[turbofan] Implement rounding of floats on x64 and ia32 without sse4.1.

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 1062 matching lines @@
 
   void psllq(XMMRegister reg, byte imm8);
   void psrlq(XMMRegister reg, byte imm8);
   void pslld(XMMRegister reg, byte imm8);
   void psrld(XMMRegister reg, byte imm8);
 
   void cvttsd2si(Register dst, const Operand& src);
   void cvttsd2si(Register dst, XMMRegister src);
   void cvttss2siq(Register dst, XMMRegister src);
   void cvttss2siq(Register dst, const Operand& src);
+  void cvtss2siq(Register dst, XMMRegister src);
+  void cvtss2siq(Register dst, const Operand& src);
   void cvttsd2siq(Register dst, XMMRegister src);
   void cvttsd2siq(Register dst, const Operand& src);
 
   void cvtlsi2sd(XMMRegister dst, const Operand& src);
   void cvtlsi2sd(XMMRegister dst, Register src);
 
   void cvtqsi2ss(XMMRegister dst, const Operand& src);
   void cvtqsi2ss(XMMRegister dst, Register src);
 
   void cvtqsi2sd(XMMRegister dst, const Operand& src);
   void cvtqsi2sd(XMMRegister dst, Register src);
 
 
   void cvtss2sd(XMMRegister dst, XMMRegister src);
   void cvtss2sd(XMMRegister dst, const Operand& src);
   void cvtsd2ss(XMMRegister dst, XMMRegister src);
   void cvtsd2ss(XMMRegister dst, const Operand& src);
 
   void cvtsd2si(Register dst, XMMRegister src);
   void cvtsd2siq(Register dst, XMMRegister src);
+  void cvtsd2siq(Register dst, const Operand& src);
 
   void addsd(XMMRegister dst, XMMRegister src);
   void addsd(XMMRegister dst, const Operand& src);
   void subsd(XMMRegister dst, XMMRegister src);
   void subsd(XMMRegister dst, const Operand& src);
   void mulsd(XMMRegister dst, XMMRegister src);
   void mulsd(XMMRegister dst, const Operand& src);
   void divsd(XMMRegister dst, XMMRegister src);
   void divsd(XMMRegister dst, const Operand& src);
 
@@ skipping 22 matching lines @@
   void extractps(Register dst, XMMRegister src, byte imm8);
 
   void pextrd(Register dst, XMMRegister src, int8_t imm8);
 
   void pinsrd(XMMRegister dst, Register src, int8_t imm8);
   void pinsrd(XMMRegister dst, const Operand& src, int8_t imm8);
 
   void roundss(XMMRegister dst, XMMRegister src, RoundingMode mode);
   void roundsd(XMMRegister dst, XMMRegister src, RoundingMode mode);
 
+  void ldmxcsr(const Operand& dst);
+  void stmxcsr(const Operand& dst);
+
   // AVX instruction
   void vfmadd132sd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vfmasd(0x99, dst, src1, src2);
   }
   void vfmadd213sd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vfmasd(0xa9, dst, src1, src2);
   }
   void vfmadd231sd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vfmasd(0xb9, dst, src1, src2);
   }
@@ skipping 196 matching lines @@
   void vpsrlq(XMMRegister dst, XMMRegister src, byte imm8) {
     XMMRegister iop = {2};
     vpd(0x73, iop, dst, src);
     emit(imm8);
   }
   void vpsllq(XMMRegister dst, XMMRegister src, byte imm8) {
     XMMRegister iop = {6};
     vpd(0x73, iop, dst, src);
     emit(imm8);
   }
+  void vpslld(XMMRegister dst, XMMRegister src, byte imm8) {
+    XMMRegister iop = {6};
+    vpd(0x72, iop, dst, src);
+    emit(imm8);
+  }
   void vcvtss2sd(XMMRegister dst, XMMRegister src1, XMMRegister src2) {
     vsd(0x5a, dst, src1, src2, kF3, k0F, kWIG);
   }
   void vcvtss2sd(XMMRegister dst, XMMRegister src1, const Operand& src2) {
     vsd(0x5a, dst, src1, src2, kF3, k0F, kWIG);
   }
   void vcvtlsi2sd(XMMRegister dst, XMMRegister src1, Register src2) {
     XMMRegister isrc2 = {src2.code()};
     vsd(0x2a, dst, src1, isrc2, kF2, k0F, kW0);
   }
@@ skipping 38 matching lines @@
     vsd(0x2c, idst, xmm0, src, kF2, k0F, kW0);
   }
   void vcvttss2siq(Register dst, XMMRegister src) {
     XMMRegister idst = {dst.code()};
     vsd(0x2c, idst, xmm0, src, kF3, k0F, kW1);
   }
   void vcvttss2siq(Register dst, const Operand& src) {
     XMMRegister idst = {dst.code()};
     vsd(0x2c, idst, xmm0, src, kF3, k0F, kW1);
   }
+  void vcvtss2siq(Register dst, XMMRegister src) {
+    XMMRegister idst = {dst.code()};
+    vsd(0x2d, idst, xmm0, src, kF3, k0F, kW1);
+  }
+  void vcvtss2siq(Register dst, const Operand& src) {
+    XMMRegister idst = {dst.code()};
+    vsd(0x2d, idst, xmm0, src, kF3, k0F, kW1);
+  }
   void vcvttsd2siq(Register dst, XMMRegister src) {
     XMMRegister idst = {dst.code()};
     vsd(0x2c, idst, xmm0, src, kF2, k0F, kW1);
   }
   void vcvttsd2siq(Register dst, const Operand& src) {
     XMMRegister idst = {dst.code()};
     vsd(0x2c, idst, xmm0, src, kF2, k0F, kW1);
   }
+  void vcvtsd2siq(Register dst, XMMRegister src) {
+    XMMRegister idst = {dst.code()};
+    vsd(0x2d, idst, xmm0, src, kF2, k0F, kW1);
+  }
+  void vcvtsd2siq(Register dst, const Operand& src) {
+    XMMRegister idst = {dst.code()};
+    vsd(0x2d, idst, xmm0, src, kF2, k0F, kW1);
+  }
   void vcvtsd2si(Register dst, XMMRegister src) {
     XMMRegister idst = {dst.code()};
     vsd(0x2d, idst, xmm0, src, kF2, k0F, kW0);
   }
   void vucomisd(XMMRegister dst, XMMRegister src) {
     vsd(0x2e, dst, xmm0, src, k66, k0F, kWIG);
   }
   void vucomisd(XMMRegister dst, const Operand& src) {
     vsd(0x2e, dst, xmm0, src, k66, k0F, kWIG);
   }
@@ skipping 768 matching lines @@
   Assembler* assembler_;
 #ifdef DEBUG
   int space_before_;
 #endif
 };
 
 }  // namespace internal
 }  // namespace v8
 
 #endif  // V8_X64_ASSEMBLER_X64_H_