flywithu/ntel-gpu-tools
1
0
Fork 0
mirror of https://github.com/tiagovignatti/intel-gpu-tools.git synced 2025-07-04 13:16:19 +00:00
Code Issues Releases Wiki Activity

benchmarks/gem_exec_nop: Include a measurement across all rings

Browse Source 
For sync, it really is just the average latency across all rings, but
for continuous we can expect to see the effect of concurrent dispatch
across rings. Hopefully.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Chris Wilson 2016-03-23 12:37:38 +00:00
parent 6867b87b73
commit 4e2a785e24
1 changed files with 30 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
benchmarks/gem_exec_nop.c
View File

@ -47,6 +47,11 @@
#define LOCAL_I915_EXEC_NO_RELOC (1<<11) #define LOCAL_I915_EXEC_NO_RELOC (1<<11)
#define LOCAL_I915_EXEC_HANDLE_LUT (1<<12) #define LOCAL_I915_EXEC_HANDLE_LUT (1<<12)
#define LOCAL_I915_EXEC_BSD_SHIFT (13)
#define LOCAL_I915_EXEC_BSD_MASK (3 << LOCAL_I915_EXEC_BSD_SHIFT)
#define ENGINE_FLAGS (I915_EXEC_RING_MASK | LOCAL_I915_EXEC_BSD_MASK)
#define SYNC 0x1 #define SYNC 0x1
static double elapsed(const struct timespec *start, static double elapsed(const struct timespec *start,
@ -67,6 +72,8 @@ static int loop(unsigned ring, int reps, unsigned flags)
{ {
struct drm_i915_gem_execbuffer2 execbuf; struct drm_i915_gem_execbuffer2 execbuf;
struct drm_i915_gem_exec_object2 gem_exec; struct drm_i915_gem_exec_object2 gem_exec;
unsigned engines[16];
unsigned nengine;
int fd; int fd;
fd = drm_open_driver(DRIVER_INTEL); fd = drm_open_driver(DRIVER_INTEL);
@ -77,15 +84,26 @@ static int loop(unsigned ring, int reps, unsigned flags)
memset(&execbuf, 0, sizeof(execbuf)); memset(&execbuf, 0, sizeof(execbuf));
execbuf.buffers_ptr = (uintptr_t)&gem_exec; execbuf.buffers_ptr = (uintptr_t)&gem_exec;
execbuf.buffer_count = 1; execbuf.buffer_count = 1;
execbuf.flags = ring;
execbuf.flags |= LOCAL_I915_EXEC_HANDLE_LUT; execbuf.flags |= LOCAL_I915_EXEC_HANDLE_LUT;
execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC; execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC;
if (__gem_execbuf(fd, &execbuf)) { if (__gem_execbuf(fd, &execbuf)) {
execbuf.flags = ring; execbuf.flags = 0;
if (__gem_execbuf(fd, &execbuf)) if (__gem_execbuf(fd, &execbuf))
return 77; return 77;
} }
nengine = 0;
if (ring == -1) {
for (ring = 1; ring < 16; ring++) {
execbuf.flags &= ~ENGINE_FLAGS;
execbuf.flags |= ring;
if (__gem_execbuf(fd, &execbuf) == 0)
engines[nengine++] = ring;
}
} else
engines[nengine++] = ring;
while (reps--) { while (reps--) {
struct timespec start, end; struct timespec start, end;
unsigned count = 0; unsigned count = 0;
@ -95,10 +113,14 @@ static int loop(unsigned ring, int reps, unsigned flags)
clock_gettime(CLOCK_MONOTONIC, &start); clock_gettime(CLOCK_MONOTONIC, &start);
do { do {
do_ioctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &execbuf); for (int inner = 0; inner < 1024; inner++) {
count++; execbuf.flags &= ~ENGINE_FLAGS;
execbuf.flags |= engines[count++ % nengine];
gem_execbuf(fd, &execbuf);
if (flags & SYNC) if (flags & SYNC)
gem_sync(fd, gem_exec.handle); gem_sync(fd, gem_exec.handle);
}
clock_gettime(CLOCK_MONOTONIC, &end); clock_gettime(CLOCK_MONOTONIC, &end);
} while (elapsed(&start, &end) < 2.); } while (elapsed(&start, &end) < 2.);
@ -128,6 +150,8 @@ int main(int argc, char **argv)
ring = I915_EXEC_BLT; ring = I915_EXEC_BLT;
else if (strcmp(optarg, "vecs") == 0) else if (strcmp(optarg, "vecs") == 0)
ring = I915_EXEC_VEBOX; ring = I915_EXEC_VEBOX;
else if (strcmp(optarg, "all") == 0)
ring = -1;
else else
ring = atoi(optarg); ring = atoi(optarg);
break; break;