From f000bb759051c3ef7061b0df0c5d6d20e31d4cbd Mon Sep 17 00:00:00 2001
From: Shawn Hoffman <godisgovernment@gmail.com>
Date: Mon, 3 Oct 2016 07:15:26 -0700
Subject: [PATCH] Jit64: use domain-appropriate instructions Original commit:
 commit 62d3423ccf090da95b0caae1652e8fbb248c4b99 Author: Tillmann Karras
 <tilkax@gmail.com> Date:   Wed Nov 18 20:32:20 2015 +0100

---
 .../Core/PowerPC/Jit64/Jit_FloatingPoint.cpp  | 30 +++++++++----------
 .../Core/Core/PowerPC/Jit64/Jit_LoadStore.cpp |  2 +-
 2 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp b/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
index 4a50a0ca82..65339e162d 100644
--- a/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
+++ b/Source/Core/Core/PowerPC/Jit64/Jit_FloatingPoint.cpp
@@ -149,18 +149,18 @@ void Jit64::HandleNaNs(UGeckoInstruction inst, X64Reg xmm_out, X64Reg xmm, X64Re
       SwitchToFarCode();
       SetJumpTarget(handle_nan);
       MOVAPD(tmp, R(clobber));
-      PANDN(clobber, R(xmm));
-      PAND(tmp, M(psGeneratedQNaN));
-      POR(tmp, R(clobber));
+      ANDNPD(clobber, R(xmm));
+      ANDPD(tmp, M(psGeneratedQNaN));
+      ORPD(tmp, R(clobber));
       MOVAPD(xmm, R(tmp));
       for (u32 x : inputs)
       {
         MOVAPD(clobber, fpr.R(x));
         CMPPD(clobber, R(clobber), CMP_ORD);
         MOVAPD(tmp, R(clobber));
-        PANDN(clobber, fpr.R(x));
-        PAND(xmm, R(tmp));
-        POR(xmm, R(clobber));
+        ANDNPD(clobber, fpr.R(x));
+        ANDPD(xmm, R(tmp));
+        ORPD(xmm, R(clobber));
       }
       FixupBranch done = J(true);
       SwitchToNearCode();
@@ -350,7 +350,7 @@ void Jit64::fmaddXX(UGeckoInstruction inst)
         ADDSD(XMM1, fpr.R(b));
     }
     if (inst.SUBOP5 == 31)  // nmadd
-      PXOR(XMM1, M(packed ? psSignBits2 : psSignBits));
+      XORPD(XMM1, M(packed ? psSignBits2 : psSignBits));
   }
   fpr.BindToRegister(d, !single);
   if (single)
@@ -384,15 +384,15 @@ void Jit64::fsign(UGeckoInstruction inst)
   switch (inst.SUBOP10)
   {
   case 40:  // neg
-    avx_op(&XEmitter::VPXOR, &XEmitter::PXOR, fpr.RX(d), src, M(packed ? psSignBits2 : psSignBits),
-           packed);
+    avx_op(&XEmitter::VXORPD, &XEmitter::XORPD, fpr.RX(d), src,
+           M(packed ? psSignBits2 : psSignBits), packed);
     break;
   case 136:  // nabs
-    avx_op(&XEmitter::VPOR, &XEmitter::POR, fpr.RX(d), src, M(packed ? psSignBits2 : psSignBits),
+    avx_op(&XEmitter::VORPD, &XEmitter::ORPD, fpr.RX(d), src, M(packed ? psSignBits2 : psSignBits),
            packed);
     break;
   case 264:  // abs
-    avx_op(&XEmitter::VPAND, &XEmitter::PAND, fpr.RX(d), src, M(packed ? psAbsMask2 : psAbsMask),
+    avx_op(&XEmitter::VANDPD, &XEmitter::ANDPD, fpr.RX(d), src, M(packed ? psAbsMask2 : psAbsMask),
            packed);
     break;
   default:
@@ -416,7 +416,7 @@ void Jit64::fselx(UGeckoInstruction inst)
   bool packed = inst.OPCD == 4;  // ps_sel
 
   fpr.Lock(a, b, c, d);
-  PXOR(XMM0, R(XMM0));
+  XORPD(XMM0, R(XMM0));
   // This condition is very tricky; there's only one right way to handle both the case of
   // negative/positive zero and NaN properly.
   // (a >= -0.0 ? c : b) transforms into (0 > a ? b : c), hence the NLE.
@@ -433,9 +433,9 @@ void Jit64::fselx(UGeckoInstruction inst)
   else
   {
     MOVAPD(XMM1, R(XMM0));
-    PAND(XMM0, fpr.R(b));
-    PANDN(XMM1, fpr.R(c));
-    POR(XMM1, R(XMM0));
+    ANDPD(XMM0, fpr.R(b));
+    ANDNPD(XMM1, fpr.R(c));
+    ORPD(XMM1, R(XMM0));
   }
 
   fpr.BindToRegister(d, !packed);
diff --git a/Source/Core/Core/PowerPC/Jit64/Jit_LoadStore.cpp b/Source/Core/Core/PowerPC/Jit64/Jit_LoadStore.cpp
index 458b143112..f834282ea9 100644
--- a/Source/Core/Core/PowerPC/Jit64/Jit_LoadStore.cpp
+++ b/Source/Core/Core/PowerPC/Jit64/Jit_LoadStore.cpp
@@ -362,7 +362,7 @@ void Jit64::dcbz(UGeckoInstruction inst)
     FixupBranch slow = J_CC(CC_Z, true);
 
     // Fast path: compute full address, then zero out 32 bytes of memory.
-    PXOR(XMM0, R(XMM0));
+    XORPS(XMM0, R(XMM0));
     MOVAPS(MComplex(RMEM, RSCRATCH, SCALE_1, 0), XMM0);
     MOVAPS(MComplex(RMEM, RSCRATCH, SCALE_1, 16), XMM0);