] = strtod( &s_z[0], &endptr );
-
-
- /* atom name and type */
- //BGF_FIX
- atom_name[4] = 0;
- //BGF_FIX
-
- strcpy( system->atoms[atom_cnt].name, atom_name );
- Trim_Spaces( element );
- system->atoms[atom_cnt].type =
- Get_Atom_Type( &(system->reaxprm), element );
-
- /* fprintf( stderr,
- "a:%3d(%1d) c:%10.5f%10.5f%10.5f q:%10.5f occ:%s temp:%s seg_id:%s element:%s\n",
- atom_cnt, system->atoms[ atom_cnt ].type,
- system->atoms[ atom_cnt ].x[0],
- system->atoms[ atom_cnt ].x[1], system->atoms[ atom_cnt ].x[2],
- system->atoms[ atom_cnt ].q, occupancy, temp_factor,
- seg_id, element ); */
-
- atom_cnt++;
- }
- else if (!strncmp( tokens[0], "CRYSTX", 6 ))
- {
- sscanf( backup, BGF_CRYSTX_FORMAT,
- &descriptor[0],
- &s_a[0],
- &s_b[0],
- &s_c[0],
- &s_alpha[0],
- &s_beta[0],
- &s_gamma[0] );
-
- /* Compute full volume tensor from the angles */
- Init_Box_From_CRYST( atof(s_a), atof(s_b), atof(s_c),
- atof(s_alpha), atof(s_beta), atof(s_gamma),
- &(system->box) );
- }
- else if (!strncmp( tokens[0], "CONECT", 6 ))
- {
- /* check number of restrictions */
- Check_Input_Range( token_cnt - 2, 0, MAX_RESTRICT,
- "CONECT line exceeds max restrictions allowed.\n" );
-
- /* read bond restrictions */
- if ( is_Valid_Serial( workspace, bgf_serial = atoi(tokens[1]) ) )
- ratom = workspace->map_serials[ bgf_serial ];
-
- workspace->restricted[ ratom ] = token_cnt - 2;
- for ( i = 2; i < token_cnt; ++i )
- if ( is_Valid_Serial( workspace, bgf_serial = atoi(tokens[i]) ) )
- workspace->restricted_list[ (ratom * MAX_RESTRICT) + (i - 2) ] =
- workspace->map_serials[ bgf_serial ];
-
- /* fprintf( stderr, "restriction on %d:", ratom );
- for( i = 0; i < workspace->restricted[ ratom ]; ++i )
- fprintf( stderr, " %d", workspace->restricted_list[ratom][i] );
- fprintf( stderr, "\n" ); */
- }
- }
-
- fclose( bgf );
-
-#if defined(DEBUG_FOCUS)
- fprintf( stderr, "bgf file read\n" );
-#endif
-
- return 1;
-}
diff --git a/PuReMD-GPU/src/print_utils.c b/PuReMD-GPU/src/print_utils.c
index 913ff617a23f9f395a300a0985dec7ad36c33fab..d0f0e1bad12720a793074fbbc1e4d194ebb8fd8e 100644
--- a/PuReMD-GPU/src/print_utils.c
+++ b/PuReMD-GPU/src/print_utils.c
@@ -19,9 +19,11 @@
----------------------------------------------------------------------*/
#include "print_utils.h"
+
+#include "geo_tools.h"
#include "list.h"
-#include "pdb_tools.h"
#include "system_props.h"
+#include "tool_box.h"
#include "vector.h"
@@ -374,18 +376,6 @@ void Init_Force_Test_Functions( )
#endif
-char *Get_Element( reax_system *system, int i )
-{
- return &( system->reaxprm.sbp[system->atoms[i].type].name[0] );
-}
-
-
-char *Get_Atom_Name( reax_system *system, int i )
-{
- return &(system->atoms[i].name[0]);
-}
-
-
/* near nbrs contain both i-j, j-i nbrhood info */
void Print_Near_Neighbors( reax_system *system, control_params *control,
static_storage *workspace, list **lists )
@@ -625,35 +615,48 @@ void Output_Results( reax_system *system, control_params *control,
data->E_vdW, data->E_Ele, data->E_Pol );
#endif
-#ifdef __PRINT_CPU_RESULTS__
+#ifndef HAVE_CUDA
t_elapsed = Get_Timing_Info( data->timing.total );
if ( data->step == data->prev_steps )
f_update = 1;
else f_update = out_control->energy_update_freq;
- fprintf( out_control->log, "%6d%10.2f%10.2f%10.2f%10.2f%10.2f%10.2f%10.2f\n",
+ fprintf( out_control->log, "%6d %10.2f %10.2f %10.2f %10.2f %10.2f %10.4f %10.4f %10.2f %10.4f %10.4f %10.4f %10.4f %10.4f %10.4f\n",
data->step, t_elapsed / f_update,
data->timing.nbrs / f_update,
data->timing.init_forces / f_update,
data->timing.bonded / f_update,
data->timing.nonb / f_update,
data->timing.QEq / f_update,
- (double)data->timing.matvecs / f_update );
+ data->timing.QEq_sort_mat_rows / f_update,
+ (double)data->timing.solver_iters / f_update,
+ data->timing.pre_comp / f_update,
+ data->timing.pre_app / f_update,
+ data->timing.solver_spmv / f_update,
+ data->timing.solver_vector_ops / f_update,
+ data->timing.solver_orthog / f_update,
+ data->timing.solver_tri_solve / f_update );
#else
t_elapsed = Get_Timing_Info( d_timing.total );
if ( data->step == data->prev_steps )
f_update = 1;
else f_update = out_control->energy_update_freq;
- fprintf( out_control->log, "%6d%10.2f%10.2f%10.2f%10.2f%10.2f%10.2f%10.2f\n",
+ fprintf( out_control->log, "%6d %10.2f %10.2f %10.2f %10.2f %10.2f %10.4f %10.4f %10.2f %10.4f %10.4f %10.4f %10.4f %10.4f %10.4f\n",
data->step, t_elapsed / f_update,
- d_timing.nbrs / f_update,
- d_timing.init_forces / f_update,
- d_timing.bonded / f_update,
- d_timing.nonb / f_update,
- d_timing.QEq / f_update,
- (double)d_timing.matvecs / f_update );
-
+ d_timing->timing.nbrs / f_update,
+ d_timing->timing.init_forces / f_update,
+ d_timing->timing.bonded / f_update,
+ d_timing->timing.nonb / f_update,
+ d_timing->timing.QEq / f_update,
+ d_timing->timing.QEq_sort_mat_rows / f_update,
+ (double)d_timing->timing.solver_iters / f_update,
+ d_timing->timing.pre_comp / f_update,
+ d_timing->timing.pre_app / f_update,
+ d_timing->timing.solver_spmv / f_update,
+ d_timing->timing.solver_vector_ops / f_update,
+ d_timing->timing.solver_orthog / f_update,
+ d_timing->timing.solver_tri_solve / f_update );
#endif
//fprintf (stderr, " total %10.5f \n", t_elapsed);
@@ -673,16 +676,32 @@ void Output_Results( reax_system *system, control_params *control,
data->timing.init_forces = 0;
data->timing.bonded = 0;
data->timing.nonb = 0;
- data->timing.QEq = 0;
- data->timing.matvecs = 0;
-
+ data->timing.QEq = ZERO;
+ data->timing.QEq_sort_mat_rows = ZERO;
+ data->timing.pre_comp = ZERO;
+ data->timing.pre_app = ZERO;
+ data->timing.solver_iters = 0;
+ data->timing.solver_spmv = ZERO;
+ data->timing.solver_vector_ops = ZERO;
+ data->timing.solver_orthog = ZERO;
+ data->timing.solver_tri_solve = ZERO;
+
+#ifdef HAVE_CUDA
d_timing.total = Get_Time( );
d_timing.nbrs = 0;
d_timing.init_forces = 0;
d_timing.bonded = 0;
d_timing.nonb = 0;
- d_timing.QEq = 0;
- d_timing.matvecs = 0;
+ d_timing->timing.QEq = ZERO;
+ d_timing->timing.QEq_sort_mat_rows = ZERO;
+ d_timing->timing.pre_comp = ZERO;
+ d_timing->timing.pre_app = ZERO;
+ d_timing->timing.solver_iters = 0;
+ d_timing->timing.solver_spmv = ZERO;
+ d_timing->timing.solver_vector_ops = ZERO;
+ d_timing->timing.solver_orthog = ZERO;
+ d_timing->timing.solver_tri_solve = ZERO;
+#endif
fflush( out_control->out );
fflush( out_control->pot );
@@ -716,16 +735,16 @@ void Output_Results( reax_system *system, control_params *control,
if ( out_control->write_steps > 0 &&
data->step % out_control->write_steps == 0 )
{
- // t_start = Get_Time( );
+ //t_start = Get_Time( );
out_control->append_traj_frame( system, control, data,
workspace, lists, out_control );
- //Write_PDB( system, control, data, workspace, *lists+BONDS, out_control );
- // t_elapsed = Get_Timing_Info( t_start );
- // fprintf(stdout, "append_frame took %.6f seconds\n", t_elapsed );
+ //Write_PDB( system, *lists+BONDS, data, control, workspace, out_control );
+ //t_elapsed = Get_Timing_Info( t_start );
+ //fprintf(stdout, "append_frame took %.6f seconds\n", t_elapsed );
}
- // fprintf( stderr, "output_results... done\n" );
+ //fprintf( stderr, "output_results... done\n" );
}
@@ -759,23 +778,46 @@ void Print_Linear_System( reax_system *system, control_params *control,
sprintf( fname, "%s.H%d.out", control->sim_name, step );
out = fopen( fname, "w" );
- H = &workspace->H;
+ H = workspace->H;
for ( i = 0; i < system->N; ++i )
{
for ( j = H->start[i]; j < H->start[i + 1] - 1; ++j )
{
fprintf( out, "%6d%6d %24.15e\n",
- workspace->orig_id[i], workspace->orig_id[H->entries[j].j],
- H->entries[j].val );
+ workspace->orig_id[i], workspace->orig_id[H->j[j]],
+ H->val[j] );
fprintf( out, "%6d%6d %24.15e\n",
- workspace->orig_id[H->entries[j].j], workspace->orig_id[i],
- H->entries[j].val );
+ workspace->orig_id[H->j[j]], workspace->orig_id[i],
+ H->val[j] );
}
// the diagonal entry
fprintf( out, "%6d%6d %24.15e\n",
- workspace->orig_id[i], workspace->orig_id[i], H->entries[j].val );
+ workspace->orig_id[i], workspace->orig_id[i], H->val[j] );
+ }
+
+ fclose( out );
+
+ sprintf( fname, "%s.H_sp%d.out", control->sim_name, step );
+ out = fopen( fname, "w" );
+ H = workspace->H_sp;
+
+ for ( i = 0; i < system->N; ++i )
+ {
+ for ( j = H->start[i]; j < H->start[i + 1] - 1; ++j )
+ {
+ fprintf( out, "%6d%6d %24.15e\n",
+ workspace->orig_id[i], workspace->orig_id[H->j[j]],
+ H->val[j] );
+
+ fprintf( out, "%6d%6d %24.15e\n",
+ workspace->orig_id[H->j[j]], workspace->orig_id[i],
+ H->val[j] );
+ }
+ // the diagonal entry
+ fprintf( out, "%6d%6d %24.15e\n",
+ workspace->orig_id[i], workspace->orig_id[i], H->val[j] );
}
fclose( out );
@@ -834,11 +876,11 @@ void Print_Sparse_Matrix( sparse_matrix *A )
{
int i, j;
- for ( i = 0; i < 10; ++i )
+ for ( i = 0; i < A->n; ++i )
{
fprintf( stderr, "i:%d j(val):", i );
for ( j = A->start[i]; j < A->start[i + 1]; ++j )
- fprintf( stderr, "%d(%.4f) ", A->entries[j].j, A->entries[j].val );
+ fprintf( stderr, "%d(%.4f) ", A->j[j], A->val[j] );
fprintf( stderr, "\n" );
}
}
@@ -850,8 +892,14 @@ void Print_Sparse_Matrix2( sparse_matrix *A, char *fname )
FILE *f = fopen( fname, "w" );
for ( i = 0; i < A->n; ++i )
+ {
for ( j = A->start[i]; j < A->start[i + 1]; ++j )
- fprintf( f, "%d%d %.15e\n", A->entries[j].j, i, A->entries[j].val );
+ {
+ //fprintf( f, "%d%d %.15e\n", A->entries[j].j, i, A->entries[j].val );
+ //Convert 0-based to 1-based (for Matlab)
+ fprintf( f, "%6d%6d %24.15e\n", i+1, A->j[j]+1, A->val[j] );
+ }
+ }
fclose(f);
}
diff --git a/PuReMD-GPU/src/print_utils.h b/PuReMD-GPU/src/print_utils.h
index 5f479bdc99fa30c518f69b5a23fa88b19af1a306..46d08516e00b002792d507b5effd7bd1ee5d551d 100644
--- a/PuReMD-GPU/src/print_utils.h
+++ b/PuReMD-GPU/src/print_utils.h
@@ -23,28 +23,25 @@
#include "mytypes.h"
+
typedef void (*print_interaction)(reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls*);
-extern print_interaction Print_Interactions[NO_OF_INTERACTIONS];
+ static_storage*, list**, output_controls*);
-char *Get_Element( reax_system*, int );
+extern print_interaction Print_Interactions[NO_OF_INTERACTIONS];
-char *Get_Atom_Name( reax_system*, int );
-void Print_Near_Neighbors( reax_system*, control_params*, static_storage*,
- list** );
+void Print_Near_Neighbors( reax_system*, control_params*, static_storage*, list** );
-void Print_Far_Neighbors( reax_system*, control_params*, static_storage*,
- list** );
+void Print_Far_Neighbors( reax_system*, control_params*, static_storage*, list** );
void Print_Total_Force( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Output_Results( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Print_Bond_Orders( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Print_Linear_System( reax_system*, control_params*, static_storage*, int );
@@ -61,23 +58,23 @@ void Print_Bond_List2( reax_system*, list*, char* );
#ifdef TEST_FORCES
void Dummy_Printer( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Print_Bond_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Print_LonePair_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
-void Print_OverUnderCoor_Forces(reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls*);
+ static_storage*, list**, output_controls* );
+void Print_OverUnderCoor_Forces(reax_system*, control_params*,
+ simulation_data*, static_storage*, list**, output_controls*);
void Print_Three_Body_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
-void Print_Hydrogen_Bond_Forces(reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls*);
+ static_storage*, list**, output_controls* );
+void Print_Hydrogen_Bond_Forces(reax_system*, control_params*,
+ simulation_data*, static_storage*, list**, output_controls*);
void Print_Four_Body_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Print_vdW_Coulomb_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Compare_Total_Forces( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
void Init_Force_Test_Functions( );
#endif
diff --git a/PuReMD-GPU/src/qeq.c b/PuReMD-GPU/src/qeq.c
new file mode 100644
index 0000000000000000000000000000000000000000..a319e89b872d34e3e3da7ca9121a8a8b63a4d83c
--- /dev/null
+++ b/PuReMD-GPU/src/qeq.c
@@ -0,0 +1,1667 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#include "qeq.h"
+
+#include "allocate.h"
+#include "index_utils.h"
+#include "list.h"
+#include "lin_alg.h"
+#include "print_utils.h"
+#include "tool_box.h"
+#if defined(HAVE_SUPERLU_MT)
+#include "slu_mt_ddefs.h"
+#endif
+
+
+#if defined(TEST_MAT)
+static sparse_matrix * create_test_mat( void )
+{
+ unsigned int i, n;
+ sparse_matrix *H_test;
+
+ if ( Allocate_Matrix( &H_test, 3, 6 ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for test matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ //3x3, SPD, store lower half
+ i = 0;
+ n = 0;
+ H_test->start[n] = i;
+ H_test->j[i] = 0;
+ H_test->val[i] = 4.;
+ ++i;
+ ++n;
+ H_test->start[n] = i;
+ H_test->j[i] = 0;
+ H_test->val[i] = 12.;
+ ++i;
+ H_test->j[i] = 1;
+ H_test->val[i] = 37.;
+ ++i;
+ ++n;
+ H_test->start[n] = i;
+ H_test->j[i] = 0;
+ H_test->val[i] = -16.;
+ ++i;
+ H_test->j[i] = 1;
+ H_test->val[i] = -43.;
+ ++i;
+ H_test->j[i] = 2;
+ H_test->val[i] = 98.;
+ ++i;
+ ++n;
+ H_test->start[n] = i;
+
+ return H_test;
+}
+#endif
+
+
+/* Routine used with qsort for sorting nonzeros within a sparse matrix row
+ *
+ * v1/v2: pointers to column indices of nonzeros within a row (unsigned int)
+ */
+static int compare_matrix_entry(const void *v1, const void *v2)
+{
+ /* larger element has larger column index */
+ return *(unsigned int *)v1 - *(unsigned int *)v2;
+}
+
+
+/* Routine used for sorting nonzeros within a sparse matrix row;
+ * internally, a combination of qsort and manual sorting is utilized
+ * (parallel calls to qsort when multithreading, rows mapped to threads)
+ *
+ * A: sparse matrix for which to sort nonzeros within a row, stored in CSR format
+ */
+static void Sort_Matrix_Rows( sparse_matrix * const A )
+{
+ unsigned int i, j, k, si, ei, *temp_j;
+ real *temp_val;
+
+ #pragma omp parallel default(none) private(i, j, k, si, ei, temp_j, temp_val) shared(stderr)
+ {
+ if ( ( temp_j = (unsigned int*) malloc( A->n * sizeof(unsigned int)) ) == NULL
+ || ( temp_val = (real*) malloc( A->n * sizeof(real)) ) == NULL )
+ {
+ fprintf( stderr, "Not enough space for matrix row sort. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* sort each row of A using column indices */
+ #pragma omp for schedule(guided)
+ for ( i = 0; i < A->n; ++i )
+ {
+ si = A->start[i];
+ ei = A->start[i + 1];
+ memcpy( temp_j, A->j + si, sizeof(unsigned int) * (ei - si) );
+ memcpy( temp_val, A->val + si, sizeof(real) * (ei - si) );
+
+ //TODO: consider implementing single custom one-pass sort instead of using qsort + manual sort
+ /* polymorphic sort in standard C library using column indices */
+ qsort( temp_j, ei - si, sizeof(unsigned int), compare_matrix_entry );
+
+ /* manually sort vals */
+ for ( j = 0; j < (ei - si); ++j )
+ {
+ for ( k = 0; k < (ei - si); ++k )
+ {
+ if ( A->j[si + j] == temp_j[k] )
+ {
+ A->val[si + k] = temp_val[j];
+ break;
+ }
+
+ }
+ }
+
+ /* copy sorted column indices */
+ memcpy( A->j + si, temp_j, sizeof(unsigned int) * (ei - si) );
+ }
+
+ free( temp_val );
+ free( temp_j );
+ }
+}
+
+
+static void Calculate_Droptol( const sparse_matrix * const A, real * const droptol,
+ const real dtol )
+{
+ int i, j, k;
+ real val;
+#ifdef _OPENMP
+ static real *droptol_local;
+ unsigned int tid;
+#endif
+
+ #pragma omp parallel default(none) private(i, j, k, val, tid), shared(droptol_local, stderr)
+ {
+#ifdef _OPENMP
+ tid = omp_get_thread_num();
+
+ #pragma omp master
+ {
+ /* keep b_local for program duration to avoid allocate/free
+ * overhead per Sparse_MatVec call*/
+ if ( droptol_local == NULL )
+ {
+ if ( (droptol_local = (real*) malloc( omp_get_num_threads() * A->n * sizeof(real))) == NULL )
+ {
+ fprintf( stderr, "Not enough space for droptol. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ }
+
+ #pragma omp barrier
+#endif
+
+ /* init droptol to 0 */
+ for ( i = 0; i < A->n; ++i )
+ {
+#ifdef _OPENMP
+ droptol_local[tid * A->n + i] = 0.0;
+#else
+ droptol[i] = 0.0;
+#endif
+ }
+
+ #pragma omp barrier
+
+ /* calculate sqaure of the norm of each row */
+ #pragma omp for schedule(static)
+ for ( i = 0; i < A->n; ++i )
+ {
+ for ( k = A->start[i]; k < A->start[i + 1] - 1; ++k )
+ {
+ j = A->j[k];
+ val = A->val[k];
+
+#ifdef _OPENMP
+ droptol_local[tid * A->n + i] += val * val;
+ droptol_local[tid * A->n + j] += val * val;
+#else
+ droptol[i] += val * val;
+ droptol[j] += val * val;
+#endif
+ }
+
+ val = A->val[k]; // diagonal entry
+#ifdef _OPENMP
+ droptol_local[tid * A->n + i] += val * val;
+#else
+ droptol[i] += val * val;
+#endif
+ }
+
+ #pragma omp barrier
+
+#ifdef _OPENMP
+ #pragma omp for schedule(static)
+ for ( i = 0; i < A->n; ++i )
+ {
+ droptol[i] = 0.0;
+ for ( k = 0; k < omp_get_num_threads(); ++k )
+ {
+ droptol[i] += droptol_local[k * A->n + i];
+ }
+ }
+#endif
+
+ #pragma omp barrier
+
+ /* calculate local droptol for each row */
+ //fprintf( stderr, "droptol: " );
+ #pragma omp for schedule(static)
+ for ( i = 0; i < A->n; ++i )
+ {
+ //fprintf( stderr, "%f-->", droptol[i] );
+ droptol[i] = SQRT( droptol[i] ) * dtol;
+ //fprintf( stderr, "%f ", droptol[i] );
+ }
+ //fprintf( stderr, "\n" );
+ }
+}
+
+
+static int Estimate_LU_Fill( const sparse_matrix * const A, const real * const droptol )
+{
+ int i, j, pj;
+ int fillin;
+ real val;
+
+ fillin = 0;
+
+ #pragma omp parallel for schedule(static) \
+ default(none) private(i, j, pj, val) reduction(+: fillin)
+ for ( i = 0; i < A->n; ++i )
+ {
+ for ( pj = A->start[i]; pj < A->start[i + 1] - 1; ++pj )
+ {
+ j = A->j[pj];
+ val = A->val[pj];
+
+ if ( FABS(val) > droptol[i] )
+ {
+ ++fillin;
+ }
+ }
+ }
+
+ return fillin + A->n;
+}
+
+
+#if defined(HAVE_SUPERLU_MT)
+static real SuperLU_Factorize( const sparse_matrix * const A,
+ sparse_matrix * const L, sparse_matrix * const U )
+{
+ unsigned int i, pj, count, *Ltop, *Utop, r;
+ sparse_matrix *A_t;
+ SuperMatrix A_S, AC_S, L_S, U_S;
+ NCformat *A_S_store;
+ SCPformat *L_S_store;
+ NCPformat *U_S_store;
+ superlumt_options_t superlumt_options;
+ pxgstrf_shared_t pxgstrf_shared;
+ pdgstrf_threadarg_t *pdgstrf_threadarg;
+ int_t nprocs;
+ fact_t fact;
+ trans_t trans;
+ yes_no_t refact, usepr;
+ real u, drop_tol;
+ real *a, *at;
+ int_t *asub, *atsub, *xa, *xat;
+ int_t *perm_c; /* column permutation vector */
+ int_t *perm_r; /* row permutations from partial pivoting */
+ void *work;
+ int_t info, lwork;
+ int_t permc_spec, panel_size, relax;
+ Gstat_t Gstat;
+ flops_t flopcnt;
+
+ /* Default parameters to control factorization. */
+#ifdef _OPENMP
+ //TODO: set as global parameter and use
+ #pragma omp parallel \
+ default(none) shared(nprocs)
+ {
+ #pragma omp master
+ {
+ /* SuperLU_MT spawns threads internally, so set and pass parameter */
+ nprocs = omp_get_num_threads();
+ }
+ }
+#else
+ nprocs = 1;
+#endif
+
+// fact = EQUILIBRATE; /* equilibrate A (i.e., scale rows & cols to have unit norm), then factorize */
+ fact = DOFACT; /* factor from scratch */
+ trans = NOTRANS;
+ refact = NO; /* first time factorization */
+ //TODO: add to control file and use the value there to set these
+ panel_size = sp_ienv(1); /* # consec. cols treated as unit task */
+ relax = sp_ienv(2); /* # cols grouped as relaxed supernode */
+ u = 1.0; /* diagonal pivoting threshold */
+ usepr = NO;
+ drop_tol = 0.0;
+ work = NULL;
+ lwork = 0;
+
+//#if defined(DEBUG)
+ fprintf( stderr, "nprocs = %d\n", nprocs );
+ fprintf( stderr, "Panel size = %d\n", panel_size );
+ fprintf( stderr, "Relax = %d\n", relax );
+//#endif
+
+ if ( !(perm_r = intMalloc(A->n)) )
+ {
+ SUPERLU_ABORT("Malloc fails for perm_r[].");
+ }
+ if ( !(perm_c = intMalloc(A->n)) )
+ {
+ SUPERLU_ABORT("Malloc fails for perm_c[].");
+ }
+ if ( !(superlumt_options.etree = intMalloc(A->n)) )
+ {
+ SUPERLU_ABORT("Malloc fails for etree[].");
+ }
+ if ( !(superlumt_options.colcnt_h = intMalloc(A->n)) )
+ {
+ SUPERLU_ABORT("Malloc fails for colcnt_h[].");
+ }
+ if ( !(superlumt_options.part_super_h = intMalloc(A->n)) )
+ {
+ SUPERLU_ABORT("Malloc fails for part_super__h[].");
+ }
+ if ( ( (a = (real*) malloc( (2 * A->start[A->n] - A->n) * sizeof(real))) == NULL )
+ || ( (asub = (int_t*) malloc( (2 * A->start[A->n] - A->n) * sizeof(int_t))) == NULL )
+ || ( (xa = (int_t*) malloc( (A->n + 1) * sizeof(int_t))) == NULL )
+ || ( (Ltop = (unsigned int*) malloc( (A->n + 1) * sizeof(unsigned int))) == NULL )
+ || ( (Utop = (unsigned int*) malloc( (A->n + 1) * sizeof(unsigned int))) == NULL ) )
+ {
+ fprintf( stderr, "Not enough space for SuperLU factorization. Terminating...\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ if ( Allocate_Matrix( &A_t, A->n, A->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ /* Set up the sparse matrix data structure for A. */
+ Transpose( A, A_t );
+
+ count = 0;
+ for ( i = 0; i < A->n; ++i )
+ {
+ xa[i] = count;
+ for ( pj = A->start[i]; pj < A->start[i + 1]; ++pj )
+ {
+ a[count] = A->entries[pj].val;
+ asub[count] = A->entries[pj].j;
+ ++count;
+ }
+ for ( pj = A_t->start[i] + 1; pj < A_t->start[i + 1]; ++pj )
+ {
+ a[count] = A_t->entries[pj].val;
+ asub[count] = A_t->entries[pj].j;
+ ++count;
+ }
+ }
+ xa[i] = count;
+
+ dCompRow_to_CompCol( A->n, A->n, 2 * A->start[A->n] - A->n, a, asub, xa,
+ &at, &atsub, &xat );
+
+ for ( i = 0; i < (2 * A->start[A->n] - A->n); ++i )
+ fprintf( stderr, "%6d", asub[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i < (2 * A->start[A->n] - A->n); ++i )
+ fprintf( stderr, "%6.1f", a[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i <= A->n; ++i )
+ fprintf( stderr, "%6d", xa[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i < (2 * A->start[A->n] - A->n); ++i )
+ fprintf( stderr, "%6d", atsub[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i < (2 * A->start[A->n] - A->n); ++i )
+ fprintf( stderr, "%6.1f", at[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i <= A->n; ++i )
+ fprintf( stderr, "%6d", xat[i] );
+ fprintf( stderr, "\n" );
+
+ A_S.Stype = SLU_NC; /* column-wise, no supernode */
+ A_S.Dtype = SLU_D; /* double-precision */
+ A_S.Mtype = SLU_GE; /* full (general) matrix -- required for parallel factorization */
+ A_S.nrow = A->n;
+ A_S.ncol = A->n;
+ A_S.Store = (void *) SUPERLU_MALLOC( sizeof(NCformat) );
+ A_S_store = (NCformat *) A_S.Store;
+ A_S_store->nnz = 2 * A->start[A->n] - A->n;
+ A_S_store->nzval = at;
+ A_S_store->rowind = atsub;
+ A_S_store->colptr = xat;
+
+ /* ------------------------------------------------------------
+ Allocate storage and initialize statistics variables.
+ ------------------------------------------------------------*/
+ StatAlloc( A->n, nprocs, panel_size, relax, &Gstat );
+ StatInit( A->n, nprocs, &Gstat );
+
+ /* ------------------------------------------------------------
+ Get column permutation vector perm_c[], according to permc_spec:
+ permc_spec = 0: natural ordering
+ permc_spec = 1: minimum degree ordering on structure of A'*A
+ permc_spec = 2: minimum degree ordering on structure of A'+A
+ permc_spec = 3: approximate minimum degree for unsymmetric matrices
+ ------------------------------------------------------------*/
+ permc_spec = 0;
+ get_perm_c( permc_spec, &A_S, perm_c );
+
+ /* ------------------------------------------------------------
+ Initialize the option structure superlumt_options using the
+ user-input parameters;
+ Apply perm_c to the columns of original A to form AC.
+ ------------------------------------------------------------*/
+ pdgstrf_init( nprocs, fact, trans, refact, panel_size, relax,
+ u, usepr, drop_tol, perm_c, perm_r,
+ work, lwork, &A_S, &AC_S, &superlumt_options, &Gstat );
+
+ for ( i = 0; i < ((NCPformat*)AC_S.Store)->nnz; ++i )
+ fprintf( stderr, "%6.1f", ((real*)(((NCPformat*)AC_S.Store)->nzval))[i] );
+ fprintf( stderr, "\n" );
+
+ /* ------------------------------------------------------------
+ Compute the LU factorization of A.
+ The following routine will create nprocs threads.
+ ------------------------------------------------------------*/
+ pdgstrf( &superlumt_options, &AC_S, perm_r, &L_S, &U_S, &Gstat, &info );
+
+ fprintf( stderr, "INFO: %d\n", info );
+
+ flopcnt = 0;
+ for (i = 0; i < nprocs; ++i)
+ {
+ flopcnt += Gstat.procstat[i].fcops;
+ }
+ Gstat.ops[FACT] = flopcnt;
+
+//#if defined(DEBUG)
+ printf("\n** Result of sparse LU **\n");
+ L_S_store = (SCPformat *) L_S.Store;
+ U_S_store = (NCPformat *) U_S.Store;
+ printf( "No of nonzeros in factor L = " IFMT "\n", L_S_store->nnz );
+ printf( "No of nonzeros in factor U = " IFMT "\n", U_S_store->nnz );
+ fflush( stdout );
+//#endif
+
+ /* convert L and R from SuperLU formats to CSR */
+ memset( Ltop, 0, (A->n + 1) * sizeof(int) );
+ memset( Utop, 0, (A->n + 1) * sizeof(int) );
+ memset( L->start, 0, (A->n + 1) * sizeof(int) );
+ memset( U->start, 0, (A->n + 1) * sizeof(int) );
+
+ for ( i = 0; i < 2 * L_S_store->nnz; ++i )
+ fprintf( stderr, "%6.1f", ((real*)(L_S_store->nzval))[i] );
+ fprintf( stderr, "\n" );
+ for ( i = 0; i < 2 * U_S_store->nnz; ++i )
+ fprintf( stderr, "%6.1f", ((real*)(U_S_store->nzval))[i] );
+ fprintf( stderr, "\n" );
+
+ printf( "No of supernodes in factor L = " IFMT "\n", L_S_store->nsuper );
+ for ( i = 0; i < A->n; ++i )
+ {
+ fprintf( stderr, "nzval_col_beg[%5d] = %d\n", i, L_S_store->nzval_colbeg[i] );
+ fprintf( stderr, "nzval_col_end[%5d] = %d\n", i, L_S_store->nzval_colend[i] );
+ //TODO: correct for SCPformat for L?
+ //for( pj = L_S_store->rowind_colbeg[i]; pj < L_S_store->rowind_colend[i]; ++pj )
+// for( pj = 0; pj < L_S_store->rowind_colend[i] - L_S_store->rowind_colbeg[i]; ++pj )
+// {
+// ++Ltop[L_S_store->rowind[L_S_store->rowind_colbeg[i] + pj] + 1];
+// }
+ fprintf( stderr, "col_beg[%5d] = %d\n", i, U_S_store->colbeg[i] );
+ fprintf( stderr, "col_end[%5d] = %d\n", i, U_S_store->colend[i] );
+ for ( pj = U_S_store->colbeg[i]; pj < U_S_store->colend[i]; ++pj )
+ {
+ ++Utop[U_S_store->rowind[pj] + 1];
+ fprintf( stderr, "Utop[%5d] = %d\n", U_S_store->rowind[pj] + 1, Utop[U_S_store->rowind[pj] + 1] );
+ }
+ }
+ for ( i = 1; i <= A->n; ++i )
+ {
+// Ltop[i] = L->start[i] = Ltop[i] + Ltop[i - 1];
+ Utop[i] = U->start[i] = Utop[i] + Utop[i - 1];
+// fprintf( stderr, "Utop[%5d] = %d\n", i, Utop[i] );
+// fprintf( stderr, "U->start[%5d] = %d\n", i, U->start[i] );
+ }
+ for ( i = 0; i < A->n; ++i )
+ {
+// for( pj = 0; pj < L_S_store->nzval_colend[i] - L_S_store->nzval_colbeg[i]; ++pj )
+// {
+// r = L_S_store->rowind[L_S_store->rowind_colbeg[i] + pj];
+// L->entries[Ltop[r]].j = r;
+// L->entries[Ltop[r]].val = ((real*)L_S_store->nzval)[L_S_store->nzval_colbeg[i] + pj];
+// ++Ltop[r];
+// }
+ for ( pj = U_S_store->colbeg[i]; pj < U_S_store->colend[i]; ++pj )
+ {
+ r = U_S_store->rowind[pj];
+ U->entries[Utop[r]].j = i;
+ U->entries[Utop[r]].val = ((real*)U_S_store->nzval)[pj];
+ ++Utop[r];
+ }
+ }
+
+ /* ------------------------------------------------------------
+ Deallocate storage after factorization.
+ ------------------------------------------------------------*/
+ pxgstrf_finalize( &superlumt_options, &AC_S );
+ Deallocate_Matrix( A_t );
+ free( xa );
+ free( asub );
+ free( a );
+ SUPERLU_FREE( perm_r );
+ SUPERLU_FREE( perm_c );
+ SUPERLU_FREE( ((NCformat *)A_S.Store)->rowind );
+ SUPERLU_FREE( ((NCformat *)A_S.Store)->colptr );
+ SUPERLU_FREE( ((NCformat *)A_S.Store)->nzval );
+ SUPERLU_FREE( A_S.Store );
+ if ( lwork == 0 )
+ {
+ Destroy_SuperNode_SCP(&L_S);
+ Destroy_CompCol_NCP(&U_S);
+ }
+ else if ( lwork > 0 )
+ {
+ SUPERLU_FREE(work);
+ }
+ StatFree(&Gstat);
+
+ free( Utop );
+ free( Ltop );
+
+ //TODO: return iters
+ return 0.;
+}
+#endif
+
+
+/* Diagonal (Jacobi) preconditioner computation */
+static real diag_pre_comp( const reax_system * const system, real * const Hdia_inv )
+{
+ unsigned int i;
+ real start;
+
+ start = Get_Time( );
+
+ #pragma omp parallel for schedule(static) \
+ default(none) private(i)
+ for ( i = 0; i < system->N; ++i )
+ {
+ Hdia_inv[i] = 1.0 / system->reaxprm.sbp[system->atoms[i].type].eta;
+ }
+
+ return Get_Timing_Info( start );
+}
+
+
+/* Incomplete Cholesky factorization with dual thresholding */
+static real ICHOLT( const sparse_matrix * const A, const real * const droptol,
+ sparse_matrix * const L, sparse_matrix * const U )
+{
+ int *tmp_j;
+ real *tmp_val;
+ int i, j, pj, k1, k2, tmptop, Ltop;
+ real val, start;
+ int *Utop;
+
+ start = Get_Time( );
+
+ if ( ( Utop = (int*) malloc((A->n + 1) * sizeof(int)) ) == NULL ||
+ ( tmp_j = (int*) malloc(A->n * sizeof(int)) ) == NULL ||
+ ( tmp_val = (real*) malloc(A->n * sizeof(real)) ) == NULL )
+ {
+ fprintf( stderr, "not enough memory for ICHOLT preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ // clear variables
+ Ltop = 0;
+ tmptop = 0;
+ memset( L->start, 0, (A->n + 1) * sizeof(unsigned int) );
+ memset( U->start, 0, (A->n + 1) * sizeof(unsigned int) );
+ memset( Utop, 0, A->n * sizeof(unsigned int) );
+
+ //fprintf( stderr, "n: %d\n", A->n );
+ for ( i = 0; i < A->n; ++i )
+ {
+ L->start[i] = Ltop;
+ tmptop = 0;
+
+ for ( pj = A->start[i]; pj < A->start[i + 1] - 1; ++pj )
+ {
+ j = A->j[pj];
+ val = A->val[pj];
+ //fprintf( stderr, "i: %d, j: %d", i, j );
+
+ if ( FABS(val) > droptol[i] )
+ {
+ k1 = 0;
+ k2 = L->start[j];
+ while ( k1 < tmptop && k2 < L->start[j + 1] )
+ {
+ if ( tmp_j[k1] < L->j[k2] )
+ {
+ ++k1;
+ }
+ else if ( tmp_j[k1] > L->j[k2] )
+ {
+ ++k2;
+ }
+ else
+ {
+ val -= (tmp_val[k1++] * L->val[k2++]);
+ }
+ }
+
+ // L matrix is lower triangular,
+ // so right before the start of next row comes jth diagonal
+ val /= L->val[L->start[j + 1] - 1];
+
+ tmp_j[tmptop] = j;
+ tmp_val[tmptop] = val;
+ ++tmptop;
+ }
+ //fprintf( stderr, " -- done\n" );
+ }
+
+ // sanity check
+ if ( A->j[pj] != i )
+ {
+ fprintf( stderr, "i=%d, badly built A matrix!\n", i );
+ exit( NUMERIC_BREAKDOWN );
+ }
+
+ // compute the ith diagonal in L
+ val = A->val[pj];
+ for ( k1 = 0; k1 < tmptop; ++k1 )
+ {
+ val -= (tmp_val[k1] * tmp_val[k1]);
+ }
+
+ tmp_j[tmptop] = i;
+ tmp_val[tmptop] = SQRT(val);
+
+ // apply the dropping rule once again
+ //fprintf( stderr, "row%d: tmptop: %d\n", i, tmptop );
+ //for( k1 = 0; k1<= tmptop; ++k1 )
+ // fprintf( stderr, "%d(%f) ", tmp[k1].j, tmp[k1].val );
+ //fprintf( stderr, "\n" );
+ //fprintf( stderr, "row(%d): droptol=%.4f\n", i+1, droptol[i] );
+ for ( k1 = 0; k1 < tmptop; ++k1 )
+ {
+ if ( FABS(tmp_val[k1]) > droptol[i] / tmp_val[tmptop] )
+ {
+ L->j[Ltop] = tmp_j[k1];
+ L->val[Ltop] = tmp_val[k1];
+ U->start[tmp_j[k1] + 1]++;
+ ++Ltop;
+ //fprintf( stderr, "%d(%.4f) ", tmp[k1].j+1, tmp[k1].val );
+ }
+ }
+ // keep the diagonal in any case
+ L->j[Ltop] = tmp_j[k1];
+ L->val[Ltop] = tmp_val[k1];
+ ++Ltop;
+ //fprintf( stderr, "%d(%.4f)\n", tmp[k1].j+1, tmp[k1].val );
+ }
+
+ L->start[i] = Ltop;
+// fprintf( stderr, "nnz(L): %d, max: %d\n", Ltop, L->n * 50 );
+
+ /* U = L^T (Cholesky factorization) */
+ Transpose( L, U );
+// for ( i = 1; i <= U->n; ++i )
+// {
+// Utop[i] = U->start[i] = U->start[i] + U->start[i - 1] + 1;
+// }
+// for ( i = 0; i < L->n; ++i )
+// {
+// for ( pj = L->start[i]; pj < L->start[i + 1]; ++pj )
+// {
+// j = L->j[pj];
+// U->j[Utop[j]] = i;
+// U->val[Utop[j]] = L->val[pj];
+// Utop[j]++;
+// }
+// }
+
+// fprintf( stderr, "nnz(U): %d, max: %d\n", Utop[U->n], U->n * 50 );
+
+ free( tmp_val );
+ free( tmp_j );
+ free( Utop );
+
+ return Get_Timing_Info( start );
+}
+
+
+/* Fine-grained (parallel) incomplete Cholesky factorization
+ *
+ * Reference:
+ * Edmond Chow and Aftab Patel
+ * Fine-Grained Parallel Incomplete LU Factorization
+ * SIAM J. Sci. Comp. */
+static real ICHOL_PAR( const sparse_matrix * const A, const unsigned int sweeps,
+ sparse_matrix * const U_t, sparse_matrix * const U )
+{
+ unsigned int i, j, k, pj, x = 0, y = 0, ei_x, ei_y;
+ real *D, *D_inv, sum, start;
+ sparse_matrix *DAD;
+ int *Utop;
+
+ start = Get_Time( );
+
+ if ( Allocate_Matrix( DAD, A->n, A->m ) == FAILURE ||
+ ( D = (real*) malloc(A->n * sizeof(real)) ) == NULL ||
+ ( D_inv = (real*) malloc(A->n * sizeof(real)) ) == NULL ||
+ ( Utop = (int*) malloc((A->n + 1) * sizeof(int)) ) == NULL )
+ {
+ fprintf( stderr, "not enough memory for ICHOL_PAR preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(D_inv, D) private(i)
+ for ( i = 0; i < A->n; ++i )
+ {
+ D_inv[i] = SQRT( A->val[A->start[i + 1] - 1] );
+ D[i] = 1. / D_inv[i];
+ }
+
+ memset( U->start, 0, sizeof(unsigned int) * (A->n + 1) );
+ memset( Utop, 0, sizeof(unsigned int) * (A->n + 1) );
+
+ /* to get convergence, A must have unit diagonal, so apply
+ * transformation DAD, where D = D(1./sqrt(D(A))) */
+ memcpy( DAD->start, A->start, sizeof(int) * (A->n + 1) );
+ #pragma omp parallel for schedule(guided) \
+ default(none) shared(DAD, D_inv, D) private(i, pj)
+ for ( i = 0; i < A->n; ++i )
+ {
+ /* non-diagonals */
+ for ( pj = A->start[i]; pj < A->start[i + 1] - 1; ++pj )
+ {
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = A->val[pj] * D[i] * D[A->j[pj]];
+ }
+ /* diagonal */
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = 1.;
+ }
+
+ /* initial guesses for U^T,
+ * assume: A and DAD symmetric and stored lower triangular */
+ memcpy( U_t->start, DAD->start, sizeof(int) * (DAD->n + 1) );
+ memcpy( U_t->j, DAD->j, sizeof(int) * (DAD->m) );
+ memcpy( U_t->val, DAD->val, sizeof(real) * (DAD->m) );
+
+ for ( i = 0; i < sweeps; ++i )
+ {
+ /* for each nonzero */
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, stderr) private(sum, ei_x, ei_y, k) firstprivate(x, y)
+ for ( j = 0; j < A->start[A->n]; ++j )
+ {
+ sum = ZERO;
+
+ /* determine row bounds of current nonzero */
+ x = 0;
+ ei_x = 0;
+ for ( k = 0; k <= A->n; ++k )
+ {
+ if ( U_t->start[k] > j )
+ {
+ x = U_t->start[k - 1];
+ ei_x = U_t->start[k];
+ break;
+ }
+ }
+ /* column bounds of current nonzero */
+ y = U_t->start[U_t->j[j]];
+ ei_y = U_t->start[U_t->j[j] + 1];
+
+ /* sparse dot product: dot( U^T(i,1:j-1), U^T(j,1:j-1) ) */
+ while ( U_t->j[x] < U_t->j[j] &&
+ U_t->j[y] < U_t->j[j] &&
+ x < ei_x && y < ei_y )
+ {
+ if ( U_t->j[x] == U_t->j[y] )
+ {
+ sum += (U_t->val[x] * U_t->val[y]);
+ ++x;
+ ++y;
+ }
+ else if ( U_t->j[x] < U_t->j[y] )
+ {
+ ++x;
+ }
+ else
+ {
+ ++y;
+ }
+ }
+
+ sum = DAD->val[j] - sum;
+
+ /* diagonal entries */
+ if ( (k - 1) == U_t->j[j] )
+ {
+ /* sanity check */
+ if ( sum < ZERO )
+ {
+ fprintf( stderr, "Numeric breakdown in ICHOL Terminating.\n");
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "A(%5d,%5d) = %10.3f\n",
+ k - 1, A->entries[j].j, A->entries[j].val );
+ fprintf( stderr, "sum = %10.3f\n", sum);
+#endif
+ exit(NUMERIC_BREAKDOWN);
+ }
+
+ U_t->val[j] = SQRT( sum );
+ }
+ /* non-diagonal entries */
+ else
+ {
+ U_t->val[j] = sum / U_t->val[ei_y - 1];
+ }
+ }
+ }
+
+ /* apply inverse transformation D^{-1}U^{T},
+ * since DAD \approx U^{T}U, so
+ * D^{-1}DADD^{-1} = A \approx D^{-1}U^{T}UD^{-1} */
+ #pragma omp parallel for schedule(guided) \
+ default(none) shared(D_inv) private(i, pj)
+ for ( i = 0; i < A->n; ++i )
+ {
+ for ( pj = A->start[i]; pj < A->start[i + 1]; ++pj )
+ {
+ U_t->val[pj] *= D_inv[i];
+ }
+ }
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "nnz(L): %d, max: %d\n", U_t->start[U_t->n], U_t->n * 50 );
+#endif
+
+ /* transpose U^{T} and copy into U */
+ Transpose( U_t, U );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "nnz(U): %d, max: %d\n", Utop[U->n], U->n * 50 );
+#endif
+
+ Deallocate_Matrix( DAD );
+ free(D_inv);
+ free(D);
+ free(Utop);
+
+ return Get_Timing_Info( start );
+}
+
+
+/* Fine-grained (parallel) incomplete LU factorization
+ *
+ * Reference:
+ * Edmond Chow and Aftab Patel
+ * Fine-Grained Parallel Incomplete LU Factorization
+ * SIAM J. Sci. Comp.
+ *
+ * A: symmetric, half-stored (lower triangular), CSR format
+ * sweeps: number of loops over non-zeros for computation
+ * L / U: factorized triangular matrices (A \approx LU), CSR format */
+static real ILU_PAR( const sparse_matrix * const A, const unsigned int sweeps,
+ sparse_matrix * const L, sparse_matrix * const U )
+{
+ unsigned int i, j, k, pj, x, y, ei_x, ei_y;
+ real *D, *D_inv, sum, start;
+ sparse_matrix *DAD;
+
+ start = Get_Time( );
+
+ if ( Allocate_Matrix( DAD, A->n, A->m ) == FAILURE ||
+ ( D = (real*) malloc(A->n * sizeof(real)) ) == NULL ||
+ ( D_inv = (real*) malloc(A->n * sizeof(real)) ) == NULL )
+ {
+ fprintf( stderr, "not enough memory for ILU_PAR preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(D, D_inv) private(i)
+ for ( i = 0; i < A->n; ++i )
+ {
+ D_inv[i] = SQRT( A->val[A->start[i + 1] - 1] );
+ D[i] = 1.0 / D_inv[i];
+ }
+
+ /* to get convergence, A must have unit diagonal, so apply
+ * transformation DAD, where D = D(1./sqrt(D(A))) */
+ memcpy( DAD->start, A->start, sizeof(int) * (A->n + 1) );
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, D) private(i, pj)
+ for ( i = 0; i < A->n; ++i )
+ {
+ /* non-diagonals */
+ for ( pj = A->start[i]; pj < A->start[i + 1] - 1; ++pj )
+ {
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = D[i] * A->val[pj] * D[A->j[pj]];
+ }
+ /* diagonal */
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = 1.0;
+ }
+
+ /* initial guesses for L and U,
+ * assume: A and DAD symmetric and stored lower triangular */
+ memcpy( L->start, DAD->start, sizeof(int) * (DAD->n + 1) );
+ memcpy( L->j, DAD->j, sizeof(int) * (DAD->start[DAD->n]) );
+ memcpy( L->val, DAD->val, sizeof(real) * (DAD->start[DAD->n]) );
+ /* store U^T in CSR for row-wise access and tranpose later */
+ memcpy( U->start, DAD->start, sizeof(int) * (DAD->n + 1) );
+ memcpy( U->j, DAD->j, sizeof(int) * (DAD->start[DAD->n]) );
+ memcpy( U->val, DAD->val, sizeof(real) * (DAD->start[DAD->n]) );
+
+ /* L has unit diagonal, by convention */
+ #pragma omp parallel for schedule(static) default(none) private(i)
+ for ( i = 0; i < A->n; ++i )
+ {
+ L->val[L->start[i + 1] - 1] = 1.0;
+ }
+
+ for ( i = 0; i < sweeps; ++i )
+ {
+ /* for each nonzero in L */
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD) private(j, k, x, y, ei_x, ei_y, sum)
+ for ( j = 0; j < DAD->start[DAD->n]; ++j )
+ {
+ sum = ZERO;
+
+ /* determine row bounds of current nonzero */
+ x = 0;
+ ei_x = 0;
+ for ( k = 1; k <= DAD->n; ++k )
+ {
+ if ( DAD->start[k] > j )
+ {
+ x = DAD->start[k - 1];
+ ei_x = DAD->start[k];
+ break;
+ }
+ }
+ /* determine column bounds of current nonzero */
+ y = DAD->start[DAD->j[j]];
+ ei_y = DAD->start[DAD->j[j] + 1];
+
+ /* sparse dot product:
+ * dot( L(i,1:j-1), U(1:j-1,j) ) */
+ while ( L->j[x] < L->j[j] &&
+ L->j[y] < L->j[j] &&
+ x < ei_x && y < ei_y )
+ {
+ if ( L->j[x] == L->j[y] )
+ {
+ sum += (L->val[x] * U->val[y]);
+ ++x;
+ ++y;
+ }
+ else if ( L->j[x] < L->j[y] )
+ {
+ ++x;
+ }
+ else
+ {
+ ++y;
+ }
+ }
+
+ if ( j != ei_x - 1 )
+ {
+ L->val[j] = ( DAD->val[j] - sum ) / U->val[ei_y - 1];
+ }
+ }
+
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD) private(j, k, x, y, ei_x, ei_y, sum)
+ for ( j = 0; j < DAD->start[DAD->n]; ++j )
+ {
+ sum = ZERO;
+
+ /* determine row bounds of current nonzero */
+ x = 0;
+ ei_x = 0;
+ for ( k = 1; k <= DAD->n; ++k )
+ {
+ if ( DAD->start[k] > j )
+ {
+ x = DAD->start[k - 1];
+ ei_x = DAD->start[k];
+ break;
+ }
+ }
+ /* determine column bounds of current nonzero */
+ y = DAD->start[DAD->j[j]];
+ ei_y = DAD->start[DAD->j[j] + 1];
+
+ /* sparse dot product:
+ * dot( L(i,1:i-1), U(1:i-1,j) ) */
+ while ( U->j[x] < U->j[j] &&
+ U->j[y] < U->j[j] &&
+ x < ei_x && y < ei_y )
+ {
+ if ( U->j[x] == U->j[y] )
+ {
+ sum += (L->val[y] * U->val[x]);
+ ++x;
+ ++y;
+ }
+ else if ( U->j[x] < U->j[y] )
+ {
+ ++x;
+ }
+ else
+ {
+ ++y;
+ }
+ }
+
+ U->val[j] = DAD->val[j] - sum;
+ }
+ }
+
+ /* apply inverse transformation:
+ * since DAD \approx LU, then
+ * D^{-1}DADD^{-1} = A \approx D^{-1}LUD^{-1} */
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, D_inv) private(i, pj)
+ for ( i = 0; i < DAD->n; ++i )
+ {
+ for ( pj = DAD->start[i]; pj < DAD->start[i + 1]; ++pj )
+ {
+ L->val[pj] = D_inv[i] * L->val[pj];
+ /* currently storing U^T, so use row index instead of column index */
+ U->val[pj] = U->val[pj] * D_inv[i];
+ }
+ }
+
+ Transpose_I( U );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "nnz(L): %d, max: %d\n", L->start[L->n], L->n * 50 );
+ fprintf( stderr, "nnz(U): %d, max: %d\n", Utop[U->n], U->n * 50 );
+#endif
+
+ Deallocate_Matrix( DAD );
+ free( D_inv );
+ free( D );
+
+ return Get_Timing_Info( start );
+}
+
+
+/* Fine-grained (parallel) incomplete LU factorization with thresholding
+ *
+ * Reference:
+ * Edmond Chow and Aftab Patel
+ * Fine-Grained Parallel Incomplete LU Factorization
+ * SIAM J. Sci. Comp.
+ *
+ * A: symmetric, half-stored (lower triangular), CSR format
+ * droptol: row-wise tolerances used for dropping
+ * sweeps: number of loops over non-zeros for computation
+ * L / U: factorized triangular matrices (A \approx LU), CSR format */
+static real ILUT_PAR( const sparse_matrix * const A, const real * droptol,
+ const unsigned int sweeps, sparse_matrix * const L, sparse_matrix * const U )
+{
+ unsigned int i, j, k, pj, x, y, ei_x, ei_y, Ltop, Utop;
+ real *D, *D_inv, sum, start;
+ sparse_matrix *DAD, *L_temp, *U_temp;
+
+ start = Get_Time( );
+
+ if ( Allocate_Matrix( DAD, A->n, A->m ) == FAILURE ||
+ Allocate_Matrix( L_temp, A->n, A->m ) == FAILURE ||
+ Allocate_Matrix( U_temp, A->n, A->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for ILUT_PAR preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ if ( ( D = (real*) malloc(A->n * sizeof(real)) ) == NULL ||
+ ( D_inv = (real*) malloc(A->n * sizeof(real)) ) == NULL )
+ {
+ fprintf( stderr, "not enough memory for ILUT_PAR preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(D, D_inv) private(i)
+ for ( i = 0; i < A->n; ++i )
+ {
+ D_inv[i] = SQRT( A->val[A->start[i + 1] - 1] );
+ D[i] = 1.0 / D_inv[i];
+ }
+
+ /* to get convergence, A must have unit diagonal, so apply
+ * transformation DAD, where D = D(1./sqrt(D(A))) */
+ memcpy( DAD->start, A->start, sizeof(int) * (A->n + 1) );
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, D) private(i, pj)
+ for ( i = 0; i < A->n; ++i )
+ {
+ /* non-diagonals */
+ for ( pj = A->start[i]; pj < A->start[i + 1] - 1; ++pj )
+ {
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = D[i] * A->val[pj] * D[A->j[pj]];
+ }
+ /* diagonal */
+ DAD->j[pj] = A->j[pj];
+ DAD->val[pj] = 1.0;
+ }
+
+ /* initial guesses for L and U,
+ * assume: A and DAD symmetric and stored lower triangular */
+ memcpy( L_temp->start, DAD->start, sizeof(int) * (DAD->n + 1) );
+ memcpy( L_temp->j, DAD->j, sizeof(int) * (DAD->start[DAD->n]) );
+ memcpy( L_temp->val, DAD->val, sizeof(real) * (DAD->start[DAD->n]) );
+ /* store U^T in CSR for row-wise access and tranpose later */
+ memcpy( U_temp->start, DAD->start, sizeof(int) * (DAD->n + 1) );
+ memcpy( U_temp->j, DAD->j, sizeof(int) * (DAD->start[DAD->n]) );
+ memcpy( U_temp->val, DAD->val, sizeof(real) * (DAD->start[DAD->n]) );
+
+ /* L has unit diagonal, by convention */
+ #pragma omp parallel for schedule(static) \
+ default(none) private(i) shared(L_temp)
+ for ( i = 0; i < A->n; ++i )
+ {
+ L_temp->val[L_temp->start[i + 1] - 1] = 1.0;
+ }
+
+ for ( i = 0; i < sweeps; ++i )
+ {
+ /* for each nonzero in L */
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, L_temp, U_temp) private(j, k, x, y, ei_x, ei_y, sum)
+ for ( j = 0; j < DAD->start[DAD->n]; ++j )
+ {
+ sum = ZERO;
+
+ /* determine row bounds of current nonzero */
+ x = 0;
+ ei_x = 0;
+ for ( k = 1; k <= DAD->n; ++k )
+ {
+ if ( DAD->start[k] > j )
+ {
+ x = DAD->start[k - 1];
+ ei_x = DAD->start[k];
+ break;
+ }
+ }
+ /* determine column bounds of current nonzero */
+ y = DAD->start[DAD->j[j]];
+ ei_y = DAD->start[DAD->j[j] + 1];
+
+ /* sparse dot product:
+ * dot( L(i,1:j-1), U(1:j-1,j) ) */
+ while ( L_temp->j[x] < L_temp->j[j] &&
+ L_temp->j[y] < L_temp->j[j] &&
+ x < ei_x && y < ei_y )
+ {
+ if ( L_temp->j[x] == L_temp->j[y] )
+ {
+ sum += (L_temp->val[x] * U_temp->val[y]);
+ ++x;
+ ++y;
+ }
+ else if ( L_temp->j[x] < L_temp->j[y] )
+ {
+ ++x;
+ }
+ else
+ {
+ ++y;
+ }
+ }
+
+ if ( j != ei_x - 1 )
+ {
+ L_temp->val[j] = ( DAD->val[j] - sum ) / U_temp->val[ei_y - 1];
+ }
+ }
+
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, L_temp, U_temp) private(j, k, x, y, ei_x, ei_y, sum)
+ for ( j = 0; j < DAD->start[DAD->n]; ++j )
+ {
+ sum = ZERO;
+
+ /* determine row bounds of current nonzero */
+ x = 0;
+ ei_x = 0;
+ for ( k = 1; k <= DAD->n; ++k )
+ {
+ if ( DAD->start[k] > j )
+ {
+ x = DAD->start[k - 1];
+ ei_x = DAD->start[k];
+ break;
+ }
+ }
+ /* determine column bounds of current nonzero */
+ y = DAD->start[DAD->j[j]];
+ ei_y = DAD->start[DAD->j[j] + 1];
+
+ /* sparse dot product:
+ * dot( L(i,1:i-1), U(1:i-1,j) ) */
+ while ( U_temp->j[x] < U_temp->j[j] &&
+ U_temp->j[y] < U_temp->j[j] &&
+ x < ei_x && y < ei_y )
+ {
+ if ( U_temp->j[x] == U_temp->j[y] )
+ {
+ sum += (L_temp->val[y] * U_temp->val[x]);
+ ++x;
+ ++y;
+ }
+ else if ( U_temp->j[x] < U_temp->j[y] )
+ {
+ ++x;
+ }
+ else
+ {
+ ++y;
+ }
+ }
+
+ U_temp->val[j] = DAD->val[j] - sum;
+ }
+ }
+
+ /* apply inverse transformation:
+ * since DAD \approx LU, then
+ * D^{-1}DADD^{-1} = A \approx D^{-1}LUD^{-1} */
+ #pragma omp parallel for schedule(static) \
+ default(none) shared(DAD, L_temp, U_temp, D_inv) private(i, pj)
+ for ( i = 0; i < DAD->n; ++i )
+ {
+ for ( pj = DAD->start[i]; pj < DAD->start[i + 1]; ++pj )
+ {
+ L_temp->val[pj] = D_inv[i] * L_temp->val[pj];
+ /* currently storing U^T, so use row index instead of column index */
+ U_temp->val[pj] = U_temp->val[pj] * D_inv[i];
+ }
+ }
+
+ /* apply the dropping rule */
+ Ltop = 0;
+ Utop = 0;
+ for ( i = 0; i < DAD->n; ++i )
+ {
+ L->start[i] = Ltop;
+ U->start[i] = Utop;
+
+ for ( pj = L_temp->start[i]; pj < L_temp->start[i + 1] - 1; ++pj )
+ {
+ if ( FABS( L_temp->val[pj] ) > FABS( droptol[i] / L_temp->val[L_temp->start[i + 1] - 1] ) )
+ {
+ L->j[Ltop] = L_temp->j[pj];
+ L->val[Ltop] = L_temp->val[pj];
+ ++Ltop;
+ }
+ }
+
+ /* diagonal */
+ L->j[Ltop] = L_temp->j[pj];
+ L->val[Ltop] = L_temp->val[pj];
+ ++Ltop;
+
+ for ( pj = U_temp->start[i]; pj < U_temp->start[i + 1] - 1; ++pj )
+ {
+ if ( FABS( U_temp->val[pj] ) > FABS( droptol[i] / U_temp->val[U_temp->start[i + 1] - 1] ) )
+ {
+ U->j[Utop] = U_temp->j[pj];
+ U->val[Utop] = U_temp->val[pj];
+ ++Utop;
+ }
+ }
+
+ /* diagonal */
+ U->j[Utop] = U_temp->j[pj];
+ U->val[Utop] = U_temp->val[pj];
+ ++Utop;
+ }
+
+ L->start[i] = Ltop;
+ U->start[i] = Utop;
+
+ Transpose_I( U );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "nnz(L): %d\n", L->start[L->n] );
+ fprintf( stderr, "nnz(U): %d\n", U->start[U->n] );
+#endif
+
+ Deallocate_Matrix( U_temp );
+ Deallocate_Matrix( L_temp );
+ Deallocate_Matrix( DAD );
+ free( D_inv );
+ free( D );
+
+ return Get_Timing_Info( start );
+}
+
+
+/* Setup routine which performs the following:
+ * 1) init storage for QEq matrices and other dependent routines
+ * 2) compute preconditioner (if sim. step matches refactor step)
+ * 3) extrapolate ficticious charges s and t
+ */
+static void Init_MatVec( const reax_system * const system, const control_params * const control,
+ simulation_data * const data, static_storage * const workspace, const list * const far_nbrs )
+{
+ int i, fillin;
+ real s_tmp, t_tmp, time;
+ sparse_matrix *Hptr;
+// char fname[100];
+
+ if (control->qeq_domain_sparsify_enabled)
+ {
+ Hptr = workspace->H_sp;
+ }
+ else
+ {
+ Hptr = workspace->H;
+ }
+
+#if defined(TEST_MAT)
+ Hptr = create_test_mat( );
+#endif
+
+ if (control->pre_comp_refactor > 0 &&
+ ((data->step - data->prev_steps) % control->pre_comp_refactor == 0 || workspace->L == NULL))
+ {
+ //Print_Linear_System( system, control, workspace, data->step );
+
+ time = Get_Time( );
+ if ( control->pre_comp_type != DIAG_PC )
+ {
+ Sort_Matrix_Rows( workspace->H );
+ if ( control->qeq_domain_sparsify_enabled == TRUE )
+ {
+ Sort_Matrix_Rows( workspace->H_sp );
+ }
+
+ if ( control->pre_app_type == TRI_SOLVE_GC_PA )
+ {
+ if ( control->qeq_domain_sparsify_enabled == TRUE )
+ {
+ Hptr = setup_graph_coloring( workspace->H_sp );
+ }
+ else
+ {
+ Hptr = setup_graph_coloring( workspace->H );
+ }
+
+ Sort_Matrix_Rows( Hptr );
+ }
+ }
+ data->timing.QEq_sort_mat_rows += Get_Timing_Info( time );
+
+#if defined(DEBUG)
+ fprintf( stderr, "H matrix sorted\n" );
+#endif
+
+ switch ( control->pre_comp_type )
+ {
+ case DIAG_PC:
+ if ( workspace->Hdia_inv == NULL )
+ {
+ if ( ( workspace->Hdia_inv = (real *) calloc( system->N, sizeof( real ) ) ) == NULL )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+ data->timing.pre_comp += diag_pre_comp( system, workspace->Hdia_inv );
+ break;
+
+ case ICHOLT_PC:
+ Calculate_Droptol( Hptr, workspace->droptol, control->pre_comp_droptol );
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "drop tolerances calculated\n" );
+#endif
+
+ if ( workspace->L == NULL )
+ {
+ fillin = Estimate_LU_Fill( Hptr, workspace->droptol );
+ if ( Allocate_Matrix( workspace->L, far_nbrs->n, fillin ) == FAILURE ||
+ Allocate_Matrix( workspace->U, far_nbrs->n, fillin ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+#if defined(DEBUG)
+ fprintf( stderr, "fillin = %d\n", fillin );
+ fprintf( stderr, "allocated memory: L = U = %ldMB\n",
+ fillin * sizeof(sparse_matrix_entry) / (1024 * 1024) );
+#endif
+ }
+
+ data->timing.pre_comp += ICHOLT( Hptr, workspace->droptol, workspace->L, workspace->U );
+ break;
+
+ case ILU_PAR_PC:
+ if ( workspace->L == NULL )
+ {
+ /* factors have sparsity pattern as H */
+ if ( Allocate_Matrix( workspace->L, Hptr->n, Hptr->m ) == FAILURE ||
+ Allocate_Matrix( workspace->U, Hptr->n, Hptr->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ data->timing.pre_comp += ILU_PAR( Hptr, control->pre_comp_sweeps, workspace->L, workspace->U );
+ break;
+
+ case ILUT_PAR_PC:
+ Calculate_Droptol( Hptr, workspace->droptol, control->pre_comp_droptol );
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "drop tolerances calculated\n" );
+#endif
+
+ if ( workspace->L == NULL )
+ {
+ /* TODO: safest storage estimate is ILU(0) (same as lower triangular portion of H), could improve later */
+ if ( Allocate_Matrix( workspace->L, Hptr->n, Hptr->m ) == FAILURE ||
+ Allocate_Matrix( workspace->U, Hptr->n, Hptr->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+ data->timing.pre_comp += ILUT_PAR( Hptr, workspace->droptol, control->pre_comp_sweeps,
+ workspace->L, workspace->U );
+ break;
+
+ case ILU_SUPERLU_MT_PC:
+ if ( workspace->L == NULL )
+ {
+ /* factors have sparsity pattern as H */
+ if ( Allocate_Matrix( workspace->L, Hptr->n, Hptr->m ) == FAILURE ||
+ Allocate_Matrix( workspace->U, Hptr->n, Hptr->m ) == FAILURE )
+ {
+ fprintf( stderr, "not enough memory for preconditioning matrices. terminating.\n" );
+ exit( INSUFFICIENT_MEMORY );
+ }
+ }
+
+#if defined(HAVE_SUPERLU_MT)
+ data->timing.pre_comp += SuperLU_Factorize( Hptr, workspace->L, workspace->U );
+#else
+ fprintf( stderr, "SuperLU MT support disabled. Re-compile before enabling. Terminating...\n" );
+ exit( INVALID_INPUT );
+#endif
+ break;
+
+ default:
+ fprintf( stderr, "Unrecognized preconditioner computation method. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ }
+
+#if defined(DEBUG)
+ fprintf( stderr, "condest = %f\n", condest(workspace->L, workspace->U) );
+#endif
+
+#if defined(DEBUG_FOCUS)
+ sprintf( fname, "%s.L%d.out", control->sim_name, data->step );
+ Print_Sparse_Matrix2( workspace->L, fname );
+ sprintf( fname, "%s.U%d.out", control->sim_name, data->step );
+ Print_Sparse_Matrix2( workspace->U, fname );
+
+ fprintf( stderr, "icholt-" );
+ //sprintf( fname, "%s.L%d.out", control->sim_name, data->step );
+ //Print_Sparse_Matrix2( workspace->L, fname );
+ //Print_Sparse_Matrix( U );
+#endif
+ }
+
+ /* extrapolation for s & t */
+ //TODO: good candidate for vectorization, avoid moving data with head pointer and circular buffer
+ #pragma omp parallel for schedule(static) \
+ default(none) private(i, s_tmp, t_tmp)
+ for ( i = 0; i < system->N; ++i )
+ {
+ // no extrapolation
+ //s_tmp = workspace->s[index_wkspace_sys(0,i,system->N)];
+ //t_tmp = workspace->t[index_wkspace_sys(0,i,system->N)];
+
+ // linear
+ //s_tmp = 2 * workspace->s[index_wkspace_sys(0,i,system->N)] - workspace->s[index_wkspace_sys(1,i,system->N)];
+ //t_tmp = 2 * workspace->t[index_wkspace_sys(0,i,system->N)] - workspace->t[index_wkspace_sys(1,i,system->N)];
+
+ // quadratic
+// s_tmp = workspace->s[index_wkspace_sys(2,i,system->N)] +
+// 3 * (workspace->s[index_wkspace_sys(0,i,system->N)]-workspace->s[index_wkspace_sys(1,i,system->N)]);
+ t_tmp = workspace->t[index_wkspace_sys(2,i,system->N)] +
+ 3 * (workspace->t[index_wkspace_sys(0,i,system->N)] -workspace->t[index_wkspace_sys(1,i,system->N)]);
+
+ // cubic
+ s_tmp = 4 * (workspace->s[index_wkspace_sys(0,i,system->N)] + workspace->s[index_wkspace_sys(2,i,system->N)]) -
+ (6 * workspace->s[index_wkspace_sys(1,i,system->N)] + workspace->s[index_wkspace_sys(3,i,system->N)]);
+ //t_tmp = 4 * (workspace->t[index_wkspace_sys(0,i,system->N)] + workspace->t[index_wkspace_sys(2,i,system->N)]) -
+ // (6 * workspace->t[index_wkspace_sys(1,i,system->N)] + workspace->t[index_wkspace_sys(3,i,system->N)] );
+
+ // 4th order
+// s_tmp = 5 * (workspace->s[index_wkspace_sys(0,i,system->N)] -
+// workspace->s[index_wkspace_sys(3,i,system->N)]) + 10 *
+// (-workspace->s[index_wkspace_sys(1,i,system->N)] +
+// workspace->s[index_wkspace_sys(2,i,system->N)] ) +
+// workspace->s[index_wkspace_sys(4,i,system->N)];
+// t_tmp = 5 * (workspace->t[index_wkspace_sys(0,i,system->N)] -
+// workspace->t[index_wkspace_sys(3,i,system->N)]) + 10 *
+// (-workspace->t[index_wkspace_sys(1,i,system->N)] +
+// workspace->t[index_wkspace_sys(2,i,system->N)] ) +
+// workspace->t[index_wkspace_sys(4,i,system->N)];
+
+ workspace->s[index_wkspace_sys(4,i,system->N)] = workspace->s[index_wkspace_sys(3,i,system->N)];
+ workspace->s[index_wkspace_sys(3,i,system->N)] = workspace->s[index_wkspace_sys(2,i,system->N)];
+ workspace->s[index_wkspace_sys(2,i,system->N)] = workspace->s[index_wkspace_sys(1,i,system->N)];
+ workspace->s[index_wkspace_sys(1,i,system->N)] = workspace->s[index_wkspace_sys(0,i,system->N)];
+ workspace->s[index_wkspace_sys(0,i,system->N)] = s_tmp;
+
+ workspace->t[index_wkspace_sys(4,i,system->N)] = workspace->t[index_wkspace_sys(3,i,system->N)];
+ workspace->t[index_wkspace_sys(3,i,system->N)] = workspace->t[index_wkspace_sys(2,i,system->N)];
+ workspace->t[index_wkspace_sys(2,i,system->N)] = workspace->t[index_wkspace_sys(1,i,system->N)];
+ workspace->t[index_wkspace_sys(1,i,system->N)] = workspace->t[index_wkspace_sys(0,i,system->N)];
+ workspace->t[index_wkspace_sys(0,i,system->N)] = t_tmp;
+ }
+}
+
+
+/* Combine ficticious charges s and t to get atomic charge q
+ */
+static void Calculate_Charges( const reax_system * const system, static_storage * const workspace )
+{
+ int i;
+ real u, s_sum, t_sum;
+
+ s_sum = t_sum = 0.;
+ for ( i = 0; i < system->N; ++i )
+ {
+ s_sum += workspace->s[index_wkspace_sys(0,i,system->N)];
+ t_sum += workspace->t[index_wkspace_sys(0,i,system->N)];
+ }
+
+ u = s_sum / t_sum;
+ for ( i = 0; i < system->N; ++i )
+ {
+ system->atoms[i].q = workspace->s[index_wkspace_sys(0,i,system->N)]
+ - u * workspace->t[index_wkspace_sys(0,i,system->N)];
+ }
+}
+
+
+/* Main driver method for QEq kernel
+ *
+ * Rough outline:
+ * 1) init / setup routines
+ * 2) perform 2 linear solves
+ * 3) compute atomic charges based on output of 2)
+ */
+void QEq( reax_system * const system, control_params * const control, simulation_data * const data,
+ static_storage * const workspace, const list * const far_nbrs,
+ const output_controls * const out_control )
+{
+ int iters;
+
+ Init_MatVec( system, control, data, workspace, far_nbrs );
+
+ switch ( control->qeq_solver_type )
+ {
+ case GMRES_S:
+ iters = GMRES( workspace, control, data, workspace->H, workspace->b_s, control->qeq_solver_q_err,
+ &workspace->s[index_wkspace_sys(0,0,system->N)], out_control->log,
+ ((data->step - data->prev_steps) % control->pre_comp_refactor == 0) ? TRUE : FALSE );
+ iters += GMRES( workspace, control, data, workspace->H, workspace->b_t, control->qeq_solver_q_err,
+ &workspace->t[index_wkspace_sys(0,0,system->N)], out_control->log, FALSE );
+ break;
+ case GMRES_H_S:
+ iters = GMRES_HouseHolder( workspace, control, data, workspace->H, workspace->b_s, control->qeq_solver_q_err,
+ &workspace->s[index_wkspace_sys(0,0,system->N)], out_control->log, (data->step - data->prev_steps) % control->pre_comp_refactor == 0 );
+ iters += GMRES_HouseHolder( workspace, control, data, workspace->H, workspace->b_t, control->qeq_solver_q_err,
+ &workspace->t[index_wkspace_sys(0,0,system->N)], out_control->log, 0 );
+ break;
+ case CG_S:
+ iters = CG( workspace, workspace->H, workspace->b_s, control->qeq_solver_q_err,
+ &workspace->s[index_wkspace_sys(0,0,system->N)], out_control->log ) + 1;
+ iters += CG( workspace, workspace->H, workspace->b_t, control->qeq_solver_q_err,
+ &workspace->t[index_wkspace_sys(0,0,system->N)], out_control->log ) + 1;
+ break;
+ case SDM_S:
+ iters = SDM( workspace, workspace->H, workspace->b_s, control->qeq_solver_q_err,
+ &workspace->s[index_wkspace_sys(0,0,system->N)], out_control->log ) + 1;
+ iters += SDM( workspace, workspace->H, workspace->b_t, control->qeq_solver_q_err,
+ &workspace->t[index_wkspace_sys(0,0,system->N)], out_control->log ) + 1;
+ break;
+ default:
+ fprintf( stderr, "Unrecognized QEq solver selection. Terminating...\n" );
+ exit( INVALID_INPUT );
+ break;
+ }
+
+ data->timing.solver_iters += iters;
+
+#if defined(DEBUG_FOCUS)
+ fprintf( stderr, "linsolve-" );
+#endif
+
+ Calculate_Charges( system, workspace );
+}
diff --git a/PuReMD-GPU/src/qeq.h b/PuReMD-GPU/src/qeq.h
new file mode 100644
index 0000000000000000000000000000000000000000..4c6c7ea2ce396f5f3cd9a538b801f2658199b7d9
--- /dev/null
+++ b/PuReMD-GPU/src/qeq.h
@@ -0,0 +1,73 @@
+/*----------------------------------------------------------------------
+ PuReMD-GPU - Reax Force Field Simulator
+
+ Copyright (2014) Purdue University
+ Sudhir Kylasa, skylasa@purdue.edu
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#ifndef __QEq_H_
+#define __QEq_H_
+
+#include "mytypes.h"
+
+
+void QEq( reax_system* const, control_params* const, simulation_data* const,
+ static_storage* const, const list* const,
+ const output_controls* const );
+
+
+//static inline HOST_DEVICE void swap( sparse_matrix_entry *array,
+// int index1, int index2 )
+//{
+// sparse_matrix_entry temp = array[index1];
+// array[index1] = array[index2];
+// array[index2] = temp;
+//}
+//
+//
+//static inline HOST_DEVICE void quick_sort( sparse_matrix_entry *array,
+// int start, int end )
+//{
+// int i = start;
+// int k = end;
+//
+// if (end - start >= 1)
+// {
+// int pivot = array[start].j;
+//
+// while (k > i)
+// {
+// while ((array[i].j <= pivot) && (i <= end) && (k > i))
+// {
+// i++;
+// }
+// while ((array[k].j > pivot) && (k >= start) && (k >= i))
+// {
+// k--;
+// }
+// if (k > i)
+// {
+// swap( array, i, k );
+// }
+// }
+// swap( array, start, k );
+// quick_sort( array, start, k - 1 );
+// quick_sort( array, k + 1, end );
+// }
+//}
+
+
+#endif
diff --git a/PuReMD-GPU/src/random.h b/PuReMD-GPU/src/random.h
index b19bc58e3dcef04a324b108be718bfbff3e5c06c..a936477278d06a989d50ab0faeafb8a737a4e5fd 100644
--- a/PuReMD-GPU/src/random.h
+++ b/PuReMD-GPU/src/random.h
@@ -58,7 +58,7 @@ static inline HOST_DEVICE double GRandom(double mean, double sigma)
rsq = v1 * v1 + v2 * v2;
}
- return mean + v1 * sigma * sqrt(-2.0 * log(rsq) / rsq);
+ return mean + v1 * sigma * SQRT(-2.0 * LOG(rsq) / rsq);
}
diff --git a/PuReMD-GPU/src/reset_utils.c b/PuReMD-GPU/src/reset_utils.c
index f79596aa9d29a65f673448d18a28c73c00444e43..ecb921bb00255081bec5470baaab070df8cb80ef 100644
--- a/PuReMD-GPU/src/reset_utils.c
+++ b/PuReMD-GPU/src/reset_utils.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -28,18 +29,20 @@ void Reset_Atoms( reax_system* system )
{
int i;
- for( i = 0; i < system->N; ++i )
- memset( system->atoms[i].f, 0.0, RVEC_SIZE );
+ for ( i = 0; i < system->N; ++i )
+ {
+ memset( system->atoms[i].f, 0.0, sizeof(rvec) );
+ }
}
void Reset_Pressures( simulation_data *data )
{
- rtensor_MakeZero( data->flex_bar.P );
+ rtensor_MakeZero( data->flex_bar.P );
data->iso_bar.P = 0;
rvec_MakeZero( data->int_press );
rvec_MakeZero( data->ext_press );
- /* fprintf( stderr, "reset: ext_press (%12.6f %12.6f %12.6f)\n",
+ /* fprintf( stderr, "reset: ext_press (%12.6f %12.6f %12.6f)\n",
data->ext_press[0], data->ext_press[1], data->ext_press[2] ); */
}
@@ -97,49 +100,57 @@ void Reset_Workspace( reax_system *system, static_storage *workspace )
}
-void Reset_Neighbor_Lists( reax_system *system, control_params *control,
+void Reset_Neighbor_Lists( reax_system *system, control_params *control,
static_storage *workspace, list **lists )
{
int i, tmp;
list *bonds = (*lists) + BONDS;
list *hbonds = (*lists) + HBONDS;
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
tmp = Start_Index( i, bonds );
Set_End_Index( i, tmp, bonds );
}
- //TODO check if this is needed
- memset (bonds->select.bond_list, 0, BOND_DATA_SIZE * bonds->num_intrs );
+ //TODO: added for GPU, verify if correct
+ memset( bonds->select.bond_list, 0, BOND_DATA_SIZE * bonds->num_intrs );
- if( control->hb_cut > 0 )
- for( i = 0; i < system->N; ++i )
- if( system->reaxprm.sbp[system->atoms[i].type].p_hbond == 1) {
+ if ( control->hb_cut > 0 )
+ {
+ for ( i = 0; i < system->N; ++i )
+ {
+ if ( system->reaxprm.sbp[system->atoms[i].type].p_hbond == 1)
+ {
tmp = Start_Index( workspace->hbond_index[i], hbonds );
Set_End_Index( workspace->hbond_index[i], tmp, hbonds );
- /* fprintf( stderr, "i:%d, hbond: %d-%d\n",
- i, Start_Index( workspace->hbond_index[i], hbonds ),
+ /* fprintf( stderr, "i:%d, hbond: %d-%d\n",
+ i, Start_Index( workspace->hbond_index[i], hbonds ),
End_Index( workspace->hbond_index[i], hbonds ) );*/
}
+ }
+ }
}
-void Reset( reax_system *system, control_params *control,
+void Reset( reax_system *system, control_params *control,
simulation_data *data, static_storage *workspace, list **lists )
{
Reset_Atoms( system );
Reset_Simulation_Data( data );
- if( control->ensemble == NPT || control->ensemble == sNPT ||
+ if ( control->ensemble == NPT || control->ensemble == sNPT ||
control->ensemble == iNPT )
+ {
Reset_Pressures( data );
+ }
- Reset_Workspace( system, workspace );
+ Reset_Workspace( system, workspace );
Reset_Neighbor_Lists( system, control, workspace, lists );
-#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "reset - ");
#endif
}
@@ -147,16 +158,18 @@ void Reset( reax_system *system, control_params *control,
void Reset_Grid( grid *g )
{
- memset (g->top, 0, INT_SIZE * g->ncell[0]*g->ncell[1]*g->ncell[2]);
+ memset( g->top, 0, INT_SIZE * g->ncell[0]*g->ncell[1]*g->ncell[2] );
}
+
void Reset_Marks( grid *g, ivec *grid_stack, int grid_top )
{
int i;
- for( i = 0; i < grid_top; ++i )
- g->mark[grid_stack[i][0] * g->ncell[1]*g->ncell[2] +
- grid_stack[i][1] * g->ncell[2] +
- grid_stack[i][2]] = 0;
+ for ( i = 0; i < grid_top; ++i )
+ {
+ g->mark[grid_stack[i][0] * g->ncell[1]*g->ncell[2]
+ + grid_stack[i][1] * g->ncell[2] + grid_stack[i][2]] = 0;
+ }
}
diff --git a/PuReMD-GPU/src/restart.c b/PuReMD-GPU/src/restart.c
index b6ccb014d91ad33cd337d7688345d06811b2c681..13abdecc8142c5f40b942b79c2e886246372576c 100644
--- a/PuReMD-GPU/src/restart.c
+++ b/PuReMD-GPU/src/restart.c
@@ -1,9 +1,10 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
@@ -19,11 +20,12 @@
----------------------------------------------------------------------*/
#include "restart.h"
+
#include "box.h"
#include "vector.h"
void Write_Binary_Restart( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace )
+ simulation_data *data, static_storage *workspace )
{
int i;
char fname[MAX_STR];
@@ -65,8 +67,8 @@ void Write_Binary_Restart( reax_system *system, control_params *control,
void Read_Binary_Restart( char *fname, reax_system *system,
- control_params *control, simulation_data *data,
- static_storage *workspace )
+ control_params *control, simulation_data *data,
+ static_storage *workspace )
{
int i;
FILE *fres;
@@ -103,14 +105,13 @@ void Read_Binary_Restart( char *fname, reax_system *system,
workspace->map_serials = (int*) calloc( MAX_ATOM_ID, sizeof(int) );
for ( i = 0; i < MAX_ATOM_ID; ++i )
+ {
workspace->map_serials[i] = -1;
+ }
workspace->orig_id = (int*) calloc( system->N, sizeof(int) );
workspace->restricted = (int*) calloc( system->N, sizeof(int) );
workspace->restricted_list = (int*) calloc( system->N * MAX_RESTRICT, sizeof(int) );
- //CHANGE
- //for( i = 0; i < system->N; ++i )
- // workspace->restricted_list[i] = (int*) calloc( MAX_RESTRICT, sizeof(int) );
for ( i = 0; i < system->N; ++i )
{
@@ -175,8 +176,7 @@ void Write_ASCII_Restart( reax_system *system, control_params *control,
void Read_ASCII_Restart( char *fname, reax_system *system,
- control_params *control, simulation_data *data,
- static_storage *workspace )
+ control_params *control, simulation_data *data, static_storage *workspace )
{
int i;
FILE *fres;
@@ -185,8 +185,7 @@ void Read_ASCII_Restart( char *fname, reax_system *system,
fres = fopen( fname, "r" );
/* header */
- //fscanf( fres, READ_RESTART_HEADER,
- fscanf( fres, RESTART_HEADER,
+ fscanf( fres, READ_RESTART_HEADER,
&data->prev_steps, &system->N, &data->therm.T, &data->therm.xi,
&data->therm.v_xi, &data->therm.v_xi_old, &data->therm.G_xi,
&system->box.box[0][0], &system->box.box[0][1], &system->box.box[0][2],
@@ -194,7 +193,7 @@ void Read_ASCII_Restart( char *fname, reax_system *system,
&system->box.box[2][0], &system->box.box[2][1], &system->box.box[2][2]);
Make_Consistent( &(system->box) );
-//#if defined(DEBUG_FOCUS)
+#if defined(DEBUG_FOCUS)
fprintf( stderr, "restart step: %d\n", data->prev_steps );
fprintf( stderr, "restart thermostat: %10.6f %10.6f %10.6f %10.6f %10.6f\n",
data->therm.T, data->therm.xi,
@@ -204,22 +203,20 @@ void Read_ASCII_Restart( char *fname, reax_system *system,
system->box.box[0][0], system->box.box[0][1], system->box.box[0][2],
system->box.box[1][0], system->box.box[1][1], system->box.box[1][2],
system->box.box[2][0], system->box.box[2][1], system->box.box[2][2] );
- fprintf ( stderr, "Total Atoms read: %d \n", system->N);
-//#endif
+#endif
/* memory allocations for atoms, atom maps, bond restrictions */
system->atoms = (reax_atom*) calloc( system->N, sizeof(reax_atom) );
workspace->map_serials = (int*) calloc( MAX_ATOM_ID, sizeof(int) );
for ( i = 0; i < MAX_ATOM_ID; ++i )
+ {
workspace->map_serials[i] = -1;
+ }
workspace->orig_id = (int*) calloc( system->N, sizeof(int) );
workspace->restricted = (int*) calloc( system->N, sizeof(int) );
workspace->restricted_list = (int*) calloc( system->N * MAX_RESTRICT, sizeof(int) );
- //CHANGE
- //for( i = 0; i < system->N; ++i )
- // workspace->restricted_list[i] = (int*) calloc( MAX_RESTRICT, sizeof(int) );
for ( i = 0; i < system->N; ++i )
{
@@ -240,11 +237,15 @@ void Read_ASCII_Restart( char *fname, reax_system *system,
void Write_Restart( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- output_controls *out_control )
+ simulation_data *data, static_storage *workspace, output_controls
+ *out_control )
{
if ( out_control->restart_format == WRITE_ASCII )
+ {
Write_ASCII_Restart( system, control, data, workspace );
+ }
else if ( out_control->restart_format == WRITE_BINARY )
+ {
Write_Binary_Restart( system, control, data, workspace );
+ }
}
diff --git a/PuReMD-GPU/src/single_body_interactions.c b/PuReMD-GPU/src/single_body_interactions.c
index b26f493e703819f066389991a4845acab113b326..4c5824dd9862770863aa3e3299ca4f1f691c561e 100644
--- a/PuReMD-GPU/src/single_body_interactions.c
+++ b/PuReMD-GPU/src/single_body_interactions.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -27,11 +28,8 @@
#include "vector.h"
-void LonePair_OverUnder_Coordination_Energy( reax_system *system,
- control_params *control,
- simulation_data *data,
- static_storage *workspace,
- list **lists,
+void LonePair_OverUnder_Coordination_Energy( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
output_controls *out_control )
{
int i, j, pj, type_i, type_j;
@@ -49,7 +47,7 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
single_body_parameters *sbp_i, *sbp_j;
two_body_parameters *twbp;
bond_data *pbond;
- bond_order_data *bo_ij;
+ bond_order_data *bo_ij;
list *bonds = (*lists) + BONDS;
/* Initialize parameters */
@@ -61,64 +59,71 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
p_ovun7 = system->reaxprm.gp.l[8];
p_ovun8 = system->reaxprm.gp.l[9];
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
/* set the parameter pointer */
type_i = system->atoms[i].type;
sbp_i = &(system->reaxprm.sbp[ type_i ]);
/* lone-pair Energy */
- p_lp2 = sbp_i->p_lp2;
+ p_lp2 = sbp_i->p_lp2;
expvd2 = EXP( -75 * workspace->Delta_lp[i] );
inv_expvd2 = 1. / (1. + expvd2 );
/* calculate the energy */
- data->E_Lp += e_lp =
- p_lp2 * workspace->Delta_lp[i] * inv_expvd2;
+ data->E_Lp += e_lp = p_lp2 * workspace->Delta_lp[i] * inv_expvd2;
- dElp = p_lp2 * inv_expvd2 +
- 75 * p_lp2 * workspace->Delta_lp[i] * expvd2 * SQR(inv_expvd2);
+ dElp = p_lp2 * inv_expvd2 +
+ 75 * p_lp2 * workspace->Delta_lp[i] * expvd2 * SQR(inv_expvd2);
CElp = dElp * workspace->dDelta_lp[i];
workspace->CdDelta[i] += CElp; // lp - 1st term
#ifdef TEST_ENERGY
- fprintf( out_control->elp, "%23.15e%23.15e%23.15e%23.15e\n",
- p_lp2, workspace->Delta_lp_temp[i], expvd2, dElp );
+ fprintf( out_control->elp, "%23.15e%23.15e%23.15e%23.15e\n",
+ p_lp2, workspace->Delta_lp_temp[i], expvd2, dElp );
fprintf( out_control->elp, "%6d%23.15e%23.15e%23.15e\n",
- workspace->orig_id[i]+1, workspace->nlp[i], e_lp, data->E_Lp );
+ workspace->orig_id[i] + 1, workspace->nlp[i], e_lp, data->E_Lp );
#endif
+
#ifdef TEST_FORCES
Add_dDelta( system, lists, i, CElp, workspace->f_lp ); // lp - 1st term
#endif
/* correction for C2 */
- if( system->reaxprm.gp.l[5] > 0.001 &&
+ if ( system->reaxprm.gp.l[5] > 0.001 &&
!strcmp( system->reaxprm.sbp[type_i].name, "C" ) )
- for( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
- if( i < bonds->select.bond_list[pj].nbr ) {
+ {
+ for ( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
+ {
+ if ( i < bonds->select.bond_list[pj].nbr )
+ {
j = bonds->select.bond_list[pj].nbr;
type_j = system->atoms[j].type;
- if( !strcmp( system->reaxprm.sbp[type_j].name, "C" ) ) {
+ if ( !strcmp( system->reaxprm.sbp[type_j].name, "C" ) )
+ {
twbp = &( system->reaxprm.tbp[ index_tbp(type_i,type_j,system->reaxprm.num_atom_types) ]);
bo_ij = &( bonds->select.bond_list[pj].bo_data );
Di = workspace->Delta[i];
- vov3 = bo_ij->BO - Di - 0.040*POW(Di, 4.);
+ vov3 = bo_ij->BO - Di - 0.040 * POW(Di, 4.);
- if( vov3 > 3. ) {
- data->E_Lp += e_lph = p_lp3 * SQR(vov3-3.0);
+ if ( vov3 > 3. )
+ {
+ data->E_Lp += e_lph = p_lp3 * SQR(vov3 - 3.0);
//estrain(i) += e_lph;
- deahu2dbo = 2.*p_lp3*(vov3 - 3.);
- deahu2dsbo = 2.*p_lp3*(vov3 - 3.)*(-1. - 0.16*POW(Di, 3.));
+ deahu2dbo = 2.*p_lp3 * (vov3 - 3.);
+ deahu2dsbo = 2.*p_lp3 * (vov3 - 3.) * (-1. - 0.16 * POW(Di, 3.));
bo_ij->Cdbo += deahu2dbo;
workspace->CdDelta[i] += deahu2dsbo;
#ifdef TEST_ENERGY
- fprintf(out_control->elp,"C2cor%6d%6d%23.15e%23.15e%23.15e\n",
+ fprintf(out_control->elp, "C2cor%6d%6d%23.15e%23.15e%23.15e\n",
// workspace->orig_id[i], workspace->orig_id[j],
- i+1, j+1, e_lph, deahu2dbo, deahu2dsbo );
+ i + 1, j + 1, e_lph, deahu2dbo, deahu2dsbo );
#endif
+
#ifdef TEST_FORCES
Add_dBO(system, lists, i, pj, deahu2dbo, workspace->f_lp);
Add_dDelta(system, lists, i, deahu2dsbo, workspace->f_lp);
@@ -127,44 +132,52 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
}
}
+ }
+ }
}
-
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
type_i = system->atoms[i].type;
sbp_i = &(system->reaxprm.sbp[ type_i ]);
/* over-coordination energy */
- if( sbp_i->mass > 21.0 )
+ if ( sbp_i->mass > 21.0 )
+ {
dfvl = 0.0;
- else dfvl = 1.0; // only for 1st-row elements
+ }
+ else
+ {
+ dfvl = 1.0; // only for 1st-row elements
+ }
p_ovun2 = sbp_i->p_ovun2;
sum_ovun1 = 0;
sum_ovun2 = 0;
- for( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj ) {
+ for ( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
+ {
j = bonds->select.bond_list[pj].nbr;
- type_j = system->atoms[j].type;
+ type_j = system->atoms[j].type;
bo_ij = &(bonds->select.bond_list[pj].bo_data);
sbp_j = &(system->reaxprm.sbp[ type_j ]);
twbp = &(system->reaxprm.tbp[ index_tbp(type_i,type_j,system->reaxprm.num_atom_types) ]);
sum_ovun1 += twbp->p_ovun1 * twbp->De_s * bo_ij->BO;
- sum_ovun2 += (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j])*
- ( bo_ij->BO_pi + bo_ij->BO_pi2 );
-
- /*fprintf( stdout, "%4d%4d%23.15e%23.15e%23.15e\n",
- i+1, j+1,
- dfvl * workspace->Delta_lp_temp[j],
- sbp_j->nlp_opt,
- workspace->nlp_temp[j] );*/
+ sum_ovun2 += (workspace->Delta[j] - dfvl * workspace->Delta_lp_temp[j]) *
+ ( bo_ij->BO_pi + bo_ij->BO_pi2 );
+
+ /*fprintf( stdout, "%4d%4d%23.15e%23.15e%23.15e\n",
+ i+1, j+1,
+ dfvl * workspace->Delta_lp_temp[j],
+ sbp_j->nlp_opt,
+ workspace->nlp_temp[j] );*/
}
exp_ovun1 = p_ovun3 * EXP( p_ovun4 * sum_ovun2 );
inv_exp_ovun1 = 1.0 / (1 + exp_ovun1);
- Delta_lpcorr = workspace->Delta[i] -
- (dfvl*workspace->Delta_lp_temp[i]) * inv_exp_ovun1;
+ Delta_lpcorr = workspace->Delta[i] -
+ (dfvl * workspace->Delta_lp_temp[i]) * inv_exp_ovun1;
exp_ovun2 = EXP( p_ovun2 * Delta_lpcorr );
inv_exp_ovun2 = 1.0 / (1.0 + exp_ovun2);
@@ -175,11 +188,11 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
data->E_Ov += e_ov = sum_ovun1 * CEover1;
CEover2 = sum_ovun1 * DlpVi * inv_exp_ovun2 *
- ( 1.0 - Delta_lpcorr*( DlpVi + p_ovun2 * exp_ovun2 * inv_exp_ovun2 ) );
+ ( 1.0 - Delta_lpcorr * ( DlpVi + p_ovun2 * exp_ovun2 * inv_exp_ovun2 ) );
- CEover3 = CEover2 * (1.0 - dfvl*workspace->dDelta_lp[i]*inv_exp_ovun1 );
+ CEover3 = CEover2 * (1.0 - dfvl * workspace->dDelta_lp[i] * inv_exp_ovun1 );
- CEover4 = CEover2 * (dfvl*workspace->Delta_lp_temp[i]) *
+ CEover4 = CEover2 * (dfvl * workspace->Delta_lp_temp[i]) *
p_ovun4 * exp_ovun1 * SQR(inv_exp_ovun1);
@@ -193,14 +206,13 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
inv_exp_ovun2n = 1.0 / (1.0 + exp_ovun2n);
inv_exp_ovun8 = 1.0 / (1.0 + exp_ovun8);
- data->E_Un += e_un =
- -p_ovun5 * (1.0 - exp_ovun6) * inv_exp_ovun2n * inv_exp_ovun8;
+ data->E_Un += e_un = -p_ovun5 * (1.0 - exp_ovun6) * inv_exp_ovun2n * inv_exp_ovun8;
- CEunder1 = inv_exp_ovun2n * ( p_ovun5*p_ovun6*exp_ovun6*inv_exp_ovun8 +
+ CEunder1 = inv_exp_ovun2n * ( p_ovun5 * p_ovun6 * exp_ovun6 * inv_exp_ovun8 +
p_ovun2 * e_un * exp_ovun2n);
CEunder2 = -e_un * p_ovun8 * exp_ovun8 * inv_exp_ovun8;
- CEunder3 = CEunder1 * (1.0 - dfvl*workspace->dDelta_lp[i]*inv_exp_ovun1);
- CEunder4 = CEunder1 * (dfvl*workspace->Delta_lp_temp[i]) *
+ CEunder3 = CEunder1 * (1.0 - dfvl * workspace->dDelta_lp[i] * inv_exp_ovun1);
+ CEunder4 = CEunder1 * (dfvl * workspace->Delta_lp_temp[i]) *
p_ovun4 * exp_ovun1 * SQR(inv_exp_ovun1) + CEunder2;
//fprintf( stdout, "%6d%23.15e%23.15e%23.15e\n",
@@ -215,100 +227,98 @@ void LonePair_OverUnder_Coordination_Energy( reax_system *system,
Add_dDelta( system, lists, i, CEunder3, workspace->f_un ); // UnCoor - 1st
#endif
-
- for( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj ){
+ for ( pj = Start_Index(i, bonds); pj < End_Index(i, bonds); ++pj )
+ {
pbond = &(bonds->select.bond_list[pj]);
j = pbond->nbr;
type_j = system->atoms[j].type;
bo_ij = &(pbond->bo_data);
twbp = &(system->reaxprm.tbp[ index_tbp(type_i,type_j,system->reaxprm.num_atom_types) ]);
-
- bo_ij->Cdbo += CEover1 * twbp->p_ovun1 * twbp->De_s; // OvCoor - 1st
- workspace->CdDelta[j] += CEover4*(1.0 - dfvl*workspace->dDelta_lp[j])*
- (bo_ij->BO_pi + bo_ij->BO_pi2); // OvCoor - 3a
- bo_ij->Cdbopi += CEover4 *
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]);//OvCoor-3b
- bo_ij->Cdbopi2 += CEover4 *
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]);//OvCoor-3b
+ bo_ij->Cdbo += CEover1 * twbp->p_ovun1 * twbp->De_s; // OvCoor - 1st
+ workspace->CdDelta[j] += CEover4 * (1.0 - dfvl * workspace->dDelta_lp[j]) *
+ (bo_ij->BO_pi + bo_ij->BO_pi2); // OvCoor - 3a
+ bo_ij->Cdbopi += CEover4 *
+ (workspace->Delta[j] - dfvl * workspace->Delta_lp_temp[j]); //OvCoor-3b
+ bo_ij->Cdbopi2 += CEover4 *
+ (workspace->Delta[j] - dfvl * workspace->Delta_lp_temp[j]); //OvCoor-3b
- workspace->CdDelta[j] += CEunder4*(1.0-dfvl*workspace->dDelta_lp[j]) *
+ workspace->CdDelta[j] += CEunder4 * (1.0 - dfvl * workspace->dDelta_lp[j]) *
(bo_ij->BO_pi + bo_ij->BO_pi2); // UnCoor - 2a
- bo_ij->Cdbopi += CEunder4 *
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]);//UnCoor-2b
- bo_ij->Cdbopi2 += CEunder4 *
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]);//UnCoor-2b
-
+ bo_ij->Cdbopi += CEunder4 *
+ (workspace->Delta[j] - dfvl * workspace->Delta_lp_temp[j]); //UnCoor-2b
+ bo_ij->Cdbopi2 += CEunder4 *
+ (workspace->Delta[j] - dfvl * workspace->Delta_lp_temp[j]); //UnCoor-2b
#ifdef TEST_ENERGY
/* fprintf( out_control->eov, "%6d%23.15e%23.15e"
- workspace->orig_id[j]+1,
- //twbp->p_ovun1,twbp->De_s,Delta_lpcorr*DlpVi*inv_exp_ovun2,
- CEover1*twbp->p_ovun1*twbp->De_s, CEover3 ); */
-
- /*fprintf( out_control->eov, "%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
- workspace->orig_id[j]+1,
- CEover4,
- CEover4*
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
- CEover4 * (bo_ij->BO_pi + bo_ij->BO_pi2),
- (1.0 - dfvl*workspace->dDelta_lp[j]),
- CEover4 * (1.0 - dfvl*workspace->dDelta_lp[j]) *
- (bo_ij->BO_pi + bo_ij->BO_pi2) );*/
+ workspace->orig_id[j]+1,
+ //twbp->p_ovun1,twbp->De_s,Delta_lpcorr*DlpVi*inv_exp_ovun2,
+ CEover1*twbp->p_ovun1*twbp->De_s, CEover3 ); */
+
+ /*fprintf( out_control->eov, "%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
+ workspace->orig_id[j]+1,
+ CEover4,
+ CEover4*
+ (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
+ CEover4 * (bo_ij->BO_pi + bo_ij->BO_pi2),
+ (1.0 - dfvl*workspace->dDelta_lp[j]),
+ CEover4 * (1.0 - dfvl*workspace->dDelta_lp[j]) *
+ (bo_ij->BO_pi + bo_ij->BO_pi2) );*/
/* fprintf( out_control->eun, "%6d%23.15e\n",
- workspace->orig_id[j]+1, CEunder3 ); */
+ workspace->orig_id[j]+1, CEunder3 ); */
/*fprintf( out_control->eun, "%6d%23.15e%23.15e%23.15e%23.15e\n",
- workspace->orig_id[j]+1,
- CEunder4,
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
- CEunder4*
- (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
- CEunder4*(1.0 - dfvl*workspace->dDelta_lp[j])*
- (bo_ij->BO_pi + bo_ij->BO_pi2) );*/
+ workspace->orig_id[j]+1,
+ CEunder4,
+ (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
+ CEunder4*
+ (workspace->Delta[j] - dfvl*workspace->Delta_lp_temp[j]),
+ CEunder4*(1.0 - dfvl*workspace->dDelta_lp[j])*
+ (bo_ij->BO_pi + bo_ij->BO_pi2) );*/
#endif
#ifdef TEST_FORCES
- Add_dBO( system, lists, i, pj, CEover1 * twbp->p_ovun1 * twbp->De_s,
- workspace->f_ov ); // OvCoor - 1st term
+ Add_dBO( system, lists, i, pj, CEover1 * twbp->p_ovun1 * twbp->De_s,
+ workspace->f_ov ); // OvCoor - 1st term
Add_dDelta( system, lists, j,
- CEover4 * (1.0 - dfvl*workspace->dDelta_lp[j]) *
- (bo_ij->BO_pi+bo_ij->BO_pi2), workspace->f_ov );//OvCoor3a
+ CEover4 * (1.0 - dfvl * workspace->dDelta_lp[j]) *
+ (bo_ij->BO_pi + bo_ij->BO_pi2), workspace->f_ov ); //OvCoor3a
- Add_dBOpinpi2( system, lists, i, pj,
- CEover4 * (workspace->Delta[j] -
- dfvl * workspace->Delta_lp_temp[j]),
- CEover4 * (workspace->Delta[j] -
- dfvl * workspace->Delta_lp_temp[j]),
- workspace->f_ov, workspace->f_ov ); // OvCoor - 3b
+ Add_dBOpinpi2( system, lists, i, pj,
+ CEover4 * (workspace->Delta[j] -
+ dfvl * workspace->Delta_lp_temp[j]),
+ CEover4 * (workspace->Delta[j] -
+ dfvl * workspace->Delta_lp_temp[j]),
+ workspace->f_ov, workspace->f_ov ); // OvCoor - 3b
Add_dDelta( system, lists, j,
- CEunder4 * (1.0 - dfvl*workspace->dDelta_lp[j]) *
- (bo_ij->BO_pi + bo_ij->BO_pi2),
- workspace->f_un ); // UnCoor - 2a
-
- Add_dBOpinpi2( system, lists, i, pj,
- CEunder4 * (workspace->Delta[j] -
- dfvl * workspace->Delta_lp_temp[j]),
- CEunder4 * (workspace->Delta[j] -
- dfvl * workspace->Delta_lp_temp[j]),
- workspace->f_un, workspace->f_un ); // UnCoor - 2b
+ CEunder4 * (1.0 - dfvl * workspace->dDelta_lp[j]) *
+ (bo_ij->BO_pi + bo_ij->BO_pi2),
+ workspace->f_un ); // UnCoor - 2a
+
+ Add_dBOpinpi2( system, lists, i, pj,
+ CEunder4 * (workspace->Delta[j] -
+ dfvl * workspace->Delta_lp_temp[j]),
+ CEunder4 * (workspace->Delta[j] -
+ dfvl * workspace->Delta_lp_temp[j]),
+ workspace->f_un, workspace->f_un ); // UnCoor - 2b
#endif
}
-#ifdef TEST_ENERGY
+#ifdef TEST_ENERGY
- fprintf( out_control->eov, "%6d%15.8f%15.8f%15.8f\n",
- i+1, DlpVi, Delta_lpcorr, sbp_i->valency );
+ fprintf( out_control->eov, "%6d%15.8f%15.8f%15.8f\n",
+ i + 1, DlpVi, Delta_lpcorr, sbp_i->valency );
- fprintf( out_control->eov, "%6d%15.8f%15.8f\n",
- i+1/*workspace->orig_id[i]+1*/, e_ov, data->E_Ov + data->E_Un );
+ fprintf( out_control->eov, "%6d%15.8f%15.8f\n",
+ i + 1/*workspace->orig_id[i]+1*/, e_ov, data->E_Ov + data->E_Un );
- fprintf( out_control->eov, "%6d%15.8f%15.8f\n",
- i+1/*workspace->orig_id[i]+1*/, e_un, data->E_Ov + data->E_Un );
+ fprintf( out_control->eov, "%6d%15.8f%15.8f\n",
+ i + 1/*workspace->orig_id[i]+1*/, e_un, data->E_Ov + data->E_Un );
#endif
}
}
diff --git a/PuReMD-GPU/src/system_props.c b/PuReMD-GPU/src/system_props.c
index 0126b86b776dce8fd30aea0c228731b95104b216..fc93a474cf378f1a382d0ae017cf15a9b23eb17a 100644
--- a/PuReMD-GPU/src/system_props.c
+++ b/PuReMD-GPU/src/system_props.c
@@ -1,67 +1,50 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
#include "system_props.h"
-
-#include "box.h"
+#include "tool_box.h"
#include "vector.h"
-HOST real Get_Time( )
-{
- struct timeval tim;
-
- gettimeofday(&tim, NULL );
- return( tim.tv_sec + (tim.tv_usec / 1000000.0) );
-}
-
-
-HOST real Get_Timing_Info( real t_start )
-{
- struct timeval tim;
- real t_end;
-
- gettimeofday(&tim, NULL );
- t_end = tim.tv_sec + (tim.tv_usec / 1000000.0);
- return (t_end - t_start);
-}
-
-
-void Temperature_Control( control_params *control, simulation_data *data,
- output_controls *out_control )
+void Temperature_Control( control_params *control, simulation_data *data,
+ output_controls *out_control )
{
real tmp;
- if( control->T_mode == 1 ) { // step-wise temperature control
- if( (data->step - data->prev_steps) %
- ((int)(control->T_freq / control->dt)) == 0 ) {
- if( fabs( control->T - control->T_final ) >= fabs( control->T_rate ) )
+ if ( control->T_mode == 1 ) // step-wise temperature control
+ {
+ if ( (data->step - data->prev_steps) %
+ ((int)(control->T_freq / control->dt)) == 0 )
+ {
+ if ( fabs( control->T - control->T_final ) >= fabs( control->T_rate ) )
control->T += control->T_rate;
- else control->T = control->T_final;
+ else control->T = control->T_final;
}
}
- else if( control->T_mode == 2 ) { // constant slope control
+ else if ( control->T_mode == 2 ) // constant slope control
+ {
tmp = control->T_rate * control->dt / control->T_freq;
- if( fabs( control->T - control->T_final ) >= fabs( tmp ) )
- control->T += tmp;
+ if ( fabs( control->T - control->T_final ) >= fabs( tmp ) )
+ control->T += tmp;
}
}
@@ -69,39 +52,34 @@ void Temperature_Control( control_params *control, simulation_data *data,
void Compute_Total_Mass( reax_system *system, simulation_data *data )
{
int i;
- int blocks;
- int block_size;
- real *partial_sums = 0;
data->M = 0;
- for( i = 0; i < system->N; i++ )
- data->M += system->reaxprm.sbp[ system->atoms[i].type ].mass;
+ for ( i = 0; i < system->N; i++ )
+ data->M += system->reaxprm.sbp[ system->atoms[i].type ].mass;
- data->inv_M = 1. / data->M;
+ //fprintf ( stderr, "Compute_total_Mass -->%f<-- \n", data->M );
+ data->inv_M = 1. / data->M;
}
-void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
- FILE *fout )
+void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
+ FILE *fout )
{
int i;
real m, xx, xy, xz, yy, yz, zz, det;
rvec tvec, diff;
rtensor mat, inv;
- int blocks;
- int block_size;
- rvec *l_xcm, *l_vcm, *l_amcm;
- real t_start, t_end;
-
rvec_MakeZero( data->xcm ); // position of CoM
rvec_MakeZero( data->vcm ); // velocity of CoM
rvec_MakeZero( data->amcm ); // angular momentum of CoM
rvec_MakeZero( data->avcm ); // angular velocity of CoM
+
/* Compute the position, velocity and angular momentum about the CoM */
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
m = system->reaxprm.sbp[ system->atoms[i].type ].mass;
rvec_ScaledAdd( data->xcm, m, system->atoms[i].x );
@@ -111,11 +89,13 @@ void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
rvec_ScaledAdd( data->amcm, m, tvec );
/*fprintf( fout,"%3d %g %g %g\n",
- i+1,
+ i+1,
system->atoms[i].v[0], system->atoms[i].v[1], system->atoms[i].v[2] );
- fprintf( fout, "vcm: %g %g %g\n",
- data->vcm[0], data->vcm[1], data->vcm[2] );
- */
+ fprintf( fout, "vcm: %g %g %g\n",
+ data->vcm[0], data->vcm[1], data->vcm[2] );
+ */
+ /* fprintf( stderr, "amcm: %12.6f %12.6f %12.6f\n",
+ data->amcm[0], data->amcm[1], data->amcm[2] ); */
}
rvec_Scale( data->xcm, data->inv_M, data->xcm );
@@ -129,7 +109,8 @@ void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
/* Calculate and then invert the inertial tensor */
xx = xy = xz = yy = yz = zz = 0;
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
m = system->reaxprm.sbp[ system->atoms[i].type ].mass;
rvec_ScaledSum( diff, 1., system->atoms[i].x, -1., data->xcm );
@@ -138,19 +119,10 @@ void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
xz += diff[0] * diff[2] * m;
yy += diff[1] * diff[1] * m;
yz += diff[1] * diff[2] * m;
- zz += diff[2] * diff[2] * m;
+ zz += diff[2] * diff[2] * m;
}
-#ifdef __DEBUG_CUDA__
- fprintf (stderr, " xx: %f \n", xx);
- fprintf (stderr, " xy: %f \n", xy);
- fprintf (stderr, " xz: %f \n", xz);
- fprintf (stderr, " yy: %f \n", yy);
- fprintf (stderr, " yz: %f \n", yz);
- fprintf (stderr, " zz: %f \n", zz);
-#endif
-
- mat[0][0] = yy + zz;
+ mat[0][0] = yy + zz;
mat[0][1] = mat[1][0] = -xy;
mat[0][2] = mat[2][0] = -xz;
mat[1][1] = xx + zz;
@@ -158,12 +130,12 @@ void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
mat[2][2] = xx + yy;
/* invert the inertial tensor */
- det = ( mat[0][0] * mat[1][1] * mat[2][2] +
- mat[0][1] * mat[1][2] * mat[2][0] +
+ det = ( mat[0][0] * mat[1][1] * mat[2][2] +
+ mat[0][1] * mat[1][2] * mat[2][0] +
mat[0][2] * mat[1][0] * mat[2][1] ) -
- ( mat[0][0] * mat[1][2] * mat[2][1] +
- mat[0][1] * mat[1][0] * mat[2][2] +
- mat[0][2] * mat[1][1] * mat[2][0] );
+ ( mat[0][0] * mat[1][2] * mat[2][1] +
+ mat[0][1] * mat[1][0] * mat[2][2] +
+ mat[0][2] * mat[1][1] * mat[2][0] );
inv[0][0] = mat[1][1] * mat[2][2] - mat[1][2] * mat[2][1];
inv[0][1] = mat[0][2] * mat[2][1] - mat[0][1] * mat[2][2];
@@ -175,33 +147,33 @@ void Compute_Center_of_Mass( reax_system *system, simulation_data *data,
inv[2][1] = mat[2][0] * mat[0][1] - mat[0][0] * mat[2][1];
inv[2][2] = mat[0][0] * mat[1][1] - mat[1][0] * mat[0][1];
- if( fabs(det) > ALMOST_ZERO )
- rtensor_Scale( inv, 1./det, inv );
- else
+ if ( fabs(det) > ALMOST_ZERO )
+ rtensor_Scale( inv, 1. / det, inv );
+ else
rtensor_MakeZero( inv );
/* Compute the angular velocity about the centre of mass */
- rtensor_MatVec( data->avcm, inv, data->amcm );
+ rtensor_MatVec( data->avcm, inv, data->amcm );
data->erot_cm = 0.5 * E_CONV * rvec_Dot( data->avcm, data->amcm );
#if defined(DEBUG)
- fprintf( stderr, "xcm: %24.15e %24.15e %24.15e\n",
- data->xcm[0], data->xcm[1], data->xcm[2] );
- fprintf( stderr, "vcm: %24.15e %24.15e %24.15e\n",
- data->vcm[0], data->vcm[1], data->vcm[2] );
- fprintf( stderr, "amcm: %24.15e %24.15e %24.15e\n",
- data->amcm[0], data->amcm[1], data->amcm[2] );
+ fprintf( stderr, "xcm: %24.15e %24.15e %24.15e\n",
+ data->xcm[0], data->xcm[1], data->xcm[2] );
+ fprintf( stderr, "vcm: %24.15e %24.15e %24.15e\n",
+ data->vcm[0], data->vcm[1], data->vcm[2] );
+ fprintf( stderr, "amcm: %24.15e %24.15e %24.15e\n",
+ data->amcm[0], data->amcm[1], data->amcm[2] );
/* fprintf( fout, "mat: %f %f %f\n %f %f %f\n %f %f %f\n",
- mat[0][0], mat[0][1], mat[0][2],
- mat[1][0], mat[1][1], mat[1][2],
+ mat[0][0], mat[0][1], mat[0][2],
+ mat[1][0], mat[1][1], mat[1][2],
mat[2][0], mat[2][1], mat[2][2] );
fprintf( fout, "inv: %g %g %g\n %g %g %g\n %g %g %g\n",
- inv[0][0], inv[0][1], inv[0][2],
- inv[1][0], inv[1][1], inv[1][2],
+ inv[0][0], inv[0][1], inv[0][2],
+ inv[1][0], inv[1][1], inv[1][2],
inv[2][0], inv[2][1], inv[2][2] );
fflush( fout ); */
- fprintf( stderr, "avcm: %24.15e %24.15e %24.15e\n",
- data->avcm[0], data->avcm[1], data->avcm[2] );
+ fprintf( stderr, "avcm: %24.15e %24.15e %24.15e\n",
+ data->avcm[0], data->avcm[1], data->avcm[2] );
#endif
}
@@ -214,7 +186,8 @@ void Compute_Kinetic_Energy( reax_system* system, simulation_data* data )
data->E_Kin = 0.0;
- for (i=0; i < system->N; i++) {
+ for (i = 0; i < system->N; i++)
+ {
m = system->reaxprm.sbp[system->atoms[i].type].mass;
rvec_Scale( p, m, system->atoms[i].v );
@@ -232,17 +205,17 @@ void Compute_Kinetic_Energy( reax_system* system, simulation_data* data )
}
-/* IMPORTANT: This function assumes that current kinetic energy and
- * the center of mass of the system is already computed before.
+/* IMPORTANT: This function assumes that current kinetic energy and
+ * the center of mass of the system is already computed before.
*
- * IMPORTANT: In Klein's paper, it is stated that a dU/dV term needs
- * to be added when there are long-range interactions or long-range
+ * IMPORTANT: In Klein's paper, it is stated that a dU/dV term needs
+ * to be added when there are long-range interactions or long-range
* corrections to short-range interactions present.
- * We may want to add that for more accuracy.
+ * We may want to add that for more accuracy.
*/
-void Compute_Pressure_Isotropic( reax_system* system, control_params *control,
- simulation_data* data,
- output_controls *out_control )
+void Compute_Pressure_Isotropic( reax_system* system, control_params *control,
+ simulation_data* data,
+ output_controls *out_control )
{
int i;
reax_atom *p_atom;
@@ -254,8 +227,10 @@ void Compute_Pressure_Isotropic( reax_system* system, control_params *control,
rvec_MakeZero( data->int_press );
// 0: both int and ext, 1: ext only, 2: int only
- if( control->press_mode == 0 || control->press_mode == 2 ) {
- for( i = 0; i < system->N; ++i ) {
+ if ( control->press_mode == 0 || control->press_mode == 2 )
+ {
+ for ( i = 0; i < system->N; ++i )
+ {
p_atom = &( system->atoms[i] );
/* transform x into unitbox coordinates */
@@ -265,13 +240,14 @@ void Compute_Pressure_Isotropic( reax_system* system, control_params *control,
rvec_Multiply( tmp, p_atom->f, tx );
rvec_Add( data->int_press, tmp );
- if( out_control->debug_level > 0 ) {
- fprintf( out_control->prs, "%-8d%8.2f%8.2f%8.2f",
- i+1, p_atom->x[0], p_atom->x[1], p_atom->x[2] );
- fprintf( out_control->prs, "%8.2f%8.2f%8.2f",
- p_atom->f[0], p_atom->f[1], p_atom->f[2] );
- fprintf( out_control->prs, "%8.2f%8.2f%8.2f\n",
- data->int_press[0],data->int_press[1],data->int_press[2]);
+ if ( out_control->debug_level > 0 )
+ {
+ fprintf( out_control->prs, "%-8d%8.2f%8.2f%8.2f",
+ i + 1, p_atom->x[0], p_atom->x[1], p_atom->x[2] );
+ fprintf( out_control->prs, "%8.2f%8.2f%8.2f",
+ p_atom->f[0], p_atom->f[1], p_atom->f[2] );
+ fprintf( out_control->prs, "%8.2f%8.2f%8.2f\n",
+ data->int_press[0], data->int_press[1], data->int_press[2]);
}
}
}
@@ -279,53 +255,53 @@ void Compute_Pressure_Isotropic( reax_system* system, control_params *control,
/* kinetic contribution */
data->kin_press = 2. * (E_CONV * data->E_Kin) / ( 3. * box->volume * P_CONV );
- /* Calculate total pressure in each direction */
- data->tot_press[0] = data->kin_press -
- ((data->int_press[0] + data->ext_press[0]) /
- (box->box_norms[1] * box->box_norms[2] * P_CONV));
+ /* Calculate total pressure in each direction */
+ data->tot_press[0] = data->kin_press -
+ ((data->int_press[0] + data->ext_press[0]) /
+ (box->box_norms[1] * box->box_norms[2] * P_CONV));
- data->tot_press[1] = data->kin_press -
- ((data->int_press[1] + data->ext_press[1])/
- (box->box_norms[0] * box->box_norms[2] * P_CONV));
+ data->tot_press[1] = data->kin_press -
+ ((data->int_press[1] + data->ext_press[1]) /
+ (box->box_norms[0] * box->box_norms[2] * P_CONV));
- data->tot_press[2] = data->kin_press -
- ((data->int_press[2] + data->ext_press[2])/
- (box->box_norms[0] * box->box_norms[1] * P_CONV));
+ data->tot_press[2] = data->kin_press -
+ ((data->int_press[2] + data->ext_press[2]) /
+ (box->box_norms[0] * box->box_norms[1] * P_CONV));
/* Average pressure for the whole box */
- data->iso_bar.P=(data->tot_press[0]+data->tot_press[1]+data->tot_press[2])/3;
+ data->iso_bar.P = (data->tot_press[0] + data->tot_press[1] + data->tot_press[2]) / 3;
}
-void Compute_Pressure_Isotropic_Klein( reax_system* system,
- simulation_data* data )
+void Compute_Pressure_Isotropic_Klein( reax_system* system,
+ simulation_data* data )
{
int i;
reax_atom *p_atom;
rvec dx;
- // IMPORTANT: This function assumes that current kinetic energy and
+ // IMPORTANT: This function assumes that current kinetic energy and
// the center of mass of the system is already computed before.
data->iso_bar.P = 2.0 * data->E_Kin;
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
{
p_atom = &( system->atoms[i] );
- rvec_ScaledSum(dx,1.0,p_atom->x,-1.0,data->xcm);
+ rvec_ScaledSum(dx, 1.0, p_atom->x, -1.0, data->xcm);
data->iso_bar.P += ( -F_CONV * rvec_Dot(p_atom->f, dx) );
}
data->iso_bar.P /= (3.0 * system->box.volume);
- // IMPORTANT: In Klein's paper, it is stated that a dU/dV term needs
- // to be added when there are long-range interactions or long-range
+ // IMPORTANT: In Klein's paper, it is stated that a dU/dV term needs
+ // to be added when there are long-range interactions or long-range
// corrections to short-range interactions present.
// We may want to add that for more accuracy.
}
-void Compute_Pressure( reax_system* system, simulation_data* data,
- static_storage *workspace )
+void Compute_Pressure( reax_system* system, simulation_data* data,
+ static_storage *workspace )
{
int i;
reax_atom *p_atom;
@@ -333,13 +309,14 @@ void Compute_Pressure( reax_system* system, simulation_data* data,
rtensor_MakeZero( data->flex_bar.P );
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
p_atom = &( system->atoms[i] );
// Distance_on_T3_Gen( data->rcm, p_atom->x, &(system->box), &dx );
rvec_OuterProduct( temp, p_atom->v, p_atom->v );
- rtensor_ScaledAdd( data->flex_bar.P,
- system->reaxprm.sbp[ p_atom->type ].mass, temp );
- // rvec_OuterProduct(temp, workspace->virial_forces[i], p_atom->x );
+ rtensor_ScaledAdd( data->flex_bar.P,
+ system->reaxprm.sbp[ p_atom->type ].mass, temp );
+ // rvec_OuterProduct(temp, workspace->virial_forces[i], p_atom->x );
rtensor_ScaledAdd( data->flex_bar.P, -F_CONV, temp );
}
diff --git a/PuReMD-GPU/src/system_props.h b/PuReMD-GPU/src/system_props.h
index 874132451d02b2d62d87c82065874f04a35b2d37..e2cc98350167a763ac6acdac5807d710403210bd 100644
--- a/PuReMD-GPU/src/system_props.h
+++ b/PuReMD-GPU/src/system_props.h
@@ -28,10 +28,6 @@
extern "C" {
#endif
-real Get_Time( );
-
-real Get_Timing_Info( real );
-
void Temperature_Control( control_params*, simulation_data*, output_controls* );
void Compute_Total_Mass( reax_system*, simulation_data* );
diff --git a/PuReMD-GPU/src/testmd.c b/PuReMD-GPU/src/testmd.c
index 57d8859df4645f982b93803415cb408c57a564f7..b5204950ee733f30cd69c13d28c5cfd6b2200246 100644
--- a/PuReMD-GPU/src/testmd.c
+++ b/PuReMD-GPU/src/testmd.c
@@ -22,20 +22,23 @@
#include "analyze.h"
#include "box.h"
+#include "control.h"
+#include "ffield.h"
#include "forces.h"
+#include "geo_tools.h"
#include "grid.h"
#include "init_md.h"
#include "integrate.h"
#include "neighbors.h"
-#include "param.h"
-#include "pdb_tools.h"
#include "print_utils.h"
#include "reset_utils.h"
#include "restart.h"
#include "system_props.h"
#include "traj.h"
+#include "tool_box.h"
#include "vector.h"
+#ifdef HAVE_CUDA
#include "cuda_environment.h"
#include "cuda_forces.h"
#include "cuda_init_md.h"
@@ -43,9 +46,9 @@
#include "cuda_post_evolve.h"
#include "cuda_reset_utils.h"
#include "cuda_system_props.h"
-
#ifdef __BUILD_DEBUG__
- #include "validation.h"
+ #include "cuda_validation.h"
+#endif
#endif
@@ -69,9 +72,12 @@ int BLOCKS, BLOCKS_POW_2, BLOCK_SIZE;
int MATVEC_BLOCKS;
-void Post_Evolve( reax_system* system, control_params* control,
- simulation_data* data, static_storage* workspace,
- list** lists, output_controls *out_control )
+static void Post_Evolve( reax_system * const system,
+ control_params * const control,
+ simulation_data * const data,
+ static_storage * const workspace,
+ list ** const lists,
+ output_controls * const out_control )
{
int i;
rvec diff, cross;
@@ -110,15 +116,27 @@ void Post_Evolve( reax_system* system, control_params* control,
}
-void Read_System( char *geof, char *ff, char *ctrlf,
- reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- output_controls *out_control )
+void static Read_System( char * const geo_file,
+ char * const ffield_file,
+ char * const control_file,
+ reax_system * const system,
+ control_params * const control,
+ simulation_data * const data,
+ static_storage * const workspace,
+ output_controls * const out_control )
{
FILE *ffield, *ctrl;
- ffield = fopen( ff, "r" );
- ctrl = fopen( ctrlf, "r" );
+ if ( (ffield = fopen( ffield_file, "r" )) == NULL )
+ {
+ fprintf( stderr, "Error opening the ffield file!\n" );
+ exit( FILE_NOT_FOUND );
+ }
+ if ( (ctrl = fopen( control_file, "r" )) == NULL )
+ {
+ fprintf( stderr, "Error opening the ffield file!\n" );
+ exit( FILE_NOT_FOUND );
+ }
/* ffield file */
Read_Force_Field( ffield, &(system->reaxprm) );
@@ -127,32 +145,31 @@ void Read_System( char *geof, char *ff, char *ctrlf,
Read_Control_File( ctrl, system, control, out_control );
/* geo file */
- if( control->geo_format == XYZ )
+ if( control->geo_format == CUSTOM )
{
- fprintf( stderr, "xyz input is not implemented yet\n" );
- exit( 1 );
+ Read_Geo( geo_file, system, control, data, workspace );
}
else if( control->geo_format == PDB )
{
- Read_PDB( geof, system, control, data, workspace );
+ Read_PDB( geo_file, system, control, data, workspace );
}
else if( control->geo_format == BGF )
{
- Read_BGF( geof, system, control, data, workspace );
+ Read_BGF( geo_file, system, control, data, workspace );
}
else if( control->geo_format == ASCII_RESTART )
{
- Read_ASCII_Restart( geof, system, control, data, workspace );
+ Read_ASCII_Restart( geo_file, system, control, data, workspace );
control->restart = 1;
}
else if( control->geo_format == BINARY_RESTART ) {
- Read_Binary_Restart( geof, system, control, data, workspace );
+ Read_Binary_Restart( geo_file, system, control, data, workspace );
control->restart = 1;
}
else
{
fprintf( stderr, "unknown geo file format. terminating!\n" );
- exit( 1 );
+ exit( INVALID_GEO );
}
#if defined(DEBUG_FOCUS)
@@ -172,7 +189,14 @@ void Init_Data_Structures( simulation_data *data )
}
-int main( int argc, char* argv[] )
+static void usage(char* argv[])
+{
+ fprintf(stderr, "usage: ./%s geometry ffield control\n", argv[0]);
+}
+
+
+#ifdef HAVE_CUDA
+static void gpu_main( int argc, char* argv[] )
{
reax_system system;
control_params control;
@@ -183,7 +207,6 @@ int main( int argc, char* argv[] )
evolve_function Evolve;
evolve_function Cuda_Evolve;
int steps;
-
real t_start, t_elapsed;
real *results = NULL;
@@ -259,8 +282,8 @@ int main( int argc, char* argv[] )
#ifdef __BUILD_DEBUG__
if( !validate_device (&system, &data, &workspace, &lists) )
{
- fprintf (stderr, " Results does not match between Device and host @ step --> %d \n", data.step);
- exit (1);
+ fprintf( stderr, " Results does not match between Device and host @ step --> %d \n", data.step );
+ exit( 1 );
}
#endif
@@ -331,6 +354,92 @@ int main( int argc, char* argv[] )
fprintf( out_control.log, "total: %.2f secs\n", data.timing.elapsed );
Cleanup_Cuda_Environment( );
+}
+
+
+#else
+static void cpu_main( int argc, char* argv[] )
+{
+ reax_system system;
+ control_params control;
+ simulation_data data;
+ static_storage workspace;
+ list *lists;
+ output_controls out_control;
+ evolve_function Evolve;
+ int steps;
+
+ if ( argc != 4 )
+ {
+ usage(argv);
+ exit( INVALID_INPUT );
+ }
+
+ lists = (list*) malloc( sizeof(list) * LIST_N );
+
+ Read_System( argv[1], argv[2], argv[3], &system, &control,
+ &data, &workspace, &out_control );
+
+ Initialize( &system, &control, &data, &workspace, &lists,
+ &out_control, &Evolve );
+
+ /* compute f_0 */
+ //if( control.restart == 0 ) {
+ Reset( &system, &control, &data, &workspace, &lists );
+ Generate_Neighbor_Lists( &system, &control, &data, &workspace,
+ &lists, &out_control );
+
+ //fprintf( stderr, "total: %.2f secs\n", data.timing.nbrs);
+ Compute_Forces(&system, &control, &data, &workspace, &lists, &out_control);
+ Compute_Kinetic_Energy( &system, &data );
+ Output_Results(&system, &control, &data, &workspace, &lists, &out_control);
+ ++data.step;
+ //}
+
+
+ for ( ; data.step <= control.nsteps; data.step++ )
+ {
+ if ( control.T_mode )
+ {
+ Temperature_Control( &control, &data, &out_control );
+ }
+ Evolve( &system, &control, &data, &workspace, &lists, &out_control );
+ Post_Evolve( &system, &control, &data, &workspace, &lists, &out_control );
+ Output_Results(&system, &control, &data, &workspace, &lists, &out_control);
+ Analysis( &system, &control, &data, &workspace, &lists, &out_control );
+
+ steps = data.step - data.prev_steps;
+ if ( steps && out_control.restart_freq &&
+ steps % out_control.restart_freq == 0 )
+ Write_Restart( &system, &control, &data, &workspace, &out_control );
+ }
+
+ if ( out_control.write_steps > 0 )
+ {
+ fclose( out_control.trj );
+ Write_PDB( &system, &(lists[BONDS]), &data, &control, &workspace, &out_control );
+ }
+
+ data.timing.end = Get_Time( );
+ data.timing.elapsed = Get_Timing_Info( data.timing.start );
+ fprintf( out_control.log, "total: %.2f secs\n", data.timing.elapsed );
+}
+#endif
+
+
+int main( int argc, char* argv[] )
+{
+ if ( argc != 4 )
+ {
+ usage(argv);
+ exit( INVALID_INPUT );
+ }
+
+#ifdef HAVE_CUDA
+ gpu_main( argc, argv );
+#else
+ cpu_main( argc, argv );
+#endif
- return 0;
+ return SUCCESS;
}
diff --git a/PuReMD-GPU/src/three_body_interactions.c b/PuReMD-GPU/src/three_body_interactions.c
index 7ac96e057c6c799ba88204f3f6339fe54b3c61da..f128d2a2749ead3d5b9a08e47f45c0538255caac 100644
--- a/PuReMD-GPU/src/three_body_interactions.c
+++ b/PuReMD-GPU/src/three_body_interactions.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -21,59 +22,63 @@
#include "three_body_interactions.h"
#include "bond_orders.h"
+#include "index_utils.h"
#include "list.h"
#include "lookup.h"
#include "vector.h"
-#include "index_utils.h"
/* calculates the theta angle between i-j-k */
-void Calculate_Theta( rvec dvec_ji, real d_ji, rvec dvec_jk, real d_jk,
+void Calculate_Theta( rvec dvec_ji, real d_ji, rvec dvec_jk, real d_jk,
real *theta, real *cos_theta )
{
- (*cos_theta) = Dot( dvec_ji, dvec_jk, 3 ) / ( d_ji * d_jk );
- if( *cos_theta > 1. ) *cos_theta = 1.0;
- if( *cos_theta < -1. ) *cos_theta = -1.0;
+ (*cos_theta) = rvec_Dot( dvec_ji, dvec_jk ) / ( d_ji * d_jk );
+ if ( *cos_theta > 1. )
+ {
+ *cos_theta = 1.0;
+ }
+ if ( *cos_theta < -1. )
+ {
+ *cos_theta = -1.0;
+ }
(*theta) = ACOS( *cos_theta );
}
/* calculates the derivative of the cosine of the angle between i-j-k */
-void Calculate_dCos_Theta( rvec dvec_ji, real d_ji, rvec dvec_jk, real d_jk,
- rvec* dcos_theta_di, rvec* dcos_theta_dj,
- rvec* dcos_theta_dk )
+void Calculate_dCos_Theta( rvec dvec_ji, real d_ji, rvec dvec_jk, real d_jk,
+ rvec* dcos_theta_di, rvec* dcos_theta_dj, rvec* dcos_theta_dk )
{
int t;
real sqr_d_ji = SQR(d_ji);
real sqr_d_jk = SQR(d_jk);
real inv_dists = 1.0 / (d_ji * d_jk);
real inv_dists3 = POW( inv_dists, 3 );
- real dot_dvecs = Dot( dvec_ji, dvec_jk, 3 );
+ real dot_dvecs = rvec_Dot( dvec_ji, dvec_jk );
real Cdot_inv3 = dot_dvecs * inv_dists3;
- for( t = 0; t < 3; ++t ) {
- (*dcos_theta_di)[t] = dvec_jk[t] * inv_dists -
- Cdot_inv3 * sqr_d_jk * dvec_ji[t];
+ for ( t = 0; t < 3; ++t )
+ {
+ (*dcos_theta_di)[t] = dvec_jk[t] * inv_dists - Cdot_inv3 * sqr_d_jk * dvec_ji[t];
- (*dcos_theta_dj)[t] = -(dvec_jk[t] + dvec_ji[t]) * inv_dists +
- Cdot_inv3 * ( sqr_d_jk * dvec_ji[t] + sqr_d_ji * dvec_jk[t] );
+ (*dcos_theta_dj)[t] = -(dvec_jk[t] + dvec_ji[t]) * inv_dists
+ + Cdot_inv3 * ( sqr_d_jk * dvec_ji[t] + sqr_d_ji * dvec_jk[t] );
- (*dcos_theta_dk)[t] = dvec_ji[t] * inv_dists -
- Cdot_inv3 * sqr_d_ji * dvec_jk[t];
+ (*dcos_theta_dk)[t] = dvec_ji[t] * inv_dists - Cdot_inv3 * sqr_d_ji * dvec_jk[t];
}
- /*fprintf( stderr,
+ /*fprintf( stderr,
"%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e\n",
dvec_jk[t] * inv_dists*/
}
-/* this is a 3-body interaction in which the main role is
+/* this is a 3-body interaction in which the main role is
played by j which sits in the middle of the other two. */
-void Three_Body_Interactions( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+void Three_Body_Interactions( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
int i, j, pi, k, pk, t;
int type_i, type_j, type_k;
@@ -123,7 +128,8 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
p_val10 = system->reaxprm.gp.l[17];
num_thb_intrs = 0;
- for( j = 0; j < system->N; ++j ) {
+ for ( j = 0; j < system->N; ++j )
+ {
// fprintf( out_control->eval, "j: %d\n", j );
type_j = system->atoms[j].type;
start_j = Start_Index(j, bonds);
@@ -133,21 +139,24 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
p_val5 = system->reaxprm.sbp[ type_j ].p_val5;
SBOp = 0, prod_SBO = 1;
- for( t = start_j; t < end_j; ++t ) {
+ for ( t = start_j; t < end_j; ++t )
+ {
bo_jt = &(bond_list[t].bo_data);
SBOp += (bo_jt->BO_pi + bo_jt->BO_pi2);
temp = SQR( bo_jt->BO );
- temp *= temp;
+ temp *= temp;
temp *= temp;
prod_SBO *= EXP( -temp );
}
/* modifications to match Adri's code - 09/01/09 */
- if( workspace->vlpex[j] >= 0 ){
+ if ( workspace->vlpex[j] >= 0 )
+ {
vlpadj = 0;
dSBO2 = prod_SBO - 1;
}
- else{
+ else
+ {
vlpadj = workspace->nlp[j];
dSBO2 = (prod_SBO - 1) * (1 - p_val8 * workspace->dDelta_lp[j]);
}
@@ -155,65 +164,67 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
SBO = SBOp + (1 - prod_SBO) * (-workspace->Delta_boc[j] - p_val8 * vlpadj);
dSBO1 = -8 * prod_SBO * ( workspace->Delta_boc[j] + p_val8 * vlpadj );
- if( SBO <= 0 )
+ if ( SBO <= 0 )
SBO2 = 0, CSBO2 = 0;
- else if( SBO > 0 && SBO <= 1 ) {
+ else if ( SBO > 0 && SBO <= 1 )
+ {
SBO2 = POW( SBO, p_val9 );
CSBO2 = p_val9 * POW( SBO, p_val9 - 1 );
}
- else if( SBO > 1 && SBO < 2 ) {
- SBO2 = 2 - POW( 2-SBO, p_val9 );
+ else if ( SBO > 1 && SBO < 2 )
+ {
+ SBO2 = 2 - POW( 2 - SBO, p_val9 );
CSBO2 = p_val9 * POW( 2 - SBO, p_val9 - 1 );
}
- else
- SBO2 = 2, CSBO2 = 0;
+ else
+ SBO2 = 2, CSBO2 = 0;
expval6 = EXP( p_val6 * workspace->Delta_boc[j] );
- /* unlike 2-body intrs where we enforce i<j, we cannot put any such
- restrictions here. such a restriction would prevent us from producing
+ /* unlike 2-body intrs where we enforce i<j, we cannot put any such
+ restrictions here. such a restriction would prevent us from producing
all 4-body intrs correctly */
- for( pi = start_j; pi < end_j; ++pi ) {
+ for ( pi = start_j; pi < end_j; ++pi )
+ {
Set_Start_Index( pi, num_thb_intrs, thb_intrs );
-
pbond_ij = &(bond_list[pi]);
bo_ij = &(pbond_ij->bo_data);
BOA_ij = bo_ij->BO - control->thb_cut;
- if( BOA_ij/*bo_ij->BO*/ > (real) 0.0 ) {
+ if ( BOA_ij/*bo_ij->BO*/ > 0.0 )
+ {
i = pbond_ij->nbr;
- r_ij = pbond_ij->d;
+ r_ij = pbond_ij->d;
type_i = system->atoms[i].type;
// fprintf( out_control->eval, "i: %d\n", i );
/* first copy 3-body intrs from previously computed ones where i>k.
- IMPORTANT: if it is less costly to compute theta and its
- derivative, we should definitely re-compute them,
+ IMPORTANT: if it is less costly to compute theta and its
+ derivative, we should definitely re-compute them,
instead of copying!
- in the second for-loop below, we compute only new 3-body intrs
+ in the second for-loop below, we compute only new 3-body intrs
where i < k */
- for( pk = start_j; pk < pi; ++pk ) {
+ for ( pk = start_j; pk < pi; ++pk )
+ {
// fprintf( out_control->eval, "pk: %d\n", pk );
start_pk = Start_Index( pk, thb_intrs );
end_pk = End_Index( pk, thb_intrs );
- for( t = start_pk; t < end_pk; ++t )
- if( thb_list[t].thb == i ) {
+ for ( t = start_pk; t < end_pk; ++t )
+ if ( thb_list[t].thb == i )
+ {
p_ijk = &(thb_list[num_thb_intrs]);
p_kji = &(thb_list[t]);
p_ijk->thb = bond_list[pk].nbr;
p_ijk->pthb = pk;
- p_ijk->theta = p_kji->theta;
+ p_ijk->theta = p_kji->theta;
rvec_Copy( p_ijk->dcos_di, p_kji->dcos_dk );
rvec_Copy( p_ijk->dcos_dj, p_kji->dcos_dj );
rvec_Copy( p_ijk->dcos_dk, p_kji->dcos_di );
- //if (j == 12)
- //fprintf (stderr, "Adding one for matched atom %d \n", i);
-
++num_thb_intrs;
break;
}
@@ -221,7 +232,8 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
/* and this is the second for loop mentioned above */
- for( pk = pi+1; pk < end_j; ++pk ) {
+ for ( pk = pi + 1; pk < end_j; ++pk )
+ {
pbond_jk = &(bond_list[pk]);
bo_jk = &(pbond_jk->bo_data);
BOA_jk = bo_jk->BO - control->thb_cut;
@@ -229,53 +241,55 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
type_k = system->atoms[k].type;
p_ijk = &( thb_list[num_thb_intrs] );
- //TODO - CHANGE ORIGINAL
+ //CHANGE ORIGINAL
if (BOA_jk <= 0) continue;
+ //CHANGE ORIGINAL
- Calculate_Theta( pbond_ij->dvec, pbond_ij->d,
- pbond_jk->dvec, pbond_jk->d,
- &theta, &cos_theta );
- Calculate_dCos_Theta( pbond_ij->dvec, pbond_ij->d,
- pbond_jk->dvec, pbond_jk->d,
- &(p_ijk->dcos_di), &(p_ijk->dcos_dj),
- &(p_ijk->dcos_dk) );
+ Calculate_Theta( pbond_ij->dvec, pbond_ij->d,
+ pbond_jk->dvec, pbond_jk->d,
+ &theta, &cos_theta );
+
+ Calculate_dCos_Theta( pbond_ij->dvec, pbond_ij->d,
+ pbond_jk->dvec, pbond_jk->d,
+ &(p_ijk->dcos_di), &(p_ijk->dcos_dj),
+ &(p_ijk->dcos_dk) );
p_ijk->thb = k;
p_ijk->pthb = pk;
p_ijk->theta = theta;
- //if (j == 12)
- //fprintf (stderr, "Adding one for the rest %d \n", k);
-
sin_theta = SIN( theta );
- if( sin_theta < 1.0e-5 )
+ if ( sin_theta < 1.0e-5 )
sin_theta = 1.0e-5;
++num_thb_intrs;
- if( BOA_jk > 0.0 &&
- (bo_ij->BO * bo_jk->BO) > SQR(control->thb_cut)/*0*/) {
- r_jk = pbond_jk->d;
+ if ( BOA_jk > 0.0 &&
+ (bo_ij->BO * bo_jk->BO) > SQR(control->thb_cut)/*0*/)
+ {
+ r_jk = pbond_jk->d;
thbh = &( system->reaxprm.thbp[ index_thbp(type_i,type_j,type_k,system->reaxprm.num_atom_types) ] );
flag = 0;
/* if( workspace->orig_id[i] < workspace->orig_id[k] )
- fprintf( stdout, "%6d %6d %6d %7.3f %7.3f %7.3f\n",
+ fprintf( stdout, "%6d %6d %6d %7.3f %7.3f %7.3f\n",
workspace->orig_id[i], workspace->orig_id[j],
workspace->orig_id[k], bo_ij->BO, bo_jk->BO, p_ijk->theta );
- else
- fprintf( stdout, "%6d %6d %6d %7.3f %7.3f %7.3f\n",
+ else
+ fprintf( stdout, "%6d %6d %6d %7.3f %7.3f %7.3f\n",
workspace->orig_id[k], workspace->orig_id[j],
workspace->orig_id[i], bo_jk->BO, bo_ij->BO, p_ijk->theta ); */
- for( cnt = 0; cnt < thbh->cnt; ++cnt ) {
- // fprintf( out_control->eval,
+ for ( cnt = 0; cnt < thbh->cnt; ++cnt )
+ {
+ // fprintf( out_control->eval,
// "%6d%6d%6d -- exists in thbp\n", i+1, j+1, k+1 );
- if( fabs(thbh->prm[cnt].p_val1) > 0.001 ) {
+ if ( fabs(thbh->prm[cnt].p_val1) > 0.001 )
+ {
thbp = &( thbh->prm[cnt] );
/* ANGLE ENERGY */
@@ -287,27 +301,27 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
exp3ij = EXP( -p_val3 * POW( BOA_ij, p_val4 ) );
f7_ij = 1.0 - exp3ij;
- Cf7ij = p_val3 * p_val4 *
- POW( BOA_ij, p_val4 - 1.0 ) * exp3ij;
+ Cf7ij = p_val3 * p_val4 *
+ POW( BOA_ij, p_val4 - 1.0 ) * exp3ij;
exp3jk = EXP( -p_val3 * POW( BOA_jk, p_val4 ) );
f7_jk = 1.0 - exp3jk;
- Cf7jk = p_val3 * p_val4 *
- POW( BOA_jk, p_val4 - 1.0 ) * exp3jk;
+ Cf7jk = p_val3 * p_val4 *
+ POW( BOA_jk, p_val4 - 1.0 ) * exp3jk;
expval7 = EXP( -p_val7 * workspace->Delta_boc[j] );
trm8 = 1.0 + expval6 + expval7;
f8_Dj = p_val5 - ( (p_val5 - 1.0) * (2.0 + expval6) / trm8 );
Cf8j = ( (1.0 - p_val5) / SQR(trm8) ) *
- (p_val6 * expval6 * trm8 -
- (2.0 + expval6) * ( p_val6 * expval6 - p_val7 * expval7 ));
+ (p_val6 * expval6 * trm8 -
+ (2.0 + expval6) * ( p_val6 * expval6 - p_val7 * expval7 ));
- theta_0 = 180.0 -
- theta_00 * (1.0 - EXP(-p_val10 * (2.0 - SBO2)));
- theta_0 = DEG2RAD( theta_0 );
+ theta_0 = 180.0 -
+ theta_00 * (1.0 - EXP(-p_val10 * (2.0 - SBO2)));
+ theta_0 = DEG2RAD( theta_0 );
- expval2theta = EXP(-p_val2 * SQR(theta_0-theta));
- if( p_val1 >= 0 )
+ expval2theta = EXP(-p_val2 * SQR(theta_0 - theta));
+ if ( p_val1 >= 0 )
expval12theta = p_val1 * (1.0 - expval2theta);
else // To avoid linear Me-H-Me angles (6/6/06)
expval12theta = p_val1 * -expval2theta;
@@ -315,11 +329,11 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
CEval1 = Cf7ij * f7_jk * f8_Dj * expval12theta;
CEval2 = Cf7jk * f7_ij * f8_Dj * expval12theta;
CEval3 = Cf8j * f7_ij * f7_jk * expval12theta;
- CEval4 = -2.0 * p_val1 * p_val2 * f7_ij * f7_jk * f8_Dj *
- expval2theta * (theta_0 - theta);
+ CEval4 = -2.0 * p_val1 * p_val2 * f7_ij * f7_jk * f8_Dj *
+ expval2theta * (theta_0 - theta);
- Ctheta_0 = p_val10 * DEG2RAD(theta_00) *
- exp( -p_val10 * (2.0 - SBO2) );
+ Ctheta_0 = p_val10 * DEG2RAD(theta_00) *
+ exp( -p_val10 * (2.0 - SBO2) );
CEval5 = -CEval4 * Ctheta_0 * CSBO2;
CEval6 = CEval5 * dSBO1;
@@ -342,13 +356,13 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
exp_pen4 = EXP( p_pen4 * workspace->Delta[j] );
trm_pen34 = 1.0 + exp_pen3 + exp_pen4;
f9_Dj = ( 2.0 + exp_pen3 ) / trm_pen34;
- Cf9j = (-p_pen3 * exp_pen3 * trm_pen34 -
+ Cf9j = (-p_pen3 * exp_pen3 * trm_pen34 -
(2.0 + exp_pen3) * ( -p_pen3 * exp_pen3 +
- p_pen4 * exp_pen4 )) /
- SQR( trm_pen34 );
+ p_pen4 * exp_pen4 )) /
+ SQR( trm_pen34 );
- data->E_Pen += e_pen =
- p_pen1 * f9_Dj * exp_pen2ij * exp_pen2jk;
+ data->E_Pen += e_pen =
+ p_pen1 * f9_Dj * exp_pen2ij * exp_pen2jk;
CEpen1 = e_pen * Cf9j / f9_Dj;
temp = -2.0 * p_pen2 * e_pen;
@@ -364,66 +378,64 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
p_coa4 = system->reaxprm.gp.l[30];
exp_coa2 = EXP( p_coa2 * workspace->Delta_boc[j] );
- data->E_Coa += e_coa =
- p_coa1 / (1. + exp_coa2) *
- EXP( -p_coa3 * SQR(total_bo[i] - BOA_ij) ) *
- EXP( -p_coa3 * SQR(total_bo[k] - BOA_jk) ) *
- EXP( -p_coa4 * SQR(BOA_ij - 1.5) ) *
- EXP( -p_coa4 * SQR(BOA_jk - 1.5) );
+ data->E_Coa += e_coa =
+ p_coa1 / (1. + exp_coa2) *
+ EXP( -p_coa3 * SQR(total_bo[i] - BOA_ij) ) *
+ EXP( -p_coa3 * SQR(total_bo[k] - BOA_jk) ) *
+ EXP( -p_coa4 * SQR(BOA_ij - 1.5) ) *
+ EXP( -p_coa4 * SQR(BOA_jk - 1.5) );
CEcoa1 = -2 * p_coa4 * (BOA_ij - 1.5) * e_coa;
CEcoa2 = -2 * p_coa4 * (BOA_jk - 1.5) * e_coa;
- CEcoa3 = -p_coa2 * exp_coa2 * e_coa / (1+exp_coa2);
- CEcoa4 = -2*p_coa3 * (total_bo[i]-BOA_ij) * e_coa;
- CEcoa5 = -2*p_coa3 * (total_bo[k]-BOA_jk) * e_coa;
+ CEcoa3 = -p_coa2 * exp_coa2 * e_coa / (1 + exp_coa2);
+ CEcoa4 = -2 * p_coa3 * (total_bo[i] - BOA_ij) * e_coa;
+ CEcoa5 = -2 * p_coa3 * (total_bo[k] - BOA_jk) * e_coa;
/* END COALITION ENERGY */
/* FORCES */
- bo_ij->Cdbo += (CEval1 + CEpen2 + (CEcoa1-CEcoa4));
- bo_jk->Cdbo += (CEval2 + CEpen3 + (CEcoa2-CEcoa5));
- workspace->CdDelta[j] += ((CEval3 + CEval7) +
- CEpen1 + CEcoa3);
+ bo_ij->Cdbo += (CEval1 + CEpen2 + (CEcoa1 - CEcoa4));
+ bo_jk->Cdbo += (CEval2 + CEpen3 + (CEcoa2 - CEcoa5));
+ workspace->CdDelta[j] += ((CEval3 + CEval7) +
+ CEpen1 + CEcoa3);
workspace->CdDelta[i] += CEcoa4;
- workspace->CdDelta[k] += CEcoa5;
+ workspace->CdDelta[k] += CEcoa5;
- for( t = start_j; t < end_j; ++t ) {
+ for ( t = start_j; t < end_j; ++t )
+ {
pbond_jt = &( bond_list[t] );
bo_jt = &(pbond_jt->bo_data);
temp_bo_jt = bo_jt->BO;
temp = CUBE( temp_bo_jt );
- pBOjt7 = temp * temp * temp_bo_jt;
+ pBOjt7 = temp * temp * temp_bo_jt;
- // fprintf( out_control->eval, "%6d%12.8f\n",
- // workspace->orig_id[ bond_list[t].nbr ],
+ // fprintf( out_control->eval, "%6d%12.8f\n",
+ // workspace->orig_id[ bond_list[t].nbr ],
// (CEval6 * pBOjt7) );
bo_jt->Cdbo += (CEval6 * pBOjt7);
bo_jt->Cdbopi += CEval5;
bo_jt->Cdbopi2 += CEval5;
- }
-
+ }
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT) {
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT)
+ {
rvec_ScaledAdd( system->atoms[i].f, CEval8, p_ijk->dcos_di );
rvec_ScaledAdd( system->atoms[j].f, CEval8, p_ijk->dcos_dj );
rvec_ScaledAdd( system->atoms[k].f, CEval8, p_ijk->dcos_dk );
-
- /*
- if (i == 0) fprintf (stderr, " atom %d adding to i (j) = 0\n", j);
- if (k == 0) fprintf (stderr, " atom %d adding to i (k) = 0\n", j);
- */
}
- else {
+ else
+ {
/* terms not related to bond order derivatives
- are added directly into
+ are added directly into
forces and pressure vector/tensor */
rvec_Scale( force, CEval8, p_ijk->dcos_di );
rvec_Add( system->atoms[i].f, force );
rvec_iMultiply( ext_press, pbond_ij->rel_box, force );
rvec_Add( data->ext_press, ext_press );
- rvec_ScaledAdd( system->atoms[j].f, CEval8, p_ijk->dcos_dj );
+ rvec_ScaledAdd( system->atoms[j].f,
+ CEval8, p_ijk->dcos_dj );
rvec_Scale( force, CEval8, p_ijk->dcos_dk );
rvec_Add( system->atoms[k].f, force );
@@ -432,96 +444,97 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
/* This part is for a fully-flexible box */
- /* rvec_OuterProduct( temp_rtensor,
+ /* rvec_OuterProduct( temp_rtensor,
p_ijk->dcos_di, system->atoms[i].x );
rtensor_Scale( total_rtensor, +CEval8, temp_rtensor );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
p_ijk->dcos_dj, system->atoms[j].x );
rtensor_ScaledAdd(total_rtensor, CEval8, temp_rtensor);
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
p_ijk->dcos_dk, system->atoms[k].x );
rtensor_ScaledAdd(total_rtensor, CEval8, temp_rtensor);
if( pbond_ij->imaginary || pbond_jk->imaginary )
- rtensor_ScaledAdd( data->flex_bar.P,
+ rtensor_ScaledAdd( data->flex_bar.P,
-1.0, total_rtensor );
else
rtensor_Add( data->flex_bar.P, total_rtensor ); */
}
#ifdef TEST_ENERGY
- fprintf( out_control->eval,
- //"%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e",
- "%6d%6d%6d%23.15e%23.15e%23.15e\n",
- i+1, j+1, k+1,
- //workspace->orig_id[i]+1,
- //workspace->orig_id[j]+1,
- //workspace->orig_id[k]+1,
- //workspace->Delta_boc[j],
- RAD2DEG(theta), /*BOA_ij, BOA_jk, */
- e_ang, data->E_Ang );
-
- /*fprintf( out_control->eval,
+ fprintf( out_control->eval,
+ //"%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e",
+ "%6d%6d%6d%23.15e%23.15e%23.15e\n",
+ i + 1, j + 1, k + 1,
+ //workspace->orig_id[i]+1,
+ //workspace->orig_id[j]+1,
+ //workspace->orig_id[k]+1,
+ //workspace->Delta_boc[j],
+ RAD2DEG(theta), /*BOA_ij, BOA_jk, */
+ e_ang, data->E_Ang );
+
+ /*fprintf( out_control->eval,
"%23.15e%23.15e%23.15e%23.15e",
p_val3, p_val4, BOA_ij, BOA_jk );
- fprintf( out_control->eval,
+ fprintf( out_control->eval,
"%23.15e%23.15e%23.15e%23.15e",
f7_ij, f7_jk, f8_Dj, expval12theta );
- fprintf( out_control->eval,
+ fprintf( out_control->eval,
"%23.15e%23.15e%23.15e%23.15e%23.15e\n",
CEval1, CEval2, CEval3, CEval4, CEval5
- //CEval6, CEval7, CEval8 );*/
+ //CEval6, CEval7, CEval8 );*/
- /*fprintf( out_control->eval,
+ /*fprintf( out_control->eval,
"%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e%23.15e\n",
- -p_ijk->dcos_di[0]/sin_theta,
- -p_ijk->dcos_di[1]/sin_theta,
- -p_ijk->dcos_di[2]/sin_theta,
- -p_ijk->dcos_dj[0]/sin_theta,
- -p_ijk->dcos_dj[1]/sin_theta,
- -p_ijk->dcos_dj[2]/sin_theta,
- -p_ijk->dcos_dk[0]/sin_theta,
- -p_ijk->dcos_dk[1]/sin_theta,
+ -p_ijk->dcos_di[0]/sin_theta,
+ -p_ijk->dcos_di[1]/sin_theta,
+ -p_ijk->dcos_di[2]/sin_theta,
+ -p_ijk->dcos_dj[0]/sin_theta,
+ -p_ijk->dcos_dj[1]/sin_theta,
+ -p_ijk->dcos_dj[2]/sin_theta,
+ -p_ijk->dcos_dk[0]/sin_theta,
+ -p_ijk->dcos_dk[1]/sin_theta,
-p_ijk->dcos_dk[2]/sin_theta );*/
- /* fprintf( out_control->epen,
- "%23.15e%23.15e%23.15e\n",
+ /* fprintf( out_control->epen,
+ "%23.15e%23.15e%23.15e\n",
CEpen1, CEpen2, CEpen3 );
- fprintf( out_control->epen,
+ fprintf( out_control->epen,
"%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
workspace->orig_id[i], workspace->orig_id[j],
- workspace->orig_id[k], RAD2DEG(theta),
+ workspace->orig_id[k], RAD2DEG(theta),
BOA_ij, BOA_jk, e_pen, data->E_Pen ); */
- fprintf( out_control->ecoa,
- "%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
- workspace->orig_id[i],
- workspace->orig_id[j],
- workspace->orig_id[k],
- RAD2DEG(theta), BOA_ij, BOA_jk,
- e_coa, data->E_Coa );
+ fprintf( out_control->ecoa,
+ "%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
+ workspace->orig_id[i],
+ workspace->orig_id[j],
+ workspace->orig_id[k],
+ RAD2DEG(theta), BOA_ij, BOA_jk,
+ e_coa, data->E_Coa );
#endif
#ifdef TEST_FORCES /* angle forces */
Add_dBO( system, lists, j, pi, CEval1, workspace->f_ang );
Add_dBO( system, lists, j, pk, CEval2, workspace->f_ang );
- Add_dDelta( system, lists,
- j, CEval3 + CEval7, workspace->f_ang );
+ Add_dDelta( system, lists,
+ j, CEval3 + CEval7, workspace->f_ang );
- for( t = start_j; t < end_j; ++t ) {
+ for ( t = start_j; t < end_j; ++t )
+ {
pbond_jt = &( bond_list[t] );
bo_jt = &(pbond_jt->bo_data);
temp_bo_jt = bo_jt->BO;
temp = CUBE( temp_bo_jt );
- pBOjt7 = temp * temp * temp_bo_jt;
+ pBOjt7 = temp * temp * temp_bo_jt;
Add_dBO( system, lists, j, t, pBOjt7 * CEval6,
- workspace->f_ang );
- Add_dBOpinpi2( system, lists, j, t,
- CEval5, CEval5,
- workspace->f_ang, workspace->f_ang );
+ workspace->f_ang );
+ Add_dBOpinpi2( system, lists, j, t,
+ CEval5, CEval5,
+ workspace->f_ang, workspace->f_ang );
}
rvec_ScaledAdd( workspace->f_ang[i], CEval8, p_ijk->dcos_di );
@@ -536,10 +549,10 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
/* end penalty forces */
/* coalition forces */
- Add_dBO( system, lists,
- j, pi, CEcoa1-CEcoa4, workspace->f_coa );
- Add_dBO( system, lists,
- j, pk, CEcoa2-CEcoa5, workspace->f_coa );
+ Add_dBO( system, lists,
+ j, pi, CEcoa1 - CEcoa4, workspace->f_coa );
+ Add_dBO( system, lists,
+ j, pk, CEcoa2 - CEcoa5, workspace->f_coa );
Add_dDelta( system, lists, j, CEcoa3, workspace->f_coa );
Add_dDelta( system, lists, i, CEcoa4, workspace->f_coa );
Add_dDelta( system, lists, k, CEcoa5, workspace->f_coa );
@@ -555,32 +568,36 @@ void Three_Body_Interactions( reax_system *system, control_params *control,
}
}
- if( num_thb_intrs >= thb_intrs->num_intrs * DANGER_ZONE ) {
+
+ if ( num_thb_intrs >= thb_intrs->num_intrs * DANGER_ZONE )
+ {
workspace->realloc.num_3body = num_thb_intrs;
- if( num_thb_intrs > thb_intrs->num_intrs ) {
+ if ( num_thb_intrs > thb_intrs->num_intrs )
+ {
fprintf( stderr, "step%d-ran out of space on angle_list: top=%d, max=%d",
- data->step, num_thb_intrs, thb_intrs->num_intrs );
- exit( INSUFFICIENT_SPACE );
+ data->step, num_thb_intrs, thb_intrs->num_intrs );
+ exit( INSUFFICIENT_MEMORY );
}
}
- //fprintf( stderr,"%d: Number of angle interactions: %d\n",
+ //fprintf( stderr,"%d: Number of angle interactions: %d\n",
// data->step, num_thb_intrs );
#ifdef TEST_ENERGY
- fprintf( stderr,"Number of angle interactions: %d\n", num_thb_intrs );
+ fprintf( stderr, "Number of angle interactions: %d\n", num_thb_intrs );
- fprintf( stderr,"Angle Energy:%g\t Penalty Energy:%g\t Coalition Energy:%g\n",
- data->E_Ang, data->E_Pen, data->E_Coa );
+ fprintf( stderr, "Angle Energy:%g\t Penalty Energy:%g\t Coalition Energy:%g\n",
+ data->E_Ang, data->E_Pen, data->E_Coa );
- fprintf( stderr,"3body: ext_press (%23.15e %23.15e %23.15e)\n",
- data->ext_press[0], data->ext_press[1], data->ext_press[2] );
+ fprintf( stderr, "3body: ext_press (%23.15e %23.15e %23.15e)\n",
+ data->ext_press[0], data->ext_press[1], data->ext_press[2] );
#endif
}
-void Hydrogen_Bonds( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+
+void Hydrogen_Bonds( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i, j, k, pi, pk, itr, top;
int type_i, type_j, type_k;
@@ -610,10 +627,11 @@ void Hydrogen_Bonds( reax_system *system, control_params *control,
/* loops below discover the Hydrogen bonds between i-j-k triplets.
here j is H atom and there has to be some bond between i and j.
Hydrogen bond is between j and k.
- so in this function i->X, j->H, k->Z when we map
+ so in this function i->X, j->H, k->Z when we map
variables onto the ones in the handout.*/
- for( j = 0; j < system->N; ++j )
- if( system->reaxprm.sbp[system->atoms[j].type].p_hbond==1 ) {// j must be H
+ for ( j = 0; j < system->N; ++j )
+ if ( system->reaxprm.sbp[system->atoms[j].type].p_hbond == 1 ) // j must be H
+ {
/*set j's variables */
type_j = system->atoms[j].type;
start_j = Start_Index(j, bonds);
@@ -622,21 +640,23 @@ void Hydrogen_Bonds( reax_system *system, control_params *control,
hb_end_j = End_Index ( workspace->hbond_index[j], hbonds );
top = 0;
- for( pi = start_j; pi < end_j; ++pi ) {
+ for ( pi = start_j; pi < end_j; ++pi )
+ {
pbond_ij = &( bond_list[pi] );
i = pbond_ij->nbr;
bo_ij = &(pbond_ij->bo_data);
type_i = system->atoms[i].type;
- if( system->reaxprm.sbp[type_i].p_hbond == 2 &&
+ if ( system->reaxprm.sbp[type_i].p_hbond == 2 &&
bo_ij->BO >= HB_THRESHOLD )
hblist[top++] = pi;
}
- // fprintf( stderr, "j: %d, top: %d, hb_start_j: %d, hb_end_j:%d\n",
+ // fprintf( stderr, "j: %d, top: %d, hb_start_j: %d, hb_end_j:%d\n",
// j, top, hb_start_j, hb_end_j );
- for( pk = hb_start_j; pk < hb_end_j; ++pk ) {
+ for ( pk = hb_start_j; pk < hb_end_j; ++pk )
+ {
/* set k's varibles */
k = hbond_list[pk].nbr;
type_k = system->atoms[k].type;
@@ -644,69 +664,59 @@ void Hydrogen_Bonds( reax_system *system, control_params *control,
r_jk = nbr_jk->d;
rvec_Scale( dvec_jk, hbond_list[pk].scl, nbr_jk->dvec );
- for( itr=0; itr < top; ++itr ) {
+ for ( itr = 0; itr < top; ++itr )
+ {
pi = hblist[itr];
pbond_ij = &( bond_list[pi] );
i = pbond_ij->nbr;
- if( i != k ) {
+ if ( i != k )
+ {
bo_ij = &(pbond_ij->bo_data);
type_i = system->atoms[i].type;
- r_ij = pbond_ij->d;
+ r_ij = pbond_ij->d;
hbp = &(system->reaxprm.hbp[ index_hbp(type_i, type_j, type_k, system->reaxprm.num_atom_types) ]);
++num_hb_intrs;
- Calculate_Theta( pbond_ij->dvec, pbond_ij->d, dvec_jk, r_jk,
- &theta, &cos_theta );
+ Calculate_Theta( pbond_ij->dvec, pbond_ij->d, dvec_jk, r_jk, &theta, &cos_theta );
/* the derivative of cos(theta) */
Calculate_dCos_Theta( pbond_ij->dvec, pbond_ij->d, dvec_jk, r_jk,
- &dcos_theta_di, &dcos_theta_dj,
- &dcos_theta_dk );
+ &dcos_theta_di, &dcos_theta_dj, &dcos_theta_dk );
/* hydrogen bond energy*/
- sin_theta2 = SIN( theta/2.0 );
+ sin_theta2 = SIN( theta / 2.0 );
sin_xhz4 = SQR(sin_theta2);
sin_xhz4 *= sin_xhz4;
cos_xhz1 = ( 1.0 - cos_theta );
exp_hb2 = EXP( -hbp->p_hb2 * bo_ij->BO );
- exp_hb3 = EXP( -hbp->p_hb3 * ( hbp->r0_hb / r_jk +
- r_jk / hbp->r0_hb - 2.0 ) );
+ exp_hb3 = EXP( -hbp->p_hb3 * ( hbp->r0_hb / r_jk
+ + r_jk / hbp->r0_hb - 2.0 ) );
- data->E_HB += e_hb =
- hbp->p_hb1 * (1.0 - exp_hb2) * exp_hb3 * sin_xhz4;
+ data->E_HB += e_hb = hbp->p_hb1 * (1.0 - exp_hb2) * exp_hb3 * sin_xhz4;
- CEhb1 = hbp->p_hb1*hbp->p_hb2 * exp_hb2*exp_hb3 * sin_xhz4;
- CEhb2 = -hbp->p_hb1/2.0*(1.0 - exp_hb2) * exp_hb3 * cos_xhz1;
- CEhb3 = -hbp->p_hb3 * e_hb * (-hbp->r0_hb / SQR(r_jk) +
- 1.0 / hbp->r0_hb);
+ CEhb1 = hbp->p_hb1 * hbp->p_hb2 * exp_hb2 * exp_hb3 * sin_xhz4;
+ CEhb2 = -hbp->p_hb1 / 2.0 * (1.0 - exp_hb2) * exp_hb3 * cos_xhz1;
+ CEhb3 = -hbp->p_hb3 * e_hb * (-hbp->r0_hb / SQR(r_jk) + 1.0 / hbp->r0_hb);
/* hydrogen bond forces */
bo_ij->Cdbo += CEhb1; // dbo term
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT ) {
- rvec_ScaledAdd( system->atoms[i].f,
- +CEhb2, dcos_theta_di ); //dcos terms
- rvec_ScaledAdd( system->atoms[j].f,
- +CEhb2, dcos_theta_dj );
-
-
-
-
- //TODO
- rvec_ScaledAdd( system->atoms[k].f,
- +CEhb2, dcos_theta_dk );
-
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT)
+ {
+ rvec_ScaledAdd( system->atoms[i].f,
+ +CEhb2, dcos_theta_di ); //dcos terms
+ rvec_ScaledAdd( system->atoms[j].f,
+ +CEhb2, dcos_theta_dj );
+ rvec_ScaledAdd( system->atoms[k].f,
+ +CEhb2, dcos_theta_dk );
//dr terms
- rvec_ScaledAdd( system->atoms[j].f, -CEhb3/r_jk, dvec_jk );
-
-
- //TODO
- rvec_ScaledAdd( system->atoms[k].f, +CEhb3/r_jk, dvec_jk );
+ rvec_ScaledAdd( system->atoms[j].f, -CEhb3 / r_jk, dvec_jk );
+ rvec_ScaledAdd( system->atoms[k].f, +CEhb3 / r_jk, dvec_jk );
}
else
{
- /* for pressure coupling, terms that are not related
- to bond order derivatives are added directly into
+ /* for pressure coupling, terms that are not related
+ to bond order derivatives are added directly into
pressure vector/tensor */
rvec_Scale( force, +CEhb2, dcos_theta_di ); // dcos terms
rvec_Add( system->atoms[i].f, force );
@@ -717,39 +727,32 @@ void Hydrogen_Bonds( reax_system *system, control_params *control,
ivec_Scale( rel_jk, hbond_list[pk].scl, nbr_jk->rel_box );
rvec_Scale( force, +CEhb2, dcos_theta_dk );
-
-
-
- //TODO
rvec_Add( system->atoms[k].f, force );
-
-
-
rvec_iMultiply( ext_press, rel_jk, force );
rvec_ScaledAdd( data->ext_press, 1.0, ext_press );
//dr terms
- rvec_ScaledAdd( system->atoms[j].f, -CEhb3/r_jk, dvec_jk );
+ rvec_ScaledAdd( system->atoms[j].f, -CEhb3 / r_jk, dvec_jk );
- rvec_Scale( force, CEhb3/r_jk, dvec_jk );
+ rvec_Scale( force, CEhb3 / r_jk, dvec_jk );
rvec_Add( system->atoms[k].f, force );
rvec_iMultiply( ext_press, rel_jk, force );
rvec_ScaledAdd( data->ext_press, 1.0, ext_press );
/* This part is intended for a fully-flexible box */
- /* rvec_OuterProduct( temp_rtensor,
+ /* rvec_OuterProduct( temp_rtensor,
dcos_theta_di, system->atoms[i].x );
rtensor_Scale( total_rtensor, -CEhb2, temp_rtensor );
rvec_ScaledSum( temp_rvec, -CEhb2, dcos_theta_dj,
-CEhb3/r_jk, pbond_jk->dvec );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[j].x );
rtensor_Add( total_rtensor, temp_rtensor );
rvec_ScaledSum( temp_rvec, -CEhb2, dcos_theta_dk,
+CEhb3/r_jk, pbond_jk->dvec );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
temp_rvec, system->atoms[k].x );
rtensor_Add( total_rtensor, temp_rtensor );
@@ -760,38 +763,38 @@ void Hydrogen_Bonds( reax_system *system, control_params *control,
}
#ifdef TEST_ENERGY
- /*fprintf( out_control->ehb,
+ /*fprintf( out_control->ehb,
"%23.15e%23.15e%23.15e\n%23.15e%23.15e%23.15e\n%23.15e%23.15e%23.15e\n",
- dcos_theta_di[0], dcos_theta_di[1], dcos_theta_di[2],
- dcos_theta_dj[0], dcos_theta_dj[1], dcos_theta_dj[2],
+ dcos_theta_di[0], dcos_theta_di[1], dcos_theta_di[2],
+ dcos_theta_dj[0], dcos_theta_dj[1], dcos_theta_dj[2],
dcos_theta_dk[0], dcos_theta_dk[1], dcos_theta_dk[2]);
fprintf( out_control->ehb, "%23.15e%23.15e%23.15e\n",
CEhb1, CEhb2, CEhb3 ); */
- fprintf( stderr, //out_control->ehb,
- "%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
- workspace->orig_id[i],
- workspace->orig_id[j],
- workspace->orig_id[k],
- r_jk, theta, bo_ij->BO, e_hb, data->E_HB );
+ fprintf( stderr, //out_control->ehb,
+ "%6d%6d%6d%23.15e%23.15e%23.15e%23.15e%23.15e\n",
+ workspace->orig_id[i],
+ workspace->orig_id[j],
+ workspace->orig_id[k],
+ r_jk, theta, bo_ij->BO, e_hb, data->E_HB );
#endif
#ifdef TEST_FORCES
// dbo term
Add_dBO( system, lists, j, pi, +CEhb1, workspace->f_hb );
// dcos terms
- rvec_ScaledAdd( workspace->f_hb[i], +CEhb2, dcos_theta_di );
+ rvec_ScaledAdd( workspace->f_hb[i], +CEhb2, dcos_theta_di );
rvec_ScaledAdd( workspace->f_hb[j], +CEhb2, dcos_theta_dj );
rvec_ScaledAdd( workspace->f_hb[k], +CEhb2, dcos_theta_dk );
// dr terms
- rvec_ScaledAdd( workspace->f_hb[j], -CEhb3/r_jk, dvec_jk );
- rvec_ScaledAdd( workspace->f_hb[k], +CEhb3/r_jk, dvec_jk );
+ rvec_ScaledAdd( workspace->f_hb[j], -CEhb3 / r_jk, dvec_jk );
+ rvec_ScaledAdd( workspace->f_hb[k], +CEhb3 / r_jk, dvec_jk );
#endif
}
}
}
}
- /* fprintf( stderr, "hydbonds: ext_press (%23.15e %23.15e %23.15e)\n",
+ /* fprintf( stderr, "hydbonds: ext_press (%23.15e %23.15e %23.15e)\n",
data->ext_press[0], data->ext_press[1], data->ext_press[2] ); */
#ifdef TEST_FORCES
diff --git a/PuReMD-GPU/src/tool_box.c b/PuReMD-GPU/src/tool_box.c
new file mode 100644
index 0000000000000000000000000000000000000000..1782e71cdac028ddb5c0947ef371cf9259bfad2b
--- /dev/null
+++ b/PuReMD-GPU/src/tool_box.c
@@ -0,0 +1,467 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#include "tool_box.h"
+
+#include <ctype.h>
+
+
+/************** taken from box.c **************/
+void Transform( rvec x1, simulation_box *box, char flag, rvec x2 )
+{
+ int i, j;
+ real tmp;
+
+ // printf(">x1: (%lf, %lf, %lf)\n",x1[0],x1[1],x1[2]);
+
+ if (flag > 0)
+ {
+ for (i = 0; i < 3; i++)
+ {
+ tmp = 0.0;
+ for (j = 0; j < 3; j++)
+ tmp += box->trans[i][j] * x1[j];
+ x2[i] = tmp;
+ }
+ }
+ else
+ {
+ for (i = 0; i < 3; i++)
+ {
+ tmp = 0.0;
+ for (j = 0; j < 3; j++)
+ tmp += box->trans_inv[i][j] * x1[j];
+ x2[i] = tmp;
+ }
+ }
+ // printf(">x2: (%lf, %lf, %lf)\n", x2[0], x2[1], x2[2]);
+}
+
+
+void Transform_to_UnitBox( rvec x1, simulation_box *box, char flag, rvec x2 )
+{
+ Transform( x1, box, flag, x2 );
+
+ x2[0] /= box->box_norms[0];
+ x2[1] /= box->box_norms[1];
+ x2[2] /= box->box_norms[2];
+}
+
+
+/* determine whether point p is inside the box */
+void Fit_to_Periodic_Box( simulation_box *box, rvec *p )
+{
+ int i;
+
+ for ( i = 0; i < 3; ++i )
+ {
+ //TODO: verify box boundary coordinates -- assuming orthogonal box pinned at origin
+ if ( (*p)[i] < 0. )
+ {
+ /* handle lower coords */
+ while ( (*p)[i] < 0. )
+ (*p)[i] += box->box_norms[i];
+ }
+ else if ( (*p)[i] >= box->box_norms[i] )
+ {
+ /* handle higher coords */
+ while ( (*p)[i] >= box->box_norms[i] )
+ (*p)[i] -= box->box_norms[i];
+ }
+// if ( (*p)[i] < box->min[i] )
+// {
+// /* handle lower coords */
+// while ( (*p)[i] < box->min[i] )
+// (*p)[i] += box->box_norms[i];
+// }
+// else if ( (*p)[i] >= box->max[i] )
+// {
+// /* handle higher coords */
+// while ( (*p)[i] >= box->max[i] )
+// (*p)[i] -= box->box_norms[i];
+// }
+ }
+}
+
+
+/* determine the touch point, tp, of a box to
+ its neighbor denoted by the relative coordinate rl */
+/*
+inline void Box_Touch_Point( simulation_box *box, ivec rl, rvec tp )
+{
+ int d;
+
+ for ( d = 0; d < 3; ++d )
+ if ( rl[d] == -1 )
+ tp[d] = box->min[d];
+ else if ( rl[d] == 0 )
+ tp[d] = NEG_INF - 1.;
+ else
+ tp[d] = box->max[d];
+}
+*/
+
+
+/* determine whether point p is inside the box */
+/* assumes orthogonal box */
+/*
+inline int is_Inside_Box( simulation_box *box, rvec p )
+{
+ if ( p[0] < box->min[0] || p[0] >= box->max[0] ||
+ p[1] < box->min[1] || p[1] >= box->max[1] ||
+ p[2] < box->min[2] || p[2] >= box->max[2] )
+ return FALSE;
+
+ return TRUE;
+}
+*/
+
+
+/*
+inline int iown_midpoint( simulation_box *box, rvec p1, rvec p2 )
+{
+ rvec midp;
+
+ midp[0] = (p1[0] + p2[0]) / 2;
+ midp[1] = (p1[1] + p2[1]) / 2;
+ midp[2] = (p1[2] + p2[2]) / 2;
+
+ if ( midp[0] < box->min[0] || midp[0] >= box->max[0] ||
+ midp[1] < box->min[1] || midp[1] >= box->max[1] ||
+ midp[2] < box->min[2] || midp[2] >= box->max[2] )
+ return FALSE;
+
+ return TRUE;
+}
+*/
+
+
+/**************** from grid.c ****************/
+/* finds the closest point of grid cell cj to ci.
+ no need to consider periodic boundary conditions as in the serial case
+ because the box of a process is not periodic in itself */
+/*
+inline void GridCell_Closest_Point( grid_cell *gci, grid_cell *gcj,
+ ivec ci, ivec cj, rvec cp )
+{
+ int d;
+
+ for ( d = 0; d < 3; d++ )
+ if ( cj[d] > ci[d] )
+ cp[d] = gcj->min[d];
+ else if ( cj[d] == ci[d] )
+ cp[d] = NEG_INF - 1.;
+ else
+ cp[d] = gcj->max[d];
+}
+
+
+inline void GridCell_to_Box_Points( grid_cell *gc, ivec rl, rvec cp, rvec fp )
+{
+ int d;
+
+ for ( d = 0; d < 3; ++d )
+ if ( rl[d] == -1 )
+ {
+ cp[d] = gc->min[d];
+ fp[d] = gc->max[d];
+ }
+ else if ( rl[d] == 0 )
+ {
+ cp[d] = fp[d] = NEG_INF - 1.;
+ }
+ else
+ {
+ cp[d] = gc->max[d];
+ fp[d] = gc->min[d];
+ }
+}
+
+
+inline real DistSqr_between_Special_Points( rvec sp1, rvec sp2 )
+{
+ int i;
+ real d_sqr = 0;
+
+ for ( i = 0; i < 3; ++i )
+ {
+ if ( sp1[i] > NEG_INF && sp2[i] > NEG_INF )
+ {
+ d_sqr += SQR( sp1[i] - sp2[i] );
+ }
+ }
+
+ return d_sqr;
+}
+
+
+inline real DistSqr_to_Special_Point( rvec cp, rvec x )
+{
+ int i;
+ real d_sqr = 0;
+
+ for ( i = 0; i < 3; ++i )
+ {
+ if ( cp[i] > NEG_INF )
+ {
+ d_sqr += SQR( cp[i] - x[i] );
+ }
+ }
+
+ return d_sqr;
+}
+
+
+inline int Relative_Coord_Encoding( ivec c )
+{
+ return 9 * (c[0] + 1) + 3 * (c[1] + 1) + (c[2] + 1);
+}
+*/
+
+
+/************** from geo_tools.c *****************/
+void Make_Point( real x, real y, real z, rvec* p )
+{
+ (*p)[0] = x;
+ (*p)[1] = y;
+ (*p)[2] = z;
+}
+
+
+int is_Valid_Serial( static_storage *workspace, int serial )
+{
+ if( workspace->map_serials[ serial ] < 0 )
+ {
+ fprintf( stderr, "CONECT line includes invalid pdb serial number %d.\n", serial );
+ fprintf( stderr, "Please correct the input file.Terminating...\n" );
+ exit( INVALID_INPUT );
+ }
+
+ return TRUE;
+}
+
+
+int Check_Input_Range( int val, int lo, int hi, char *message )
+{
+ if ( val < lo || val > hi )
+ {
+ fprintf( stderr, "%s\nInput %d - Out of range %d-%d. Terminating...\n",
+ message, val, lo, hi );
+ exit( INVALID_INPUT );
+ }
+
+ return SUCCESS;
+}
+
+
+void Trim_Spaces( char *element )
+{
+ int i, j;
+
+ for ( i = 0; element[i] == ' '; ++i ); // skip initial space chars
+
+ for ( j = i; j < (int)(strlen(element)) && element[j] != ' '; ++j )
+ {
+ element[j - i] = toupper( element[j] ); // make uppercase, offset to 0
+ }
+ element[j - i] = 0; // finalize the string
+}
+
+
+/************ from system_props.c *************/
+real Get_Time( )
+{
+ gettimeofday(&tim, NULL );
+ return ( tim.tv_sec + (tim.tv_usec / 1000000.0) );
+}
+
+
+real Get_Timing_Info( real t_start )
+{
+ gettimeofday(&tim, NULL );
+ t_end = tim.tv_sec + (tim.tv_usec / 1000000.0);
+ return (t_end - t_start);
+}
+
+
+void Update_Timing_Info( real *t_start, real *timing )
+{
+ gettimeofday(&tim, NULL );
+ t_end = tim.tv_sec + (tim.tv_usec / 1000000.0);
+ *timing += (t_end - *t_start);
+ *t_start = t_end;
+}
+
+
+/*********** from io_tools.c **************/
+int Get_Atom_Type( reax_interaction *reax_param, char *s )
+{
+ int i;
+
+ for ( i = 0; i < reax_param->num_atom_types; ++i )
+ {
+ if ( !strcmp( reax_param->sbp[i].name, s ) )
+ {
+ return i;
+ }
+ }
+
+ fprintf( stderr, "Unknown atom type %s. Terminating...\n", s );
+ exit( UNKNOWN_ATOM_TYPE );
+
+ return FAILURE;
+}
+
+
+char *Get_Element( reax_system *system, int i )
+{
+ return &( system->reaxprm.sbp[system->atoms[i].type].name[0] );
+}
+
+
+char *Get_Atom_Name( reax_system *system, int i )
+{
+ return &(system->atoms[i].name[0]);
+}
+
+
+int Allocate_Tokenizer_Space( char **line, char **backup, char ***tokens )
+{
+ int i;
+
+ if ( (*line = (char*) malloc( sizeof(char) * MAX_LINE )) == NULL )
+ {
+ return FAILURE;
+ }
+
+ if ( (*backup = (char*) malloc( sizeof(char) * MAX_LINE )) == NULL )
+ {
+ return FAILURE;
+ }
+
+ if ( (*tokens = (char**) malloc( sizeof(char*) * MAX_TOKENS )) == NULL )
+ {
+ return FAILURE;
+ }
+
+ for ( i = 0; i < MAX_TOKENS; i++ )
+ {
+ if ( ((*tokens)[i] = (char*) malloc(sizeof(char) * MAX_TOKEN_LEN)) == NULL )
+ {
+ return FAILURE;
+ }
+ }
+
+ return SUCCESS;
+}
+
+
+int Tokenize( char* s, char*** tok )
+{
+ char test[MAX_LINE];
+ char *sep = "\t \n!=";
+ char *word;
+ int count = 0;
+
+ strncpy( test, s, MAX_LINE );
+
+ for ( word = strtok(test, sep); word; word = strtok(NULL, sep) )
+ {
+ strncpy( (*tok)[count], word, MAX_LINE );
+ count++;
+ }
+
+ return count;
+}
+
+
+/***************** taken from lammps ************************/
+/* safe malloc */
+void *smalloc( long n, char *name )
+{
+ void *ptr;
+
+ if ( n <= 0 )
+ {
+ fprintf( stderr, "WARNING: trying to allocate %ld bytes for array %s. ",
+ n, name );
+ fprintf( stderr, "returning NULL.\n" );
+ return NULL;
+ }
+
+ ptr = malloc( n );
+ if ( ptr == NULL )
+ {
+ fprintf( stderr, "ERROR: failed to allocate %ld bytes for array %s",
+ n, name );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ return ptr;
+}
+
+
+/* safe calloc */
+void *scalloc( int n, int size, char *name )
+{
+ void *ptr;
+
+ if ( n <= 0 )
+ {
+ fprintf( stderr, "WARNING: trying to allocate %d elements for array %s. ",
+ n, name );
+ fprintf( stderr, "returning NULL.\n" );
+ return NULL;
+ }
+
+ if ( size <= 0 )
+ {
+ fprintf( stderr, "WARNING: elements size for array %s is %d. ",
+ name, size );
+ fprintf( stderr, "returning NULL.\n" );
+ return NULL;
+ }
+
+ ptr = calloc( n, size );
+ if ( ptr == NULL )
+ {
+ fprintf( stderr, "ERROR: failed to allocate %d bytes for array %s",
+ n * size, name );
+ exit( INSUFFICIENT_MEMORY );
+ }
+
+ return ptr;
+}
+
+
+/* safe free */
+void sfree( void *ptr, char *name )
+{
+ if ( ptr == NULL )
+ {
+ fprintf( stderr, "WARNING: trying to free the already NULL pointer %s!\n",
+ name );
+ return;
+ }
+
+ free( ptr );
+ ptr = NULL;
+}
diff --git a/PuReMD-GPU/src/tool_box.h b/PuReMD-GPU/src/tool_box.h
new file mode 100644
index 0000000000000000000000000000000000000000..db97076149a5f5c8868d02d299a25581d3b5a934
--- /dev/null
+++ b/PuReMD-GPU/src/tool_box.h
@@ -0,0 +1,72 @@
+/*----------------------------------------------------------------------
+ SerialReax - Reax Force Field Simulator
+
+ Copyright (2010) Purdue University
+ Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
+ Ananth Y Grama, ayg@cs.purdue.edu
+
+ This program is free software; you can redistribute it and/or
+ modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation; either version 2 of
+ the License, or (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ See the GNU General Public License for more details:
+ <http://www.gnu.org/licenses/>.
+ ----------------------------------------------------------------------*/
+
+#ifndef __TOOL_BOX_H_
+#define __TOOL_BOX_H_
+
+#include "mytypes.h"
+
+struct timeval tim;
+real t_end;
+
+
+/* from box.h */
+void Transform( rvec, simulation_box*, char, rvec );
+void Transform_to_UnitBox( rvec, simulation_box*, char, rvec );
+void Fit_to_Periodic_Box( simulation_box*, rvec* );
+//void Box_Touch_Point( simulation_box*, ivec, rvec );
+//int is_Inside_Box( simulation_box*, rvec );
+//int iown_midpoint( simulation_box*, rvec, rvec );
+
+/* from grid.h */
+/*
+void GridCell_Closest_Point( grid_cell*, grid_cell*, ivec, ivec, rvec );
+void GridCell_to_Box_Points( grid_cell*, ivec, rvec, rvec );
+real DistSqr_between_Special_Points( rvec, rvec );
+real DistSqr_to_Special_Point( rvec, rvec );
+int Relative_Coord_Encoding( ivec );
+*/
+
+/* from geo_tools.h */
+void Make_Point( real, real, real, rvec* );
+int is_Valid_Serial( static_storage*, int );
+int Check_Input_Range( int, int, int, char* );
+void Trim_Spaces( char* );
+
+/* from system_props.h */
+real Get_Time( );
+real Get_Timing_Info( real );
+void Update_Timing_Info( real*, real* );
+
+/* from io_tools.h */
+int Get_Atom_Type( reax_interaction*, char* );
+char *Get_Element( reax_system*, int );
+char *Get_Atom_Name( reax_system*, int );
+int Allocate_Tokenizer_Space( char**, char**, char*** );
+int Tokenize( char*, char*** );
+
+/* from lammps */
+void *smalloc( long, char* );
+void *scalloc( int, int, char* );
+void sfree( void*, char* );
+
+
+#endif
diff --git a/PuReMD-GPU/src/traj.c b/PuReMD-GPU/src/traj.c
index 2844c370ee79702ed0c75d090afe545149aae185..f8852d5d2cc2e425b67f6ffa6871b52c2f755046 100644
--- a/PuReMD-GPU/src/traj.c
+++ b/PuReMD-GPU/src/traj.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -30,7 +31,8 @@
/************************************************/
/* CUSTOM FORMAT ROUTINES */
/************************************************/
-int Write_Custom_Header(reax_system *system, control_params *control,
+
+int Write_Custom_Header(reax_system *system, control_params *control,
static_storage *workspace, output_controls *out_control)
{
int i, header_len, control_block_len, frame_format_len;
@@ -40,119 +42,129 @@ int Write_Custom_Header(reax_system *system, control_params *control,
char atom_format[100], bond_format[100], angle_format[100];
sprintf( control_block, CONTROL_BLOCK,
- system->N,
- control->restart,
- control->restart_from,
- control->random_vel,
- out_control->restart_freq,
- control->ensemble,
- control->nsteps,
- control->dt,
- control->reposition_atoms,
- control->restrict_bonds,
- control->tabulate,
- control->nbr_cut,
- control->r_cut,
- control->bg_cut,
- control->bo_cut,
- control->thb_cut,
- control->hb_cut,
- control->q_err,
- control->T_init,
- control->T_final,
- control->Tau_T,
- control->T_mode,
- control->T_rate,
- control->T_freq,
- control->P[0], control->P[1], control->P[2],
- control->Tau_P[0], control->Tau_P[1], control->Tau_P[2],
- control->compressibility,
- control->press_mode,
- control->remove_CoM_vel,
- out_control->write_steps,
- out_control->traj_compress,
- out_control->traj_format,
- out_control->atom_format,
- out_control->bond_info,
- out_control->angle_info,
- out_control->energy_update_freq,
- control->molec_anal,
- control->freq_molec_anal );
-
- control_block_len = strlen( control_block );
-
-
- sprintf( frame_format, "Frame Format: %d\n%s\n%s\n",
- NUM_FRAME_GLOBALS, FRAME_GLOBALS_FORMAT, FRAME_GLOBAL_NAMES );
-
- atom_format[0] = OPT_NOATOM;
- switch( out_control->atom_format )
- {
- case OPT_ATOM_BASIC: sprintf( atom_format, "Atom_Basic: %s", ATOM_BASIC );
- break;
- case OPT_ATOM_wF: sprintf( atom_format, "Atom_wF: %s", ATOM_wF );
- break;
- case OPT_ATOM_wV: sprintf( atom_format, "Atom_wV: %s", ATOM_wV );
- break;
- case OPT_ATOM_FULL: sprintf( atom_format, "Atom_Full: %s", ATOM_FULL );
- break;
- }
- strcat( frame_format, atom_format );
-
- bond_format[0] = OPT_NOBOND;
- if( out_control->bond_info == OPT_BOND_BASIC )
- sprintf( bond_format, "Bond_Line: %s", BOND_BASIC );
- else if( out_control->bond_info == OPT_BOND_FULL )
- sprintf( bond_format, "Bond_Line_Full: %s", BOND_FULL );
- strcat( frame_format, bond_format );
+ system->N,
+ control->restart,
+ control->restart_from,
+ control->random_vel,
+ out_control->restart_freq,
+ control->ensemble,
+ control->nsteps,
+ control->dt,
+ control->reposition_atoms,
+ control->restrict_bonds,
+ control->tabulate,
+ control->nbr_cut,
+ control->r_cut,
+ control->bg_cut,
+ control->bo_cut,
+ control->thb_cut,
+ control->hb_cut,
+ control->qeq_solver_q_err,
+ control->T_init,
+ control->T_final,
+ control->Tau_T,
+ control->T_mode,
+ control->T_rate,
+ control->T_freq,
+ control->P[0], control->P[1], control->P[2],
+ control->Tau_P[0], control->Tau_P[1], control->Tau_P[2],
+ control->compressibility,
+ control->press_mode,
+ control->remove_CoM_vel,
+ out_control->write_steps,
+ out_control->traj_compress,
+ out_control->traj_format,
+ out_control->atom_format,
+ out_control->bond_info,
+ out_control->angle_info,
+ out_control->energy_update_freq,
+ control->molec_anal,
+ control->freq_molec_anal );
+
+ control_block_len = strlen( control_block );
+
+ sprintf( frame_format, "Frame Format: %d\n%s\n%s\n",
+ NUM_FRAME_GLOBALS, FRAME_GLOBALS_FORMAT, FRAME_GLOBAL_NAMES );
+
+ atom_format[0] = OPT_NOATOM;
+ switch ( out_control->atom_format )
+ {
+ case OPT_ATOM_BASIC:
+ sprintf( atom_format, "Atom_Basic: %s", ATOM_BASIC );
+ break;
+ case OPT_ATOM_wF:
+ sprintf( atom_format, "Atom_wF: %s", ATOM_wF );
+ break;
+ case OPT_ATOM_wV:
+ sprintf( atom_format, "Atom_wV: %s", ATOM_wV );
+ break;
+ case OPT_ATOM_FULL:
+ sprintf( atom_format, "Atom_Full: %s", ATOM_FULL );
+ break;
+ }
+ strcat( frame_format, atom_format );
- angle_format[0] = OPT_NOANGLE;
- if( out_control->angle_info == OPT_ANGLE_BASIC )
- sprintf( angle_format, "Angle_Line: %s", ANGLE_BASIC );
- strcat( frame_format, angle_format );
+ bond_format[0] = OPT_NOBOND;
+ if ( out_control->bond_info == OPT_BOND_BASIC )
+ {
+ sprintf( bond_format, "Bond_Line: %s", BOND_BASIC );
+ }
+ else if ( out_control->bond_info == OPT_BOND_FULL )
+ {
+ sprintf( bond_format, "Bond_Line_Full: %s", BOND_FULL );
+ }
+ strcat( frame_format, bond_format );
- frame_format_len = strlen( frame_format );
+ angle_format[0] = OPT_NOANGLE;
+ if ( out_control->angle_info == OPT_ANGLE_BASIC )
+ {
+ sprintf( angle_format, "Angle_Line: %s", ANGLE_BASIC );
+ }
+ strcat( frame_format, angle_format );
+ frame_format_len = strlen( frame_format );
- header_len = HEADER_INIT_LEN + (control_block_len + SIZE_INFO_LEN2)+
- (frame_format_len + SIZE_INFO_LEN2) +
- (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2);
+ header_len = HEADER_INIT_LEN + (control_block_len + SIZE_INFO_LEN2) +
+ (frame_format_len + SIZE_INFO_LEN2) +
+ (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2);
- out_control->write( out_control->trj, HEADER_INIT,
- header_len, HEADER_INIT_LEN, out_control->traj_title );
+ out_control->write( out_control->trj, HEADER_INIT,
+ header_len, HEADER_INIT_LEN, out_control->traj_title );
- out_control->write( out_control->trj, SIZE_INFO_LINE2,
- control_block_len + (frame_format_len + SIZE_INFO_LEN2) +
- (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2),
- control_block_len );
- out_control->write( out_control->trj, "%s", control_block );
+ out_control->write( out_control->trj, SIZE_INFO_LINE2,
+ control_block_len + (frame_format_len + SIZE_INFO_LEN2) +
+ (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2),
+ control_block_len );
+ out_control->write( out_control->trj, "%s", control_block );
- out_control->write( out_control->trj, SIZE_INFO_LINE2,
- frame_format_len +
- (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2),
- frame_format_len );
- out_control->write( out_control->trj, "%s", frame_format );
+ out_control->write( out_control->trj, SIZE_INFO_LINE2,
+ frame_format_len +
+ (ATOM_MAPPING_LEN * system->N + SIZE_INFO_LEN2),
+ frame_format_len );
+ out_control->write( out_control->trj, "%s", frame_format );
- out_control->write( out_control->trj, SIZE_INFO_LINE2,
- ATOM_MAPPING_LEN * system->N,
- ATOM_MAPPING_LEN * system->N );
+ out_control->write( out_control->trj, SIZE_INFO_LINE2,
+ ATOM_MAPPING_LEN * system->N,
+ ATOM_MAPPING_LEN * system->N );
- for( i = 0; i < system->N; ++i )
- out_control->write( out_control->trj, ATOM_MAPPING,
- workspace->orig_id[i],
- system->atoms[i].type,
- system->atoms[i].name,
- system->reaxprm.sbp[ system->atoms[i].type ].mass );
+ for ( i = 0; i < system->N; ++i )
+ {
+ out_control->write( out_control->trj, ATOM_MAPPING,
+ workspace->orig_id[i],
+ system->atoms[i].type,
+ system->atoms[i].name,
+ system->reaxprm.sbp[ system->atoms[i].type ].mass );
+ }
- fflush( out_control->trj );
+ fflush( out_control->trj );
- return 0;
+ return 0;
}
-int Append_Custom_Frame( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+int Append_Custom_Frame( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
int i, j, pi, pk, pk_j;
int write_atoms, write_bonds, write_angles;
@@ -166,278 +178,326 @@ int Append_Custom_Frame( reax_system *system, control_params *control,
/* IMPORTANT: This whole part will go to init_trj after finalized! */
- switch( out_control->atom_format )
+ switch ( out_control->atom_format )
{
- case OPT_ATOM_BASIC:
- atom_line_len = ATOM_BASIC_LEN;
- write_atoms = 1;
- break;
- case OPT_ATOM_wF:
- atom_line_len = ATOM_wF_LEN;
- write_atoms = 1;
- break;
- case OPT_ATOM_wV:
- atom_line_len = ATOM_wV_LEN;
- write_atoms = 1;
- break;
- case OPT_ATOM_FULL:
- atom_line_len = ATOM_FULL_LEN;
- write_atoms = 1;
- break;
- default:
- atom_line_len = 0;
- write_atoms = 0;
+ case OPT_ATOM_BASIC:
+ atom_line_len = ATOM_BASIC_LEN;
+ write_atoms = 1;
+ break;
+ case OPT_ATOM_wF:
+ atom_line_len = ATOM_wF_LEN;
+ write_atoms = 1;
+ break;
+ case OPT_ATOM_wV:
+ atom_line_len = ATOM_wV_LEN;
+ write_atoms = 1;
+ break;
+ case OPT_ATOM_FULL:
+ atom_line_len = ATOM_FULL_LEN;
+ write_atoms = 1;
+ break;
+ default:
+ atom_line_len = 0;
+ write_atoms = 0;
}
-
/* bond preparations */
bond_line_len = write_bonds = 0;
- if( out_control->bond_info == OPT_BOND_BASIC )
+ if ( out_control->bond_info == OPT_BOND_BASIC )
{
bond_line_len = BOND_BASIC_LEN;
write_bonds = 1;
}
- else if( out_control->bond_info == OPT_BOND_FULL )
+ else if ( out_control->bond_info == OPT_BOND_FULL )
{
bond_line_len = BOND_FULL_LEN;
write_bonds = 1;
}
#ifdef __DEBUG_CUDA__
- fprintf (stderr, "Append Custom Frame -- write_bonds --> %d \n", write_bonds);
+ fprintf( stderr, "Append Custom Frame -- write_bonds --> %d \n", write_bonds );
#endif
num_bonds = 0;
- if( write_bonds )
+ if ( write_bonds )
{
-
#ifdef __PRINT_CPU_RESULTS__
- //fprintf (stderr, "Synching bonds from device for printing ....\n");
+#ifdef __DEBUG_CUDA__
+ fprintf( stderr, "Synching bonds from device for printing ....\n" );
+#endif
Sync_Host_Device_List( bonds, (dev_lists + BONDS), TYP_BOND );
#endif
- for( i = 0; i < system->N; ++i )
- for( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
- if( i < bonds->select.bond_list[j].nbr &&
+ for ( i = 0; i < system->N; ++i )
+ {
+ for ( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
+ {
+ if ( i < bonds->select.bond_list[j].nbr &&
bonds->select.bond_list[j].bo_data.BO >= control->bg_cut )
+ {
++num_bonds;
+ }
+ }
+ }
}
-
/* angle preparations */
- if( out_control->angle_info == OPT_ANGLE_BASIC )
+ if ( out_control->angle_info == OPT_ANGLE_BASIC )
{
angle_line_len = ANGLE_BASIC_LEN;
write_angles = 1;
}
- else
+ else
{
angle_line_len = 0;
write_angles = 0;
}
#ifdef __DEBUG_CUDA__
- fprintf (stderr, "Append Custom Frame -- write-angles --> %d \n", write_angles );
+ fprintf( stderr, "Append Custom Frame -- write-angles --> %d \n", write_angles );
#endif
num_thb_intrs = 0;
- if( write_angles ) {
-
+ if ( write_angles )
+ {
#ifdef __PRINT_CPU_RESULTS__
- //fprintf (stderr, "Synching three bodies from deivce for printing ... \n");
+#ifdef __DEBUG_CUDA__
+ fprintf( stderr, "Synching three bodies from deivce for printing ... \n" );
+#endif
Sync_Host_Device_List( thb_intrs, dev_lists + THREE_BODIES, TYP_THREE_BODY );
- if ( !write_bonds) {
- //fprintf (stderr, "Synching bonds for three bodies from device for printing ... \n");
+ if ( !write_bonds )
+ {
+#ifdef __DEBUG_CUDA__
+ fprintf( stderr, "Synching bonds for three bodies from device for printing ... \n" );
+#endif
Sync_Host_Device_List( bonds, (dev_lists + BONDS), TYP_BOND );
}
#endif
- for( j = 0; j < system->N; ++j )
- for( pi = Start_Index(j, bonds); pi < End_Index(j, bonds); ++pi )
- if( bonds->select.bond_list[pi].bo_data.BO >= control->bg_cut )
+ for ( j = 0; j < system->N; ++j )
+ {
+ for ( pi = Start_Index(j, bonds); pi < End_Index(j, bonds); ++pi )
+ {
+ if ( bonds->select.bond_list[pi].bo_data.BO >= control->bg_cut )
+ {
// physical j&i bond
- for( pk = Start_Index( pi, thb_intrs );
+ for ( pk = Start_Index( pi, thb_intrs );
pk < End_Index( pi, thb_intrs ); ++pk )
- if( bonds->select.bond_list[pi].nbr <
- thb_intrs->select.three_body_list[pk].thb ) {
+ {
+ if ( bonds->select.bond_list[pi].nbr <
+ thb_intrs->select.three_body_list[pk].thb )
+ {
// get k's pointer on j's bond list
pk_j = thb_intrs->select.three_body_list[pk].pthb;
- if( bonds->select.bond_list[pk_j].bo_data.BO >= control->bg_cut )
+ if ( bonds->select.bond_list[pk_j].bo_data.BO >= control->bg_cut )
// physical j&k bond
++num_thb_intrs;
}
+ }
+ }
+ }
+ }
}
-
/* get correct pressure */
- if( control->ensemble == NPT || control->ensemble == sNPT )
+ if ( control->ensemble == NPT || control->ensemble == sNPT )
+ {
P = data->flex_bar.P_scalar;
- else if( control->ensemble == iNPT )
+ }
+ else if ( control->ensemble == iNPT )
+ {
P = data->iso_bar.P;
- else P = 0;
-
+ }
+ else
+ {
+ P = 0;
+ }
/* calculate total frame length*/
sprintf( buffer, FRAME_GLOBALS,
- data->step, data->time,
- data->E_Tot, data->E_Pot, E_CONV * data->E_Kin, data->therm.T,
- P, system->box.volume,
- system->box.box_norms[0],
- system->box.box_norms[1],
- system->box.box_norms[2],
- 90.0, 90.0, 90.0, // IMPORTANT: need to rewrite for flexible boxes!
- data->E_BE,
- data->E_Ov, data->E_Un, data->E_Lp,
- data->E_Ang, data->E_Pen, data->E_Coa, data->E_HB,
- data->E_Tor, data->E_Con,
- data->E_vdW, data->E_Ele, data->E_Pol );
+ data->step, data->time,
+ data->E_Tot, data->E_Pot, E_CONV * data->E_Kin, data->therm.T,
+ P, system->box.volume,
+ system->box.box_norms[0],
+ system->box.box_norms[1],
+ system->box.box_norms[2],
+ 90.0, 90.0, 90.0, // IMPORTANT: need to rewrite for flexible boxes!
+ data->E_BE,
+ data->E_Ov, data->E_Un, data->E_Lp,
+ data->E_Ang, data->E_Pen, data->E_Coa, data->E_HB,
+ data->E_Tor, data->E_Con,
+ data->E_vdW, data->E_Ele, data->E_Pol );
frame_globals_len = strlen( buffer );
- frame_len = frame_globals_len +
- write_atoms * SIZE_INFO_LEN3 + system->N * atom_line_len +
- write_bonds * SIZE_INFO_LEN3 + num_bonds * bond_line_len +
- write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
+ frame_len = frame_globals_len +
+ write_atoms * SIZE_INFO_LEN3 + system->N * atom_line_len +
+ write_bonds * SIZE_INFO_LEN3 + num_bonds * bond_line_len +
+ write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
/* write size info & frame globals */
- out_control->write( out_control->trj, SIZE_INFO_LINE2,
- frame_len, frame_globals_len );
+ out_control->write( out_control->trj, SIZE_INFO_LINE2,
+ frame_len, frame_globals_len );
out_control->write( out_control->trj, "%s", buffer );
- /* write size info & atom lines */
- if( write_atoms )
+ /* write size info & atom lines */
+ if ( write_atoms )
{
rest_of_frame_len = system->N * atom_line_len +
- write_bonds * SIZE_INFO_LEN3 + num_bonds * bond_line_len +
- write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
+ write_bonds * SIZE_INFO_LEN3 + num_bonds * bond_line_len +
+ write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
- out_control->write( out_control->trj, SIZE_INFO_LINE3,
- rest_of_frame_len, system->N * atom_line_len,
- system->N );
+ out_control->write( out_control->trj, SIZE_INFO_LINE3,
+ rest_of_frame_len, system->N * atom_line_len,
+ system->N );
}
- switch( out_control->atom_format )
+ switch ( out_control->atom_format )
{
- case 4:
- for( i = 0; i < system->N; ++i )
- out_control->write( out_control->trj, ATOM_BASIC,
- workspace->orig_id[i],
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2],
- system->atoms[i].q );
- break;
- case 5:
- for( i = 0; i < system->N; ++i )
- out_control->write( out_control->trj, ATOM_wF,
- workspace->orig_id[i],
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2],
- system->atoms[i].f[0],
- system->atoms[i].f[1],
- system->atoms[i].f[2],
- system->atoms[i].q );
- break;
- case 6:
- for( i = 0; i < system->N; ++i )
- out_control->write( out_control->trj, ATOM_wV,
- workspace->orig_id[i],
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2],
- system->atoms[i].v[0],
- system->atoms[i].v[1],
- system->atoms[i].v[2],
- system->atoms[i].q );
- break;
- case 7:
- for( i = 0; i < system->N; ++i )
- out_control->write( out_control->trj, ATOM_FULL,
- workspace->orig_id[i],
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2],
- system->atoms[i].v[0],
- system->atoms[i].v[1],
- system->atoms[i].v[2],
- system->atoms[i].f[0],
- system->atoms[i].f[1],
- system->atoms[i].f[2],
- system->atoms[i].q );
- break;
+ case 4:
+ for ( i = 0; i < system->N; ++i )
+ out_control->write( out_control->trj, ATOM_BASIC,
+ workspace->orig_id[i],
+ system->atoms[i].x[0],
+ system->atoms[i].x[1],
+ system->atoms[i].x[2],
+ system->atoms[i].q );
+ break;
+ case 5:
+ for ( i = 0; i < system->N; ++i )
+ out_control->write( out_control->trj, ATOM_wF,
+ workspace->orig_id[i],
+ system->atoms[i].x[0],
+ system->atoms[i].x[1],
+ system->atoms[i].x[2],
+ system->atoms[i].f[0],
+ system->atoms[i].f[1],
+ system->atoms[i].f[2],
+ system->atoms[i].q );
+ break;
+ case 6:
+ for ( i = 0; i < system->N; ++i )
+ out_control->write( out_control->trj, ATOM_wV,
+ workspace->orig_id[i],
+ system->atoms[i].x[0],
+ system->atoms[i].x[1],
+ system->atoms[i].x[2],
+ system->atoms[i].v[0],
+ system->atoms[i].v[1],
+ system->atoms[i].v[2],
+ system->atoms[i].q );
+ break;
+ case 7:
+ for ( i = 0; i < system->N; ++i )
+ out_control->write( out_control->trj, ATOM_FULL,
+ workspace->orig_id[i],
+ system->atoms[i].x[0],
+ system->atoms[i].x[1],
+ system->atoms[i].x[2],
+ system->atoms[i].v[0],
+ system->atoms[i].v[1],
+ system->atoms[i].v[2],
+ system->atoms[i].f[0],
+ system->atoms[i].f[1],
+ system->atoms[i].f[2],
+ system->atoms[i].q );
+ break;
}
fflush( out_control->trj );
/* write size info & bond lines */
- if( write_bonds )
+ if ( write_bonds )
{
rest_of_frame_len = num_bonds * bond_line_len +
- write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
+ write_angles * SIZE_INFO_LEN3 + num_thb_intrs * angle_line_len;
- out_control->write( out_control->trj, SIZE_INFO_LINE3,
- rest_of_frame_len, num_bonds * bond_line_len,
- num_bonds );
+ out_control->write( out_control->trj, SIZE_INFO_LINE3,
+ rest_of_frame_len, num_bonds * bond_line_len,
+ num_bonds );
}
- if( out_control->bond_info == 1 ) {
- for( i = 0; i < system->N; ++i )
- for( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
- if( i < bonds->select.bond_list[j].nbr &&
- bonds->select.bond_list[j].bo_data.BO >= control->bg_cut ) {
+ if ( out_control->bond_info == 1 )
+ {
+ for ( i = 0; i < system->N; ++i )
+ {
+ for ( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
+ {
+ if ( i < bonds->select.bond_list[j].nbr &&
+ bonds->select.bond_list[j].bo_data.BO >= control->bg_cut )
+ {
bo_ij = &( bonds->select.bond_list[j] );
- out_control->write( out_control->trj, BOND_BASIC,
- workspace->orig_id[i],
- workspace->orig_id[bo_ij->nbr],
- bo_ij->d, bo_ij->bo_data.BO );
+ out_control->write( out_control->trj, BOND_BASIC,
+ workspace->orig_id[i],
+ workspace->orig_id[bo_ij->nbr],
+ bo_ij->d, bo_ij->bo_data.BO );
}
+ }
+ }
}
- else if( out_control->bond_info == 2 ) {
- for( i = 0; i < system->N; ++i )
- for( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
- if( i < bonds->select.bond_list[j].nbr &&
- bonds->select.bond_list[j].bo_data.BO >= control->bg_cut ) {
+ else if ( out_control->bond_info == 2 )
+ {
+ for ( i = 0; i < system->N; ++i )
+ {
+ for ( j = Start_Index( i, bonds ); j < End_Index( i, bonds ); ++j )
+ {
+ if ( i < bonds->select.bond_list[j].nbr &&
+ bonds->select.bond_list[j].bo_data.BO >= control->bg_cut )
+ {
bo_ij = &( bonds->select.bond_list[j] );
- out_control->write( out_control->trj, BOND_FULL,
- workspace->orig_id[i],
- workspace->orig_id[bo_ij->nbr],
- bo_ij->d, bo_ij->bo_data.BO, bo_ij->bo_data.BO_s,
- bo_ij->bo_data.BO_pi, bo_ij->bo_data.BO_pi2 );
+ out_control->write( out_control->trj, BOND_FULL,
+ workspace->orig_id[i],
+ workspace->orig_id[bo_ij->nbr],
+ bo_ij->d, bo_ij->bo_data.BO, bo_ij->bo_data.BO_s,
+ bo_ij->bo_data.BO_pi, bo_ij->bo_data.BO_pi2 );
}
+ }
+ }
}
fflush( out_control->trj );
/* write size info & angle lines */
- if( out_control->angle_info ) {
+ if ( out_control->angle_info )
+ {
out_control->write( out_control->trj, SIZE_INFO_LINE3,
- num_thb_intrs * angle_line_len,
- num_thb_intrs * angle_line_len, num_thb_intrs );
+ num_thb_intrs * angle_line_len,
+ num_thb_intrs * angle_line_len, num_thb_intrs );
- for( j = 0; j < system->N; ++j )
- for( pi = Start_Index(j, bonds); pi < End_Index(j, bonds); ++pi )
- if( bonds->select.bond_list[pi].bo_data.BO >= control->bg_cut )
+ for ( j = 0; j < system->N; ++j )
+ {
+ for ( pi = Start_Index(j, bonds); pi < End_Index(j, bonds); ++pi )
+ {
+ if ( bonds->select.bond_list[pi].bo_data.BO >= control->bg_cut )
+ {
// physical j&i bond
- for( pk = Start_Index( pi, thb_intrs );
+ for ( pk = Start_Index( pi, thb_intrs );
pk < End_Index( pi, thb_intrs ); ++pk )
- if( bonds->select.bond_list[pi].nbr <
- thb_intrs->select.three_body_list[pk].thb ) {
- pk_j = thb_intrs->select.three_body_list[pk].pthb;
+ {
+ if ( bonds->select.bond_list[pi].nbr <
+ thb_intrs->select.three_body_list[pk].thb )
+ {
+ pk_j = thb_intrs->select.three_body_list[pk].pthb;
// get k's pointer on j's bond list
- if( bonds->select.bond_list[pk_j].bo_data.BO >= control->bg_cut )
+ if ( bonds->select.bond_list[pk_j].bo_data.BO >= control->bg_cut )
+ {
// physical j&k bond
out_control->write( out_control->trj, ANGLE_BASIC,
- workspace->orig_id[bonds->select.bond_list[pi].nbr],
- workspace->orig_id[j],
- workspace->orig_id[thb_intrs->select.three_body_list[pk].thb],
- RAD2DEG(thb_intrs->select.three_body_list[pk].theta) );
+ workspace->orig_id[bonds->select.bond_list[pi].nbr],
+ workspace->orig_id[j],
+ workspace->orig_id[thb_intrs->select.three_body_list[pk].thb],
+ RAD2DEG(thb_intrs->select.three_body_list[pk].theta) );
+ }
}
+ }
+ }
+ }
+ }
}
fflush( out_control->trj );
@@ -445,45 +505,47 @@ int Append_Custom_Frame( reax_system *system, control_params *control,
return 0;
}
-/*
- void Read_Traj( output_controls *out_control, char *traj_name )
- {
- int skip_all, skip_part, n;
- char size_buffer[50];
-// char read_buffer[2048];
-out_control->trj = (FILE *)gzopen( traj_name, "r" );
+void Read_Traj( output_controls *out_control, char *traj_name )
+{
+ int skip_all, skip_part, n;
+ char size_buffer[50];
-fprintf( stderr, "file opened!\n" );
+ out_control->trj = gzopen( traj_name, "r" );
-while( !gzeof( out_control->trj ) )
-{
-if( gzgets( out_control->trj, size_buffer, 50 ) == Z_NULL )
-break;
+ fprintf( stderr, "file opened!\n" );
+
+ while ( !gzeof( out_control->trj ) )
+ {
+ if ( gzgets( out_control->trj, size_buffer, 50 ) == Z_NULL )
+ {
+ break;
+ }
-fprintf( stderr, "read line\n" );
+ fprintf( stderr, "read line\n" );
-if( strlen( size_buffer ) >= SIZE_INFO_LEN3 )
-sscanf( size_buffer, "%d %d %d", &skip_all, &skip_part, &n );
-else
-sscanf( size_buffer, "%d %d", &skip_all, &skip_part );
+ if ( strlen( size_buffer ) >= SIZE_INFO_LEN3 )
+ {
+ sscanf( size_buffer, "%d %d %d", &skip_all, &skip_part, &n );
+ }
+ else
+ {
+ sscanf( size_buffer, "%d %d", &skip_all, &skip_part );
+ }
-fprintf( stderr, "%d %d\n", skip_all, skip_part );
+ fprintf( stderr, "%d %d\n", skip_all, skip_part );
-gzseek( out_control->trj, skip_part, SEEK_CUR );
-}
+ gzseek( out_control->trj, skip_part, SEEK_CUR );
+ }
-gzclose( out_control->trj );
+ gzclose( out_control->trj );
}
- */
-
/********************************************************/
/************ XYZ FORMAT ROUTINES ***************/
/********************************************************/
-
-int Write_xyz_Header( reax_system *system, control_params *control,
+int Write_xyz_Header( reax_system *system, control_params *control,
static_storage* workspace, output_controls *out_control )
{
fflush( out_control->trj );
@@ -492,25 +554,27 @@ int Write_xyz_Header( reax_system *system, control_params *control,
}
-int Append_xyz_Frame( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+int Append_xyz_Frame( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
int i;
out_control->write( out_control->trj, "%d\n", system->N );
out_control->write( out_control->trj, "%d\t%8.3f\t%8.3f\t%8.3f\t%8.3f\n",
- data->step,
- data->E_Tot, data->E_Pot,
- E_CONV*data->E_Kin, data->therm.T );
+ data->step,
+ data->E_Tot, data->E_Pot,
+ E_CONV * data->E_Kin, data->therm.T );
- for( i = 0; i < system->N; ++i )
+ for ( i = 0; i < system->N; ++i )
+ {
out_control->write( out_control->trj, "%3s %10.5f %10.5f %10.5f\n",
- system->reaxprm.sbp[ system->atoms[i].type ].name,
- system->atoms[i].x[0],
- system->atoms[i].x[1],
- system->atoms[i].x[2] );
+ system->reaxprm.sbp[ system->atoms[i].type ].name,
+ system->atoms[i].x[0],
+ system->atoms[i].x[1],
+ system->atoms[i].x[2] );
+ }
fflush( out_control->trj );
diff --git a/PuReMD-GPU/src/traj.h b/PuReMD-GPU/src/traj.h
index 35d92602eee7c2d0b5ee83889623df2cb2106c71..200f67711e60285f67f32ecb238f81e95d3f9b0d 100644
--- a/PuReMD-GPU/src/traj.h
+++ b/PuReMD-GPU/src/traj.h
@@ -25,6 +25,7 @@
#include <zlib.h>
+
#define BLOCK_MARK "REAX_BLOCK_MARK "
#define BLOCK_MARK_LEN 16
@@ -74,11 +75,27 @@
#define SIZE_INFO_LEN3 33
-enum ATOM_LINE_OPTS {OPT_NOATOM = 0, OPT_ATOM_BASIC = 4, OPT_ATOM_wF = 5,
- OPT_ATOM_wV = 6, OPT_ATOM_FULL = 7
- };
-enum BOND_LINE_OPTS {OPT_NOBOND, OPT_BOND_BASIC, OPT_BOND_FULL};
-enum ANGLE_LINE_OPTS {OPT_NOANGLE, OPT_ANGLE_BASIC};
+enum ATOM_LINE_OPTS
+{
+ OPT_NOATOM = 0,
+ OPT_ATOM_BASIC = 4,
+ OPT_ATOM_wF = 5,
+ OPT_ATOM_wV = 6,
+ OPT_ATOM_FULL = 7,
+};
+
+enum BOND_LINE_OPTS
+{
+ OPT_NOBOND = 0,
+ OPT_BOND_BASIC = 1,
+ OPT_BOND_FULL = 2,
+};
+
+enum ANGLE_LINE_OPTS
+{
+ OPT_NOANGLE = 0,
+ OPT_ANGLE_BASIC = 1,
+};
struct
@@ -143,10 +160,8 @@ int Skip_Next_Block( gzFile, int*);
No. of torsion entries (int)
Torsion info lines as per torsion format.
*/
-int Write_Custom_Header( reax_system*, control_params*,
- static_storage*, output_controls* );
-int Write_xyz_Header ( reax_system*, control_params*,
- static_storage*, output_controls* );
+int Write_Custom_Header( reax_system*, control_params*, static_storage*, output_controls* );
+int Write_xyz_Header ( reax_system*, control_params*, static_storage*, output_controls* );
/*
Write_Traj_Header( gzfile file,
@@ -168,7 +183,7 @@ char Write_Traj_Header( FILE*, int, char**, char**, control_params* );
char** various flags);
*/
int Push_Traj_Frame( /*gzfile*/ FILE*, reax_system*, control_params*,
- simulation_data*, static_storage*, list**, char** );
+ simulation_data*, static_storage*, list**, char** );
/*
Append_Traj_Frame( gzfile file,
@@ -180,11 +195,11 @@ int Push_Traj_Frame( /*gzfile*/ FILE*, reax_system*, control_params*,
char** various flags);
*/
int Append_Custom_Frame( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
+ static_storage*, list**, output_controls* );
int Append_xyz_Frame ( reax_system*, control_params*, simulation_data*,
- static_storage*, list**, output_controls* );
-
+ static_storage*, list**, output_controls* );
void Read_Traj( output_controls*, char * );
+
#endif
diff --git a/PuReMD-GPU/src/two_body_interactions.c b/PuReMD-GPU/src/two_body_interactions.c
index 2e7a6daf9039ea26c22b2fcfda5913e46255ad75..d5b53a05e607d043ba5f59f96f51a1065438c1d4 100644
--- a/PuReMD-GPU/src/two_body_interactions.c
+++ b/PuReMD-GPU/src/two_body_interactions.c
@@ -1,19 +1,20 @@
/*----------------------------------------------------------------------
- PuReMD-GPU - Reax Force Field Simulator
+ SerialReax - Reax Force Field Simulator
- Copyright (2014) Purdue University
- Sudhir Kylasa, skylasa@purdue.edu
+ Copyright (2010) Purdue University
Hasan Metin Aktulga, haktulga@cs.purdue.edu
+ Joseph Fogarty, jcfogart@mail.usf.edu
+ Sagar Pandit, pandit@usf.edu
Ananth Y Grama, ayg@cs.purdue.edu
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of
+ published by the Free Software Foundation; either version 2 of
the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU General Public License for more details:
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
@@ -21,14 +22,14 @@
#include "two_body_interactions.h"
#include "bond_orders.h"
+#include "index_utils.h"
#include "list.h"
#include "lookup.h"
#include "vector.h"
-#include "index_utils.h"
-void Bond_Energy( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
+void Bond_Energy( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace,
list **lists, output_controls *out_control )
{
int i, j, pj;
@@ -50,12 +51,14 @@ void Bond_Energy( reax_system *system, control_params *control,
gp10 = system->reaxprm.gp.l[10];
gp37 = (int) system->reaxprm.gp.l[37];
- for( i=0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
start_i = Start_Index(i, bonds);
end_i = End_Index(i, bonds);
//fprintf( stderr, "i=%d start=%d end=%d\n", i, start_i, end_i );
- for( pj = start_i; pj < end_i; ++pj )
- if( i < bonds->select.bond_list[pj].nbr ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ if ( i < bonds->select.bond_list[pj].nbr )
+ {
/* set the pointers */
j = bonds->select.bond_list[pj].nbr;
type_i = system->atoms[i].type;
@@ -68,15 +71,12 @@ void Bond_Energy( reax_system *system, control_params *control,
/* calculate the constants */
pow_BOs_be2 = POW( bo_ij->BO_s, twbp->p_be2 );
exp_be12 = EXP( twbp->p_be1 * ( 1.0 - pow_BOs_be2 ) );
- CEbo = -twbp->De_s * exp_be12 *
- ( 1.0 - twbp->p_be1 * twbp->p_be2 * pow_BOs_be2 );
+ CEbo = -twbp->De_s * exp_be12 *
+ ( 1.0 - twbp->p_be1 * twbp->p_be2 * pow_BOs_be2 );
/* calculate the Bond Energy */
- ebond =
- -twbp->De_s * bo_ij->BO_s * exp_be12
- -twbp->De_p * bo_ij->BO_pi
- -twbp->De_pp * bo_ij->BO_pi2;
-
+ ebond = -twbp->De_s * bo_ij->BO_s * exp_be12
+ - twbp->De_p * bo_ij->BO_pi - twbp->De_pp * bo_ij->BO_pi2;
data->E_BE += ebond;
/* calculate derivatives of Bond Orders */
@@ -85,34 +85,36 @@ void Bond_Energy( reax_system *system, control_params *control,
bo_ij->Cdbopi2 -= (CEbo + twbp->De_pp);
#ifdef TEST_ENERGY
- fprintf( out_control->ebond, "%6d%6d%24.15e%24.15e\n",
- workspace->orig_id[i], workspace->orig_id[j],
- // i+1, j+1,
- bo_ij->BO, ebond/*, data->E_BE*/ );
- /* fprintf( out_control->ebond, "%6d%6d%12.6f%12.6f%12.6f\n",
- workspace->orig_id[i], workspace->orig_id[j],
+ fprintf( out_control->ebond, "%6d%6d%24.15e%24.15e\n",
+ workspace->orig_id[i], workspace->orig_id[j],
+ // i+1, j+1,
+ bo_ij->BO, ebond/*, data->E_BE*/ );
+ /* fprintf( out_control->ebond, "%6d%6d%12.6f%12.6f%12.6f\n",
+ workspace->orig_id[i], workspace->orig_id[j],
CEbo, -twbp->De_p, -twbp->De_pp );*/
#endif
#ifdef TEST_FORCES
Add_dBO( system, lists, i, pj, CEbo, workspace->f_be );
- Add_dBOpinpi2( system, lists, i, pj,
- -(CEbo + twbp->De_p), -(CEbo + twbp->De_pp),
- workspace->f_be, workspace->f_be );
+ Add_dBOpinpi2( system, lists, i, pj,
+ -(CEbo + twbp->De_p), -(CEbo + twbp->De_pp),
+ workspace->f_be, workspace->f_be );
#endif
/* Stabilisation terminal triple bond */
- if( bo_ij->BO >= 1.00 ) {
- if( gp37 == 2 ||
- (sbp_i->mass == 12.0000 && sbp_j->mass == 15.9990) ||
- (sbp_j->mass == 12.0000 && sbp_i->mass == 15.9990) ) {
+ if ( bo_ij->BO >= 1.00 )
+ {
+ if ( gp37 == 2 ||
+ (sbp_i->mass == 12.0000 && sbp_j->mass == 15.9990) ||
+ (sbp_j->mass == 12.0000 && sbp_i->mass == 15.9990) )
+ {
// ba = SQR(bo_ij->BO - 2.50);
exphu = EXP( -gp7 * SQR(bo_ij->BO - 2.50) );
//oboa=abo(j1)-boa;
//obob=abo(j2)-boa;
- exphua1 = EXP(-gp3*(workspace->total_bond_order[i]-bo_ij->BO));
- exphub1 = EXP(-gp3*(workspace->total_bond_order[j]-bo_ij->BO));
+ exphua1 = EXP(-gp3 * (workspace->total_bond_order[i] - bo_ij->BO));
+ exphub1 = EXP(-gp3 * (workspace->total_bond_order[j] - bo_ij->BO));
//ovoab=abo(j1)-aval(it1)+abo(j2)-aval(it2);
- exphuov = EXP(gp4*(workspace->Delta[i] + workspace->Delta[j]));
+ exphuov = EXP(gp4 * (workspace->Delta[i] + workspace->Delta[j]));
hulpov = 1.0 / (1.0 + 25.0 * exphuov);
estriph = gp10 * exphu * hulpov * (exphua1 + exphub1);
@@ -120,24 +122,22 @@ void Bond_Energy( reax_system *system, control_params *control,
//estrain(j2) = estrain(j2) + 0.50*estriph;
data->E_BE += estriph;
- decobdbo = gp10 * exphu * hulpov * (exphua1 + exphub1) *
- ( gp3 - 2.0 * gp7 * (bo_ij->BO-2.50) );
- decobdboua = -gp10 * exphu * hulpov *
- (gp3*exphua1 + 25.0*gp4*exphuov*hulpov*(exphua1+exphub1));
- decobdboub = -gp10 * exphu * hulpov *
- (gp3*exphub1 + 25.0*gp4*exphuov*hulpov*(exphua1+exphub1));
+ decobdbo = gp10 * exphu * hulpov * (exphua1 + exphub1) *
+ ( gp3 - 2.0 * gp7 * (bo_ij->BO - 2.50) );
+ decobdboua = -gp10 * exphu * hulpov *
+ (gp3 * exphua1 + 25.0 * gp4 * exphuov * hulpov * (exphua1 + exphub1));
+ decobdboub = -gp10 * exphu * hulpov *
+ (gp3 * exphub1 + 25.0 * gp4 * exphuov * hulpov * (exphua1 + exphub1));
bo_ij->Cdbo += decobdbo;
workspace->CdDelta[i] += decobdboua;
workspace->CdDelta[j] += decobdboub;
- //loop_j ++;
- //fprintf (stderr, "incrementing loopj %d \n", loop_j);
#ifdef TEST_ENERGY
- fprintf( out_control->ebond,
- "%6d%6d%24.15e%24.15e%24.15e%24.15e\n",
- workspace->orig_id[i], workspace->orig_id[j],
- //i+1, j+1,
- estriph, decobdbo, decobdboua, decobdboub );
+ fprintf( out_control->ebond,
+ "%6d%6d%24.15e%24.15e%24.15e%24.15e\n",
+ workspace->orig_id[i], workspace->orig_id[j],
+ //i+1, j+1,
+ estriph, decobdbo, decobdboua, decobdboub );
#endif
#ifdef TEST_FORCES
Add_dBO( system, lists, i, pj, decobdbo, workspace->f_be );
@@ -151,9 +151,9 @@ void Bond_Energy( reax_system *system, control_params *control,
}
-void vdW_Coulomb_Energy( reax_system *system, control_params *control,
- simulation_data *data, static_storage *workspace,
- list **lists, output_controls *out_control )
+void vdW_Coulomb_Energy( reax_system *system, control_params *control,
+ simulation_data *data, static_storage *workspace, list **lists,
+ output_controls *out_control )
{
int i, j, pj;
int start_i, end_i;
@@ -172,20 +172,22 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
p_vdW1 = system->reaxprm.gp.l[28];
p_vdW1i = 1.0 / p_vdW1;
- far_nbrs = (*lists) + FAR_NBRS;
+ far_nbrs = (*lists) + FAR_NBRS;
e_ele = 0;
e_vdW = 0;
e_core = 0;
de_core = 0;
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
start_i = Start_Index(i, far_nbrs);
end_i = End_Index(i, far_nbrs);
// fprintf( stderr, "i: %d, start: %d, end: %d\n",
// i, start_i, end_i );
- for( pj = start_i; pj < end_i; ++pj )
- if( far_nbrs->select.far_nbr_list[pj].d <= control->r_cut ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ if ( far_nbrs->select.far_nbr_list[pj].d <= control->r_cut )
+ {
nbr_pj = &( far_nbrs->select.far_nbr_list[pj] );
j = nbr_pj->nbr;
r_ij = nbr_pj->d;
@@ -202,15 +204,16 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
Tap = Tap * r_ij + control->Tap1;
Tap = Tap * r_ij + control->Tap0;
- dTap = 7*control->Tap7 * r_ij + 6*control->Tap6;
- dTap = dTap * r_ij + 5*control->Tap5;
- dTap = dTap * r_ij + 4*control->Tap4;
- dTap = dTap * r_ij + 3*control->Tap3;
- dTap = dTap * r_ij + 2*control->Tap2;
- dTap += control->Tap1/r_ij;
+ dTap = 7 * control->Tap7 * r_ij + 6 * control->Tap6;
+ dTap = dTap * r_ij + 5 * control->Tap5;
+ dTap = dTap * r_ij + 4 * control->Tap4;
+ dTap = dTap * r_ij + 3 * control->Tap3;
+ dTap = dTap * r_ij + 2 * control->Tap2;
+ dTap += control->Tap1 / r_ij;
/*vdWaals Calculations*/
- if(system->reaxprm.gp.vdw_type==1 || system->reaxprm.gp.vdw_type==3) {
+ if (system->reaxprm.gp.vdw_type == 1 || system->reaxprm.gp.vdw_type == 3)
+ {
// shielding
powr_vdW1 = POW(r_ij, p_vdW1);
powgi_vdW1 = POW( 1.0 / twbp->gamma_w, p_vdW1);
@@ -219,35 +222,37 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
exp1 = EXP( twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
exp2 = EXP( 0.5 * twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
- data->E_vdW += e_vdW =
- self_coef * Tap * twbp->D * (exp1 - 2.0 * exp2);
+ data->E_vdW += e_vdW =
+ self_coef * Tap * twbp->D * (exp1 - 2.0 * exp2);
- dfn13 = POW( powr_vdW1 + powgi_vdW1, p_vdW1i - 1.0) *
- POW(r_ij, p_vdW1 - 2.0);
+ dfn13 = POW( powr_vdW1 + powgi_vdW1, p_vdW1i - 1.0) *
+ POW(r_ij, p_vdW1 - 2.0);
- CEvd = self_coef * ( dTap * twbp->D * (exp1 - 2 * exp2) -
- Tap * twbp->D * (twbp->alpha / twbp->r_vdW) *
- (exp1 - exp2) * dfn13 );
+ CEvd = self_coef * ( dTap * twbp->D * (exp1 - 2 * exp2) -
+ Tap * twbp->D * (twbp->alpha / twbp->r_vdW) *
+ (exp1 - exp2) * dfn13 );
}
- else{ // no shielding
+ else // no shielding
+ {
exp1 = EXP( twbp->alpha * (1.0 - r_ij / twbp->r_vdW) );
exp2 = EXP( 0.5 * twbp->alpha * (1.0 - r_ij / twbp->r_vdW) );
- data->E_vdW += e_vdW =
- self_coef * Tap * twbp->D * (exp1 - 2.0 * exp2);
+ data->E_vdW += e_vdW =
+ self_coef * Tap * twbp->D * (exp1 - 2.0 * exp2);
- CEvd = self_coef * ( dTap * twbp->D * (exp1 - 2.0 * exp2) -
- Tap * twbp->D * (twbp->alpha / twbp->r_vdW) *
- (exp1 - exp2) );
+ CEvd = self_coef * ( dTap * twbp->D * (exp1 - 2.0 * exp2) -
+ Tap * twbp->D * (twbp->alpha / twbp->r_vdW) *
+ (exp1 - exp2) );
}
- if(system->reaxprm.gp.vdw_type==2 || system->reaxprm.gp.vdw_type==3) {
+ if (system->reaxprm.gp.vdw_type == 2 || system->reaxprm.gp.vdw_type == 3)
+ {
// innner wall
- e_core = twbp->ecore * EXP(twbp->acore * (1.0-(r_ij/twbp->rcore)));
+ e_core = twbp->ecore * EXP(twbp->acore * (1.0 - (r_ij / twbp->rcore)));
e_vdW += self_coef * Tap * e_core;
data->E_vdW += self_coef * Tap * e_core;
- de_core = -(twbp->acore/twbp->rcore) * e_core;
+ de_core = -(twbp->acore / twbp->rcore) * e_core;
CEvd += self_coef * ( dTap * e_core + Tap * de_core );
}
@@ -257,24 +262,26 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
tmp = Tap / dr3gamij_3;
//tmp = Tap * nbr_pj->inv_dr3gamij_3; -- precomputed during compte_H
- data->E_Ele += e_ele =
- self_coef * C_ele * system->atoms[i].q * system->atoms[j].q * tmp;
+ data->E_Ele += e_ele =
+ self_coef * C_ele * system->atoms[i].q * system->atoms[j].q * tmp;
CEclmb = self_coef * C_ele * system->atoms[i].q * system->atoms[j].q *
- ( dTap - Tap * r_ij / dr3gamij_1 ) / dr3gamij_3;
- /*CEclmb = self_coef*C_ele*system->atoms[i].q*system->atoms[j].q*
+ ( dTap - Tap * r_ij / dr3gamij_1 ) / dr3gamij_3;
+ /*CEclmb = self_coef*C_ele*system->atoms[i].q*system->atoms[j].q*
( dTap- Tap*r_ij*nbr_pj->inv_dr3gamij_1 )*nbr_pj->inv_dr3gamij_3;*/
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT) {
- rvec_ScaledAdd( system->atoms[i].f,
- -(CEvd+CEclmb), nbr_pj->dvec );
- rvec_ScaledAdd( system->atoms[j].f,
- +(CEvd+CEclmb), nbr_pj->dvec );
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT )
+ {
+ rvec_ScaledAdd( system->atoms[i].f,
+ -(CEvd + CEclmb), nbr_pj->dvec );
+ rvec_ScaledAdd( system->atoms[j].f,
+ +(CEvd + CEclmb), nbr_pj->dvec );
}
- else { // NPT, iNPT or sNPT
- /* for pressure coupling, terms not related to bond order
+ else // NPT, iNPT or sNPT
+ {
+ /* for pressure coupling, terms not related to bond order
derivatives are added directly into pressure vector/tensor */
rvec_Scale( temp, CEvd + CEclmb, nbr_pj->dvec );
@@ -284,47 +291,47 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
rvec_iMultiply( ext_press, nbr_pj->rel_box, temp );
rvec_Add( data->ext_press, ext_press );
- /*fprintf( stderr, "nonbonded(%d,%d): rel_box (%f %f %f)",
+ /*fprintf( stderr, "nonbonded(%d,%d): rel_box (%f %f %f)",
i,j,nbr_pj->rel_box[0],nbr_pj->rel_box[1],nbr_pj->rel_box[2] );
fprintf( stderr, "force(%f %f %f)", temp[0], temp[1], temp[2] );
- fprintf( stderr, "ext_press (%12.6f %12.6f %12.6f)\n",
+ fprintf( stderr, "ext_press (%12.6f %12.6f %12.6f)\n",
data->ext_press[0], data->ext_press[1], data->ext_press[2] );*/
- /* This part is intended for a fully-flexible box */
- /* rvec_OuterProduct( temp_rtensor, nbr_pj->dvec,
+ /* This part is intended for a fully-flexible box */
+ /* rvec_OuterProduct( temp_rtensor, nbr_pj->dvec,
system->atoms[i].x );
- rtensor_Scale( total_rtensor,
+ rtensor_Scale( total_rtensor,
F_C * -(CEvd + CEclmb), temp_rtensor );
- rvec_OuterProduct( temp_rtensor,
+ rvec_OuterProduct( temp_rtensor,
nbr_pj->dvec, system->atoms[j].x );
- rtensor_ScaledAdd( total_rtensor,
+ rtensor_ScaledAdd( total_rtensor,
F_C * +(CEvd + CEclmb), temp_rtensor );
if( nbr_pj->imaginary )
- // This is an external force due to an imaginary nbr
- rtensor_ScaledAdd( data->flex_bar.P, -1.0, total_rtensor );
- else
- // This interaction is completely internal
- rtensor_Add( data->flex_bar.P, total_rtensor ); */
+ // This is an external force due to an imaginary nbr
+ rtensor_ScaledAdd( data->flex_bar.P, -1.0, total_rtensor );
+ else
+ // This interaction is completely internal
+ rtensor_Add( data->flex_bar.P, total_rtensor ); */
}
#ifdef TEST_ENERGY
rvec_MakeZero( temp );
rvec_ScaledAdd( temp, +CEvd, nbr_pj->dvec );
fprintf( out_control->evdw,
- "%6d%6d%24.15e%24.15e%24.15e%24.15e%24.15e\n",
- //i+1, j+1,
- MIN( workspace->orig_id[i], workspace->orig_id[j] ),
- MAX( workspace->orig_id[i], workspace->orig_id[j] ),
- r_ij, e_vdW, temp[0], temp[1], temp[2]/*, data->E_vdW*/ );
+ "%6d%6d%24.15e%24.15e%24.15e%24.15e%24.15e\n",
+ //i+1, j+1,
+ MIN( workspace->orig_id[i], workspace->orig_id[j] ),
+ MAX( workspace->orig_id[i], workspace->orig_id[j] ),
+ r_ij, e_vdW, temp[0], temp[1], temp[2]/*, data->E_vdW*/ );
fprintf( out_control->ecou, "%6d%6d%24.15e%24.15e%24.15e%24.15e\n",
- MIN( workspace->orig_id[i], workspace->orig_id[j] ),
- MAX( workspace->orig_id[i], workspace->orig_id[j] ),
- r_ij, system->atoms[i].q, system->atoms[j].q,
- e_ele/*, data->E_Ele*/ );
+ MIN( workspace->orig_id[i], workspace->orig_id[j] ),
+ MAX( workspace->orig_id[i], workspace->orig_id[j] ),
+ r_ij, system->atoms[i].q, system->atoms[j].q,
+ e_ele/*, data->E_Ele*/ );
#endif
#ifdef TEST_FORCES
rvec_ScaledAdd( workspace->f_vdw[i], -CEvd, nbr_pj->dvec );
@@ -337,13 +344,13 @@ void vdW_Coulomb_Energy( reax_system *system, control_params *control,
// fclose( fout );
- // fprintf( stderr, "nonbonded: ext_press (%24.15e %24.15e %24.15e)\n",
+ // fprintf( stderr, "nonbonded: ext_press (%24.15e %24.15e %24.15e)\n",
// data->ext_press[0], data->ext_press[1], data->ext_press[2] );
}
-void LR_vdW_Coulomb( reax_system *system, control_params *control,
- int i, int j, real r_ij, LR_data *lr )
+void LR_vdW_Coulomb( reax_system *system, control_params *control,
+ int i, int j, real r_ij, LR_data *lr )
{
real p_vdW1 = system->reaxprm.gp.l[28];
real p_vdW1i = 1.0 / p_vdW1;
@@ -367,12 +374,12 @@ void LR_vdW_Coulomb( reax_system *system, control_params *control,
Tap = Tap * r_ij + control->Tap1;
Tap = Tap * r_ij + control->Tap0;
- dTap = 7*control->Tap7 * r_ij + 6*control->Tap6;
- dTap = dTap * r_ij + 5*control->Tap5;
- dTap = dTap * r_ij + 4*control->Tap4;
- dTap = dTap * r_ij + 3*control->Tap3;
- dTap = dTap * r_ij + 2*control->Tap2;
- dTap += control->Tap1/r_ij;
+ dTap = 7 * control->Tap7 * r_ij + 6 * control->Tap6;
+ dTap = dTap * r_ij + 5 * control->Tap5;
+ dTap = dTap * r_ij + 4 * control->Tap4;
+ dTap = dTap * r_ij + 3 * control->Tap3;
+ dTap = dTap * r_ij + 2 * control->Tap2;
+ dTap += control->Tap1 / r_ij;
/* vdWaals calculations */
@@ -383,20 +390,21 @@ void LR_vdW_Coulomb( reax_system *system, control_params *control,
exp1 = EXP( twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
exp2 = EXP( 0.5 * twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
- lr->e_vdW = Tap * twbp->D * (exp1 - 2.0 * exp2);
+ lr->e_vdW = Tap * twbp->D * (exp1 - 2.0 * exp2);
/* fprintf(stderr,"vdW: Tap:%f, r: %f, f13:%f, D:%f, Energy:%f,\
-Gamma_w:%f, p_vdw: %f, alpha: %f, r_vdw: %f, %lf %lf\n",
-Tap, r_ij, fn13, twbp->D, Tap * twbp->D * (exp1 - 2.0 * exp2),
-powgi_vdW1, p_vdW1, twbp->alpha, twbp->r_vdW, exp1, exp2); */
+ Gamma_w:%f, p_vdw: %f, alpha: %f, r_vdw: %f, %lf %lf\n",
+ Tap, r_ij, fn13, twbp->D, Tap * twbp->D * (exp1 - 2.0 * exp2),
+ powgi_vdW1, p_vdW1, twbp->alpha, twbp->r_vdW, exp1, exp2); */
dfn13 = POW( powr_vdW1 + powgi_vdW1, p_vdW1i - 1.0) * POW(r_ij, p_vdW1 - 2.0);
- lr->CEvd = dTap * twbp->D * (exp1 - 2 * exp2) -
- Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2) * dfn13;
+ lr->CEvd = dTap * twbp->D * (exp1 - 2 * exp2) -
+ Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2) * dfn13;
/*vdWaals Calculations*/
- if(system->reaxprm.gp.vdw_type==1 || system->reaxprm.gp.vdw_type==3)
- { // shielding
+ if (system->reaxprm.gp.vdw_type == 1 || system->reaxprm.gp.vdw_type == 3)
+ {
+ // shielding
powr_vdW1 = POW(r_ij, p_vdW1);
powgi_vdW1 = POW( 1.0 / twbp->gamma_w, p_vdW1);
@@ -404,30 +412,32 @@ powgi_vdW1, p_vdW1, twbp->alpha, twbp->r_vdW, exp1, exp2); */
exp1 = EXP( twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
exp2 = EXP( 0.5 * twbp->alpha * (1.0 - fn13 / twbp->r_vdW) );
- lr->e_vdW = Tap * twbp->D * (exp1 - 2.0 * exp2);
+ lr->e_vdW = Tap * twbp->D * (exp1 - 2.0 * exp2);
- dfn13 = POW( powr_vdW1 + powgi_vdW1, p_vdW1i - 1.0) *
- POW(r_ij, p_vdW1 - 2.0);
+ dfn13 = POW( powr_vdW1 + powgi_vdW1, p_vdW1i - 1.0) *
+ POW(r_ij, p_vdW1 - 2.0);
- lr->CEvd = dTap * twbp->D * (exp1 - 2.0 * exp2) -
- Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2) * dfn13;
+ lr->CEvd = dTap * twbp->D * (exp1 - 2.0 * exp2) -
+ Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2) * dfn13;
}
- else{ // no shielding
+ else // no shielding
+ {
exp1 = EXP( twbp->alpha * (1.0 - r_ij / twbp->r_vdW) );
exp2 = EXP( 0.5 * twbp->alpha * (1.0 - r_ij / twbp->r_vdW) );
lr->e_vdW = Tap * twbp->D * (exp1 - 2.0 * exp2);
- lr->CEvd = dTap * twbp->D * (exp1 - 2.0 * exp2) -
- Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2);
+ lr->CEvd = dTap * twbp->D * (exp1 - 2.0 * exp2) -
+ Tap * twbp->D * (twbp->alpha / twbp->r_vdW) * (exp1 - exp2);
}
- if(system->reaxprm.gp.vdw_type==2 || system->reaxprm.gp.vdw_type==3)
- { // innner wall
- e_core = twbp->ecore * EXP(twbp->acore * (1.0-(r_ij/twbp->rcore)));
+ if (system->reaxprm.gp.vdw_type == 2 || system->reaxprm.gp.vdw_type == 3)
+ {
+ // innner wall
+ e_core = twbp->ecore * EXP(twbp->acore * (1.0 - (r_ij / twbp->rcore)));
lr->e_vdW += Tap * e_core;
- de_core = -(twbp->acore/twbp->rcore) * e_core;
+ de_core = -(twbp->acore / twbp->rcore) * e_core;
lr->CEvd += dTap * e_core + Tap * de_core;
}
@@ -439,10 +449,10 @@ powgi_vdW1, p_vdW1, twbp->alpha, twbp->r_vdW, exp1, exp2); */
lr->H = EV_to_KCALpMOL * tmp;
lr->e_ele = C_ele * tmp;
/* fprintf( stderr,"i:%d(%d), j:%d(%d), gamma:%f,\
-Tap:%f, dr3gamij_3:%f, qi: %f, qj: %f\n",
-i, system->atoms[i].type, j, system->atoms[j].type,
-twbp->gamma, Tap, dr3gamij_3,
-system->atoms[i].q, system->atoms[j].q ); */
+ Tap:%f, dr3gamij_3:%f, qi: %f, qj: %f\n",
+ i, system->atoms[i].type, j, system->atoms[j].type,
+ twbp->gamma, Tap, dr3gamij_3,
+ system->atoms[i].q, system->atoms[j].q ); */
lr->CEclmb = C_ele * ( dTap - Tap * r_ij / dr3gamij_1 ) / dr3gamij_3;
/* fprintf( stdout, "%d %d\t%g\t%g %g\t%g %g\t%g %g\n",
@@ -454,10 +464,9 @@ system->atoms[i].q, system->atoms[j].q ); */
}
-void Tabulated_vdW_Coulomb_Energy( reax_system *system, control_params *control,
- simulation_data *data,
- static_storage *workspace, list **lists,
- output_controls *out_control )
+void Tabulated_vdW_Coulomb_Energy( reax_system *system,
+ control_params *control, simulation_data *data, static_storage *workspace,
+ list **lists, output_controls *out_control )
{
int i, j, pj, r, steps, update_freq, update_energies;
int type_i, type_j, tmin, tmax;
@@ -474,13 +483,16 @@ void Tabulated_vdW_Coulomb_Energy( reax_system *system, control_params *control,
update_freq = out_control->energy_update_freq;
update_energies = update_freq > 0 && steps % update_freq == 0;
- for( i = 0; i < system->N; ++i ) {
+ for ( i = 0; i < system->N; ++i )
+ {
type_i = system->atoms[i].type;
- start_i = Start_Index(i,far_nbrs);
- end_i = End_Index(i,far_nbrs);
+ start_i = Start_Index(i, far_nbrs);
+ end_i = End_Index(i, far_nbrs);
- for( pj = start_i; pj < end_i; ++pj )
- if( far_nbrs->select.far_nbr_list[pj].d <= control->r_cut ) {
+ for ( pj = start_i; pj < end_i; ++pj )
+ {
+ if ( far_nbrs->select.far_nbr_list[pj].d <= control->r_cut )
+ {
nbr_pj = &( far_nbrs->select.far_nbr_list[pj] );
j = nbr_pj->nbr;
type_j = system->atoms[j].type;
@@ -488,43 +500,46 @@ void Tabulated_vdW_Coulomb_Energy( reax_system *system, control_params *control,
self_coef = (i == j) ? 0.5 : 1.0;
tmin = MIN( type_i, type_j );
tmax = MAX( type_i, type_j );
- t = &( LR[ index_lr (tmin,tmax,system->reaxprm.num_atom_types) ] );
+ t = &( LR[ index_lr(tmin,tmax,system->reaxprm.num_atom_types) ] );
/* Cubic Spline Interpolation */
r = (int)(r_ij * t->inv_dx);
- if( r == 0 ) ++r;
- base = (real)(r+1) * t->dx;
+ if ( r == 0 ) ++r;
+ base = (real)(r + 1) * t->dx;
dif = r_ij - base;
//fprintf(stderr, "r: %f, i: %d, base: %f, dif: %f\n", r, i, base, dif);
- if( update_energies ) {
- e_vdW = ((t->vdW[r].d*dif + t->vdW[r].c)*dif + t->vdW[r].b)*dif +
- t->vdW[r].a;
+ if ( update_energies )
+ {
+ e_vdW = ((t->vdW[r].d * dif + t->vdW[r].c) * dif + t->vdW[r].b) * dif +
+ t->vdW[r].a;
e_vdW *= self_coef;
- e_ele = ((t->ele[r].d*dif + t->ele[r].c)*dif + t->ele[r].b)*dif +
- t->ele[r].a;
+ e_ele = ((t->ele[r].d * dif + t->ele[r].c) * dif + t->ele[r].b) * dif +
+ t->ele[r].a;
e_ele *= self_coef * system->atoms[i].q * system->atoms[j].q;
data->E_vdW += e_vdW;
data->E_Ele += e_ele;
- }
+ }
- CEvd = ((t->CEvd[r].d*dif + t->CEvd[r].c)*dif + t->CEvd[r].b)*dif +
- t->CEvd[r].a;
+ CEvd = ((t->CEvd[r].d * dif + t->CEvd[r].c) * dif + t->CEvd[r].b) * dif +
+ t->CEvd[r].a;
CEvd *= self_coef;
//CEvd = (3*t->vdW[r].d*dif + 2*t->vdW[r].c)*dif + t->vdW[r].b;
- CEclmb = ((t->CEclmb[r].d*dif+t->CEclmb[r].c)*dif+t->CEclmb[r].b)*dif +
- t->CEclmb[r].a;
+ CEclmb = ((t->CEclmb[r].d * dif + t->CEclmb[r].c) * dif + t->CEclmb[r].b) * dif +
+ t->CEclmb[r].a;
CEclmb *= self_coef * system->atoms[i].q * system->atoms[j].q;
- if( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT) {
+ if ( control->ensemble == NVE || control->ensemble == NVT || control->ensemble == bNVT)
+ {
rvec_ScaledAdd( system->atoms[i].f, -(CEvd + CEclmb), nbr_pj->dvec );
rvec_ScaledAdd( system->atoms[j].f, +(CEvd + CEclmb), nbr_pj->dvec );
}
- else { // NPT, iNPT or sNPT
- /* for pressure coupling, terms not related to bond order
+ else // NPT, iNPT or sNPT
+ {
+ /* for pressure coupling, terms not related to bond order
derivatives are added directly into pressure vector/tensor */
rvec_Scale( temp, CEvd + CEclmb, nbr_pj->dvec );
rvec_ScaledAdd( system->atoms[i].f, -1., temp );
@@ -535,11 +550,11 @@ void Tabulated_vdW_Coulomb_Energy( reax_system *system, control_params *control,
#ifdef TEST_ENERGY
fprintf(out_control->evdw, "%6d%6d%24.15e%24.15e%24.15e\n",
- workspace->orig_id[i], workspace->orig_id[j],
+ workspace->orig_id[i], workspace->orig_id[j],
r_ij, e_vdW, data->E_vdW );
- fprintf(out_control->ecou,"%6d%6d%24.15e%24.15e%24.15e%24.15e%24.15e\n",
+ fprintf(out_control->ecou, "%6d%6d%24.15e%24.15e%24.15e%24.15e%24.15e\n",
workspace->orig_id[i], workspace->orig_id[j],
- r_ij, system->atoms[i].q, system->atoms[j].q,
+ r_ij, system->atoms[i].q, system->atoms[j].q,
e_ele, data->E_Ele );
#endif
#ifdef TEST_FORCES
@@ -549,23 +564,24 @@ void Tabulated_vdW_Coulomb_Energy( reax_system *system, control_params *control,
rvec_ScaledAdd( workspace->f_ele[j], +CEclmb, nbr_pj->dvec );
#endif
}
+ }
}
}
#if defined(OLD)
- /* Linear extrapolation */
- /*p = (r_ij * t->inv_dx;
- r = (int) p;
- prev = &( t->y[r] );
- next = &( t->y[r+1] );
-
- tmp = p - r;
- e_vdW = self_coef * (prev->e_vdW + tmp*(next->e_vdW - prev->e_vdW ));
- CEvd = self_coef * (prev->CEvd + tmp*(next->CEvd - prev->CEvd ));
-
- e_ele = self_coef * (prev->e_ele + tmp*(next->e_ele - prev->e_ele ));
- e_ele = e_ele * system->atoms[i].q * system->atoms[j].q;
- CEclmb = self_coef * (prev->CEclmb+tmp*(next->CEclmb - prev->CEclmb));
- CEclmb = CEclmb * system->atoms[i].q * system->atoms[j].q;*/
+/* Linear extrapolation */
+/*p = (r_ij * t->inv_dx;
+ r = (int) p;
+ prev = &( t->y[r] );
+ next = &( t->y[r+1] );
+
+ tmp = p - r;
+ e_vdW = self_coef * (prev->e_vdW + tmp*(next->e_vdW - prev->e_vdW ));
+ CEvd = self_coef * (prev->CEvd + tmp*(next->CEvd - prev->CEvd ));
+
+ e_ele = self_coef * (prev->e_ele + tmp*(next->e_ele - prev->e_ele ));
+ e_ele = e_ele * system->atoms[i].q * system->atoms[j].q;
+ CEclmb = self_coef * (prev->CEclmb+tmp*(next->CEclmb - prev->CEclmb));
+ CEclmb = CEclmb * system->atoms[i].q * system->atoms[j].q;*/
#endif
diff --git a/PuReMD-GPU/src/vector.c b/PuReMD-GPU/src/vector.c
index 7cf06eb8e6cb1560b651b8b16a091f3a387cdb6c..e396344d173a6d5343faf9d675f48a8ea4e0ca04 100644
--- a/PuReMD-GPU/src/vector.c
+++ b/PuReMD-GPU/src/vector.c
@@ -21,53 +21,90 @@
#include "vector.h"
-int Vector_isZero( real* v, int k )
+inline int Vector_isZero( const real * const v, const unsigned int k )
{
- for( --k; k>=0; --k )
- if( fabs( v[k] ) > ALMOST_ZERO )
- return 0;
+ unsigned int i;
- return 1;
+ #pragma omp master
+ {
+ ret = TRUE;
+ }
+
+ #pragma omp barrier
+
+ #pragma omp for reduction(&&: ret) schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ if ( FABS( v[i] ) > ALMOST_ZERO )
+ {
+ ret = FALSE;
+ }
+ }
+
+ return ret;
}
-void Vector_MakeZero( real *v, int k )
+inline void Vector_MakeZero( real * const v, const unsigned int k )
{
- for( --k; k>=0; --k )
- v[k] = 0;
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ v[i] = ZERO;
+ }
}
-void Vector_Copy( real* dest, real* v, int k )
+inline void Vector_Copy( real * const dest, const real * const v, const unsigned int k )
{
- for( --k; k>=0; --k )
- dest[k] = v[k];
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ dest[i] = v[i];
+ }
}
-void Vector_Print( FILE *fout, char *vname, real *v, int k )
+void Vector_Print( FILE * const fout, const char * const vname, const real * const v,
+ const unsigned int k )
{
- int i;
+ unsigned int i;
fprintf( fout, "%s:\n", vname );
- for( i = 0; i < k; ++i )
+ for ( i = 0; i < k; ++i )
+ {
fprintf( fout, "%24.15e\n", v[i] );
+ }
fprintf( fout, "\n" );
}
-real Norm( real* v1, int k )
+inline real Norm( const real * const v1, const unsigned int k )
{
- real ret = 0;
+ unsigned int i;
+
+ #pragma omp master
+ {
+ ret2 = ZERO;
+ }
- for( --k; k>=0; --k )
- ret += SQR( v1[k] );
+ #pragma omp barrier
+
+ #pragma omp for reduction(+: ret2) schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ ret2 += SQR( v1[i] );
+ }
- return SQRT( ret );
+ return SQRT( ret2 );
}
-void rvec_Sum( rvec ret, rvec v1 ,rvec v2 )
+inline void rvec_Sum( rvec ret, const rvec v1 , const rvec v2 )
{
ret[0] = v1[0] + v2[0];
ret[1] = v1[1] + v2[1];
@@ -75,13 +112,14 @@ void rvec_Sum( rvec ret, rvec v1 ,rvec v2 )
}
-real rvec_ScaledDot( real c1, rvec v1, real c2, rvec v2 )
+inline real rvec_ScaledDot( const real c1, const rvec v1,
+ const real c2, const rvec v2 )
{
- return (c1*c2) * (v1[0]*v2[0] + v1[1]*v2[1] + v1[2]*v2[2]);
+ return (c1 * c2) * (v1[0] * v2[0] + v1[1] * v2[1] + v1[2] * v2[2]);
}
-void rvec_Multiply( rvec r, rvec v1, rvec v2 )
+inline void rvec_Multiply( rvec r, const rvec v1, const rvec v2 )
{
r[0] = v1[0] * v2[0];
r[1] = v1[1] * v2[1];
@@ -89,7 +127,7 @@ void rvec_Multiply( rvec r, rvec v1, rvec v2 )
}
-void rvec_Divide( rvec r, rvec v1, rvec v2 )
+inline void rvec_Divide( rvec r, const rvec v1, const rvec v2 )
{
r[0] = v1[0] / v2[0];
r[1] = v1[1] / v2[1];
@@ -97,7 +135,7 @@ void rvec_Divide( rvec r, rvec v1, rvec v2 )
}
-void rvec_iDivide( rvec r, rvec v1, ivec v2 )
+inline void rvec_iDivide( rvec r, const rvec v1, const ivec v2 )
{
r[0] = v1[0] / v2[0];
r[1] = v1[1] / v2[1];
@@ -105,7 +143,7 @@ void rvec_iDivide( rvec r, rvec v1, ivec v2 )
}
-void rvec_Invert( rvec r, rvec v )
+inline void rvec_Invert( rvec r, const rvec v )
{
r[0] = 1. / v[0];
r[1] = 1. / v[1];
@@ -113,154 +151,189 @@ void rvec_Invert( rvec r, rvec v )
}
-void rvec_OuterProduct( rtensor r, rvec v1, rvec v2 )
+inline void rvec_OuterProduct( rtensor r, const rvec v1, const rvec v2 )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
r[i][j] = v1[i] * v2[j];
+ }
+ }
}
-
-int rvec_isZero( rvec v )
+inline int rvec_isZero( const rvec v )
{
- if( fabs(v[0]) > ALMOST_ZERO ||
- fabs(v[1]) > ALMOST_ZERO ||
+ if ( fabs(v[0]) > ALMOST_ZERO ||
+ fabs(v[1]) > ALMOST_ZERO ||
fabs(v[2]) > ALMOST_ZERO )
- return 0;
- return 1;
+ {
+ return FALSE;
+ }
+ return TRUE;
}
-void rtensor_Multiply( rtensor ret, rtensor m1, rtensor m2 )
+inline void rtensor_Multiply( rtensor ret, rtensor m1, rtensor m2 )
{
- int i, j, k;
+ unsigned int i, j, k;
rtensor temp;
// check if the result matrix is the same as one of m1, m2.
- // if so, we cannot modify the contents of m1 or m2, so
+ // if so, we cannot modify the contents of m1 or m2, so
// we have to use a temp matrix.
- if( ret == m1 || ret == m2 )
+ if ( ret == m1 || ret == m2 )
{
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ for ( j = 0; j < 3; ++j )
{
- temp[i][j] = 0;
- for( k = 0; k < 3; ++k )
+ temp[i][j] = 0;
+ for ( k = 0; k < 3; ++k )
temp[i][j] += m1[i][k] * m2[k][j];
}
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
- ret[i][j] = temp[i][j];
+ for ( i = 0; i < 3; ++i )
+ for ( j = 0; j < 3; ++j )
+ ret[i][j] = temp[i][j];
}
else
{
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ for ( j = 0; j < 3; ++j )
{
- ret[i][j] = 0;
- for( k = 0; k < 3; ++k )
+ ret[i][j] = 0;
+ for ( k = 0; k < 3; ++k )
ret[i][j] += m1[i][k] * m2[k][j];
}
}
}
-void rtensor_MatVec( rvec ret, rtensor m, rvec v )
+inline void rtensor_MatVec( rvec ret, rtensor m, const rvec v )
{
- int i;
+ unsigned int i;
rvec temp;
- // if ret is the same vector as v, we cannot modify the
+ // if ret is the same vector as v, we cannot modify the
// contents of v until all computation is finished.
- if( ret == v )
+ if ( ret == v )
{
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
+ {
temp[i] = m[i][0] * v[0] + m[i][1] * v[1] + m[i][2] * v[2];
+ }
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
+ {
ret[i] = temp[i];
+ }
}
else
{
- for( i = 0; i < 3; ++i )
+ for ( i = 0; i < 3; ++i )
+ {
ret[i] = m[i][0] * v[0] + m[i][1] * v[1] + m[i][2] * v[2];
+ }
}
}
-void rtensor_Scale( rtensor ret, real c, rtensor m )
+inline void rtensor_Scale( rtensor ret, const real c, rtensor m )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] = c * m[i][j];
+ }
+ }
}
-void rtensor_Add( rtensor ret, rtensor t )
+inline void rtensor_Add( rtensor ret, rtensor t )
{
int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] += t[i][j];
+ }
+ }
}
-void rtensor_ScaledAdd( rtensor ret, real c, rtensor t )
+inline void rtensor_ScaledAdd( rtensor ret, const real c, rtensor t )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] += c * t[i][j];
+ }
+ }
}
-void rtensor_Sum( rtensor ret, rtensor t1, rtensor t2 )
+inline void rtensor_Sum( rtensor ret, rtensor t1, rtensor t2 )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] = t1[i][j] + t2[i][j];
+ }
+ }
}
-void rtensor_ScaledSum( rtensor ret, real c1, rtensor t1,
- real c2, rtensor t2 )
+inline void rtensor_ScaledSum( rtensor ret, const real c1, rtensor t1,
+ const real c2, rtensor t2 )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] = c1 * t1[i][j] + c2 * t2[i][j];
+ }
+ }
}
-void rtensor_Copy( rtensor ret, rtensor t )
+inline void rtensor_Copy( rtensor ret, rtensor t )
{
- int i, j;
+ unsigned int i, j;
- for( i = 0; i < 3; ++i )
- for( j = 0; j < 3; ++j )
+ for ( i = 0; i < 3; ++i )
+ {
+ for ( j = 0; j < 3; ++j )
+ {
ret[i][j] = t[i][j];
+ }
+ }
}
-void rtensor_Identity( rtensor t )
+inline void rtensor_Identity( rtensor t )
{
t[0][0] = t[1][1] = t[2][2] = 1;
t[0][1] = t[0][2] = t[1][0] = t[1][2] = t[2][0] = t[2][1] = ZERO;
}
-void rtensor_MakeZero( rtensor t )
+inline void rtensor_MakeZero( rtensor t )
{
t[0][0] = t[0][1] = t[0][2] = ZERO;
t[1][0] = t[1][1] = t[1][2] = ZERO;
@@ -268,50 +341,58 @@ void rtensor_MakeZero( rtensor t )
}
-void rtensor_Transpose( rtensor ret, rtensor t )
+inline void rtensor_Transpose( rtensor ret, rtensor t )
{
- ret[0][0] = t[0][0], ret[1][1] = t[1][1], ret[2][2] = t[2][2];
- ret[0][1] = t[1][0], ret[0][2] = t[2][0];
- ret[1][0] = t[0][1], ret[1][2] = t[2][1];
- ret[2][0] = t[0][2], ret[2][1] = t[1][2];
+ ret[0][0] = t[0][0];
+ ret[1][1] = t[1][1];
+ ret[2][2] = t[2][2];
+
+ ret[0][1] = t[1][0];
+ ret[0][2] = t[2][0];
+
+ ret[1][0] = t[0][1];
+ ret[1][2] = t[2][1];
+
+ ret[2][0] = t[0][2];
+ ret[2][1] = t[1][2];
}
-real rtensor_Det( rtensor t )
+inline real rtensor_Det( rtensor t )
{
return ( t[0][0] * (t[1][1] * t[2][2] - t[1][2] * t[2][1] ) +
- t[0][1] * (t[1][2] * t[2][0] - t[1][0] * t[2][2] ) +
- t[0][2] * (t[1][0] * t[2][1] - t[1][1] * t[2][0] ) );
+ t[0][1] * (t[1][2] * t[2][0] - t[1][0] * t[2][2] ) +
+ t[0][2] * (t[1][0] * t[2][1] - t[1][1] * t[2][0] ) );
}
-real rtensor_Trace( rtensor t )
+inline real rtensor_Trace( rtensor t )
{
return (t[0][0] + t[1][1] + t[2][2]);
}
-void Print_rTensor(FILE* fp, rtensor t)
+void Print_rTensor(FILE * const fp, rtensor t)
{
- int i, j;
+ unsigned int i, j;
- for (i=0; i < 3; i++)
+ for (i = 0; i < 3; i++)
{
- fprintf(fp,"[");
- for (j=0; j < 3; j++)
- fprintf(fp,"%8.3f,\t",t[i][j]);
- fprintf(fp,"]\n");
+ fprintf(fp, "[");
+ for (j = 0; j < 3; j++)
+ fprintf(fp, "%8.3f,\t", t[i][j]);
+ fprintf(fp, "]\n");
}
}
-void ivec_MakeZero( ivec v )
+inline void ivec_MakeZero( ivec v )
{
v[0] = v[1] = v[2] = 0;
}
-void ivec_rScale( ivec dest, real C, rvec src )
+inline void ivec_rScale( ivec dest, const real C, const rvec src )
{
dest[0] = (int)(C * src[0]);
dest[1] = (int)(C * src[1]);
@@ -319,20 +400,22 @@ void ivec_rScale( ivec dest, real C, rvec src )
}
-int ivec_isZero( ivec v )
+inline int ivec_isZero( const ivec v )
{
- if( v[0]==0 && v[1]==0 && v[2]==0 )
- return 1;
- return 0;
+ if ( v[0] == 0 && v[1] == 0 && v[2] == 0 )
+ {
+ return TRUE;
+ }
+ return FALSE;
}
-int ivec_isEqual( ivec v1, ivec v2 )
+inline int ivec_isEqual( const ivec v1, const ivec v2 )
{
- if( v1[0]==v2[0] && v1[1]==v2[1] && v1[2]==v2[2] )
- return 1;
+ if ( v1[0] == v2[0] && v1[1] == v2[1] && v1[2] == v2[2] )
+ {
+ return TRUE;
+ }
- return 0;
+ return FALSE;
}
-
-
diff --git a/PuReMD-GPU/src/vector.h b/PuReMD-GPU/src/vector.h
index e1111e514928e79fc79197a0f2486d5eefb1cfa3..79748544fb8349bb797efe246e3430561fefef14 100644
--- a/PuReMD-GPU/src/vector.h
+++ b/PuReMD-GPU/src/vector.h
@@ -26,72 +26,85 @@
#include "random.h"
+/* global to make OpenMP shared (Vector_isZero) */
+unsigned int ret;
+/* global to make OpenMP shared (Dot, Norm) */
+real ret2;
+
+
#ifdef __cplusplus
extern "C" {
#endif
-int Vector_isZero( real*, int );
-void Vector_MakeZero( real*, int );
-void Vector_Copy( real*, real*, int );
-//void Vector_Scale( real*, real, real*, int );
-//void Vector_Sum( real*, real, real*, real, real*, int );
-//void Vector_Add( real*, real, real*, int );
-void Vector_Print( FILE*, char*, real*, int );
-real Norm( real*, int );
-
-void rvec_Sum( rvec, rvec, rvec );
-real rvec_ScaledDot( real, rvec, real, rvec );
-void rvec_Multiply( rvec, rvec, rvec );
-void rvec_Divide( rvec, rvec, rvec );
-void rvec_iDivide( rvec, rvec, ivec );
-void rvec_Invert( rvec, rvec );
-void rvec_OuterProduct( rtensor, rvec, rvec );
-int rvec_isZero( rvec );
+int Vector_isZero( const real * const, const unsigned int );
+void Vector_MakeZero( real * const, const unsigned int );
+void Vector_Copy( real * const, const real * const, const unsigned int );
+void Vector_Print( FILE * const, const char * const, const real * const, const unsigned int );
+real Norm( const real * const, const unsigned int );
+
+void rvec_Sum( rvec, const rvec, const rvec );
+real rvec_ScaledDot( const real, const rvec, const real, const rvec );
+void rvec_Multiply( rvec, const rvec, const rvec );
+void rvec_Divide( rvec, const rvec, const rvec );
+void rvec_iDivide( rvec, const rvec, const ivec );
+void rvec_Invert( rvec, const rvec );
+void rvec_OuterProduct( rtensor, const rvec, const rvec );
+int rvec_isZero( const rvec );
void rtensor_MakeZero( rtensor );
void rtensor_Multiply( rtensor, rtensor, rtensor );
-void rtensor_MatVec( rvec, rtensor, rvec );
-void rtensor_Scale( rtensor, real, rtensor );
+void rtensor_MatVec( rvec, rtensor, const rvec );
+void rtensor_Scale( rtensor, const real, rtensor );
void rtensor_Add( rtensor, rtensor );
-void rtensor_ScaledAdd( rtensor, real, rtensor );
+void rtensor_ScaledAdd( rtensor, const real, rtensor );
void rtensor_Sum( rtensor, rtensor, rtensor );
-void rtensor_ScaledSum( rtensor, real, rtensor, real, rtensor );
-void rtensor_Scale( rtensor, real, rtensor );
+void rtensor_ScaledSum( rtensor, const real, rtensor, const real, rtensor );
+void rtensor_Scale( rtensor, const real, rtensor );
void rtensor_Copy( rtensor, rtensor );
void rtensor_Identity( rtensor );
void rtensor_Transpose( rtensor, rtensor );
real rtensor_Det( rtensor );
real rtensor_Trace( rtensor );
-void Print_rTensor(FILE*, rtensor);
+void Print_rTensor(FILE * const, rtensor);
-int ivec_isZero( ivec );
-int ivec_isEqual( ivec, ivec );
+int ivec_isZero( const ivec );
+int ivec_isEqual( const ivec, const ivec );
void ivec_MakeZero( ivec );
-void ivec_rScale( ivec, real, rvec );
+void ivec_rScale( ivec, const real, const rvec );
-static inline HOST_DEVICE real Dot( real* v1, real* v2, int k )
+static inline HOST_DEVICE real Dot( const real * const v1, const real * const v2, const unsigned int k )
{
- real ret = 0;
+ unsigned int i;
+
+ #pragma omp master
+ {
+ ret2 = ZERO;
+ }
+
+ #pragma omp barrier
- for ( --k; k >= 0; --k )
- ret += v1[k] * v2[k];
- return ret;
+ #pragma omp for reduction(+: ret2) schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ ret2 += v1[i] * v2[i];
+ }
+
+ return ret2;
}
-/////////////////////////////
-//rvec functions
-/////////////////////////////
static inline HOST_DEVICE void rvec_MakeZero( rvec v )
{
- v[0] = v[1] = v[2] = ZERO;
+ v[0] = ZERO;
+ v[1] = ZERO;
+ v[2] = ZERO;
}
-static inline HOST_DEVICE void rvec_Add( rvec ret, rvec v )
+static inline HOST_DEVICE void rvec_Add( rvec ret, const rvec v )
{
ret[0] += v[0];
ret[1] += v[1];
@@ -99,13 +112,15 @@ static inline HOST_DEVICE void rvec_Add( rvec ret, rvec v )
}
-static inline HOST_DEVICE void rvec_Copy( rvec dest, rvec src )
+static inline HOST_DEVICE void rvec_Copy( rvec dest, const rvec src )
{
- dest[0] = src[0], dest[1] = src[1], dest[2] = src[2];
+ dest[0] = src[0];
+ dest[1] = src[1];
+ dest[2] = src[2];
}
-static inline HOST_DEVICE void rvec_Cross( rvec ret, rvec v1, rvec v2 )
+static inline HOST_DEVICE void rvec_Cross( rvec ret, const rvec v1, const rvec v2 )
{
ret[0] = v1[1] * v2[2] - v1[2] * v2[1];
ret[1] = v1[2] * v2[0] - v1[0] * v2[2];
@@ -113,13 +128,16 @@ static inline HOST_DEVICE void rvec_Cross( rvec ret, rvec v1, rvec v2 )
}
-static inline HOST_DEVICE void rvec_ScaledAdd( rvec ret, real c, rvec v )
+static inline HOST_DEVICE void rvec_ScaledAdd( rvec ret, const real c, const rvec v )
{
- ret[0] += c * v[0], ret[1] += c * v[1], ret[2] += c * v[2];
+ ret[0] += c * v[0];
+ ret[1] += c * v[1];
+ ret[2] += c * v[2];
}
-static inline HOST_DEVICE void rvec_ScaledSum( rvec ret, real c1, rvec v1 , real c2, rvec v2 )
+static inline HOST_DEVICE void rvec_ScaledSum( rvec ret, const real c1, const rvec v1,
+ const real c2, const rvec v2 )
{
ret[0] = c1 * v1[0] + c2 * v2[0];
ret[1] = c1 * v1[1] + c2 * v2[1];
@@ -135,25 +153,27 @@ static inline HOST_DEVICE void rvec_Random( rvec v )
}
-static inline HOST_DEVICE real rvec_Norm_Sqr( rvec v )
+static inline HOST_DEVICE real rvec_Norm_Sqr( const rvec v )
{
return SQR(v[0]) + SQR(v[1]) + SQR(v[2]);
}
-static inline HOST_DEVICE void rvec_Scale( rvec ret, real c, rvec v )
+static inline HOST_DEVICE void rvec_Scale( rvec ret, const real c, const rvec v )
{
- ret[0] = c * v[0], ret[1] = c * v[1], ret[2] = c * v[2];
+ ret[0] = c * v[0];
+ ret[1] = c * v[1];
+ ret[2] = c * v[2];
}
-static inline HOST_DEVICE real rvec_Dot( rvec v1, rvec v2 )
+static inline HOST_DEVICE real rvec_Dot( const rvec v1, const rvec v2 )
{
return v1[0] * v2[0] + v1[1] * v2[1] + v1[2] * v2[2];
}
-static inline HOST_DEVICE void rvec_iMultiply( rvec r, ivec v1, rvec v2 )
+static inline HOST_DEVICE void rvec_iMultiply( rvec r, const ivec v1, const rvec v2 )
{
r[0] = v1[0] * v2[0];
r[1] = v1[1] * v2[1];
@@ -161,22 +181,21 @@ static inline HOST_DEVICE void rvec_iMultiply( rvec r, ivec v1, rvec v2 )
}
-static inline HOST_DEVICE real rvec_Norm( rvec v )
+static inline HOST_DEVICE real rvec_Norm( const rvec v )
{
return SQRT( SQR(v[0]) + SQR(v[1]) + SQR(v[2]) );
}
-/////////////////
-//ivec functions
-/////////////////
-static inline HOST_DEVICE void ivec_Copy( ivec dest , ivec src )
+static inline HOST_DEVICE void ivec_Copy( ivec dest , const ivec src )
{
- dest[0] = src[0], dest[1] = src[1], dest[2] = src[2];
+ dest[0] = src[0];
+ dest[1] = src[1];
+ dest[2] = src[2];
}
-static inline HOST_DEVICE void ivec_Scale( ivec dest, real C, ivec src )
+static inline HOST_DEVICE void ivec_Scale( ivec dest, const real C, const ivec src )
{
dest[0] = C * src[0];
dest[1] = C * src[1];
@@ -184,7 +203,7 @@ static inline HOST_DEVICE void ivec_Scale( ivec dest, real C, ivec src )
}
-static inline HOST_DEVICE void ivec_Sum( ivec dest, ivec v1, ivec v2 )
+static inline HOST_DEVICE void ivec_Sum( ivec dest, const ivec v1, const ivec v2 )
{
dest[0] = v1[0] + v2[0];
dest[1] = v1[1] + v2[1];
@@ -192,27 +211,43 @@ static inline HOST_DEVICE void ivec_Sum( ivec dest, ivec v1, ivec v2 )
}
-/////////////////
-//vector functions
-/////////////////
-static inline HOST_DEVICE void Vector_Sum( real* dest, real c, real* v, real d, real* y, int k )
+static inline HOST_DEVICE void Vector_Sum( real * const dest, const real c,
+ const real * const v, const real d, const real * const y,
+ const unsigned int k )
{
- for (k--; k >= 0; k--)
- dest[k] = c * v[k] + d * y[k];
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ dest[i] = c * v[i] + d * y[i];
+ }
}
-static inline HOST_DEVICE void Vector_Scale( real* dest, real c, real* v, int k )
+static inline HOST_DEVICE void Vector_Scale( real * const dest, const real c,
+ const real * const v, const unsigned int k )
{
- for (k--; k >= 0; k--)
- dest[k] = c * v[k];
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ dest[i] = c * v[i];
+ }
}
-static inline HOST_DEVICE void Vector_Add( real* dest, real c, real* v, int k )
+static inline HOST_DEVICE void Vector_Add( real * const dest, const real c,
+ const real * const v, const unsigned int k )
{
- for (k--; k >= 0; k--)
- dest[k] += c * v[k];
+ unsigned int i;
+
+ #pragma omp for schedule(static)
+ for ( i = 0; i < k; ++i )
+ {
+ dest[i] += c * v[i];
+ }
}
#ifdef __cplusplus
diff --git a/README.md b/README.md
index 442146a67a8a5c20732075c6daf7c541aa72e5f6..abdadbab72c5d297de9a8559471a036ee229c001 100644
--- a/README.md
+++ b/README.md
@@ -7,5 +7,5 @@ Files from the [Purdue Reactive Molecular Dynamics](https://www.cs.purdue.edu/pu
Roughly by target platform
- [Serial](https://www.cs.purdue.edu/puremd/docs/80859.pdf)
- [MPI (message passing interface)](https://www.cs.purdue.edu/puremd/docs/Parallel-Reactive-Molecular-Dynamics.pdf)
-- [CUDA (single GPU)](http://dx.doi.org/10.1016/j.jcp.2014.04.035) (single GPU)
+- [CUDA (single GPU)](http://dx.doi.org/10.1016/j.jcp.2014.04.035)
- [CUDA+MPI (multi-GPU)](https://www.cs.purdue.edu/puremd/docs/pgpuremd.pdf)
diff --git a/configure.ac b/configure.ac
index 2488af52296535e6ec37ca3c3d687e7f905d78a7..659ab4575e6565b17a6624d586145077ee98a47a 100644
--- a/configure.ac
+++ b/configure.ac
@@ -46,23 +46,25 @@ AC_ARG_ENABLE([mpi-gpu],
[enable MPI+CUDA (multi GPU) support @<:@default: no@:>@])],
[pack_mpi_gpu_enabled=${enableval}], [pack_mpi_gpu_enabled=no])
-if test "x${pack_serial_enabled}" = "xyes" || test "x${pack_openmp_enabled}" = "xyes"; then
- AC_CONFIG_SUBDIRS([sPuReMD])
+if test "x${pack_serial_enabled}" = "xyes" || test "x${pack_openmp_enabled}" = "xyes" || test "x${pack_gpu_enabled}" = "xyes"; then
+ AC_CONFIG_SUBDIRS([PuReMD-GPU])
if test "x${pack_serial_enabled}" = "xyes" || test "x${pack_openmp_enabled}" != "xyes"; then
export BUILD_OPENMP="no"
else
- export BUILD_OPENMP="yes"
+ if test "x${pack_gpu_enabled}" = "xyes"; then
+ export BUILD_GPU="yes"
+ else
+ export BUILD_OPENMP="yes"
+ fi
fi
fi
AM_CONDITIONAL([BUILD_S_OMP], [test "x${pack_serial_enabled}" = "xyes" || test "x${pack_openmp_enabled}" = "xyes"])
+AM_CONDITIONAL([BUILD_GPU], [test "x${pack_gpu_enabled}" = "xyes"])
+
if test "x${pack_mpi_enabled}" = "xyes"; then
AC_CONFIG_SUBDIRS([PuReMD])
fi
AM_CONDITIONAL([BUILD_MPI], [test "x${pack_mpi_enabled}" = "xyes"])
-if test "x${pack_gpu_enabled}" = "xyes"; then
- AC_CONFIG_SUBDIRS([PuReMD-GPU])
-fi
-AM_CONDITIONAL([BUILD_GPU], [test "x${pack_gpu_enabled}" = "xyes"])
if test "x${pack_mpi_not_gpu_enabled}" = "xyes" || test "x${pack_mpi_gpu_enabled}" = "xyes"; then
AC_CONFIG_SUBDIRS([PG-PuReMD])
if test "x${pack_mpi_not_gpu_enabled}" = "xyes" || test "x${pack_mpi_gpu_enabled}" != "xyes"; then
diff --git a/sPuReMD/Makefile.am b/sPuReMD/Makefile.am
index 7c986471c7759f44ea3cd9aea126c246929f1647..12cffee9894677d37875b1d6dc9576d723447f21 100644
--- a/sPuReMD/Makefile.am
+++ b/sPuReMD/Makefile.am
@@ -4,7 +4,7 @@ bin_PROGRAMS = bin/spuremd
bin_spuremd_SOURCES = src/ffield.c src/grid.c src/list.c src/lookup.c src/print_utils.c \
src/reset_utils.c src/restart.c src/random.c src/tool_box.c src/traj.c \
src/vector.c src/allocate.c src/analyze.c src/box.c src/system_props.c src/control.c \
- src/geo_tools.c src/neighbors.c src/lin_alg.c src/QEq.c src/bond_orders.c \
+ src/geo_tools.c src/neighbors.c src/lin_alg.c src/qeq.c src/bond_orders.c \
src/single_body_interactions.c src/two_body_interactions.c \
src/three_body_interactions.c src/four_body_interactions.c src/forces.c \
src/integrate.c src/init_md.c src/testmd.c
@@ -12,7 +12,7 @@ bin_spuremd_SOURCES = src/ffield.c src/grid.c src/list.c src/lookup.c src/print_
include_HEADERS = src/mytypes.h src/ffield.h src/grid.h src/list.h src/lookup.h src/print_utils.h \
src/reset_utils.h src/restart.h src/random.h src/tool_box.h src/traj.h \
src/vector.h src/allocate.h src/analyze.h src/box.h src/system_props.h src/control.h \
- src/geo_tools.h src/neighbors.h src/lin_alg.h src/QEq.h src/bond_orders.h \
+ src/geo_tools.h src/neighbors.h src/lin_alg.h src/qeq.h src/bond_orders.h \
src/single_body_interactions.h src/two_body_interactions.h \
src/three_body_interactions.h src/four_body_interactions.h src/forces.h \
src/integrate.h src/init_md.h
diff --git a/sPuReMD/src/forces.c b/sPuReMD/src/forces.c
index 9108a8dd4026e001716f707302e40d671d48b838..5a0585685b7af523c9b129d6ec57cd3589675edc 100644
--- a/sPuReMD/src/forces.c
+++ b/sPuReMD/src/forces.c
@@ -20,6 +20,7 @@
----------------------------------------------------------------------*/
#include "forces.h"
+
#include "box.h"
#include "bond_orders.h"
#include "single_body_interactions.h"
@@ -29,7 +30,7 @@
#include "list.h"
#include "print_utils.h"
#include "system_props.h"
-#include "QEq.h"
+#include "qeq.h"
#include "vector.h"
diff --git a/sPuReMD/src/integrate.c b/sPuReMD/src/integrate.c
index 142863cad18291a04f4f56fc364c73cf8fdabd3b..f9768c49bc89915346645f6134e9e346c845c2d6 100644
--- a/sPuReMD/src/integrate.c
+++ b/sPuReMD/src/integrate.c
@@ -20,13 +20,14 @@
----------------------------------------------------------------------*/
#include "integrate.h"
+
#include "allocate.h"
#include "box.h"
#include "forces.h"
#include "grid.h"
#include "neighbors.h"
#include "print_utils.h"
-#include "QEq.h"
+#include "qeq.h"
#include "reset_utils.h"
#include "restart.h"
#include "system_props.h"
@@ -34,7 +35,6 @@
#include "list.h"
-
void Velocity_Verlet_NVE(reax_system* system, control_params* control,
simulation_data *data, static_storage *workspace,
list **lists, output_controls *out_control )
diff --git a/sPuReMD/src/lin_alg.h b/sPuReMD/src/lin_alg.h
index fe2d644cae630be6414944989b52adae8a6e1d61..e5a468ffb1669e53e638972b82daad9f7fcc07d0 100644
--- a/sPuReMD/src/lin_alg.h
+++ b/sPuReMD/src/lin_alg.h
@@ -19,8 +19,8 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#ifndef __GMRES_H_
-#define __GMRES_H_
+#ifndef __LIN_ALG_H_
+#define __LIN_ALG_H_
#include "mytypes.h"
diff --git a/sPuReMD/src/mytypes.h b/sPuReMD/src/mytypes.h
index 69441bb72b7499b663a506dc8c95d40e56473a78..9d741f256799b99dc012f44a30070d3893546782 100644
--- a/sPuReMD/src/mytypes.h
+++ b/sPuReMD/src/mytypes.h
@@ -37,7 +37,7 @@
#include "zlib.h"
#ifdef _OPENMP
-#include <omp.h>
+ #include <omp.h>
#endif
//#define DEBUG_FOCUS
@@ -51,6 +51,7 @@
#define TRUE 1
#define FALSE 0
+#define LOG log
#define EXP exp
#define SQRT sqrt
#define POW pow
@@ -71,10 +72,10 @@
/* NaN IEEE 754 representation for C99 in math.h
* Note: function choice must match REAL typedef below */
#ifdef NAN
-#define IS_NAN_REAL(a) (isnan(a))
+ #define IS_NAN_REAL(a) (isnan(a))
#else
-#warn "No support for NaN"
-#define NAN_REAL(a) (0)
+ #warn "No support for NaN"
+ #define NAN_REAL(a) (0)
#endif
#define PI 3.14159265
@@ -948,20 +949,21 @@ typedef struct
typedef void (*interaction_function)(reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls*);
+ simulation_data*, static_storage*, list**, output_controls*);
+
interaction_function Interaction_Functions[NO_OF_INTERACTIONS];
typedef void (*evolve_function)(reax_system*, control_params*,
- simulation_data*, static_storage*,
- list**, output_controls*);
+ simulation_data*, static_storage*,
+ list**, output_controls*);
typedef real (*lookup_function)(real);
+
lookup_table Exp, Sqrt, Cube_Root, Four_Third_Root, Cos, Sin, ACos;
LR_lookup_table **LR;
-
typedef void (*get_far_neighbors_function)(rvec, rvec, simulation_box*,
- control_params*, far_neighbor_data*,
- int*);
+ control_params*, far_neighbor_data*, int*);
+
+
#endif
diff --git a/sPuReMD/src/QEq.c b/sPuReMD/src/qeq.c
similarity index 99%
rename from sPuReMD/src/QEq.c
rename to sPuReMD/src/qeq.c
index 026a3ae1a84e4912958737161818aea77080da44..be99a1fd42194fc18b467a913f7a74485c125bec 100644
--- a/sPuReMD/src/QEq.c
+++ b/sPuReMD/src/qeq.c
@@ -19,7 +19,7 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#include "QEq.h"
+#include "qeq.h"
#include "allocate.h"
#include "list.h"
diff --git a/sPuReMD/src/QEq.h b/sPuReMD/src/qeq.h
similarity index 100%
rename from sPuReMD/src/QEq.h
rename to sPuReMD/src/qeq.h
diff --git a/sPuReMD/src/random.c b/sPuReMD/src/random.c
index f3a5096c65485111fb5cba7321976518a2e42972..9b09e7526b7a8418470cbf8c1b45bd1940dcbfa9 100644
--- a/sPuReMD/src/random.c
+++ b/sPuReMD/src/random.c
@@ -19,7 +19,8 @@
<http://www.gnu.org/licenses/>.
----------------------------------------------------------------------*/
-#include <random.h>
+#include "random.h"
+
/* System random number generator used linear congruance method with
large periodicity for generation of pseudo random number. function
@@ -53,5 +54,5 @@ double GRandom(double mean, double sigma)
rsq = v1 * v1 + v2 * v2;
}
- return mean + v1 * sigma * sqrt(-2.0 * log(rsq) / rsq);
+ return mean + v1 * sigma * SQRT(-2.0 * LOG(rsq) / rsq);
}