Skip to content

out-of-place native FFT: move buffer alloc inside timing #5

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

Merged
oleksandr-pavlyk merged 3 commits into from
Jul 24, 2019
Merged

out-of-place native FFT: move buffer alloc inside timing #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
639dc5b
out-of-place native FFT: move buffer alloc inside timing
bibikar Jul 24, 2019
20294a5
Fix indentation in Makefile
bibikar Jul 24, 2019
8003dde
Don't remove printing of small buffers
bibikar Jul 24, 2019
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
14 changes: 7 additions & 7 deletions native/Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,14 @@
BENCHMARKS = fft_cdp-in-c fft_cdp-in-nc fft_cdp-out-c \
fft_cdp-out-nc fft_rdp-out-c fft_rdp-out-nc \
rfft_rdp-out-c rfft_rdp-out-nc \
fft_rdp2-out-c fft2_cdp-in-c fft2_cdp-in-nc \
fft3_cdp-in-c fft3_cdp-in-nc fft2_cdp-out-c \
fft2_cdp-out-nc fft3_cdp-out-c fft3_cdp-out-nc
fft_cdp-out-nc fft_rdp-out-c fft_rdp-out-nc \
rfft_rdp-out-c rfft_rdp-out-nc \
fft_rdp2-out-c fft2_cdp-in-c fft2_cdp-in-nc \
fft3_cdp-in-c fft3_cdp-in-nc fft2_cdp-out-c \
fft2_cdp-out-nc fft3_cdp-out-c fft3_cdp-out-nc

CC = icc
CFLAGS = -m64 -fPIC -fp-model strict -O3 -g -fomit-frame-pointer \
-DNDEBUG -qopenmp -xSSE4.2 -axCORE-AVX2,COMMON-AVX512 \
-lmkl_rt
-DNDEBUG -qopenmp -xSSE4.2 -axCORE-AVX2,COMMON-AVX512 \
-lmkl_rt

CINCLUDE := $(addprefix -I,$(CINCLUDE))

Expand Down
16 changes: 9 additions & 7 deletions native/fft2_cdp-out-c.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,15 +59,15 @@ int main() {
mkl_free(re_vec);
mkl_free(im_vec);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

warm_up_threads();

for(si=0; si < samps; time_tot=0, si++) {

t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -115,16 +115,18 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
time_tot += t1 - t0;

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

mkl_free(buf);

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

Expand Down
16 changes: 9 additions & 7 deletions native/fft2_cdp-out-nc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,16 +63,16 @@ int main() {
strides[1] = N2;
strides[2] = 1;

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

warm_up_threads();

for(si=0; si < samps; time_tot=0, si++) {

for(it = -1; it <reps; it++) {
t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -104,6 +104,12 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0 && it == -1) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();

if (it >= 0) time_tot += t1 - t0;
Expand All @@ -112,10 +118,6 @@ int main() {
printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

mkl_free(buf);

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

Expand Down
10 changes: 6 additions & 4 deletions native/fft3_cdp-out-c.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,16 +115,18 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
time_tot += t1 - t0;

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

mkl_free(buf);

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

Expand Down
16 changes: 9 additions & 7 deletions native/fft3_cdp-out-nc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,9 +59,6 @@ int main() {
mkl_free(re_vec);
mkl_free(im_vec);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

strides[0] = 0;
strides[1] = N3 * N2;
strides[2] = N3;
Expand All @@ -75,6 +72,9 @@ int main() {

t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -105,17 +105,19 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0 && it == -1) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
if(it >= 0) time_tot += t1 - t0;
}

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

mkl_free(buf);

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

Expand Down
15 changes: 9 additions & 6 deletions native/fft_cdp-out-c.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,15 +51,15 @@ int main() {
mkl_free(re_vec);
mkl_free(im_vec);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

warm_up_threads();

for(si=0; si < samps; time_tot=0, si++) {

t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -94,18 +94,21 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
time_tot += t1 - t0;

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

mkl_free(buf);
mkl_free(x);

return 0;
Expand Down
15 changes: 9 additions & 6 deletions native/fft_cdp-out-nc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,9 +51,6 @@ int main() {
mkl_free(re_vec);
mkl_free(im_vec);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

warm_up_threads();

for(si=0; si < samps; time_tot=0, si++) {
Expand All @@ -62,6 +59,9 @@ int main() {

t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand All @@ -82,19 +82,22 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0 && it == -1) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
if (it >= 0) time_tot += t1 - t0;
}

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

mkl_free(buf);
mkl_free(x);

return 0;
Expand Down
15 changes: 9 additions & 6 deletions native/fft_rdp-out-c.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,14 +35,14 @@ int main() {
err = vdRngGaussian(VSL_RNG_METHOD_GAUSSIAN_ICDF, stream, N, x, d_zero, d_one);
assert(err == VSL_STATUS_OK);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

warm_up_threads();

for(si = 0; si < samps; time_tot=0, si++) {
t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -89,18 +89,21 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();
time_tot += t1 - t0;

printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

mkl_free(buf);
mkl_free(x);

return 0;
Expand Down
21 changes: 12 additions & 9 deletions native/fft_rdp-out-nc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,17 +26,14 @@ int main() {
#define DESCRIPTION_STR "Real double, not in-place, not cached"
#include "read_n_echo_env.inc"

err = vslNewStream(&stream, VSL_BRNG_MT19937, SEED);
assert(err == VSL_STATUS_OK);

x = (double *) mkl_malloc(N * sizeof(double), 64);
assert(x);

err = vdRngGaussian(VSL_RNG_METHOD_GAUSSIAN_ICDF, stream, N, x, d_zero, d_one);
err = vslNewStream(&stream, VSL_BRNG_MT19937, SEED);
assert(err == VSL_STATUS_OK);

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);
err = vdRngGaussian(VSL_RNG_METHOD_GAUSSIAN_ICDF, stream, N, x, d_zero, d_one);
assert(err == VSL_STATUS_OK);

warm_up_threads();

Expand All @@ -46,6 +43,9 @@ int main() {

t0 = moment_now();

buf = (MKL_Complex16 *) mkl_malloc(N * sizeof(MKL_Complex16), 64);
assert(buf);

status = DftiCreateDescriptor(
&hand,
DFTI_DOUBLE,
Expand Down Expand Up @@ -77,6 +77,12 @@ int main() {
status = DftiFreeDescriptor(&hand);
assert(status == 0);

if (si == 0 && it == -1) {
#include "print_buf.inc"
}

mkl_free(buf);

t1 = moment_now();

if (it >= 0) time_tot += t1 - t0;
Expand All @@ -85,12 +91,9 @@ int main() {
printf("%.5g\n", seconds_from_moment(time_tot));
}

#include "print_buf.inc"

err = vslDeleteStream(&stream);
assert(err == VSL_STATUS_OK);

mkl_free(buf);
mkl_free(x);

return 0;
Expand Down
Loading