Skip to content

Sync with the recent update in gpu-opt-guide #2766

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
hansangbae wants to merge 1 commit into
base: development
Choose a base branch
from
Open

Sync with the recent update in gpu-opt-guide #2766

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 24 additions & 21 deletions Publications/GPU-Opt-Guide/OpenMP/01_collapse/test_collapse_2levels.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,32 +44,35 @@ int main(void) {
/* map data to device */
#pragma omp target enter data map(to: u[0:SIZE], dx[0:P * P])

start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(2) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
/* run 5 times and report the timing from the last iteration */
for (int r = 0; r < 5; r++) {
start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(2) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}
}
}
}

end = omp_get_wtime();
end = omp_get_wtime();
}

#pragma omp target exit data map(from: w[0:SIZE])

Expand Down
762 changes: 410 additions & 352 deletions Publications/GPU-Opt-Guide/OpenMP/01_collapse/test_collapse_2levels.debug
View file
Open in desktop

Large diffs are not rendered by default.

45 changes: 24 additions & 21 deletions Publications/GPU-Opt-Guide/OpenMP/01_collapse/test_collapse_3levels.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,32 +44,35 @@ int main(void) {
/* map data to device */
#pragma omp target enter data map(to: u[0:SIZE], dx[0:P * P])

start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(3) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
/* run 5 times and report the timing from the last iteration */
for (int r = 0; r < 5; r++) {
start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(3) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}
}
}
}

end = omp_get_wtime();
end = omp_get_wtime();
}

#pragma omp target exit data map(from: w[0:SIZE])

Expand Down
778 changes: 419 additions & 359 deletions Publications/GPU-Opt-Guide/OpenMP/01_collapse/test_collapse_3levels.debug
View file
Open in desktop

Large diffs are not rendered by default.

45 changes: 24 additions & 21 deletions Publications/GPU-Opt-Guide/OpenMP/01_collapse/test_collapse_4levels.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,32 +44,35 @@ int main(void) {
/* map data to device */
#pragma omp target enter data map(to: u[0:SIZE], dx[0:P * P])

start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(4) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
/* run 5 times and report the timing from the last iteration */
for (int r = 0; r < 5; r++) {
start = omp_get_wtime();

/* offload the kernel with collapse clause */
#pragma omp target teams distribute parallel for collapse(4) \
private(b, i, j, k, l)
for (b = 0; b < BLOCKS; b++) {
for (i = 0; i < P; i++) {
for (j = 0; j < P; j++) {
for (k = 0; k < P; k++) {
double ur = 0.;
double us = 0.;
double ut = 0.;

for (l = 0; l < P; l++) {
ur += dx[IDX2(i, l)] * u[IDX4(b, l, j, k)];
us += dx[IDX2(k, l)] * u[IDX4(b, i, l, k)];
ut += dx[IDX2(j, l)] * u[IDX4(b, i, j, l)];
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}

w[IDX4(b, i, j, k)] = ur * us * ut;
}
}
}
}

end = omp_get_wtime();
end = omp_get_wtime();
}

#pragma omp target exit data map(from: w[0:SIZE])

Expand Down
Loading