diff options
author | bunnei <bunneidev@gmail.com> | 2019-01-26 05:42:14 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-01-26 05:42:14 +0100 |
commit | 1f4ca1e841cd0b0427218d787efe10a3fa62df33 (patch) | |
tree | 00cc1743c6a6ba593e3b56897b13c2272a71d779 /src/video_core/shader/decode/hfma2.cpp | |
parent | Merge pull request #2054 from bunnei/scope-context-refactor (diff) | |
parent | shader_ir: Fixup clang build (diff) | |
download | yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar.gz yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar.bz2 yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar.lz yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar.xz yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.tar.zst yuzu-1f4ca1e841cd0b0427218d787efe10a3fa62df33.zip |
Diffstat (limited to 'src/video_core/shader/decode/hfma2.cpp')
-rw-r--r-- | src/video_core/shader/decode/hfma2.cpp | 76 |
1 files changed, 76 insertions, 0 deletions
diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp new file mode 100644 index 000000000..4a6b945f9 --- /dev/null +++ b/src/video_core/shader/decode/hfma2.cpp @@ -0,0 +1,76 @@ +// Copyright 2018 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <tuple> + +#include "common/assert.h" +#include "common/common_types.h" +#include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/shader_ir.h" + +namespace VideoCommon::Shader { + +using Tegra::Shader::HalfPrecision; +using Tegra::Shader::HalfType; +using Tegra::Shader::Instruction; +using Tegra::Shader::OpCode; + +u32 ShaderIR::DecodeHfma2(BasicBlock& bb, const BasicBlock& code, u32 pc) { + const Instruction instr = {program_code[pc]}; + const auto opcode = OpCode::Decode(instr); + + if (opcode->get().GetId() == OpCode::Id::HFMA2_RR) { + UNIMPLEMENTED_IF(instr.hfma2.rr.precision != HalfPrecision::None); + } else { + UNIMPLEMENTED_IF(instr.hfma2.precision != HalfPrecision::None); + } + + constexpr auto identity = HalfType::H0_H1; + + const HalfType type_a = instr.hfma2.type_a; + const Node op_a = GetRegister(instr.gpr8); + + bool neg_b{}, neg_c{}; + auto [saturate, type_b, op_b, type_c, + op_c] = [&]() -> std::tuple<bool, HalfType, Node, HalfType, Node> { + switch (opcode->get().GetId()) { + case OpCode::Id::HFMA2_CR: + neg_b = instr.hfma2.negate_b; + neg_c = instr.hfma2.negate_c; + return {instr.hfma2.saturate, instr.hfma2.type_b, + GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset), instr.hfma2.type_reg39, + GetRegister(instr.gpr39)}; + case OpCode::Id::HFMA2_RC: + neg_b = instr.hfma2.negate_b; + neg_c = instr.hfma2.negate_c; + return {instr.hfma2.saturate, instr.hfma2.type_reg39, GetRegister(instr.gpr39), + instr.hfma2.type_b, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)}; + case OpCode::Id::HFMA2_RR: + neg_b = instr.hfma2.rr.negate_b; + neg_c = instr.hfma2.rr.negate_c; + return {instr.hfma2.rr.saturate, instr.hfma2.type_b, GetRegister(instr.gpr20), + instr.hfma2.rr.type_c, GetRegister(instr.gpr39)}; + case OpCode::Id::HFMA2_IMM_R: + neg_c = instr.hfma2.negate_c; + return {instr.hfma2.saturate, identity, UnpackHalfImmediate(instr, true), + instr.hfma2.type_reg39, GetRegister(instr.gpr39)}; + default: + return {false, identity, Immediate(0), identity, Immediate(0)}; + } + }(); + UNIMPLEMENTED_IF_MSG(saturate, "HFMA2 saturation is not implemented"); + + op_b = GetOperandAbsNegHalf(op_b, false, neg_b); + op_c = GetOperandAbsNegHalf(op_c, false, neg_c); + + MetaHalfArithmetic meta{true, {type_a, type_b, type_c}}; + Node value = Operation(OperationCode::HFma, meta, op_a, op_b, op_c); + value = HalfMerge(GetRegister(instr.gpr0), value, instr.hfma2.merge); + + SetRegister(bb, instr.gpr0, value); + + return pc; +} + +} // namespace VideoCommon::Shader
\ No newline at end of file |