flywithu/ntel-gpu-tools
1
0
Fork 0
mirror of https://github.com/tiagovignatti/intel-gpu-tools.git synced 2025-06-29 02:36:29 +00:00
Code Issues Releases Wiki Activity

igt/gem_streaming_writes: Trim number of batch buffers allocated

Browse Source 
Reduce memory usage for batches by a factor of 64 - which we immediately
spend some of in increasing the stress.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Chris Wilson 2015-05-11 17:35:35 +01:00
parent 5f932c4dad
commit 357073c257
1 changed files with 60 additions and 48 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

108
tests/gem_streaming_writes.c
View File

@ -43,7 +43,7 @@
#include "igt_aux.h" #include "igt_aux.h"
#include "intel_chipset.h" #include "intel_chipset.h"
#define OBJECT_SIZE 256*1024 #define OBJECT_SIZE 1024*1024
#define CHUNK_SIZE 32 #define CHUNK_SIZE 32
#define COPY_BLT_CMD (2<<29|0x53<<22|0x6) #define COPY_BLT_CMD (2<<29|0x53<<22|0x6)
@ -58,7 +58,7 @@ static void test_streaming(int fd, int mode)
const int has_64bit_reloc = intel_gen(intel_get_drm_devid(fd)) >= 8; const int has_64bit_reloc = intel_gen(intel_get_drm_devid(fd)) >= 8;
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[2]; struct drm_i915_gem_relocation_entry reloc[128];
uint32_t tmp[] = { MI_BATCH_BUFFER_END }; uint32_t tmp[] = { MI_BATCH_BUFFER_END };
uint64_t __src_offset, __dst_offset; uint64_t __src_offset, __dst_offset;
uint32_t *s, *d; uint32_t *s, *d;
@ -67,7 +67,7 @@ static void test_streaming(int fd, int mode)
uint32_t handle; uint32_t handle;
uint64_t offset; uint64_t offset;
} *batch; } *batch;
int i; int i, n;
memset(exec, 0, sizeof(exec)); memset(exec, 0, sizeof(exec));
exec[0].handle = gem_create(fd, OBJECT_SIZE); exec[0].handle = gem_create(fd, OBJECT_SIZE);
@ -104,61 +104,67 @@ static void test_streaming(int fd, int mode)
__dst_offset = dst_offset; __dst_offset = dst_offset;
memset(reloc, 0, sizeof(reloc)); memset(reloc, 0, sizeof(reloc));
reloc[0].offset = 4 * sizeof(uint32_t); for (i = 0; i < 64; i++) {
reloc[0].delta = 0; reloc[2*i+0].offset = 64*i + 4 * sizeof(uint32_t);
reloc[0].target_handle = dst; reloc[2*i+0].delta = 0;
reloc[0].presumed_offset = dst_offset; reloc[2*i+0].target_handle = dst;
reloc[0].read_domains = I915_GEM_DOMAIN_RENDER; reloc[2*i+0].presumed_offset = dst_offset;
reloc[0].write_domain = I915_GEM_DOMAIN_RENDER; reloc[2*i+0].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[2*i+0].write_domain = I915_GEM_DOMAIN_RENDER;
reloc[1].offset = 7 * sizeof(uint32_t); reloc[2*i+1].offset = 64*i + 7 * sizeof(uint32_t);
if (has_64bit_reloc) if (has_64bit_reloc)
reloc[1].offset += sizeof(uint32_t); reloc[2*i+1].offset += sizeof(uint32_t);
reloc[1].delta = 0; reloc[2*i+1].delta = 0;
reloc[1].target_handle = src; reloc[2*i+1].target_handle = src;
reloc[1].presumed_offset = src_offset; reloc[2*i+1].presumed_offset = src_offset;
reloc[1].read_domains = I915_GEM_DOMAIN_RENDER; reloc[2*i+1].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[1].write_domain = 0; reloc[2*i+1].write_domain = 0;
}
exec[2].relocation_count = 2; exec[2].relocation_count = 2;
exec[2].relocs_ptr = (uintptr_t)reloc;
execbuf.buffer_count = 3; execbuf.buffer_count = 3;
execbuf.flags = I915_EXEC_NO_RELOC; execbuf.flags = I915_EXEC_NO_RELOC;
if (gem_has_blt(fd)) if (gem_has_blt(fd))
execbuf.flags |= I915_EXEC_BLT; execbuf.flags |= I915_EXEC_BLT;
batch = malloc(sizeof(*batch) * (OBJECT_SIZE / CHUNK_SIZE)); batch = malloc(sizeof(*batch) * (OBJECT_SIZE / CHUNK_SIZE / 64));
for (i = 0; i < OBJECT_SIZE / CHUNK_SIZE; i++) { for (i = n = 0; i < OBJECT_SIZE / CHUNK_SIZE / 64; i++) {
unsigned x = (i * CHUNK_SIZE) % 4096 >> 2; uint32_t *base;
unsigned y = (i * CHUNK_SIZE) / 4096;
uint32_t *b;
int j;
batch[i].handle = gem_create(fd, 4096); batch[i].handle = gem_create(fd, 4096);
batch[i].offset = 0; batch[i].offset = 0;
b = gem_mmap__cpu(fd, batch[i].handle, 0, 4096, PROT_WRITE); base = gem_mmap__cpu(fd, batch[i].handle, 0, 4096, PROT_WRITE);
igt_assert(b); igt_assert(base);
j = 0; for (int j = 0; j < 64; j++) {
b[j] = COPY_BLT_CMD | BLT_WRITE_ARGB; unsigned x = (n * CHUNK_SIZE) % 4096 >> 2;
if (has_64bit_reloc) unsigned y = (n * CHUNK_SIZE) / 4096;
b[j] += 2; uint32_t *b = base + 16 * j;
j++; int k = 0;
b[j++] = 0xcc << 16 | 1 << 25 | 1 << 24 | 4096;
b[j++] = (y << 16) | x;
b[j++] = ((y+1) << 16) | (x + (CHUNK_SIZE >> 2));
b[j++] = dst_offset;
if (has_64bit_reloc)
b[j++] = dst_offset >> 32;
b[j++] = (y << 16) | x;
b[j++] = 4096;
b[j++] = src_offset;
if (has_64bit_reloc)
b[j++] = src_offset >> 32;
b[j++] = MI_BATCH_BUFFER_END;
munmap(b, 4096); b[k] = COPY_BLT_CMD | BLT_WRITE_ARGB;
if (has_64bit_reloc)
b[k] += 2;
k++;
b[k++] = 0xcc << 16 | 1 << 25 | 1 << 24 | 4096;
b[k++] = (y << 16) | x;
b[k++] = ((y+1) << 16) | (x + (CHUNK_SIZE >> 2));
b[k++] = dst_offset;
if (has_64bit_reloc)
b[k++] = dst_offset >> 32;
b[k++] = (y << 16) | x;
b[k++] = 4096;
b[k++] = src_offset;
if (has_64bit_reloc)
b[k++] = src_offset >> 32;
b[k++] = MI_BATCH_BUFFER_END;
n++;
}
munmap(base, 4096);
} }
for (int pass = 0; pass < 256; pass++) { for (int pass = 0; pass < 256; pass++) {
@ -167,17 +173,23 @@ static void test_streaming(int fd, int mode)
/* Now copy from the src to the dst in 32byte chunks */ /* Now copy from the src to the dst in 32byte chunks */
for (offset = 0; offset < OBJECT_SIZE; offset += CHUNK_SIZE) { for (offset = 0; offset < OBJECT_SIZE; offset += CHUNK_SIZE) {
int b;
for (i = 0; i < CHUNK_SIZE/4; i++) for (i = 0; i < CHUNK_SIZE/4; i++)
s[offset/4 + i] = (OBJECT_SIZE*pass + offset)/4 + i; s[offset/4 + i] = (OBJECT_SIZE*pass + offset)/4 + i;
exec[2].handle = batch[offset/CHUNK_SIZE].handle; b = offset / CHUNK_SIZE / 64;
exec[2].offset = batch[offset/CHUNK_SIZE].offset; n = offset / CHUNK_SIZE % 64;
exec[2].relocs_ptr = (uintptr_t)(reloc + 2*n);
exec[2].handle = batch[b].handle;
exec[2].offset = batch[b].offset;
execbuf.batch_start_offset = 64*n;
gem_execbuf(fd, &execbuf); gem_execbuf(fd, &execbuf);
igt_assert(__src_offset == src_offset); igt_assert(__src_offset == src_offset);
igt_assert(__dst_offset == dst_offset); igt_assert(__dst_offset == dst_offset);
batch[offset/CHUNK_SIZE].offset = exec[2].offset; batch[b].offset = exec[2].offset;
} }
gem_set_domain(fd, dst, I915_GEM_DOMAIN_CPU, 0); gem_set_domain(fd, dst, I915_GEM_DOMAIN_CPU, 0);
@ -185,7 +197,7 @@ static void test_streaming(int fd, int mode)
igt_assert_eq(pass*OBJECT_SIZE/4 + offset, d[offset]); igt_assert_eq(pass*OBJECT_SIZE/4 + offset, d[offset]);
} }
for (i = 0; i < OBJECT_SIZE / CHUNK_SIZE; i++) for (i = 0; i < OBJECT_SIZE / CHUNK_SIZE / 64; i++)
gem_close(fd, batch[i].handle); gem_close(fd, batch[i].handle);
free(batch); free(batch);