Merge branch 'misc-speedups'

* misc-speedups:
  fixed and reenabled and slightly optimized the JIT version of fcmpo/fcmpu.
  slightly more precise speed percent display (this is really minor)
  a small thread synchronization speedup for dual core mode. it's most noticeable in games where the CPU is running behind compared to the GPU.

Conflicts:
	Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp

The Fifo.cpp changes from rdaefb3b550e2 was not merged as there was no performance benefit.

Source/Core/Common/Src/StdThread.h

@@ -281,7 +281,7 @@ namespace this_thread
 inline void yield()
 {
 #ifdef _WIN32
-	Sleep(0);
+	SwitchToThread();
 #else
 	sleep(0);
 #endif

Source/Core/Core/Src/Core.cpp

@@ -585,7 +585,7 @@ void VideoThrottle()
 
 		u32 FPS = Common::AtomicLoad(DrawnFrame) * 1000 / ElapseTime;
 		u32 VPS = DrawnVideo * 1000 / ElapseTime;
-		u32 Speed = VPS * 100 / VideoInterface::TargetRefreshRate;
+		u32 Speed = DrawnVideo * (100 * 1000) / (VideoInterface::TargetRefreshRate * ElapseTime);
 		
 		// Settings are shown the same for both extended and summary info
 		std::string SSettings = StringFromFormat("%s %s", cpu_core_base->GetName(),	_CoreParameter.bCPUThread ? "DC" : "SC");

Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp

@@ -191,8 +191,7 @@ void Jit64::Init()
 		else
 			jo.enableBlocklink = !Core::g_CoreStartupParameter.bMMU;
 	}
-
+	jo.fpAccurateFcmp = Core::g_CoreStartupParameter.bEnableFPRF;
-	jo.fpAccurateFcmp = true; // Fallback to Interpreter
 	jo.optimizeGatherPipe = true;
 	jo.fastInterrupts = false;
 	jo.accurateSinglePrecision = true;

Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp

@@ -229,8 +229,6 @@ void Jit64::fmrx(UGeckoInstruction inst)
  
 void Jit64::fcmpx(UGeckoInstruction inst)
 {
-	// TODO : This still causes crashes in Nights, and broken graphics
-	// in Paper Mario, Super Paper Mario as well as SoulCalibur 2 prolly others too.. :(
 	INSTRUCTION_START
 	JITDISABLE(FloatingPoint)
 	if (jo.fpAccurateFcmp) {
@@ -243,36 +241,59 @@ void Jit64::fcmpx(UGeckoInstruction inst)
 	int crf	= inst.CRFD;
 
 	fpr.Lock(a,b);
-	if (a != b) fpr.BindToRegister(a, true);
+	fpr.BindToRegister(b, true);
 
 	// Are we masking sNaN invalid floating point exceptions? If not this could crash if we don't handle the exception?
-	UCOMISD(fpr.R(a).GetSimpleReg(), fpr.R(b));
+	UCOMISD(fpr.R(b).GetSimpleReg(), fpr.R(a));
 
-	FixupBranch pNaN	 = J_CC(CC_P);
+	FixupBranch pNaN, pLesser, pGreater;
-	FixupBranch pLesser  = J_CC(CC_B);
+	FixupBranch continue1, continue2, continue3;
-	FixupBranch pGreater = J_CC(CC_A);
+
+	if (a != b)
+	{
+		// if B > A, goto Lesser's jump target
+		pLesser  = J_CC(CC_A);
+	}
+
+	// if (B != B) or (A != A), goto NaN's jump target
+	pNaN    	 = J_CC(CC_P);
+
+	if (a != b)
+	{
+		// if B < A, goto Greater's jump target
+		// JB can't precede the NaN check because it doesn't test ZF
+		pGreater = J_CC(CC_B);
+	}
 
 	// Equal
 	MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x2));
-	FixupBranch continue1 = J();
+	continue1 = J();
-	
+
-	// Greater Than
-	SetJumpTarget(pGreater);
-	MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x4));
-	FixupBranch continue2 = J();
-	
-	// Less Than
-	SetJumpTarget(pLesser);
-	MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x8));
-	FixupBranch continue3 = J();
-	
 	// NAN
 	SetJumpTarget(pNaN);
 	MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x1));
+
+	if (a != b)
+	{
+		continue2 = J();
+
+		// Greater Than
+		SetJumpTarget(pGreater);
+		MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x4));
+		continue3 = J();
+	
+		// Less Than
+		SetJumpTarget(pLesser);
+		MOV(8, M(&PowerPC::ppcState.cr_fast[crf]), Imm8(0x8));
+	}
 	
 	SetJumpTarget(continue1);
-	SetJumpTarget(continue2);
+	if (a != b)
-	SetJumpTarget(continue3);
+	{
+		SetJumpTarget(continue2);
+		SetJumpTarget(continue3);
+	}
+
 	fpr.UnlockAll();
 }