diff options
author | bunnei <bunneidev@gmail.com> | 2019-12-22 04:50:28 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-12-22 04:50:28 +0100 |
commit | 16dcfacbfca912ee281f62b37eafb6258935fa55 (patch) | |
tree | 130f49d287ac8dd3fd1d775e0d1a8c401ca7a289 /src/video_core | |
parent | Merge pull request #3234 from ReinUsesLisp/i2f-u8-selector (diff) | |
parent | shader/memory: Implement LDG.U8 and unaligned U8 loads (diff) | |
download | yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar.gz yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar.bz2 yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar.lz yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar.xz yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.tar.zst yuzu-16dcfacbfca912ee281f62b37eafb6258935fa55.zip |
Diffstat (limited to 'src/video_core')
-rw-r--r-- | src/video_core/shader/decode/memory.cpp | 38 |
1 files changed, 32 insertions, 6 deletions
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp index 78e92f52e..c934d0719 100644 --- a/src/video_core/shader/decode/memory.cpp +++ b/src/video_core/shader/decode/memory.cpp @@ -22,7 +22,23 @@ using Tegra::Shader::Register; namespace { -u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) { +u32 GetLdgMemorySize(Tegra::Shader::UniformType uniform_type) { + switch (uniform_type) { + case Tegra::Shader::UniformType::UnsignedByte: + case Tegra::Shader::UniformType::Single: + return 1; + case Tegra::Shader::UniformType::Double: + return 2; + case Tegra::Shader::UniformType::Quad: + case Tegra::Shader::UniformType::UnsignedQuad: + return 4; + default: + UNIMPLEMENTED_MSG("Unimplemented size={}!", static_cast<u32>(uniform_type)); + return 1; + } +} + +u32 GetStgMemorySize(Tegra::Shader::UniformType uniform_type) { switch (uniform_type) { case Tegra::Shader::UniformType::Single: return 1; @@ -170,7 +186,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { const auto [real_address_base, base_address, descriptor] = TrackGlobalMemory(bb, instr, false); - const u32 count = GetUniformTypeElementsCount(type); + const u32 count = GetLdgMemorySize(type); if (!real_address_base || !base_address) { // Tracking failed, load zeroes. for (u32 i = 0; i < count; ++i) { @@ -181,12 +197,22 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { for (u32 i = 0; i < count; ++i) { const Node it_offset = Immediate(i * 4); - const Node real_address = - Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset); - const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); + const Node real_address = Operation(OperationCode::UAdd, real_address_base, it_offset); + Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); + + if (type == Tegra::Shader::UniformType::UnsignedByte) { + // To handle unaligned loads get the byte used to dereferenced global memory + // and extract that byte from the loaded uint32. + Node byte = Operation(OperationCode::UBitwiseAnd, real_address, Immediate(3)); + byte = Operation(OperationCode::ULogicalShiftLeft, std::move(byte), Immediate(3)); + + gmem = Operation(OperationCode::UBitfieldExtract, std::move(gmem), std::move(byte), + Immediate(8)); + } SetTemporary(bb, i, gmem); } + for (u32 i = 0; i < count; ++i) { SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i)); } @@ -276,7 +302,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { break; } - const u32 count = GetUniformTypeElementsCount(type); + const u32 count = GetStgMemorySize(type); for (u32 i = 0; i < count; ++i) { const Node it_offset = Immediate(i * 4); const Node real_address = Operation(OperationCode::UAdd, real_address_base, it_offset); |