raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-21 11:26:56 +02:00
Code Issues Packages Projects Releases Wiki Activity

shader: share scalar condition results with direct subgroup votes

Browse Source 
Co-authored-by: fxpw <fxpw@outlook.com>
This commit is contained in:
nmzik
2026-10-15 00:07:35 +02:00
parent 50844badbf
commit bb2d6dbeb3
7 changed files with 109 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/shader/recompiler/ir/ShaderIR.h
+1
View File
@@ -506,6 +506,7 @@ struct ShaderInfo {
bool uses_dispatch_threads = false;
bool bvh = false;
bool subgroup_ballot = false;
bool subgroup_vote = false;
bool subgroup_barrier = false;
bool subgroup_shuffle = false;
bool subgroup_arithmetic = false;
src/graphics/shader/recompiler/ir/passes/ShaderInfoCollection.cpp
+3 -2
View File
@@ -510,7 +510,7 @@ void Visit(Program& program, ShaderStageInputInfo input_info, InputUsage& inputs
case ValueOpcode::StoreCompletion: info.subgroup_barrier = true; break;
case ValueOpcode::BvhIntersect: info.bvh = true; break;
case ValueOpcode::ConditionRef:
info.subgroup_ballot |=
info.subgroup_vote |=
inst.Flags<CFG::BranchCondition>() != CFG::BranchCondition::ScalarInstruction;
break;
case ValueOpcode::Ballot: info.subgroup_ballot = true; break;
@@ -572,7 +572,8 @@ void CollectShaderInfo(Program& program, ShaderStageInputInfo input_info) {
for (auto& buffer: info.buffers) buffer.live = false;
info.has_bitwise_xor = info.uses_lds = info.uses_gds = info.uses_flattened_srt =
info.uses_dispatch_threads = info.uses_swizzle = false;
info.bvh = info.subgroup_ballot = info.subgroup_barrier = info.subgroup_shuffle = info.subgroup_arithmetic =
info.bvh = info.subgroup_ballot = info.subgroup_vote = info.subgroup_barrier =
info.subgroup_shuffle = info.subgroup_arithmetic =
info.subgroup_local_invocation_id = info.compute_derivatives = info.image_gather_extended =
info.function_lds = info.function_scratch = info.pixel_valid_mask = info.buffer_int64_atomics =
info.buffer_u8 = info.buffer_u16 = info.shared_int64_atomics = info.coherent_buffers = info.float64 = false;