moebiusband/MD-Bench
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Small fixes into GROMACS GPU code

Browse Source 
Signed-off-by: Rafael Ravedutti <rafaelravedutti@gmail.com>
This commit is contained in:
Rafael Ravedutti 2022-11-14 18:21:14 +01:00
parent 93188d1383
commit 6eedf1776e
3 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
gromacs/cuda/force_lj.cu
View File

@ -56,8 +56,8 @@ void initDevice(Atom *atom, Neighbor *neighbor) {
cuda_PBCz = (int *) allocateGPU(atom->Nclusters_max * sizeof(int));
cuda_numneigh = (int *) allocateGPU(atom->Nclusters_max * sizeof(int));
cuda_neighbors = (int *) allocateGPU(atom->Nclusters_max * neighbor->maxneighs * sizeof(int));
natoms = (int *) malloc(atom->Nclusters_max);
ngatoms = (int *) malloc(atom->Nclusters_max);
natoms = (int *) malloc(atom->Nclusters_max * sizeof(int));
ngatoms = (int *) malloc(atom->Nclusters_max * sizeof(int));
isReneighboured = 1;
}

2
gromacs/main.c
View File

@ -65,9 +65,9 @@ double setup(Parameter *param, Eam *eam, Atom *atom, Neighbor *neighbor, Stats *
buildClusters(atom);
defineJClusters(atom);
setupPbc(atom, param);
initDevice(atom, neighbor);
binClusters(atom);
buildNeighbor(atom, neighbor);
initDevice(atom, neighbor);
E = getTimeStamp();
return E-S;
}

2
include_NVCC.mk
View File

@ -10,7 +10,7 @@ ANSI_CFLAGS += -Wextra
# A100 + Native
CFLAGS = -O3 -arch=sm_80 -march=native -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
# A40 + Native
#CFLAGS = -O3 -arch=sm_80 -march=native -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
#CFLAGS = -O3 -arch=sm_86 -march=native -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
# Cascade Lake
#CFLAGS = -O3 -march=cascadelake -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
# For GROMACS kernels, we need at least sm_61 due to atomicAdd with doubles