flywithu/ntel-gpu-tools
1
0
Fork 0
mirror of https://github.com/tiagovignatti/intel-gpu-tools.git synced 2025-06-11 01:46:14 +00:00
Code Issues Releases Wiki Activity

igt/gem_busy: Try to slow down execution

Browse Source 
With lockdep enabled, the driver overhead is enormous and we need to
slow down the GPU to compenstate (otherwise the GPU is already idle when
we expect busyness). So do more work per batch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Chris Wilson 2016-04-01 16:52:37 +01:00
parent 92ee3cc1f6
commit aec81019e1
1 changed files with 41 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

69
tests/gem_busy.c
View File

@ -23,6 +23,8 @@
#include "igt.h"
#define LOCAL_EXEC_NO_RELOC (1<<11)
/* Exercise the busy-ioctl, ensuring the ABI is never broken */
IGT_TEST_DESCRIPTION("Basic check of busy-ioctl ABI.");
@ -224,54 +226,65 @@ static void store(int fd, unsigned ring, uint32_t flags)
{
const int gen = intel_gen(intel_get_drm_devid(fd));
struct drm_i915_gem_exec_object2 obj[2];
struct drm_i915_gem_relocation_entry reloc;
struct drm_i915_gem_relocation_entry reloc[1024];
struct drm_i915_gem_execbuffer2 execbuf;
unsigned size = ALIGN(ARRAY_SIZE(reloc)*16 + 4, 4096);
uint32_t read[2], write[2];
struct timespec tv;
uint32_t batch[16];
uint32_t *batch;
int i, count;
gem_require_ring(fd, ring | flags);
igt_skip_on_f(gen == 6 && (ring & ~(3<<13)) == I915_EXEC_BSD,
"MI_STORE_DATA broken on gen6 bsd\n");
gem_quiescent_gpu(fd);
memset(&execbuf, 0, sizeof(execbuf));
execbuf.buffers_ptr = (uintptr_t)obj;
execbuf.buffer_count = 2;
execbuf.flags = ring | flags;
execbuf.flags = ring | flags | LOCAL_EXEC_NO_RELOC;
if (gen < 6)
execbuf.flags |= I915_EXEC_SECURE;
memset(obj, 0, sizeof(obj));
obj[0].handle = gem_create(fd, 4096);
obj[1].handle = gem_create(fd, 4096);
obj[0].flags = EXEC_OBJECT_WRITE;
obj[1].handle = gem_create(fd, size);
memset(&reloc, 0, sizeof(reloc));
reloc.target_handle = obj[0].handle;
reloc.presumed_offset = 0;
reloc.offset = sizeof(uint32_t);
reloc.delta = 0;
reloc.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
reloc.write_domain = I915_GEM_DOMAIN_INSTRUCTION;
obj[1].relocs_ptr = (uintptr_t)&reloc;
obj[1].relocation_count = 1;
memset(reloc, 0, sizeof(reloc));
obj[1].relocs_ptr = (uintptr_t)reloc;
obj[1].relocation_count = ARRAY_SIZE(reloc);
batch = gem_mmap__cpu(fd, obj[1].handle, 0, size, PROT_WRITE);
gem_set_domain(fd, obj[1].handle,
I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
i = 0;
batch[i] = MI_STORE_DWORD_IMM | (gen < 6 ? 1 << 22 : 0);
if (gen >= 8) {
batch[++i] = 0;
batch[++i] = 0;
} else if (gen >= 4) {
batch[++i] = 0;
batch[++i] = 0;
reloc.offset += sizeof(uint32_t);
} else {
batch[i]--;
batch[++i] = 0;
for (count = 0; count < ARRAY_SIZE(reloc); count++) {
reloc[count].target_handle = obj[0].handle;
reloc[count].presumed_offset = 0;
reloc[count].offset = sizeof(uint32_t) * (i + 1);
reloc[count].delta = 0;
reloc[count].read_domains = I915_GEM_DOMAIN_INSTRUCTION;
reloc[count].write_domain = I915_GEM_DOMAIN_INSTRUCTION;
batch[i] = MI_STORE_DWORD_IMM | (gen < 6 ? 1 << 22 : 0);
if (gen >= 8) {
batch[++i] = 0;
batch[++i] = 0;
} else if (gen >= 4) {
batch[++i] = 0;
batch[++i] = 0;
reloc[count].offset += sizeof(uint32_t);
} else {
batch[i]--;
batch[++i] = 0;
}
batch[++i] = 0xc0ffee;
}
batch[++i] = 0xc0ffee;
batch[++i] = MI_BATCH_BUFFER_END;
gem_write(fd, obj[1].handle, 0, batch, sizeof(batch));
munmap(batch, size);
igt_require(__gem_execbuf(fd, &execbuf) == 0);
count = 16;
do {
@ -281,7 +294,7 @@ static void store(int fd, unsigned ring, uint32_t flags)
__gem_busy(fd, obj[1].handle, &read[1], &write[1]);
igt_debug("After %d cycles: read[0]=%x read[1]=%x\n",
count, read[0], read[1]);
igt_require(count <= 1 << 12);
igt_require(count <= 1 << 10);
count <<= 1;
} while (read[0] == 0 || read[1] == 0);
@ -294,7 +307,7 @@ static void store(int fd, unsigned ring, uint32_t flags)
/* Calling busy in a loop should be enough to flush the rendering */
memset(&tv, 0, sizeof(tv));
while (gem_busy(fd, obj[1].handle))
igt_assert(igt_seconds_elapsed(&tv) < 2);
igt_assert(igt_seconds_elapsed(&tv) < 10);
igt_assert(!gem_busy(fd, obj[0].handle));
gem_close(fd, obj[1].handle);