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

schedule options for force kernels and to print in main fn

Browse Source 
Signed-off-by: TejeshPala <tejesh.pala@fau.de>
This commit is contained in:
TejeshPala
2024-01-11 17:09:18 +01:00
parent 1837403326
commit 4cfa664533
4 changed files with 41 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
lammps/force_lj.c
View File

@@ -41,7 +41,7 @@ double computeForceLJFullNeigh_plain_c(Parameter *param, Atom *atom, Neighbor *n
{
LIKWID_MARKER_START("force");
#pragma omp for
#pragma omp for schedule(runtime)
for(int i = 0; i < Nlocal; i++) {
neighs = &neighbor->neighbors[i * neighbor->maxneighs];
int numneighs = neighbor->numneigh[i];
@@ -131,7 +131,7 @@ double computeForceLJHalfNeigh(Parameter *param, Atom *atom, Neighbor *neighbor,
{
LIKWID_MARKER_START("forceLJ-halfneigh");
#pragma omp for
#pragma omp for schedule(runtime)
for(int i = 0; i < Nlocal; i++) {
neighs = &neighbor->neighbors[i * neighbor->maxneighs];
int numneighs = neighbor->numneigh[i];
@@ -227,7 +227,7 @@ double computeForceLJFullNeigh_simd(Parameter *param, Atom *atom, Neighbor *neig
{
LIKWID_MARKER_START("force");
#pragma omp for
#pragma omp for schedule(runtime)
for(int i = 0; i < Nlocal; i++) {
neighs = &neighbor->neighbors[i * neighbor->maxneighs];
int numneighs = neighbor->numneigh[i];

19
lammps/main.c
View File

@@ -295,13 +295,28 @@ int main(int argc, char** argv) {
printf(HLINE);
int nthreads = 0;
int chunkSize = 0;
omp_sched_t schedKind;
char schedType[10];
#pragma omp parallel
#pragma omp master
{
nthreads = omp_get_num_threads();
omp_get_schedule(&schedKind, &chunkSize);
switch (schedKind)
{
case omp_sched_static: strcpy(schedType, "static"); break;
case omp_sched_dynamic: strcpy(schedType, "dynamic"); break;
case omp_sched_guided: strcpy(schedType, "guided"); break;
case omp_sched_auto: strcpy(schedType, "auto"); break;
}
nthreads = omp_get_num_threads();
}
printf("Num threads: %d\n", nthreads);
printf("Schedule: (%s,%d)\n", schedType, chunkSize);
printf("Performance: %.2f million atom updates per second\n",
1e-6 * (double) atom.Natoms * param.ntimes / timer[TOTAL]);
#ifdef COMPUTE_STATS