Mirrors/yuzu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

gl_shader_decompiler: Reimplement shuffles with platform agnostic intrinsics

Browse source
This commit is contained in:
ReinUsesLisp 2019-11-02 23:44:13 -03:00
parent b6ae48966d
commit 08b2b1080a
No known key found for this signature in database
GPG key ID: 2DFC508897B39CFE
5 changed files with 59 additions and 132 deletions
Download patch file Download diff file Expand all files Collapse all files

82
src/video_core/shader/decode/warp.cpp
View file

@ -17,6 +17,7 @@ using Tegra::Shader::ShuffleOperation;
using Tegra::Shader::VoteOperation;
namespace {
OperationCode GetOperationCode(VoteOperation vote_op) {
switch (vote_op) {
case VoteOperation::All:
@ -30,6 +31,7 @@ OperationCode GetOperationCode(VoteOperation vote_op) {
return OperationCode::VoteAll;
}
}
} // Anonymous namespace
u32 ShaderIR::DecodeWarp(NodeBlock& bb, u32 pc) {
@ -46,50 +48,50 @@ u32 ShaderIR::DecodeWarp(NodeBlock& bb, u32 pc) {
break;
}
case OpCode::Id::SHFL: {
Node width = [this, instr] {
Node mask = instr.shfl.is_mask_imm ? Immediate(static_cast<u32>(instr.shfl.mask_imm))
: GetRegister(instr.gpr39);
// Convert the obscure SHFL mask back into GL_NV_shader_thread_shuffle's width. This has
// been done reversing Nvidia's math. It won't work on all cases due to SHFL having
// different parameters that don't properly map to GLSL's interface, but it should work
// for cases emitted by Nvidia's compiler.
if (instr.shfl.operation == ShuffleOperation::Up) {
return Operation(
OperationCode::ILogicalShiftRight,
Operation(OperationCode::IAdd, std::move(mask), Immediate(-0x2000)),
Immediate(8));
} else {
return Operation(OperationCode::ILogicalShiftRight,
Operation(OperationCode::IAdd, Immediate(0x201F),
Operation(OperationCode::INegate, std::move(mask))),
Immediate(8));
}
}();
const auto [operation, in_range] = [instr]() -> std::pair<OperationCode, OperationCode> {
switch (instr.shfl.operation) {
case ShuffleOperation::Idx:
return {OperationCode::ShuffleIndexed, OperationCode::InRangeShuffleIndexed};
case ShuffleOperation::Up:
return {OperationCode::ShuffleUp, OperationCode::InRangeShuffleUp};
case ShuffleOperation::Down:
return {OperationCode::ShuffleDown, OperationCode::InRangeShuffleDown};
case ShuffleOperation::Bfly:
return {OperationCode::ShuffleButterfly, OperationCode::InRangeShuffleButterfly};
}
UNREACHABLE_MSG("Invalid SHFL operation: {}",
static_cast<u64>(instr.shfl.operation.Value()));
return {};
}();
// Setting the predicate before the register is intentional to avoid overwriting.
Node mask = instr.shfl.is_mask_imm ? Immediate(static_cast<u32>(instr.shfl.mask_imm))
: GetRegister(instr.gpr39);
Node index = instr.shfl.is_index_imm ? Immediate(static_cast<u32>(instr.shfl.index_imm))
: GetRegister(instr.gpr20);
SetPredicate(bb, instr.shfl.pred48, Operation(in_range, index, width));
Node thread_id = Operation(OperationCode::ThreadId);
Node clamp = Operation(OperationCode::IBitwiseAnd, mask, Immediate(0x1FU));
Node seg_mask = BitfieldExtract(mask, 8, 16);
Node neg_seg_mask = Operation(OperationCode::IBitwiseNot, seg_mask);
Node min_thread_id = Operation(OperationCode::IBitwiseAnd, thread_id, seg_mask);
Node max_thread_id = Operation(OperationCode::IBitwiseOr, min_thread_id,
Operation(OperationCode::IBitwiseAnd, clamp, neg_seg_mask));
Node src_thread_id = [instr, index, neg_seg_mask, min_thread_id, thread_id] {
switch (instr.shfl.operation) {
case ShuffleOperation::Idx:
return Operation(OperationCode::IBitwiseOr,
Operation(OperationCode::IBitwiseAnd, index, neg_seg_mask),
min_thread_id);
case ShuffleOperation::Down:
return Operation(OperationCode::IAdd, thread_id, index);
case ShuffleOperation::Up:
return Operation(OperationCode::IAdd, thread_id,
Operation(OperationCode::INegate, index));
case ShuffleOperation::Bfly:
return Operation(OperationCode::IBitwiseXor, thread_id, index);
}
UNREACHABLE();
return Immediate(0U);
}();
Node in_bounds = [instr, src_thread_id, min_thread_id, max_thread_id] {
if (instr.shfl.operation == ShuffleOperation::Up) {
return Operation(OperationCode::LogicalIGreaterEqual, src_thread_id, min_thread_id);
} else {
return Operation(OperationCode::LogicalILessEqual, src_thread_id, max_thread_id);
}
}();
SetPredicate(bb, instr.shfl.pred48, in_bounds);
SetRegister(
bb, instr.gpr0,
Operation(operation, GetRegister(instr.gpr8), std::move(index), std::move(width)));
Operation(OperationCode::ShuffleIndexed, GetRegister(instr.gpr8), src_thread_id));
break;
}
default: