diff options
| author | Samuel Pitoiset <[email protected]> | 2019-07-12 12:17:13 +0200 |
|---|---|---|
| committer | Samuel Pitoiset <[email protected]> | 2019-07-12 17:47:11 +0200 |
| commit | 3f76c0f47c9196556449c0e6e0bc5339028aff2a (patch) | |
| tree | 74889695c1c8a7ef7bde735b0732636cda01a7a1 /src/amd | |
| parent | e631d65fc6155020e9f90eff676bf7f61b234d3a (diff) | |
radv/gfx10: launch 2 compute waves per CU before going onto the next CU
Signed-off-by: Samuel Pitoiset <[email protected]>
Reviewed-by: Bas Nieuwenhuizen <[email protected]>
Diffstat (limited to 'src/amd')
| -rw-r--r-- | src/amd/vulkan/radv_pipeline.c | 13 |
1 files changed, 9 insertions, 4 deletions
diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c index efb94cdcd23..a58b0d6d006 100644 --- a/src/amd/vulkan/radv_pipeline.c +++ b/src/amd/vulkan/radv_pipeline.c @@ -4378,6 +4378,7 @@ radv_compute_generate_pm4(struct radv_pipeline *pipeline) { struct radv_shader_variant *compute_shader; struct radv_device *device = pipeline->device; + unsigned threads_per_threadgroup; unsigned threadgroups_per_cu = 1; unsigned waves_per_threadgroup; unsigned max_waves_per_sh = 0; @@ -4402,10 +4403,14 @@ radv_compute_generate_pm4(struct radv_pipeline *pipeline) S_00B860_WAVESIZE(pipeline->scratch_bytes_per_wave >> 10)); /* Calculate best compute resource limits. */ - waves_per_threadgroup = - DIV_ROUND_UP(compute_shader->info.cs.block_size[0] * - compute_shader->info.cs.block_size[1] * - compute_shader->info.cs.block_size[2], 64); + threads_per_threadgroup = compute_shader->info.cs.block_size[0] * + compute_shader->info.cs.block_size[1] * + compute_shader->info.cs.block_size[2]; + waves_per_threadgroup = DIV_ROUND_UP(threads_per_threadgroup, 64); + + if (device->physical_device->rad_info.chip_class >= GFX10 && + waves_per_threadgroup == 1) + threadgroups_per_cu = 2; radeon_set_sh_reg(&pipeline->cs, R_00B854_COMPUTE_RESOURCE_LIMITS, ac_get_compute_resource_limits(&device->physical_device->rad_info, |