This decreases eviction by up to 20%, by improving the fragmentation quality. No harm in normal cases that fit VRAM fully (PTS gaming suite).
In some cases, even the VRAM-fitting cases improved slightly (openarena, urban terror).
512kb was measured as the most optimal threshold for 3d workloads common to radeon. Other drivers may need different thresholds according to their workloads.
Signed-off-by: Lauri Kasanen cand@gmx.com --- drivers/gpu/drm/radeon/radeon_object.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 1375ff8..6251456 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -104,7 +104,7 @@ bool radeon_ttm_bo_is_radeon_bo(struct ttm_buffer_object *bo)
void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) { - u32 c = 0; + u32 c = 0, i;
rbo->placement.fpfn = 0; rbo->placement.lpfn = 0; @@ -131,6 +131,15 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; rbo->placement.num_placement = c; rbo->placement.num_busy_placement = c; + + /* + * Use two-ended allocation depending on the buffer size to + * improve fragmentation quality. + * 512kb was measured as the most optimal number. + */ + if (rbo->tbo.mem.size > 512 * 1024) for (i = 0; i < c; i++) { + rbo->placements[i] |= TTM_PL_FLAG_TOPDOWN; + } }
int radeon_bo_create(struct radeon_device *rdev,