[turbofan] Implemented the optional Float32RoundDown operator.

src/compiler/ia32/code-generator-ia32.cc

 // Copyright 2013 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/compiler/code-generator.h"
 
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
 #include "src/compiler/osr.h"
@@ skipping 618 matching lines @@
       __ andps(i.OutputDoubleRegister(), kScratchDoubleReg);
       break;
     }
     case kSSEFloat32Neg: {
       // TODO(bmeurer): Use 128-bit constants.
       __ pcmpeqd(kScratchDoubleReg, kScratchDoubleReg);
       __ psllq(kScratchDoubleReg, 31);
       __ xorps(i.OutputDoubleRegister(), kScratchDoubleReg);
       break;
     }
+    case kSSEFloat32Round: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      RoundingMode const mode =
+          static_cast<RoundingMode>(MiscField::decode(instr->opcode()));
+      __ roundss(i.OutputDoubleRegister(), i.InputDoubleRegister(0), mode);
+      break;
+    }
     case kSSEFloat64Cmp:
       __ ucomisd(i.InputDoubleRegister(0), i.InputOperand(1));
       break;
     case kSSEFloat64Add:
       __ addsd(i.InputDoubleRegister(0), i.InputOperand(1));
       break;
     case kSSEFloat64Sub:
       __ subsd(i.InputDoubleRegister(0), i.InputOperand(1));
       break;
     case kSSEFloat64Mul:
@@ skipping 974 matching lines @@
     int padding_size = last_lazy_deopt_pc_ + space_needed - current_pc;
     __ Nop(padding_size);
   }
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8