Skip to content

Revive micro-benchmarks for vector functions #571

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
blapie merged 3 commits into from
Aug 29, 2024
Merged

Revive micro-benchmarks for vector functions #571

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
01ba4c9
Revive micro-benchmarks for vector functions
joanaxcruz Aug 27, 2024
f18d925
Remove warmup
joanaxcruz Aug 29, 2024
3475d69
Update bench.h
joanaxcruz Aug 29, 2024
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
77 changes: 44 additions & 33 deletions src/libm-benchmarks/bench.h
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,57 +1,68 @@
#define NITER1 100000
#define NITER2 10000
#define NITER (NITER1 * NITER2)
// ARRAY SIZE: N * veclen
#define N 10000
// NUM ITERATIONS
#define ITER 1000
// TOTAL NUMBER OF ITERATIONS
#define NITER (N * ITER)

#define callFuncSLEEF1_1(funcName, name, xmin, xmax, ulp, arg, type) ({ \
printf("%s\n", #funcName); \
uint64_t t = Sleef_currentTimeMicros(); \
for(int j=0;j<NITER2;j++) { \
type *p = (type *)(arg); \
for(int i=0;i<NITER1;i++) funcName(*p++); \
} \
fprintf(fp, name ", %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, ulp, (double)(Sleef_currentTimeMicros() - t) / NITER); \
#define callFuncSLEEF1_1(funcName, name, xmin, xmax, ulp, arg, type_in, type_out) ({ \
printf("%s\n", #funcName); \
uint64_t t0 = Sleef_currentTimeMicros(); \
for(int j=0;j<ITER;j++) { \
type_in *p = (type_in *)(arg); \
for(int i=0;i<N;i++){ \
__attribute__((unused)) type_out out = funcName(*p++); \
} \
} \
uint64_t t1 = Sleef_currentTimeMicros(); \
uint64_t dt = t1-t0; \
fprintf(fp, name ", %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, ulp, (double) dt / NITER); \
})

#define callFuncSLEEF1_2(funcName, name, xmin, xmax, ymin, ymax, ulp, arg1, arg2, type) ({ \
printf("%s\n", #funcName); \
uint64_t t = Sleef_currentTimeMicros(); \
for(int j=0;j<NITER2;j++) { \
type *p1 = (type *)(arg1), *p2 = (type *)(arg2); \
for(int i=0;i<NITER1;i++) funcName(*p1++, *p2++); \
} \
fprintf(fp, name ", %.3g, %.3g, %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, (double)ymin, (double)ymax, ulp, (double)(Sleef_currentTimeMicros() - t) / NITER); \
#define callFuncSLEEF1_2(funcName, name, xmin, xmax, ymin, ymax, ulp, arg1, arg2, type_in, type_out) ({ \
printf("%s\n", #funcName); \
uint64_t t0 = Sleef_currentTimeMicros(); \
for(int j=0;j<ITER;j++) { \
type_in *p1 = (type_in *)(arg1), *p2 = (type_in *)(arg2); \
for(int i=0;i<N;i++){ \
__attribute__((unused)) type_out out = funcName(*p1++, *p2++); \
} \
} \
uint64_t t1 = Sleef_currentTimeMicros(); \
uint64_t dt = t1-t0; \
fprintf(fp, name ", %.3g, %.3g, %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, (double)ymin, (double)ymax, ulp, (double) dt / NITER); \
})

#define callFuncSVML1_1(funcName, name, xmin, xmax, arg, type) ({ \
#define callFuncSVML1_1(funcName, name, xmin, xmax, arg, type_in) ({ \
printf("%s\n", #funcName); \
uint64_t t = Sleef_currentTimeMicros(); \
for(int j=0;j<NITER2;j++) { \
type *p = (type *)(arg); \
for(int i=0;i<NITER1;i++) funcName(*p++); \
for(int j=0;j<ITER;j++) { \
type_in *p = (type_in *)(arg); \
for(int i=0;i<N;i++) funcName(*p++); \
} \
fprintf(fp, name ", %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, (double)SVMLULP, (double)(Sleef_currentTimeMicros() - t) / NITER); \
})

#define callFuncSVML2_1(funcName, name, xmin, xmax, arg, type) ({ \
#define callFuncSVML2_1(funcName, name, xmin, xmax, arg, type_in) ({ \
printf("%s\n", #funcName); \
uint64_t t = Sleef_currentTimeMicros(); \
for(int j=0;j<NITER2;j++) { \
type *p = (type *)(arg), c; \
for(int i=0;i<NITER1;i++) funcName(&c, *p++); \
for(int j=0;j<ITER;j++) { \
type_in *p = (type_in *)(arg), c; \
for(int i=0;i<N;i++) funcName(&c, *p++); \
} \
fprintf(fp, name ", %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, (double)SVMLULP, (double)(Sleef_currentTimeMicros() - t) / NITER); \
})

#define callFuncSVML1_2(funcName, name, xmin, xmax, ymin, ymax, arg1, arg2, type) ({ \
#define callFuncSVML1_2(funcName, name, xmin, xmax, ymin, ymax, arg1, arg2, type_in) ({ \
printf("%s\n", #funcName); \
uint64_t t = Sleef_currentTimeMicros(); \
for(int j=0;j<NITER2;j++) { \
type *p1 = (type *)(arg1), *p2 = (type *)(arg2); \
for(int i=0;i<NITER1;i++) funcName(*p1++, *p2++); \
for(int j=0;j<ITER;j++) { \
type_in *p1 = (type_in *)(arg1), *p2 = (type_in *)(arg2); \
for(int i=0;i<N;i++) funcName(*p1++, *p2++); \
} \
fprintf(fp, name ", %.3g, %.3g, %.3g, %.3g, %gulps, %g\n", \
(double)xmin, (double)xmax, (double)ymin, (double)ymax, (double)SVMLULP, (double)(Sleef_currentTimeMicros() - t) / NITER); \
Expand Down
8 changes: 4 additions & 4 deletions src/libm-benchmarks/benchsleef.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,14 +39,14 @@ int cpuSupportsAVX512F() {
#endif

void fillDP(double *buf, double min, double max) {
for(int i=0;i<NITER1*veclen;i++) {
for(int i=0;i<N*veclen;i++) {
double r = ((double)random() + RAND_MAX * (double)random()) / (RAND_MAX * (double)RAND_MAX);
buf[i] = r * (max - min) + min;
}
}

void fillSP(float *buf, double min, double max) {
for(int i=0;i<NITER1*veclen;i++) {
for(int i=0;i<N*veclen;i++) {
double r = ((double)random() + RAND_MAX * (double)random()) / (RAND_MAX * (double)RAND_MAX);
buf[i] = r * (max - min) + min;
}
Expand Down Expand Up @@ -86,8 +86,8 @@ int main(int argc, char **argv) {
#error Unsupported architecture
#endif

posix_memalign((void **)&abufdp, veclen*sizeof(double), NITER1*veclen*sizeof(double));
posix_memalign((void **)&bbufdp, veclen*sizeof(double), NITER1*veclen*sizeof(double));
posix_memalign((void **)&abufdp, veclen*sizeof(double), N*veclen*sizeof(double));
posix_memalign((void **)&bbufdp, veclen*sizeof(double), N*veclen*sizeof(double));

abufsp = (float *)abufdp;
bbufsp = (float *)bbufdp;
Expand Down
Loading
⚡