raphiaura/KytyPS5
1
0
Fork 0
mirror of https://github.com/KytyPS5/KytyPS5.git synced 2026-10-20 10:56:58 +02:00
Code Issues Packages Projects Releases Wiki Activity

shader: complete signed 64-bit integer comparisons

Browse Source
This commit is contained in:
TheCruZ
2026-10-10 03:35:20 +02:00
parent 93452b24e3
commit d247d3d60c
8 changed files with 218 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.h
+6 -1
View File
@@ -487,11 +487,14 @@ enum class Opcode {
V_CMP_NE_U32,
V_CMP_GE_U32,
V_CMP_T_U32,
// Keep V_CMP_F_I64 to V_CMPX_T_U64 contiguous: VectorAluOps.cpp tests this range.
V_CMP_F_I64,
V_CMP_EQ_I64,
V_CMP_LT_I64,
V_CMP_EQ_I64,
V_CMP_LE_I64,
V_CMP_GT_I64,
V_CMP_NE_I64,
V_CMP_GE_I64,
V_CMP_T_I64,
V_CMP_F_U64,
V_CMP_LT_U64,
@@ -505,7 +508,9 @@ enum class Opcode {
V_CMPX_LT_I64,
V_CMPX_EQ_I64,
V_CMPX_LE_I64,
V_CMPX_GT_I64,
V_CMPX_NE_I64,
V_CMPX_GE_I64,
V_CMPX_T_I64,
V_CMPX_F_U64,
V_CMPX_LT_U64,
src/graphics/shader/recompiler/frontend/decode/VectorAluOps.cpp
+24 -20
View File
@@ -278,25 +278,26 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
{0xc1u, Opcode::V_CMP_LT_U32}, {0xc2u, Opcode::V_CMP_EQ_U32},
{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},
{0xa5u, Opcode::V_CMP_NE_I64, false},
{0xa0u, Opcode::V_CMP_F_I64, false}, {0xa7u, Opcode::V_CMP_T_I64, false},
{0xb0u, Opcode::V_CMPX_F_I64, false}, {0xb1u, Opcode::V_CMPX_LT_I64, false},
{0xb2u, Opcode::V_CMPX_EQ_I64, false}, {0xb3u, Opcode::V_CMPX_LE_I64, false},
{0xb7u, Opcode::V_CMPX_T_I64, false},
{0xb5u, Opcode::V_CMPX_NE_I64, false}, {0xd1u, Opcode::V_CMPX_LT_U32},
{0xc7u, Opcode::V_CMP_T_U32}, {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},
{0xe0u, Opcode::V_CMP_F_U64, false}, {0xe7u, Opcode::V_CMP_T_U64, false},
{0xf0u, Opcode::V_CMPX_F_U64, false}, {0xf1u, Opcode::V_CMPX_LT_U64, false},
{0xf2u, Opcode::V_CMPX_EQ_U64, false}, {0xf4u, Opcode::V_CMPX_GT_U64, false},
{0xf6u, Opcode::V_CMPX_GE_U64, false}, {0xf7u, Opcode::V_CMPX_T_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},
{0xd6u, Opcode::V_CMPX_GE_U32}, {0xa0u, Opcode::V_CMP_F_I64, false},
{0xa1u, Opcode::V_CMP_LT_I64, false}, {0xa2u, Opcode::V_CMP_EQ_I64, false},
{0xa3u, Opcode::V_CMP_LE_I64, false}, {0xa4u, Opcode::V_CMP_GT_I64, false},
{0xa5u, Opcode::V_CMP_NE_I64, false}, {0xa6u, Opcode::V_CMP_GE_I64, false},
{0xa7u, Opcode::V_CMP_T_I64, false}, {0xb0u, Opcode::V_CMPX_F_I64, false},
{0xb1u, Opcode::V_CMPX_LT_I64, false}, {0xb2u, Opcode::V_CMPX_EQ_I64, false},
{0xb3u, Opcode::V_CMPX_LE_I64, false}, {0xb4u, Opcode::V_CMPX_GT_I64, false},
{0xb5u, Opcode::V_CMPX_NE_I64, false}, {0xb6u, Opcode::V_CMPX_GE_I64, false},
{0xb7u, Opcode::V_CMPX_T_I64, false}, {0xe0u, Opcode::V_CMP_F_U64, false},
{0xe1u, Opcode::V_CMP_LT_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},
{0xe7u, Opcode::V_CMP_T_U64, false}, {0xf0u, Opcode::V_CMPX_F_U64, false},
{0xf1u, Opcode::V_CMPX_LT_U64, false}, {0xf2u, Opcode::V_CMPX_EQ_U64, false},
{0xf3u, Opcode::V_CMPX_LE_U64, false}, {0xf4u, Opcode::V_CMPX_GT_U64, false},
{0xf5u, Opcode::V_CMPX_NE_U64, false}, {0xf6u, Opcode::V_CMPX_GE_U64, false},
{0xf7u, Opcode::V_CMPX_T_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}, {0xe9u, Opcode::V_CMP_NGE_F16},
@@ -1245,9 +1246,10 @@ VopcSdwaFields DecodeVopcSdwaFields(uint32_t modifier) {
}
bool SupportsVopcSdwa(Opcode opcode) {
// F64 SDWA operand handling is not implemented.
return opcode != Opcode::UNSUPPORTED && opcode != Opcode::V_CMP_NE_I64 &&
(opcode < Opcode::V_CMP_F_F64 || opcode > Opcode::V_CMPX_TRU_F64);
// SDWA operand handling is not implemented for 64-bit compares.
return opcode != Opcode::UNSUPPORTED &&
(opcode < Opcode::V_CMP_F_F64 || opcode > Opcode::V_CMPX_TRU_F64) &&
(opcode < Opcode::V_CMP_F_I64 || opcode > Opcode::V_CMPX_T_U64);
}
void DecodeVopcSdwa(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
@@ -1628,7 +1630,9 @@ bool IsVopcCompareExec(Opcode opcode) {
case Opcode::V_CMPX_LT_I64:
case Opcode::V_CMPX_EQ_I64:
case Opcode::V_CMPX_LE_I64:
case Opcode::V_CMPX_GT_I64:
case Opcode::V_CMPX_NE_I64:
case Opcode::V_CMPX_GE_I64:
case Opcode::V_CMPX_T_I64:
case Opcode::V_CMPX_F_U64:
case Opcode::V_CMPX_LT_U64:
src/graphics/shader/recompiler/frontend/translate/Compare.cpp
+9 -3
View File
@@ -1,6 +1,7 @@
#include "graphics/shader/recompiler/frontend/translate/Translator.h"
#include <bit>
#include <utility>
namespace Libs::Graphics::ShaderRecompiler::Frontend {
@@ -27,7 +28,8 @@ 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) {
IR::Type type, bool scalar, bool cmpx, bool swap_operands) {
// The equality compares share their IR opcode with the unsigned ones.
const bool signed_64 = opcode == IR::ValueOpcode::SLessThan64 ||
opcode == IR::ValueOpcode::SLessThanEqual64 ||
inst.opcode == Decoder::Opcode::V_CMP_EQ_I64 ||
@@ -42,8 +44,12 @@ void Translator::EmitIntegerCompare(const Decoder::Instruction& inst, IR::ValueO
}
return ReadOperand(operand, type);
};
EmitCompareResult(inst, IR::U1(ir.Emit(opcode, {read(inst.src0), read(inst.src1)})), scalar,
cmpx);
auto lhs = read(inst.src0);
auto rhs = read(inst.src1);
if (swap_operands) {
std::swap(lhs, rhs);
}
EmitCompareResult(inst, IR::U1(ir.Emit(opcode, {lhs, rhs})), scalar, cmpx);
}
void Translator::EmitInteger16Compare(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
src/graphics/shader/recompiler/frontend/translate/Translate.cpp
+13 -1
View File
@@ -907,26 +907,38 @@ void IncludeInstructionVectorRegisters(const Decoder::Instruction& inst, uint32_
case Decoder::Opcode::V_CMPX_NEQ_F64:
case Decoder::Opcode::V_CMPX_NLT_F64:
case Decoder::Opcode::V_CMPX_TRU_F64:
case Decoder::Opcode::V_CMP_EQ_I64:
case Decoder::Opcode::V_CMP_F_I64:
case Decoder::Opcode::V_CMP_LT_I64:
case Decoder::Opcode::V_CMP_EQ_I64:
case Decoder::Opcode::V_CMP_LE_I64:
case Decoder::Opcode::V_CMP_GT_I64:
case Decoder::Opcode::V_CMP_NE_I64:
case Decoder::Opcode::V_CMP_GE_I64:
case Decoder::Opcode::V_CMP_T_I64:
case Decoder::Opcode::V_CMP_F_U64:
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_CMP_T_U64:
case Decoder::Opcode::V_CMPX_F_I64:
case Decoder::Opcode::V_CMPX_LT_I64:
case Decoder::Opcode::V_CMPX_EQ_I64:
case Decoder::Opcode::V_CMPX_LE_I64:
case Decoder::Opcode::V_CMPX_GT_I64:
case Decoder::Opcode::V_CMPX_NE_I64:
case Decoder::Opcode::V_CMPX_GE_I64:
case Decoder::Opcode::V_CMPX_T_I64:
case Decoder::Opcode::V_CMPX_F_U64:
case Decoder::Opcode::V_CMPX_LT_U64:
case Decoder::Opcode::V_CMPX_EQ_U64:
case Decoder::Opcode::V_CMPX_LE_U64:
case Decoder::Opcode::V_CMPX_GT_U64:
case Decoder::Opcode::V_CMPX_NE_U64:
case Decoder::Opcode::V_CMPX_GE_U64:
case Decoder::Opcode::V_CMPX_T_U64:
include_vector(inst.src0, 2u);
include_vector(inst.src1, 2u);
break;
src/graphics/shader/recompiler/frontend/translate/Translator.h
+1 -1
View File
@@ -122,7 +122,7 @@ private:
void EmitCompareResult(const Decoder::Instruction& inst, IR::U1 value, bool scalar, bool cmpx);
void EmitCompareConstant(const Decoder::Instruction& inst, bool value, bool scalar, bool cmpx);
void EmitIntegerCompare(const Decoder::Instruction& inst, IR::ValueOpcode opcode, IR::Type type,
bool scalar, bool cmpx);
bool scalar, bool cmpx, bool swap_operands = false);
void EmitInteger16Compare(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
bool signed_value, bool cmpx);
void EmitFloatCompare(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool half,
src/graphics/shader/recompiler/frontend/translate/Vector.cpp
+17
View File
@@ -145,6 +145,23 @@ void Translator::EmitVector(const Decoder::Instruction& inst) {
case O::V_CMPX_LE_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual64, IR::Type::U64, false, true);
return;
// The IR has no signed 64-bit greater-than: a > b is b < a, and a >= b is b <= a.
case O::V_CMP_GT_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThan64, IR::Type::U64, false, false,
true);
return;
case O::V_CMPX_GT_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThan64, IR::Type::U64, false, true,
true);
return;
case O::V_CMP_GE_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual64, IR::Type::U64, false, false,
true);
return;
case O::V_CMPX_GE_I64:
EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual64, IR::Type::U64, false, true,
true);
return;
case O::V_CMP_LE_U64:
EmitIntegerCompare(inst, IR::ValueOpcode::ULessThanEqual64, IR::Type::U64, false, false);
return;