yuzu-emu
/
yuzu
Archived
1
0
Fork 0
Code

Shader_IR: Implement FLO instruction.

This commit is contained in:
Fernando Sahmkow 2019-11-12 10:07:22 -04:00 committed by FernandoS27
parent 95137a04e1
commit f3d1b370aa
5 changed files with 35 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
src/video_core/engines/shader_bytecode.h
View File

@ -799,6 +799,12 @@ union Instruction {
BitField<40, 1, u64> invert;
} popc;
union {
BitField<41, 1, u64> sh;
BitField<40, 1, u64> invert;
BitField<48, 1, u64> is_signed;
} flo;
union {
BitField<39, 3, u64> pred;
BitField<42, 1, u64> neg_pred;

7
src/video_core/renderer_opengl/gl_shader_decompiler.cpp
View File

@ -1472,6 +1472,11 @@ private:
return GenerateUnary(operation, "bitCount", type, type);
}
template <Type type>
Expression BitMSB(Operation operation) {
return GenerateUnary(operation, "findMSB", type, type);
}
Expression HNegate(Operation operation) {
const auto GetNegate = [&](std::size_t index) {
return VisitOperand(operation, index).AsBool() + " ? -1 : 1";
@ -2043,6 +2048,7 @@ private:
&GLSLDecompiler::BitfieldInsert<Type::Int>,
&GLSLDecompiler::BitfieldExtract<Type::Int>,
&GLSLDecompiler::BitCount<Type::Int>,
&GLSLDecompiler::BitMSB<Type::Int>,
&GLSLDecompiler::Add<Type::Uint>,
&GLSLDecompiler::Mul<Type::Uint>,
@ -2061,6 +2067,7 @@ private:
&GLSLDecompiler::BitfieldInsert<Type::Uint>,
&GLSLDecompiler::BitfieldExtract<Type::Uint>,
&GLSLDecompiler::BitCount<Type::Uint>,
&GLSLDecompiler::BitMSB<Type::Uint>,
&GLSLDecompiler::Add<Type::HalfFloat>,
&GLSLDecompiler::Mul<Type::HalfFloat>,

2
src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
View File

@ -1390,6 +1390,7 @@ private:
&SPIRVDecompiler::Quaternary<&Module::OpBitFieldInsert, Type::Int>,
&SPIRVDecompiler::Ternary<&Module::OpBitFieldSExtract, Type::Int>,
&SPIRVDecompiler::Unary<&Module::OpBitCount, Type::Int>,
&SPIRVDecompiler::Unary<&Module::OpFindSMsb, Type::Int>,
&SPIRVDecompiler::Binary<&Module::OpIAdd, Type::Uint>,
&SPIRVDecompiler::Binary<&Module::OpIMul, Type::Uint>,
@ -1408,6 +1409,7 @@ private:
&SPIRVDecompiler::Quaternary<&Module::OpBitFieldInsert, Type::Uint>,
&SPIRVDecompiler::Ternary<&Module::OpBitFieldUExtract, Type::Uint>,
&SPIRVDecompiler::Unary<&Module::OpBitCount, Type::Uint>,
&SPIRVDecompiler::Unary<&Module::OpFindUMsb, Type::Uint>,
&SPIRVDecompiler::Binary<&Module::OpFAdd, Type::HalfFloat>,
&SPIRVDecompiler::Binary<&Module::OpFMul, Type::HalfFloat>,

18
src/video_core/shader/decode/arithmetic_integer.cpp
View File

@ -130,6 +130,24 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
SetRegister(bb, instr.gpr0, value);
break;
}
case OpCode::Id::FLO_R:
case OpCode::Id::FLO_C:
case OpCode::Id::FLO_IMM: {
Node value;
if (instr.flo.invert) {
op_b = Operation(OperationCode::IBitwiseNot, NO_PRECISE, op_b);
}
if (instr.flo.is_signed) {
value = Operation(OperationCode::IBitMSB, NO_PRECISE, op_b);
} else {
value = Operation(OperationCode::UBitMSB, NO_PRECISE, op_b);
}
if (instr.flo.sh) {
value = Operation(OperationCode::UBitwiseXor, NO_PRECISE, value, Immediate(31));
}
SetRegister(bb, instr.gpr0, value);
break;
}
case OpCode::Id::SEL_C:
case OpCode::Id::SEL_R:
case OpCode::Id::SEL_IMM: {

2
src/video_core/shader/node.h
View File

@ -68,6 +68,7 @@ enum class OperationCode {
IBitfieldInsert, /// (MetaArithmetic, int base, int insert, int offset, int bits) -> int
IBitfieldExtract, /// (MetaArithmetic, int value, int offset, int offset) -> int
IBitCount, /// (MetaArithmetic, int) -> int
IBitMSB, /// (MetaArithmetic, int) -> int
UAdd, /// (MetaArithmetic, uint a, uint b) -> uint
UMul, /// (MetaArithmetic, uint a, uint b) -> uint
@ -86,6 +87,7 @@ enum class OperationCode {
UBitfieldInsert, /// (MetaArithmetic, uint base, uint insert, int offset, int bits) -> uint
UBitfieldExtract, /// (MetaArithmetic, uint value, int offset, int offset) -> uint
UBitCount, /// (MetaArithmetic, uint) -> uint
UBitMSB, /// (MetaArithmetic, uint) -> uint
HAdd, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2
HMul, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2