[x64] Replace movaps with appropriate vmov* instructions when AVX is enabled.

src/x64/assembler-x64.h

 // Copyright (c) 1994-2006 Sun Microsystems Inc.
 // All Rights Reserved.
 //
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 // - Redistributions of source code must retain the above copyright notice,
 // this list of conditions and the following disclaimer.
 //
@@ skipping 990 matching lines @@
   void maxss(XMMRegister dst, const Operand& src);
   void minss(XMMRegister dst, XMMRegister src);
   void minss(XMMRegister dst, const Operand& src);
 
   void sqrtss(XMMRegister dst, XMMRegister src);
   void sqrtss(XMMRegister dst, const Operand& src);
 
   void ucomiss(XMMRegister dst, XMMRegister src);
   void ucomiss(XMMRegister dst, const Operand& src);
   void movaps(XMMRegister dst, XMMRegister src);
+
+  // Don't use this unless it's important to keep the
+  // top half of the destination register unchanged.
+  // Use movaps when moving float values and movd for integer
+  // values in xmm registers.
+  void movss(XMMRegister dst, XMMRegister src);
+
   void movss(XMMRegister dst, const Operand& src);
   void movss(const Operand& dst, XMMRegister src);
   void shufps(XMMRegister dst, XMMRegister src, byte imm8);
 
   void cvttss2si(Register dst, const Operand& src);
   void cvttss2si(Register dst, XMMRegister src);
   void cvtlsi2ss(XMMRegister dst, Register src);
 
   void andps(XMMRegister dst, XMMRegister src);
   void andps(XMMRegister dst, const Operand& src);
@@ skipping 16 matching lines @@
   // SSE2 instructions
   void movd(XMMRegister dst, Register src);
   void movd(XMMRegister dst, const Operand& src);
   void movd(Register dst, XMMRegister src);
   void movq(XMMRegister dst, Register src);
   void movq(Register dst, XMMRegister src);
   void movq(XMMRegister dst, XMMRegister src);
 
   // Don't use this unless it's important to keep the
   // top half of the destination register unchanged.
-  // Used movaps when moving double values and movq for integer
+  // Use movapd when moving double values and movq for integer
   // values in xmm registers.
   void movsd(XMMRegister dst, XMMRegister src);
 
   void movsd(const Operand& dst, XMMRegister src);
   void movsd(XMMRegister dst, const Operand& src);
 
   void movdqa(const Operand& dst, XMMRegister src);
   void movdqa(XMMRegister dst, const Operand& src);
 
   void movdqu(const Operand& dst, XMMRegister src);
@@ skipping 216 matching lines @@
   void vfmass(byte op, XMMRegister dst, XMMRegister src1, XMMRegister src2);
   void vfmass(byte op, XMMRegister dst, XMMRegister src1, const Operand& src2);
 
   void vmovd(XMMRegister dst, Register src);
   void vmovd(XMMRegister dst, const Operand& src);
   void vmovd(Register dst, XMMRegister src);
   void vmovq(XMMRegister dst, Register src);
   void vmovq(XMMRegister dst, const Operand& src);
   void vmovq(Register dst, XMMRegister src);
 
+  void vmovsd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
+    vsd(0x10, dst, src1, src2);
+  }
   void vmovsd(XMMRegister dst, const Operand& src) {
     vsd(0x10, dst, xmm0, src);
   }
-  void vmovsd(XMMRegister dst, XMMRegister src) { vsd(0x10, dst, xmm0, src); }
   void vmovsd(const Operand& dst, XMMRegister src) {
     vsd(0x11, src, xmm0, dst);
   }
   void vaddsd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vsd(0x58, dst, src1, src2);
   }
   void vaddsd(XMMRegister dst, XMMRegister src1, const Operand& src2) {
     vsd(0x58, dst, src1, src2);
   }
   void vsubsd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
@@ skipping 108 matching lines @@
   }
   void vmaxss(XMMRegister dst, XMMRegister src1, const Operand& src2) {
     vss(0x5f, dst, src1, src2);
   }
   void vminss(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vss(0x5d, dst, src1, src2);
   }
   void vminss(XMMRegister dst, XMMRegister src1, const Operand& src2) {
     vss(0x5d, dst, src1, src2);
   }
+  void vmovss(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
+    vss(0x10, dst, src1, src2);
+  }
   void vmovss(XMMRegister dst, const Operand& src) {
     vss(0x10, dst, xmm0, src);
   }
   void vmovss(const Operand& dst, XMMRegister src) {
     vss(0x11, src, xmm0, dst);
   }
   void vucomiss(XMMRegister dst, XMMRegister src);
   void vucomiss(XMMRegister dst, const Operand& src);
   void vss(byte op, XMMRegister dst, XMMRegister src1, XMMRegister src2);
   void vss(byte op, XMMRegister dst, XMMRegister src1, const Operand& src2);
@@ skipping 168 matching lines @@
     bmi2l(kF2, 0xf7, dst, src2, src1);
   }
   void shrxl(Register dst, const Operand& src1, Register src2) {
     bmi2l(kF2, 0xf7, dst, src2, src1);
   }
   void rorxq(Register dst, Register src, byte imm8);
   void rorxq(Register dst, const Operand& src, byte imm8);
   void rorxl(Register dst, Register src, byte imm8);
   void rorxl(Register dst, const Operand& src, byte imm8);
 
+  void vmovaps(XMMRegister dst, XMMRegister src) { vps(0x28, dst, xmm0, src); }
   void vmovapd(XMMRegister dst, XMMRegister src) { vpd(0x28, dst, xmm0, src); }
   void vmovmskpd(Register dst, XMMRegister src) {
     XMMRegister idst = {dst.code()};
     vpd(0x50, idst, xmm0, src);
   }
 
 #define PACKED_OP_LIST(V) \
   V(and, 0x54)            \
   V(xor, 0x57)
 
@@ skipping 566 matching lines @@
   Assembler* assembler_;
 #ifdef DEBUG
   int space_before_;
 #endif
 };
 
 }  // namespace internal
 }  // namespace v8
 
 #endif  // V8_X64_ASSEMBLER_X64_H_