griefith/test
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Metal: TF more optimal for hair refinement

Browse Source 
Patch prefers usage of Transform Feedback for hair refinement
as opposed to compute, as vertex work can be pipelined with
existing rendering work which is in-flight.

This approach is ~20% faster depending on the scene. Note that
the current implementation only uses TF, as storage buffer support
is disabled. Though once storage buffer support is added, we should
still use the TF path.

Authored by Apple: Michael Parkin-White

Pull Request: https://projects.blender.org/blender/blender/pulls/107166
This commit is contained in:
Jason Fielder
2023-04-21 12:35:21 +02:00
committed by Jeroen Bakker
parent bebb17a973
commit 3650b36141
2 changed files with 12 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
source/blender/draw/intern/draw_curves.cc
View File

@@ -33,7 +33,12 @@
BLI_INLINE eParticleRefineShaderType drw_curves_shader_type_get()
{
if (GPU_compute_shader_support() && GPU_shader_storage_buffer_objects_support()) {
/* NOTE: Curve refine is faster using transform feedback via vertex processing pipeline with
* Metal and Apple Silicon GPUs. This is also because vertex work can more easily be executed in
* parallel with fragment work, whereas compute inserts an explicit dependency,
* due to switching of command encoder types. */
if (GPU_compute_shader_support() && GPU_shader_storage_buffer_objects_support() &&
(GPU_backend_get_type() != GPU_BACKEND_METAL)) {
return PART_REFINE_SHADER_COMPUTE;
}
if (GPU_transform_feedback_support()) {

7
source/blender/draw/intern/draw_hair.cc
View File

@@ -36,7 +36,12 @@
BLI_INLINE eParticleRefineShaderType drw_hair_shader_type_get()
{
if (GPU_compute_shader_support() && GPU_shader_storage_buffer_objects_support()) {
/* NOTE: Hair refine is faster using transform feedback via vertex processing pipeline with Metal
* and Apple Silicon GPUs. This is also because vertex work can more easily be executed in
* parallel with fragment work, whereas compute inserts an explicit dependency,
* due to switching of command encoder types. */
if (GPU_compute_shader_support() && GPU_shader_storage_buffer_objects_support() &&
(GPU_backend_get_type() != GPU_BACKEND_METAL)) {
return PART_REFINE_SHADER_COMPUTE;
}
if (GPU_transform_feedback_support()) {