Fixed the safe write path of the stfd instruction in the JIT. Fixes issue 4001.

This commit is contained in:
skidau 2012-01-03 01:17:52 +11:00
parent 6b0d21efa7
commit 67e38fb6c6
1 changed files with 41 additions and 58 deletions

View File

@ -172,80 +172,63 @@ void Jit64::stfd(UGeckoInstruction inst)
int s = inst.RS;
int a = inst.RA;
if (!a)
{
Default(inst);
return;
u32 mem_mask = Memory::ADDR_MASK_HW_ACCESS;
if (Core::g_CoreStartupParameter.bMMU ||
Core::g_CoreStartupParameter.iTLBHack) {
mem_mask |= Memory::ADDR_MASK_MEM1;
}
s32 offset = (s32)(s16)inst.SIMM_16;
#ifdef ENABLE_MEM_CHECK
if (Core::g_CoreStartupParameter.bEnableDebugging)
{
mem_mask |= Memory::EXRAM_MASK;
}
#endif
gpr.FlushLockX(ABI_PARAM1);
gpr.Lock(a);
fpr.Lock(s);
gpr.BindToRegister(a, true, false);
s32 offset = (s32)(s16)inst.SIMM_16;
LEA(32, ABI_PARAM1, MDisp(gpr.R(a).GetSimpleReg(), offset));
TEST(32, R(ABI_PARAM1), Imm32(0x0c000000));
FixupBranch not_ram = J_CC(CC_Z);
// Assume that any hardware writes using this instruction will go to the FIFO.
// Star Wars - The Force Unleashed uses this trick.
TEST(32, R(ABI_PARAM1), Imm32(mem_mask));
FixupBranch safe = J_CC(CC_NZ);
// Fast routine
if (cpu_info.bSSSE3) {
MOVAPD(XMM0, fpr.R(s));
PSHUFB(XMM0, M((void *)bswapShuffle1x8));
CALL(asm_routines.fifoDirectWriteXmm64);
} else {
// This ain't working yet
MOVAPD(XMM0, fpr.R(s));
MOVQ_xmm(M(&temp64), XMM0);
MOV(32, R(EAX), M(&temp64));
MOV(32, R(ABI_PARAM1), M((void*)((u8 *)&temp64 + 4)));
BSWAP(32, EAX);
BSWAP(32, ABI_PARAM1);
MOV(32, M(((u8 *)&temp64) + 4), R(EAX));
MOV(32, M((u8 *)&temp64), R(ABI_PARAM1));
MOVQ_xmm(XMM0, M(&temp64));
CALL(asm_routines.fifoDirectWriteXmm64);
}
FixupBranch quit = J(false);
SetJumpTarget(not_ram);
#ifdef _M_IX86
AND(32, R(ABI_PARAM1), Imm32(Memory::MEMVIEW32_MASK));
#endif
if (cpu_info.bSSSE3) {
MOVAPD(XMM0, fpr.R(s));
PSHUFB(XMM0, M((void *)bswapShuffle1x8));
PSHUFB(XMM0, M((void*)bswapShuffle1x8));
#ifdef _M_X64
MOVQ_xmm(MComplex(RBX, ABI_PARAM1, SCALE_1, 0), XMM0);
#else
AND(32, R(ECX), Imm32(Memory::MEMVIEW32_MASK));
MOVQ_xmm(MDisp(ABI_PARAM1, (u32)Memory::base), XMM0);
#endif
} else {
#ifdef _M_X64
fpr.BindToRegister(s, true, false);
MOVSD(M(&temp64), fpr.RX(s));
MOVAPD(XMM0, fpr.R(s));
MOVD_xmm(R(EAX), XMM0);
UnsafeWriteRegToReg(EAX, ABI_PARAM1, 32, 4);
MEMCHECK_START
MOV(64, R(EAX), M(&temp64));
BSWAP(64, EAX);
MOV(64, MComplex(RBX, ABI_PARAM1, SCALE_1, 0), R(EAX));
MEMCHECK_END
#else
fpr.BindToRegister(s, true, false);
MOVSD(M(&temp64), fpr.RX(s));
MEMCHECK_START
MOV(32, R(EAX), M(&temp64));
BSWAP(32, EAX);
MOV(32, MDisp(ABI_PARAM1, (u32)Memory::base + 4), R(EAX));
MOV(32, R(EAX), M((void*)((u8 *)&temp64 + 4)));
BSWAP(32, EAX);
MOV(32, MDisp(ABI_PARAM1, (u32)Memory::base), R(EAX));
MEMCHECK_END
#endif
PSRLQ(XMM0, 32);
MOVD_xmm(R(EAX), XMM0);
UnsafeWriteRegToReg(EAX, ABI_PARAM1, 32, 0);
}
SetJumpTarget(quit);
FixupBranch exit = J();
SetJumpTarget(safe);
// Safe but slow routine
MOVAPD(XMM0, fpr.R(s));
MOVD_xmm(R(EAX), XMM0);
SafeWriteRegToReg(EAX, ABI_PARAM1, 32, 4);
PSRLQ(XMM0, 32);
MOVD_xmm(R(EAX), XMM0);
LEA(32, ABI_PARAM1, MDisp(gpr.R(a).GetSimpleReg(), offset));
SafeWriteRegToReg(EAX, ABI_PARAM1, 32, 0);
SetJumpTarget(exit);
gpr.UnlockAll();
gpr.UnlockAllX();
fpr.UnlockAll();