raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-21 11:26:56 +02:00
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/decode/ImageOps.cpp
+2
View File
@@ -193,7 +193,9 @@ constexpr MimgGatherInfo MIMG_GATHER_OPCODE_LIST[] = {
constexpr Detail::OpcodeMap MIMG_ATOMIC_OPCODE_LIST[] = {
{0x0fu, Opcode::IMAGE_ATOMIC_SWAP},
{0x11u, Opcode::IMAGE_ATOMIC_ADD},
{0x14u, Opcode::IMAGE_ATOMIC_SMIN},
{0x15u, Opcode::IMAGE_ATOMIC_UMIN},
{0x16u, Opcode::IMAGE_ATOMIC_SMAX},
{0x17u, Opcode::IMAGE_ATOMIC_UMAX},
{0x18u, Opcode::IMAGE_ATOMIC_AND},
{0x19u, Opcode::IMAGE_ATOMIC_OR},
src/graphics/shader/recompiler/frontend/decode/MemoryOps.cpp
+3 -1
View File
@@ -61,6 +61,7 @@ constexpr MemoryOpcodeInfo MUBUF_OPCODE_LIST[] = {
{0x3fu, Opcode::BUFFER_ATOMIC_FMIN, 1, 32},
{0x40u, Opcode::BUFFER_ATOMIC_FMAX, 1, 32},
{0x50u, Opcode::BUFFER_ATOMIC_SWAP_X2, 2, 32},
{0x59u, Opcode::BUFFER_ATOMIC_AND_X2, 2, 32},
{0x5au, Opcode::BUFFER_ATOMIC_OR_X2, 2, 32},
};
@@ -106,7 +107,7 @@ constexpr MemoryOpcodeInfo DS_OPCODE_LIST[] = {
{0x39u, Opcode::DS_READ_I8, 1, 8, true}, {0x3au, Opcode::DS_READ_U8, 1, 8},
{0x3bu, Opcode::DS_READ_I16, 1, 16, true}, {0x3cu, Opcode::DS_READ_U16, 1, 16},
{0x3du, Opcode::DS_CONSUME, 1, 32}, {0x3eu, Opcode::DS_APPEND, 1, 32},
{0x4au, Opcode::DS_OR_B64, 2, 32},
{0x40u, Opcode::DS_ADD_U64, 2, 32}, {0x4au, Opcode::DS_OR_B64, 2, 32},
{0x4du, Opcode::DS_WRITE_B64, 2, 32}, {0x4eu, Opcode::DS_WRITE2_B64, 4, 32},
{0x4fu, Opcode::DS_WRITE2ST64_B64, 4, 32}, {0x76u, Opcode::DS_READ_B64, 2, 32},
{0x77u, Opcode::DS_READ2_B64, 4, 32}, {0x78u, Opcode::DS_READ2ST64_B64, 4, 32},
@@ -168,6 +169,7 @@ bool IsDsWriteOpcode(Opcode opcode) {
bool IsDsAtomicOpcode(Opcode opcode) {
switch (opcode) {
case Opcode::DS_ADD_U32:
case Opcode::DS_ADD_U64:
case Opcode::DS_ADD_RTN_U32:
case Opcode::DS_SUB_U32:
case Opcode::DS_SUB_RTN_U32:
src/graphics/shader/recompiler/frontend/decode/ScalarAluOps.cpp
+1
View File
@@ -46,6 +46,7 @@ constexpr OpcodeMap SOP1_OPCODE_LIST[] = {
{0x09u, Opcode::S_WQM_B32},
{0x0au, Opcode::S_WQM_B64},
{0x0bu, Opcode::S_BREV_B32},
{0x0cu, Opcode::S_BREV_B64},
{0x0fu, Opcode::S_BCNT1_I32_B32},
{0x10u, Opcode::S_BCNT1_I32_B64},
{0x13u, Opcode::S_FF1_I32_B32},
src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.cpp
+5
View File
@@ -504,6 +504,7 @@ std::string InstructionToString(const Instruction& inst) {
OperandToString(inst.src0).c_str()));
case Opcode::S_ABS_I32:
case Opcode::S_BREV_B32:
case Opcode::S_BREV_B64:
case Opcode::S_BCNT1_I32_B32:
case Opcode::S_FLBIT_I32_B32:
case Opcode::S_FF1_I32_B32:
@@ -575,7 +576,9 @@ std::string InstructionToString(const Instruction& inst) {
case Opcode::IMAGE_STORE_MIP:
case Opcode::IMAGE_ATOMIC_SWAP:
case Opcode::IMAGE_ATOMIC_ADD:
case Opcode::IMAGE_ATOMIC_SMIN:
case Opcode::IMAGE_ATOMIC_UMIN:
case Opcode::IMAGE_ATOMIC_SMAX:
case Opcode::IMAGE_ATOMIC_UMAX:
case Opcode::IMAGE_ATOMIC_AND:
case Opcode::IMAGE_ATOMIC_OR:
@@ -642,6 +645,7 @@ std::string InstructionToString(const Instruction& inst) {
case Opcode::BUFFER_ATOMIC_SMAX:
case Opcode::BUFFER_ATOMIC_UMAX:
case Opcode::BUFFER_ATOMIC_AND:
case Opcode::BUFFER_ATOMIC_AND_X2:
case Opcode::BUFFER_ATOMIC_OR:
case Opcode::BUFFER_ATOMIC_OR_X2:
case Opcode::BUFFER_ATOMIC_XOR:
@@ -664,6 +668,7 @@ std::string InstructionToString(const Instruction& inst) {
case Opcode::FLAT_STORE_DWORDX3:
case Opcode::FLAT_STORE_DWORDX4:
case Opcode::DS_ADD_U32:
case Opcode::DS_ADD_U64:
case Opcode::DS_ADD_RTN_U32:
case Opcode::DS_SUB_U32:
case Opcode::DS_SUB_RTN_U32:
src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.h
+20
View File
@@ -47,6 +47,7 @@ enum class Opcode {
S_ABS_I32,
S_ABSDIFF_I32,
S_BREV_B32,
S_BREV_B64,
S_BCNT1_I32_B32,
S_BCNT1_I32_B64,
S_FF1_I32_B32,
@@ -244,6 +245,7 @@ enum class Opcode {
V_MAD_U32_U24,
V_MAD_U64_U32,
V_FMA_F64,
V_ADD_F64,
V_MUL_F64,
V_FMA_F32,
V_FMA_F16,
@@ -360,6 +362,9 @@ enum class Opcode {
V_CMP_NLT_F32,
V_CMP_TRU_F32,
V_CMP_EQ_F64,
V_CMP_LE_F64,
V_CMPX_LE_F64,
V_CMPX_GE_F64,
V_CMPX_LT_F32,
V_CMPX_EQ_F32,
V_CMPX_LE_F32,
@@ -390,12 +395,18 @@ enum class Opcode {
V_CMP_NE_I16,
V_CMP_GE_I16,
V_CMPX_LT_I16,
V_CMPX_EQ_I16,
V_CMPX_LE_I16,
V_CMPX_GT_I16,
V_CMPX_NE_I16,
V_CMPX_GE_I16,
V_CMP_LT_F16,
V_CMP_EQ_F16,
V_CMP_LE_F16,
V_CMP_GT_F16,
V_CMP_LG_F16,
V_CMP_GE_F16,
V_CMP_NGE_F16,
V_CMP_NGT_F16,
V_CMP_NEQ_F16,
V_CMP_NLT_F16,
@@ -432,11 +443,16 @@ enum class Opcode {
V_CMP_GE_U32,
V_CMP_T_U32,
V_CMP_EQ_I64,
V_CMP_LT_I64,
V_CMP_LE_I64,
V_CMP_LT_U64,
V_CMP_EQ_U64,
V_CMP_LE_U64,
V_CMP_GT_U64,
V_CMP_NE_U64,
V_CMP_GE_U64,
V_CMPX_NE_I64,
V_CMPX_LE_U64,
V_CMPX_NE_U64,
V_CMPX_LT_U32,
V_CMPX_EQ_U32,
@@ -496,6 +512,7 @@ enum class Opcode {
BUFFER_ATOMIC_SMAX,
BUFFER_ATOMIC_UMAX,
BUFFER_ATOMIC_AND,
BUFFER_ATOMIC_AND_X2,
BUFFER_ATOMIC_OR,
BUFFER_ATOMIC_OR_X2,
BUFFER_ATOMIC_XOR,
@@ -516,6 +533,7 @@ enum class Opcode {
FLAT_STORE_DWORDX3,
FLAT_STORE_DWORDX4,
DS_ADD_U32,
DS_ADD_U64,
DS_ADD_RTN_U32,
DS_SUB_U32,
DS_SUB_RTN_U32,
@@ -582,7 +600,9 @@ enum class Opcode {
IMAGE_STORE_MIP,
IMAGE_ATOMIC_SWAP,
IMAGE_ATOMIC_ADD,
IMAGE_ATOMIC_SMIN,
IMAGE_ATOMIC_UMIN,
IMAGE_ATOMIC_SMAX,
IMAGE_ATOMIC_UMAX,
IMAGE_ATOMIC_AND,
IMAGE_ATOMIC_OR,
src/graphics/shader/recompiler/frontend/decode/VectorAluOps.cpp
+29 -6
View File
@@ -228,7 +228,8 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
{0x19u, Opcode::V_CMPX_NGE_F32}, {0x1au, Opcode::V_CMPX_NLG_F32},
{0x1bu, Opcode::V_CMPX_NGT_F32}, {0x1cu, Opcode::V_CMPX_NLE_F32},
{0x1du, Opcode::V_CMPX_NEQ_F32}, {0x1eu, Opcode::V_CMPX_NLT_F32},
{0x22u, Opcode::V_CMP_EQ_F64, false},
{0x22u, Opcode::V_CMP_EQ_F64, false}, {0x23u, Opcode::V_CMP_LE_F64, false},
{0x33u, Opcode::V_CMPX_LE_F64, false}, {0x36u, Opcode::V_CMPX_GE_F64, false},
{0x80u, Opcode::V_CMP_F_I32}, {0x81u, Opcode::V_CMP_LT_I32},
{0x82u, Opcode::V_CMP_EQ_I32}, {0x83u, Opcode::V_CMP_LE_I32},
{0x84u, Opcode::V_CMP_GT_I32}, {0x85u, Opcode::V_CMP_NE_I32},
@@ -240,7 +241,9 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
{0x92u, Opcode::V_CMPX_EQ_I32}, {0x93u, Opcode::V_CMPX_LE_I32},
{0x94u, Opcode::V_CMPX_GT_I32}, {0x95u, Opcode::V_CMPX_NE_I32},
{0x96u, Opcode::V_CMPX_GE_I32}, {0x98u, Opcode::V_CMPX_CLASS_F32},
{0x99u, Opcode::V_CMPX_LT_I16, false},
{0x99u, Opcode::V_CMPX_LT_I16, false}, {0x9au, Opcode::V_CMPX_EQ_I16, false},
{0x9bu, Opcode::V_CMPX_LE_I16, false}, {0x9cu, Opcode::V_CMPX_GT_I16, false},
{0x9du, Opcode::V_CMPX_NE_I16, false}, {0x9eu, Opcode::V_CMPX_GE_I16, false},
{0xa9u, Opcode::V_CMP_LT_U16}, {0xaau, Opcode::V_CMP_EQ_U16},
{0xabu, Opcode::V_CMP_LE_U16}, {0xacu, Opcode::V_CMP_GT_U16},
{0xadu, Opcode::V_CMP_NE_U16}, {0xaeu, Opcode::V_CMP_GE_U16},
@@ -250,16 +253,19 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
{0xc3u, Opcode::V_CMP_LE_U32}, {0xc4u, Opcode::V_CMP_GT_U32},
{0xc5u, Opcode::V_CMP_NE_U32}, {0xc6u, Opcode::V_CMP_GE_U32},
{0xc7u, Opcode::V_CMP_T_U32}, {0xa2u, Opcode::V_CMP_EQ_I64, false},
{0xa1u, Opcode::V_CMP_LT_I64, false}, {0xa3u, Opcode::V_CMP_LE_I64, false},
{0xb5u, Opcode::V_CMPX_NE_I64, false}, {0xd1u, Opcode::V_CMPX_LT_U32},
{0xd2u, Opcode::V_CMPX_EQ_U32}, {0xd3u, Opcode::V_CMPX_LE_U32},
{0xd4u, Opcode::V_CMPX_GT_U32}, {0xd5u, Opcode::V_CMPX_NE_U32},
{0xd6u, Opcode::V_CMPX_GE_U32}, {0xe1u, Opcode::V_CMP_LT_U64, false},
{0xe2u, Opcode::V_CMP_EQ_U64, false}, {0xe4u, Opcode::V_CMP_GT_U64, false},
{0xe5u, Opcode::V_CMP_NE_U64, false},
{0xe2u, Opcode::V_CMP_EQ_U64, false}, {0xe3u, Opcode::V_CMP_LE_U64, false},
{0xe4u, Opcode::V_CMP_GT_U64, false}, {0xe5u, Opcode::V_CMP_NE_U64, false},
{0xe6u, Opcode::V_CMP_GE_U64, false}, {0xf3u, Opcode::V_CMPX_LE_U64, false},
{0xf5u, Opcode::V_CMPX_NE_U64, false}, {0xc9u, Opcode::V_CMP_LT_F16},
{0xcau, Opcode::V_CMP_EQ_F16}, {0xcbu, Opcode::V_CMP_LE_F16},
{0xccu, Opcode::V_CMP_GT_F16}, {0xcdu, Opcode::V_CMP_LG_F16},
{0xceu, Opcode::V_CMP_GE_F16}, {0xebu, Opcode::V_CMP_NGT_F16},
{0xceu, Opcode::V_CMP_GE_F16}, {0xe9u, Opcode::V_CMP_NGE_F16},
{0xebu, Opcode::V_CMP_NGT_F16},
{0xedu, Opcode::V_CMP_NEQ_F16}, {0xeeu, Opcode::V_CMP_NLT_F16},
{0xd9u, Opcode::V_CMPX_LT_F16}, {0xdau, Opcode::V_CMPX_EQ_F16},
{0xdbu, Opcode::V_CMPX_LE_F16}, {0xdcu, Opcode::V_CMPX_GT_F16},
@@ -281,6 +287,7 @@ constexpr OpcodeMap VOP3_OPCODE_LIST[] = {
{0x147u, Opcode::V_CUBEMA_F32},
{0x14bu, Opcode::V_FMA_F32},
{0x14cu, Opcode::V_FMA_F64},
{0x164u, Opcode::V_ADD_F64},
{0x165u, Opcode::V_MUL_F64},
{0x148u, Opcode::V_BFE_U32},
{0x149u, Opcode::V_BFE_I32},
@@ -831,6 +838,9 @@ bool IsVopcFloatCompareOpcode(Opcode opcode) {
case Opcode::V_CMP_LT_F32:
case Opcode::V_CMP_EQ_F32:
case Opcode::V_CMP_EQ_F64:
case Opcode::V_CMP_LE_F64:
case Opcode::V_CMPX_LE_F64:
case Opcode::V_CMPX_GE_F64:
case Opcode::V_CMP_LE_F32:
case Opcode::V_CMP_GT_F32:
case Opcode::V_CMP_LG_F32:
@@ -863,6 +873,7 @@ bool IsVopcFloatCompareOpcode(Opcode opcode) {
case Opcode::V_CMP_GT_F16:
case Opcode::V_CMP_LG_F16:
case Opcode::V_CMP_GE_F16:
case Opcode::V_CMP_NGE_F16:
case Opcode::V_CMP_NGT_F16:
case Opcode::V_CMP_NEQ_F16:
case Opcode::V_CMP_NLT_F16:
@@ -1141,7 +1152,9 @@ VopcSdwaFields DecodeVopcSdwaFields(uint32_t modifier) {
}
bool SupportsVopcSdwa(Opcode opcode) {
return opcode != Opcode::UNSUPPORTED && opcode != Opcode::V_CMP_EQ_F64;
return opcode != Opcode::UNSUPPORTED && opcode != Opcode::V_CMP_EQ_F64 &&
opcode != Opcode::V_CMP_LE_F64 && opcode != Opcode::V_CMPX_LE_F64 &&
opcode != Opcode::V_CMPX_GE_F64;
}
void DecodeVopcSdwa(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
@@ -1206,6 +1219,7 @@ void DecodeVopcDpp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_in
uint32_t NativeVop3SourceCount(Opcode opcode) {
switch (opcode) {
case Opcode::V_ADD_F64:
case Opcode::V_MUL_F64:
case Opcode::V_MUL_LO_U32:
case Opcode::V_MUL_HI_U32:
@@ -1357,6 +1371,7 @@ bool SupportsNativeVop3SourceModifiers(Opcode opcode) {
case Opcode::V_MAX_F32:
case Opcode::V_MAC_F32:
case Opcode::V_MAD_F32:
case Opcode::V_ADD_F64:
case Opcode::V_MUL_F64:
case Opcode::V_FMA_F64:
case Opcode::V_FMA_F32:
@@ -1462,6 +1477,8 @@ void ApplyNativeVop3SourceModifiers(Instruction& inst, uint32_t abs, uint32_t ne
bool IsVopcCompareExec(Opcode opcode) {
switch (opcode) {
case Opcode::V_CMPX_LE_F64:
case Opcode::V_CMPX_GE_F64:
case Opcode::V_CMPX_LT_F32:
case Opcode::V_CMPX_EQ_F32:
case Opcode::V_CMPX_LE_F32:
@@ -1489,8 +1506,14 @@ bool IsVopcCompareExec(Opcode opcode) {
case Opcode::V_CMPX_NE_U32:
case Opcode::V_CMPX_GE_U32:
case Opcode::V_CMPX_NE_I64:
case Opcode::V_CMPX_LE_U64:
case Opcode::V_CMPX_NE_U64:
case Opcode::V_CMPX_LT_I16:
case Opcode::V_CMPX_EQ_I16:
case Opcode::V_CMPX_LE_I16:
case Opcode::V_CMPX_GT_I16:
case Opcode::V_CMPX_NE_I16:
case Opcode::V_CMPX_GE_I16:
case Opcode::V_CMPX_LT_U16:
case Opcode::V_CMPX_EQ_U16:
case Opcode::V_CMPX_GT_U16:
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);