Merge pull request #4518 from hrydgard/fpcond

ARMJIT: Cache fpcond in a register to avoid store/load between compare and branch
hrydgard · Nov 12, 2013 · f4b5e8a · f4b5e8a
2 parents 4d4a32f + 17074f5
commit f4b5e8a
Show file tree

Hide file tree

Showing 5 changed files with 26 additions and 29 deletions.
diff --git a/Core/MIPS/ARM/ArmCompBranch.cpp b/Core/MIPS/ARM/ArmCompBranch.cpp
@@ -287,8 +287,8 @@ void Jit::BranchFPFlag(MIPSOpcode op, ArmGen::CCFlags cc, bool likely)
 	if (!likely && delaySlotIsNice)
 		CompileDelaySlot(DELAYSLOT_NICE);
 
-	LDR(R0, CTXREG, offsetof(MIPSState, fpcond));
-	TST(R0, Operand2(1, TYPE_IMM));
+	gpr.MapReg(MIPS_REG_FPCOND, 0);
+	TST(gpr.R(MIPS_REG_FPCOND), Operand2(1, TYPE_IMM));
 
 	ArmGen::FixupBranch ptr;
 	if (!likely)

diff --git a/Core/MIPS/ARM/ArmCompFPU.cpp b/Core/MIPS/ARM/ArmCompFPU.cpp
@@ -192,76 +192,72 @@ void Jit::Comp_FPUComp(MIPSOpcode op) {
 	int opc = op & 0xF;
 	if (opc >= 8) opc -= 8; // alias
 	if (opc == 0) {  // f, sf (signalling false)
-		MOVI2R(R0, 0);
-		STR(R0, CTXREG, offsetof(MIPSState, fpcond));
+		gpr.SetImm(MIPS_REG_FPCOND, 0);
 		return;
 	}
 
 	int fs = _FS;
 	int ft = _FT;
+	gpr.MapReg(MIPS_REG_FPCOND, MAP_DIRTY | MAP_NOINIT);
 	fpr.MapInIn(fs, ft);
 	VCMP(fpr.R(fs), fpr.R(ft));
 	VMRS_APSR(); // Move FP flags from FPSCR to APSR (regular flags).
 	switch(opc)
 	{
 	case 1:      // un,  ngle (unordered)
 		SetCC(CC_VS);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_VC);
 		break;
 	case 2:      // eq,  seq (equal, ordered)
 		SetCC(CC_EQ);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_NEQ);
 		break;
 	case 3:      // ueq, ngl (equal, unordered)
 		SetCC(CC_EQ);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_NEQ);
-		MOVI2R(R0, 0);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 0);
 		SetCC(CC_VS);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_AL);
-		STR(R0, CTXREG, offsetof(MIPSState, fpcond));
 		return;
 	case 4:      // olt, lt (less than, ordered)
 		SetCC(CC_LO);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_HS);
 		break;
 	case 5:      // ult, nge (less than, unordered)
 		SetCC(CC_LT);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_GE);
 		break;
 	case 6:      // ole, le (less equal, ordered)
 		SetCC(CC_LS);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_HI);
 		break;
 	case 7:      // ule, ngt (less equal, unordered)
 		SetCC(CC_LE);
-		MOVI2R(R0, 1);
+		MOVI2R(gpr.R(MIPS_REG_FPCOND), 1);
 		SetCC(CC_GT);
 		break;
 	default:
 		Comp_Generic(op);
 		return;
 	}
-	MOVI2R(R0, 0);
+	MOVI2R(gpr.R(MIPS_REG_FPCOND), 0);
 	SetCC(CC_AL);
-	STR(R0, CTXREG, offsetof(MIPSState, fpcond));
 }
 
-void Jit::Comp_FPU2op(MIPSOpcode op)
-{
+void Jit::Comp_FPU2op(MIPSOpcode op) {
 	CONDITIONAL_DISABLE;
 
 	int fs = _FS;
 	int fd = _FD;
 
-	switch (op & 0x3f) 
-	{
+	switch (op & 0x3f) {
 	case 4:	//F(fd)	   = sqrtf(F(fs));            break; //sqrt
 		fpr.MapDirtyIn(fd, fs);
 		VSQRT(fpr.R(fd), fpr.R(fs));
@@ -345,13 +341,12 @@ void Jit::Comp_mxc1(MIPSOpcode op)
 	case 2: //cfc1
 		if (fs == 31)
 		{
-			gpr.MapReg(rt, MAP_DIRTY | MAP_NOINIT);
-			LDR(R0, CTXREG, offsetof(MIPSState, fpcond));
+			gpr.MapDirtyIn(rt, MIPS_REG_FPCOND);
 			LDR(gpr.R(rt), CTXREG, offsetof(MIPSState, fcr31));
 			if (cpu_info.bArmV7) {
-				BFI(gpr.R(rt), R0, 23, 1);
+				BFI(gpr.R(rt), gpr.R(MIPS_REG_FPCOND), 23, 1);
 			} else {
-				AND(R0, R0, Operand2(1)); // Just in case
+				AND(R0, gpr.R(MIPS_REG_FPCOND), Operand2(1)); // Just in case
 				ANDI2R(gpr.R(rt), gpr.R(rt), ~(0x1 << 23), R1);  // R1 won't be used, this turns into a simple BIC.
 				ORR(gpr.R(rt), gpr.R(rt), Operand2(R0, ST_LSL, 23));
 			}
@@ -372,7 +367,7 @@ void Jit::Comp_mxc1(MIPSOpcode op)
 	case 6: //ctc1
 		if (fs == 31)
 		{
-			gpr.MapReg(rt, 0);
+			gpr.MapDirtyIn(MIPS_REG_FPCOND, rt);
 			// Hardware rounding method.
 			// Left here in case it is faster than conditional method.
 			/*
@@ -393,12 +388,11 @@ void Jit::Comp_mxc1(MIPSOpcode op)
 			// Update MIPS state
 			STR(gpr.R(rt), CTXREG, offsetof(MIPSState, fcr31));
 			if (cpu_info.bArmV7) {
-				UBFX(R0, gpr.R(rt), 23, 1);
+				UBFX(gpr.R(MIPS_REG_FPCOND), gpr.R(rt), 23, 1);
 			} else {
 				MOV(R0, Operand2(gpr.R(rt), ST_LSR, 23));
-				AND(R0, R0, Operand2(1));
+				AND(gpr.R(MIPS_REG_FPCOND), R0, Operand2(1));
 			}
-			STR(R0, CTXREG, offsetof(MIPSState, fpcond));
 		}
 		return;
 	}

diff --git a/Core/MIPS/ARM/ArmRegCache.cpp b/Core/MIPS/ARM/ArmRegCache.cpp
@@ -569,6 +569,8 @@ int ArmRegCache::GetMipsRegOffset(MIPSGPReg r) {
 		return offsetof(MIPSState, hi);
 	case MIPS_REG_LO:
 		return offsetof(MIPSState, lo);
+	case MIPS_REG_FPCOND:
+		return offsetof(MIPSState, fpcond);
 	default:
 		ERROR_LOG_REPORT(JIT, "bad mips register %i", r);
 		return 0;  // or what?

diff --git a/Core/MIPS/ARM/ArmRegCache.h b/Core/MIPS/ARM/ArmRegCache.h
@@ -31,7 +31,7 @@ using namespace ArmGen;
 // R11 = base pointer
 
 enum {
-	TOTAL_MAPPABLE_MIPSREGS = 34,
+	TOTAL_MAPPABLE_MIPSREGS = 35,
 };
 
 typedef int MIPSReg;

diff --git a/Core/MIPS/MIPS.h b/Core/MIPS/MIPS.h
@@ -68,6 +68,7 @@ enum MIPSGPReg
 	// Not real regs, just for convenience/jit mapping.
 	MIPS_REG_HI = 32,
 	MIPS_REG_LO = 33,
+	MIPS_REG_FPCOND = 34,
 };
 
 enum