static int radeon_benchmark_do_move(struct radeon_device *rdev, unsigned size,
uint64_t saddr, uint64_t daddr,
int flag, int n)
{
unsigned long start_jiffies;
unsigned long end_jiffies;
struct radeon_fence *fence = NULL;
int i, r;
start_jiffies = jiffies;
for (i = 0; i < n; i++) {
switch (flag) {
case RADEON_BENCHMARK_COPY_DMA:
r = radeon_fence_create(rdev, &fence, radeon_copy_dma_ring_index(rdev));
if (r)
return r;
r = radeon_copy_dma(rdev, saddr, daddr,
size / RADEON_GPU_PAGE_SIZE,
fence);
break;
case RADEON_BENCHMARK_COPY_BLIT:
r = radeon_fence_create(rdev, &fence, radeon_copy_blit_ring_index(rdev));
if (r)
return r;
r = radeon_copy_blit(rdev, saddr, daddr,
size / RADEON_GPU_PAGE_SIZE,
fence);
break;
default:
DRM_ERROR("Unknown copy method\n");
r = -EINVAL;
}
if (r)
goto exit_do_move;
r = radeon_fence_wait(fence, false);
if (r)
goto exit_do_move;
radeon_fence_unref(&fence);
}
end_jiffies = jiffies;
r = jiffies_to_msecs(end_jiffies - start_jiffies);
exit_do_move:
if (fence)
radeon_fence_unref(&fence);
return r;
}
/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */
static void radeon_do_test_moves(struct radeon_device *rdev, int flag)
{
struct radeon_bo *vram_obj = NULL;
struct radeon_bo **gtt_obj = NULL;
uint64_t gtt_addr, vram_addr;
unsigned n, size;
int i, r, ring;
switch (flag) {
case RADEON_TEST_COPY_DMA:
ring = radeon_copy_dma_ring_index(rdev);
break;
case RADEON_TEST_COPY_BLIT:
ring = radeon_copy_blit_ring_index(rdev);
break;
default:
DRM_ERROR("Unknown copy method\n");
return;
}
size = 1024 * 1024;
/* Number of tests =
* (Total GTT - IB pool - writeback page - ring buffers) / test size
*/
n = rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024;
for (i = 0; i < RADEON_NUM_RINGS; ++i)
n -= rdev->ring[i].ring_size;
if (rdev->wb.wb_obj)
n -= RADEON_GPU_PAGE_SIZE;
if (rdev->ih.ring_obj)
n -= rdev->ih.ring_size;
n /= size;
gtt_obj = kzalloc(n * sizeof(*gtt_obj), GFP_KERNEL);
if (!gtt_obj) {
DRM_ERROR("Failed to allocate %d pointers\n", n);
r = 1;
goto out_cleanup;
}
r = radeon_bo_create(rdev, size, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM,
NULL, &vram_obj);
if (r) {
DRM_ERROR("Failed to create VRAM object\n");
goto out_cleanup;
}
r = radeon_bo_reserve(vram_obj, false);
if (unlikely(r != 0))
goto out_unref;
r = radeon_bo_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr);
if (r) {
DRM_ERROR("Failed to pin VRAM object\n");
goto out_unres;
}
for (i = 0; i < n; i++) {
void *gtt_map, *vram_map;
void **gtt_start, **gtt_end;
void **vram_start, **vram_end;
struct radeon_fence *fence = NULL;
r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
RADEON_GEM_DOMAIN_GTT, NULL, gtt_obj + i);
if (r) {
DRM_ERROR("Failed to create GTT object %d\n", i);
goto out_lclean;
}
r = radeon_bo_reserve(gtt_obj[i], false);
if (unlikely(r != 0))
goto out_lclean_unref;
r = radeon_bo_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, >t_addr);
if (r) {
DRM_ERROR("Failed to pin GTT object %d\n", i);
goto out_lclean_unres;
}
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size;
gtt_start < gtt_end;
gtt_start++)
*gtt_start = gtt_start;
radeon_bo_kunmap(gtt_obj[i]);
if (ring == R600_RING_TYPE_DMA_INDEX)
r = radeon_copy_dma(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
else
r = radeon_copy_blit(rdev, gtt_addr, vram_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
if (r) {
DRM_ERROR("Failed GTT->VRAM copy %d\n", i);
goto out_lclean_unpin;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i);
goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(vram_obj, &vram_map);
if (r) {
DRM_ERROR("Failed to map VRAM object after copy %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
vram_start < vram_end;
gtt_start++, vram_start++) {
if (*vram_start != gtt_start) {
DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, "
"expected 0x%p (GTT/VRAM offset "
"0x%16llx/0x%16llx)\n",
i, *vram_start, gtt_start,
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)gtt_start - gtt_map),
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)gtt_start - gtt_map));
radeon_bo_kunmap(vram_obj);
goto out_lclean_unpin;
}
*vram_start = vram_start;
}
radeon_bo_kunmap(vram_obj);
if (ring == R600_RING_TYPE_DMA_INDEX)
r = radeon_copy_dma(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
else
r = radeon_copy_blit(rdev, vram_addr, gtt_addr, size / RADEON_GPU_PAGE_SIZE, &fence);
if (r) {
DRM_ERROR("Failed VRAM->GTT copy %d\n", i);
goto out_lclean_unpin;
}
r = radeon_fence_wait(fence, false);
if (r) {
DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i);
goto out_lclean_unpin;
}
radeon_fence_unref(&fence);
r = radeon_bo_kmap(gtt_obj[i], >t_map);
if (r) {
DRM_ERROR("Failed to map GTT object after copy %d\n", i);
goto out_lclean_unpin;
}
for (gtt_start = gtt_map, gtt_end = gtt_map + size,
vram_start = vram_map, vram_end = vram_map + size;
gtt_start < gtt_end;
gtt_start++, vram_start++) {
if (*gtt_start != vram_start) {
DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, "
"expected 0x%p (VRAM/GTT offset "
"0x%16llx/0x%16llx)\n",
i, *gtt_start, vram_start,
(unsigned long long)
(vram_addr - rdev->mc.vram_start +
(void*)vram_start - vram_map),
(unsigned long long)
(gtt_addr - rdev->mc.gtt_start +
(void*)vram_start - vram_map));
radeon_bo_kunmap(gtt_obj[i]);
goto out_lclean_unpin;
}
}
radeon_bo_kunmap(gtt_obj[i]);
DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n",
gtt_addr - rdev->mc.gtt_start);
continue;
out_lclean_unpin:
radeon_bo_unpin(gtt_obj[i]);
out_lclean_unres:
radeon_bo_unreserve(gtt_obj[i]);
out_lclean_unref:
radeon_bo_unref(>t_obj[i]);
out_lclean:
for (--i; i >= 0; --i) {
radeon_bo_unpin(gtt_obj[i]);
radeon_bo_unreserve(gtt_obj[i]);
radeon_bo_unref(>t_obj[i]);
}
if (fence)
radeon_fence_unref(&fence);
break;
}
radeon_bo_unpin(vram_obj);
out_unres:
radeon_bo_unreserve(vram_obj);
out_unref:
radeon_bo_unref(&vram_obj);
out_cleanup:
kfree(gtt_obj);
if (r) {
printk(KERN_WARNING "Error while testing BO move.\n");
}
}
void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize,
unsigned sdomain, unsigned ddomain)
{
struct radeon_bo *dobj = NULL;
struct radeon_bo *sobj = NULL;
struct radeon_fence *fence = NULL;
uint64_t saddr, daddr;
unsigned long start_jiffies;
unsigned long end_jiffies;
unsigned long time;
unsigned i, n, size;
int r;
size = bsize;
n = 1024;
r = radeon_bo_create(rdev, NULL, size, PAGE_SIZE, true, sdomain, &sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(sobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(sobj, sdomain, &saddr);
radeon_bo_unreserve(sobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_create(rdev, NULL, size, PAGE_SIZE, true, ddomain, &dobj);
if (r) {
goto out_cleanup;
}
r = radeon_bo_reserve(dobj, false);
if (unlikely(r != 0))
goto out_cleanup;
r = radeon_bo_pin(dobj, ddomain, &daddr);
radeon_bo_unreserve(dobj);
if (r) {
goto out_cleanup;
}
/* r100 doesn't have dma engine so skip the test */
if (rdev->asic->copy_dma) {
start_jiffies = jiffies;
for (i = 0; i < n; i++) {
r = radeon_fence_create(rdev, &fence);
if (r) {
goto out_cleanup;
}
r = radeon_copy_dma(rdev, saddr, daddr,
size / RADEON_GPU_PAGE_SIZE, fence);
if (r) {
goto out_cleanup;
}
r = radeon_fence_wait(fence, false);
if (r) {
goto out_cleanup;
}
radeon_fence_unref(&fence);
}
end_jiffies = jiffies;
time = end_jiffies - start_jiffies;
time = jiffies_to_msecs(time);
if (time > 0) {
i = ((n * size) >> 10) / time;
printk(KERN_INFO "radeon: dma %u bo moves of %ukb from"
" %d to %d in %lums (%ukb/ms %ukb/s %uM/s)\n",
n, size >> 10,
sdomain, ddomain, time,
i, i * 1000, (i * 1000) / 1024);
}