raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-28 13:56:59 +01:00
Code Issues Packages Projects Releases Wiki Activity

shader: port 64-bit ALU and atomic opcodes

Browse Source
This commit is contained in:
Senaxx
2026-10-01 18:09:22 +04:00
committed by nmzik
parent 11001c97f7
commit 0d0b215fab
20 changed files with 687 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/shader/recompiler/frontend/translate/Compare.cpp
+16 -3
View File
@@ -1,5 +1,7 @@
#include "graphics/shader/recompiler/frontend/translate/Translator.h"
#include <bit>
namespace Libs::Graphics::ShaderRecompiler::Frontend {
void Translator::EmitCompareResult(const Decoder::Instruction& inst, IR::U1 value, bool scalar,
@@ -26,9 +28,20 @@ void Translator::EmitCompareConstant(const Decoder::Instruction& inst, bool valu
void Translator::EmitIntegerCompare(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
IR::Type type, bool scalar, bool cmpx) {
const auto lhs = ReadOperand(inst.src0, type);
const auto rhs = ReadOperand(inst.src1, type);
EmitCompareResult(inst, IR::U1(ir.Emit(opcode, {lhs, rhs})), scalar, cmpx);
const bool signed_64 = opcode == IR::ValueOpcode::SLessThan64 ||
opcode == IR::ValueOpcode::SLessThanEqual64 ||
inst.opcode == Decoder::Opcode::V_CMP_EQ_I64 ||
inst.opcode == Decoder::Opcode::V_CMPX_NE_I64;
const auto read = [&](const Decoder::Operand& operand) {
// RDNA2 expands signed 64-bit integer literals by sign extension.
if (signed_64 && operand.kind == Decoder::OperandKind::LiteralConstant) {
return IR::Value(static_cast<uint64_t>(
static_cast<int64_t>(std::bit_cast<int32_t>(operand.value))));
}
return ReadOperand(operand, type);
};
EmitCompareResult(inst, IR::U1(ir.Emit(opcode, {read(inst.src0), read(inst.src1)})), scalar,
cmpx);
}
void Translator::EmitInteger16Compare(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
src/graphics/shader/recompiler/frontend/translate/Memory.cpp
+8
View File
@@ -966,6 +966,8 @@ void Translator::EmitMemory(const Decoder::Instruction& inst) {
return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicUMax32);
case Decoder::Opcode::BUFFER_ATOMIC_AND:
return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicAnd32);
case Decoder::Opcode::BUFFER_ATOMIC_AND_X2:
return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicAnd64);
case Decoder::Opcode::BUFFER_ATOMIC_OR:
return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicOr32);
case Decoder::Opcode::BUFFER_ATOMIC_OR_X2:
@@ -979,6 +981,8 @@ void Translator::EmitMemory(const Decoder::Instruction& inst) {
case Decoder::Opcode::DS_ADD_U32:
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicIAdd32, false);
case Decoder::Opcode::DS_ADD_U64:
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicIAdd64, false);
case Decoder::Opcode::DS_ADD_RTN_U32:
return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicIAdd32, true);
case Decoder::Opcode::DS_SUB_U32:
@@ -1030,8 +1034,12 @@ void Translator::EmitMemory(const Decoder::Instruction& inst) {
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicSwap32);
case Decoder::Opcode::IMAGE_ATOMIC_ADD:
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicIAdd32);
case Decoder::Opcode::IMAGE_ATOMIC_SMIN:
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicSMin32);
case Decoder::Opcode::IMAGE_ATOMIC_UMIN:
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicUMin32);
case Decoder::Opcode::IMAGE_ATOMIC_SMAX:
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicSMax32);
case Decoder::Opcode::IMAGE_ATOMIC_UMAX:
return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicUMax32);
case Decoder::Opcode::IMAGE_ATOMIC_AND:
src/graphics/shader/recompiler/frontend/translate/Scalar.cpp
+6
View File
@@ -155,6 +155,12 @@ void Translator::EmitScalar(const Decoder::Instruction& inst) {
case O::S_BREV_B32:
return SimpleInteger(inst, IR::ValueOpcode::BitReverse32, IR::Type::U32, false, false,
false);
case O::S_BREV_B64: {
const auto value = ReadU32Pair(inst.src0);
WriteU32Pair(inst.dst, {IR::U32(ir.Emit(IR::ValueOpcode::BitReverse32, {value[1]})),
IR::U32(ir.Emit(IR::ValueOpcode::BitReverse32, {value[0]}))});
return;
}
case O::S_BCNT1_I32_B32:
return SimpleInteger(inst, IR::ValueOpcode::BitCount32, IR::Type::U32, false, false,
true);
src/graphics/shader/recompiler/frontend/translate/Translate.cpp
+18
View File
@@ -894,10 +894,26 @@ void IncludeInstructionVectorRegisters(const Decoder::Instruction& inst, uint32_
case Decoder::Opcode::V_CVT_F64_F32:
case Decoder::Opcode::V_CVT_F64_U32: include_vector(inst.dst, 2u); break;
case Decoder::Opcode::V_FMA_F64: include_vector(inst.src2, 2u); [[fallthrough]];
case Decoder::Opcode::V_ADD_F64:
case Decoder::Opcode::V_MUL_F64: include_vector(inst.src1, 2u); [[fallthrough]];
case Decoder::Opcode::V_RCP_F64: include_vector(inst.dst, 2u); [[fallthrough]];
case Decoder::Opcode::V_CVT_F32_F64: include_vector(inst.src0, 2u); break;
case Decoder::Opcode::V_CMP_EQ_F64:
case Decoder::Opcode::V_CMP_LE_F64:
case Decoder::Opcode::V_CMPX_LE_F64:
case Decoder::Opcode::V_CMPX_GE_F64:
case Decoder::Opcode::V_CMP_EQ_I64:
case Decoder::Opcode::V_CMP_LT_I64:
case Decoder::Opcode::V_CMP_LE_I64:
case Decoder::Opcode::V_CMP_LT_U64:
case Decoder::Opcode::V_CMP_EQ_U64:
case Decoder::Opcode::V_CMP_LE_U64:
case Decoder::Opcode::V_CMP_GT_U64:
case Decoder::Opcode::V_CMP_NE_U64:
case Decoder::Opcode::V_CMP_GE_U64:
case Decoder::Opcode::V_CMPX_NE_I64:
case Decoder::Opcode::V_CMPX_LE_U64:
case Decoder::Opcode::V_CMPX_NE_U64:
include_vector(inst.src0, 2u);
include_vector(inst.src1, 2u);
break;
@@ -905,6 +921,8 @@ void IncludeInstructionVectorRegisters(const Decoder::Instruction& inst, uint32_
}
if (inst.family == Decoder::Family::DS) {
switch (inst.opcode) {
case Decoder::Opcode::DS_ADD_U64:
case Decoder::Opcode::DS_OR_B64:
case Decoder::Opcode::DS_WRITE_B64:
case Decoder::Opcode::DS_WRITE_B96:
case Decoder::Opcode::DS_WRITE_B128: include_vector(inst.src1, inst.data_dwords); break;
src/graphics/shader/recompiler/frontend/translate/Vector.cpp
+44
View File
@@ -114,18 +114,40 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
EmitIntegerCompare(inst, IR::ValueOpcode::INotEqual64, IR::Type::U64, false, true);
return;
case O::V_CMP_LT_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThan64, IR::Type::U64, false, false);
return;
case O::V_CMP_LE_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual64, IR::Type::U64, false, false);
return;
case O::V_CMP_LE_U64:
EmitIntegerCompare(inst, IR::ValueOpcode::ULessThanEqual64, IR::Type::U64, false, false);
return;
case O::V_CMPX_LE_U64:
EmitIntegerCompare(inst, IR::ValueOpcode::ULessThanEqual64, IR::Type::U64, false, true);
return;
case O::V_CMP_GE_U64:
EmitIntegerCompare(inst, IR::ValueOpcode::UGreaterThanEqual64, IR::Type::U64, false, false);
return;
case O::V_CMP_EQ_U16:
EmitInteger16Compare(inst, IR::ValueOpcode::IEqual32, false, false);
return;
case O::V_CMP_EQ_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::IEqual32, true, false);
return;
case O::V_CMPX_EQ_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::IEqual32, true, true);
return;
case O::V_CMP_NE_U16:
EmitInteger16Compare(inst, IR::ValueOpcode::INotEqual32, false, false);
return;
case O::V_CMP_NE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::INotEqual32, true, false);
return;
case O::V_CMPX_NE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::INotEqual32, true, true);
return;
case O::V_CMP_GT_U16:
EmitInteger16Compare(inst, IR::ValueOpcode::UGreaterThan32, false, false);
return;
@@ -150,9 +172,15 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_CMP_GT_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SGreaterThan32, true, false);
return;
case O::V_CMPX_GT_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SGreaterThan32, true, true);
return;
case O::V_CMP_GE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SGreaterThanEqual32, true, false);
return;
case O::V_CMPX_GE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SGreaterThanEqual32, true, true);
return;
case O::V_CMP_LT_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SLessThan32, true, false);
return;
@@ -162,6 +190,9 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_CMP_LE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SLessThanEqual32, true, false);
return;
case O::V_CMPX_LE_I16:
EmitInteger16Compare(inst, IR::ValueOpcode::SLessThanEqual32, true, true);
return;
case O::V_CMP_F_F32: EmitCompareConstant(inst, false, false, false); return;
case O::V_CMP_TRU_F32: EmitCompareConstant(inst, true, false, false); return;
@@ -171,6 +202,15 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_CMP_EQ_F64:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdEqual64, false, false);
return;
case O::V_CMP_LE_F64:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdLessThanEqual64, false, false);
return;
case O::V_CMPX_LE_F64:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdLessThanEqual64, false, true);
return;
case O::V_CMPX_GE_F64:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdGreaterThanEqual64, false, true);
return;
case O::V_CMPX_EQ_F32:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdEqual32, false, true);
return;
@@ -273,6 +313,9 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_CMPX_LE_F16:
EmitFloatCompare(inst, IR::ValueOpcode::FPOrdLessThanEqual32, true, true);
return;
case O::V_CMP_NGE_F16:
EmitFloatCompare(inst, IR::ValueOpcode::FPUnordLessThan32, true, false);
return;
case O::V_CMP_NGT_F16:
EmitFloatCompare(inst, IR::ValueOpcode::FPUnordLessThanEqual32, true, false);
return;
@@ -424,6 +467,7 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_SUB_F32: return FloatBinary(inst, IR::ValueOpcode::FPSub32, false);
case O::V_SUBREV_F32: return FloatBinary(inst, IR::ValueOpcode::FPSub32, true);
case O::V_MUL_F32: return FloatBinary(inst, IR::ValueOpcode::FPMul32, false);
case O::V_ADD_F64: return FloatBinary(inst, IR::ValueOpcode::FPAdd64, false);
case O::V_MUL_F64: return FloatBinary(inst, IR::ValueOpcode::FPMul64, false);
case O::V_MIN_F32: return FloatBinary(inst, IR::ValueOpcode::FPMin32, false);
case O::V_MAX_F32: return FloatBinary(inst, IR::ValueOpcode::FPMax32, false);