mirrors/shadPS4
1
0
Fork 0
mirror of https://github.com/shadps4-emu/shadPS4.git synced 2025-01-19 13:08:27 +00:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Preserve flags on some patched instructions (#720)

Browse source 
* Preserve flags on some patched instructions

* Move flag saving to Save/RestoreContext
This commit is contained in:
Paris Oplopoios 2024-09-03 20:41:35 +03:00 committed by GitHub
parent f4ff0d493c
commit cfe49fbfba
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 16 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

25
src/core/cpu_patches.cpp
View file

@ -212,31 +212,38 @@ static void RestoreRegisters(Xbyak::CodeGenerator& c,
} }
/// Switches to the patch stack and stores all registers. /// Switches to the patch stack and stores all registers.
static void SaveContext(Xbyak::CodeGenerator& c) { static void SaveContext(Xbyak::CodeGenerator& c, bool save_flags = false) {
SaveStack(c); SaveStack(c);
for (int reg = Xbyak::Operand::RAX; reg <= Xbyak::Operand::R15; reg++) { for (int reg = Xbyak::Operand::RAX; reg <= Xbyak::Operand::R15; reg++) {
c.push(Xbyak::Reg64(reg)); c.push(Xbyak::Reg64(reg));
} }
for (int reg = 0; reg <= 7; reg++) { for (int reg = 0; reg <= 7; reg++) {
c.sub(rsp, 32); c.lea(rsp, ptr[rsp - 32]);
c.vmovdqu(ptr[rsp], Xbyak::Ymm(reg)); c.vmovdqu(ptr[rsp], Xbyak::Ymm(reg));
} }
if (save_flags) {
c.pushfq();
}
} }
/// Restores all registers and restores the original stack. /// Restores all registers and restores the original stack.
/// If the destination is a register, it is not restored to preserve the output. /// If the destination is a register, it is not restored to preserve the output.
static void RestoreContext(Xbyak::CodeGenerator& c, const Xbyak::Operand& dst) { static void RestoreContext(Xbyak::CodeGenerator& c, const Xbyak::Operand& dst,
bool restore_flags = false) {
if (restore_flags) {
c.popfq();
}
for (int reg = 7; reg >= 0; reg--) { for (int reg = 7; reg >= 0; reg--) {
if ((!dst.isXMM() && !dst.isYMM()) || dst.getIdx() != reg) { if ((!dst.isXMM() && !dst.isYMM()) || dst.getIdx() != reg) {
c.vmovdqu(Xbyak::Ymm(reg), ptr[rsp]); c.vmovdqu(Xbyak::Ymm(reg), ptr[rsp]);
} }
c.add(rsp, 32); c.lea(rsp, ptr[rsp + 32]);
} }
for (int reg = Xbyak::Operand::R15; reg >= Xbyak::Operand::RAX; reg--) { for (int reg = Xbyak::Operand::R15; reg >= Xbyak::Operand::RAX; reg--) {
if (!dst.isREG() || dst.getIdx() != reg) { if (!dst.isREG() || dst.getIdx() != reg) {
c.pop(Xbyak::Reg64(reg)); c.pop(Xbyak::Reg64(reg));
} else { } else {
c.add(rsp, 8); c.lea(rsp, ptr[rsp + 8]);
} }
} }
RestoreStack(c); RestoreStack(c);
@ -376,7 +383,7 @@ static void GenerateVCVTPH2PS(const ZydisDecodedOperand* operands, Xbyak::CodeGe
const auto float_count = dst.getBit() / 32; const auto float_count = dst.getBit() / 32;
const auto byte_count = float_count * 4; const auto byte_count = float_count * 4;
SaveContext(c); SaveContext(c, true);
// Allocate stack space for outputs and load into first parameter. // Allocate stack space for outputs and load into first parameter.
c.sub(rsp, byte_count); c.sub(rsp, byte_count);
@ -412,7 +419,7 @@ static void GenerateVCVTPH2PS(const ZydisDecodedOperand* operands, Xbyak::CodeGe
} }
c.add(rsp, byte_count); c.add(rsp, byte_count);
RestoreContext(c, dst); RestoreContext(c, dst, true);
} }
using SingleToHalfFloatConverter = half_float::half (*)(float); using SingleToHalfFloatConverter = half_float::half (*)(float);
@ -440,7 +447,7 @@ static void GenerateVCVTPS2PH(const ZydisDecodedOperand* operands, Xbyak::CodeGe
const auto float_count = src.getBit() / 32; const auto float_count = src.getBit() / 32;
const auto byte_count = float_count * 4; const auto byte_count = float_count * 4;
SaveContext(c); SaveContext(c, true);
if (dst->isXMM()) { if (dst->isXMM()) {
// Allocate stack space for outputs and load into first parameter. // Allocate stack space for outputs and load into first parameter.
@ -487,7 +494,7 @@ static void GenerateVCVTPS2PH(const ZydisDecodedOperand* operands, Xbyak::CodeGe
c.add(rsp, byte_count); c.add(rsp, byte_count);
} }
RestoreContext(c, *dst); RestoreContext(c, *dst, true);
} }
static bool FilterRosetta2Only(const ZydisDecodedOperand*) { static bool FilterRosetta2Only(const ZydisDecodedOperand*) {