flywithu/ntel-gpu-tools
1
0
Fork 0
mirror of https://github.com/tiagovignatti/intel-gpu-tools.git synced 2025-06-10 17:36:11 +00:00
Code Issues Releases Wiki Activity

benchmarks/gem_blt: Fix compilation after rebase and add batch-size

Browse Source 
Add an option to do more than one copy per batch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Chris Wilson 2015-10-06 17:03:02 +01:00
parent 8253e7dc84
commit d878e18dfd
1 changed files with 63 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

91
benchmarks/gem_blt.c
View File

@ -52,14 +52,17 @@
#define BLT_SRC_TILED (1<<15) #define BLT_SRC_TILED (1<<15)
#define BLT_DST_TILED (1<<11) #define BLT_DST_TILED (1<<11)
static int has_64bit_reloc;
static int gem_linear_blt(int fd, static int gem_linear_blt(int fd,
uint32_t *batch, uint32_t *batch,
int offset,
uint32_t src, uint32_t src,
uint32_t dst, uint32_t dst,
uint32_t length, uint32_t length,
struct drm_i915_gem_relocation_entry *reloc) struct drm_i915_gem_relocation_entry *reloc)
{ {
uint32_t *b = batch; uint32_t *b = batch + offset/4;
int height = length / (16 * 1024); int height = length / (16 * 1024);
igt_assert_lte(height, 1 << 16); igt_assert_lte(height, 1 << 16);
@ -67,7 +70,7 @@ static int gem_linear_blt(int fd,
if (height) { if (height) {
int i = 0; int i = 0;
b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB; b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i-1]+=2; b[i-1]+=2;
b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024); b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024);
b[i++] = 0; b[i++] = 0;
@ -80,14 +83,14 @@ static int gem_linear_blt(int fd,
reloc->write_domain = I915_GEM_DOMAIN_RENDER; reloc->write_domain = I915_GEM_DOMAIN_RENDER;
reloc->presumed_offset = 0; reloc->presumed_offset = 0;
reloc++; reloc++;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i++] = 0; /* FIXME */ b[i++] = 0; /* FIXME */
b[i++] = 0; b[i++] = 0;
b[i++] = 16*1024; b[i++] = 16*1024;
b[i++] = 0; b[i++] = 0;
reloc->offset = (b-batch+7) * sizeof(uint32_t); reloc->offset = (b-batch+7) * sizeof(uint32_t);
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
reloc->offset += sizeof(uint32_t); reloc->offset += sizeof(uint32_t);
reloc->delta = 0; reloc->delta = 0;
reloc->target_handle = src; reloc->target_handle = src;
@ -95,7 +98,7 @@ static int gem_linear_blt(int fd,
reloc->write_domain = 0; reloc->write_domain = 0;
reloc->presumed_offset = 0; reloc->presumed_offset = 0;
reloc++; reloc++;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i++] = 0; /* FIXME */ b[i++] = 0; /* FIXME */
b += i; b += i;
@ -105,7 +108,7 @@ static int gem_linear_blt(int fd,
if (length) { if (length) {
int i = 0; int i = 0;
b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB; b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i-1]+=2; b[i-1]+=2;
b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024); b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024);
b[i++] = height << 16; b[i++] = height << 16;
@ -118,14 +121,14 @@ static int gem_linear_blt(int fd,
reloc->write_domain = I915_GEM_DOMAIN_RENDER; reloc->write_domain = I915_GEM_DOMAIN_RENDER;
reloc->presumed_offset = 0; reloc->presumed_offset = 0;
reloc++; reloc++;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i++] = 0; /* FIXME */ b[i++] = 0; /* FIXME */
b[i++] = height << 16; b[i++] = height << 16;
b[i++] = 16*1024; b[i++] = 16*1024;
b[i++] = 0; b[i++] = 0;
reloc->offset = (b-batch+7) * sizeof(uint32_t); reloc->offset = (b-batch+7) * sizeof(uint32_t);
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
reloc->offset += sizeof(uint32_t); reloc->offset += sizeof(uint32_t);
reloc->delta = 0; reloc->delta = 0;
reloc->target_handle = src; reloc->target_handle = src;
@ -133,7 +136,7 @@ static int gem_linear_blt(int fd,
reloc->write_domain = 0; reloc->write_domain = 0;
reloc->presumed_offset = 0; reloc->presumed_offset = 0;
reloc++; reloc++;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
b[i++] = 0; /* FIXME */ b[i++] = 0; /* FIXME */
b += i; b += i;
@ -160,38 +163,43 @@ static int __gem_execbuf(int fd, struct drm_i915_gem_execbuffer2 *execbuf)
return err; return err;
} }
static int run(int size, int count, int reps) static int run(int object, int batch, int count, 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];
struct drm_i915_gem_relocation_entry reloc[4]; struct drm_i915_gem_relocation_entry *reloc;
uint32_t buf[20]; uint32_t *buf, handle, src, dst;
uint32_t handle, src, dst; int fd, len, gen, size;
int fd, len;
int ring; int ring;
fd = drm_open_any(); size = ALIGN(batch * 64, 4096);
handle = gem_create(fd, 4096); reloc = malloc(sizeof(*reloc)*size/32*2);
src = gem_create(fd, size); fd = drm_open_driver(DRIVER_INTEL);
dst = gem_create(fd, size); handle = gem_create(fd, size);
buf = gem_mmap__cpu(fd, handle, 0, size, PROT_WRITE);
len = gem_linear_blt(fd, buf, 0, 1, size, reloc); gen = intel_gen(intel_get_drm_devid(fd));
gem_write(fd, handle, 0, buf, len); has_64bit_reloc = gen >= 8;
src = gem_create(fd, object);
dst = gem_create(fd, object);
len = gem_linear_blt(fd, buf, 0, 0, 1, object, reloc);
memset(exec, 0, sizeof(exec)); memset(exec, 0, sizeof(exec));
exec[0].handle = src; exec[0].handle = src;
exec[1].handle = dst; exec[1].handle = dst;
exec[2].handle = handle; exec[2].handle = handle;
if (intel_gen(intel_get_drm_devid(fd)) >= 8) if (has_64bit_reloc)
exec[2].relocation_count = len > 56 ? 4 : 2; exec[2].relocation_count = len > 56 ? 4 : 2;
else else
exec[2].relocation_count = len > 40 ? 4 : 2; exec[2].relocation_count = len > 40 ? 4 : 2;
exec[2].relocs_ptr = (uintptr_t)reloc; exec[2].relocs_ptr = (uintptr_t)reloc;
ring = 0; ring = 0;
if (HAS_BLT_RING(intel_get_drm_devid(fd))) if (gen >= 6)
ring = I915_EXEC_BLT; ring = I915_EXEC_BLT;
memset(&execbuf, 0, sizeof(execbuf)); memset(&execbuf, 0, sizeof(execbuf));
@ -203,14 +211,35 @@ static int run(int size, int count, int reps)
execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC; execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC;
if (__gem_execbuf(fd, &execbuf)) { if (__gem_execbuf(fd, &execbuf)) {
len = gem_linear_blt(fd, buf, src, dst, size, reloc); gem_set_domain(fd, handle, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
len = gem_linear_blt(fd, buf, 0, src, dst, object, reloc);
igt_assert(len == execbuf.batch_len); igt_assert(len == execbuf.batch_len);
gem_write(fd, handle, 0, buf, len);
execbuf.flags = ring; execbuf.flags = ring;
gem_execbuf(fd, &execbuf); gem_execbuf(fd, &execbuf);
} }
gem_sync(fd, handle); gem_sync(fd, handle);
if (batch > 1) {
int nreloc = exec[2].relocation_count;
int i;
if (execbuf.flags & LOCAL_I915_EXEC_HANDLE_LUT) {
src = 0;
dst = 1;
}
for (i = 1; i < batch; i++) {
len = gem_linear_blt(fd, buf, len - 8,
src, dst, object,
reloc + exec[2].relocation_count);
exec[2].relocation_count += nreloc;
}
execbuf.batch_len = len;
gem_execbuf(fd, &execbuf);
gem_sync(fd, handle);
}
while (reps--) { while (reps--) {
struct timespec start, end; struct timespec start, end;
@ -222,10 +251,9 @@ static int run(int size, int count, int reps)
gem_sync(fd, handle); gem_sync(fd, handle);
clock_gettime(CLOCK_MONOTONIC, &end); clock_gettime(CLOCK_MONOTONIC, &end);
printf("%7.3f\n", size/elapsed(&start, &end, count)/(1024*1024)); printf("%7.3f\n", object/elapsed(&start, &end, batch*count)/(1024*1024));
} }
gem_close(fd, handle);
close(fd); close(fd);
return 0; return 0;
} }
@ -235,9 +263,10 @@ 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 batch = 1;
int c; int c;
while ((c = getopt (argc, argv, "c:r:s:")) != -1) { while ((c = getopt (argc, argv, "c:r:s:b:")) != -1) {
switch (c) { switch (c) {
case 'c': case 'c':
count = atoi(optarg); count = atoi(optarg);
@ -257,10 +286,16 @@ int main(int argc, char **argv)
reps = 1; reps = 1;
break; break;
case 'b':
batch = atoi(optarg);
if (batch < 1)
batch = 1;
break;
default: default:
break; break;
} }
} }
return run(size, count, reps); return run(size, batch, count, reps);
} }