flywithu/ntel-gpu-tools
1
0
Fork 0
mirror of https://github.com/tiagovignatti/intel-gpu-tools.git synced 2025-06-11 18:06:13 +00:00
Code Issues Releases Wiki Activity

benchmarks/gem_blt: Report peak throughput

Browse Source 
Report the highest throughput measured from a large set of runs to
improve sensitivity.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Chris Wilson 2015-11-10 13:40:40 +00:00
parent 3b0e336081
commit 5cabb8c543
1 changed files with 29 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
benchmarks/gem_blt.c
View File

@ -34,6 +34,7 @@
#include <fcntl.h> #include <fcntl.h>
#include <inttypes.h> #include <inttypes.h>
#include <errno.h> #include <errno.h>
#include <math.h>
#include <sys/stat.h> #include <sys/stat.h>
#include <sys/ioctl.h> #include <sys/ioctl.h>
#include <sys/time.h> #include <sys/time.h>
@ -146,11 +147,9 @@ static int gem_linear_blt(int fd,
return (b+2 - batch) * sizeof(uint32_t); return (b+2 - batch) * sizeof(uint32_t);
} }
static double elapsed(const struct timespec *start, static double elapsed(const struct timespec *start, const struct timespec *end)
const struct timespec *end,
int loop)
{ {
return ((end->tv_sec - start->tv_sec) + 1e-9*(end->tv_nsec - start->tv_nsec))/loop; return (end->tv_sec - start->tv_sec) + 1e-9*(end->tv_nsec - start->tv_nsec);
} }
static int __gem_execbuf(int fd, struct drm_i915_gem_execbuffer2 *execbuf) static int __gem_execbuf(int fd, struct drm_i915_gem_execbuffer2 *execbuf)
@ -161,7 +160,7 @@ static int __gem_execbuf(int fd, struct drm_i915_gem_execbuffer2 *execbuf)
return err; return err;
} }
static int run(int object, int batch, int count, int reps) static int run(int object, int batch, int count, int set, int reps)
{ {
struct drm_i915_gem_execbuffer2 execbuf; struct drm_i915_gem_execbuffer2 execbuf;
struct drm_i915_gem_exec_object2 exec[3]; struct drm_i915_gem_exec_object2 exec[3];
@ -219,14 +218,13 @@ static int run(int object, int batch, int count, int reps)
if (batch > 1) { if (batch > 1) {
int nreloc = exec[2].relocation_count; int nreloc = exec[2].relocation_count;
int i;
if (execbuf.flags & LOCAL_I915_EXEC_HANDLE_LUT) { if (execbuf.flags & LOCAL_I915_EXEC_HANDLE_LUT) {
src = 0; src = 0;
dst = 1; dst = 1;
} }
for (i = 1; i < batch; i++) { for (int i = 1; i < batch; i++) {
len = gem_linear_blt(fd, buf, len - 8, len = gem_linear_blt(fd, buf, len - 8,
src, dst, object, src, dst, object,
reloc + exec[2].relocation_count); reloc + exec[2].relocation_count);
@ -239,17 +237,24 @@ static int run(int object, int batch, int count, int reps)
} }
while (reps--) { while (reps--) {
struct timespec start, end; double min = HUGE_VAL;
sleep(1); for (int s = 0; s < set; s++) {
struct timespec start, end;
double t;
clock_gettime(CLOCK_MONOTONIC, &start); clock_gettime(CLOCK_MONOTONIC, &start);
for (int loop = 0; loop < count; loop++) for (int loop = 0; loop < count; loop++)
gem_execbuf(fd, &execbuf); gem_execbuf(fd, &execbuf);
gem_sync(fd, handle); gem_sync(fd, handle);
clock_gettime(CLOCK_MONOTONIC, &end); clock_gettime(CLOCK_MONOTONIC, &end);
printf("%7.3f\n", object/elapsed(&start, &end, batch*count)/(1024*1024)); t = elapsed(&start, &end);
if (t < min)
min = t;
}
printf("%7.3f\n", object/(1024*1024.)*batch*count/min);
} }
close(fd); close(fd);
@ -261,10 +266,11 @@ int main(int argc, char **argv)
int size = 1024*1024; int size = 1024*1024;
int count = 1; int count = 1;
int reps = 13; int reps = 13;
int set = 30;
int batch = 1; int batch = 1;
int c; int c;
while ((c = getopt (argc, argv, "c:r:s:b:")) != -1) { while ((c = getopt (argc, argv, "c:r:s:b:S:")) != -1) {
switch (c) { switch (c) {
case 'c': case 'c':
count = atoi(optarg); count = atoi(optarg);
@ -278,6 +284,12 @@ int main(int argc, char **argv)
size = 4096; size = 4096;
break; break;
case 'S':
set = atoi(optarg);
if (set < 1)
set = 1;
break;
case 'r': case 'r':
reps = atoi(optarg); reps = atoi(optarg);
if (reps < 1) if (reps < 1)
@ -295,5 +307,5 @@ int main(int argc, char **argv)
} }
} }
return run(size, batch, count, reps); return run(size, batch, count, set, reps);
} }