Add floor, ceil, round (truncate) instructions for ia32, x64 (if SSE4.1) and

src/compiler/x64/instruction-selector-x64.cc

 // Copyright 2014 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/generic-node-inl.h"
 #include "src/compiler/instruction-selector-impl.h"
 #include "src/compiler/node-matchers.h"
 
 namespace v8 {
 namespace internal {
@@ skipping 102 matching lines @@
   }
 
   static const int kMaxInputCount = 3;
   InstructionOperand* base_operand_;
   InstructionOperand* index_operand_;
   InstructionOperand* displacement_operand_;
   AddressingMode mode_;
 };
 
 
+static void VisitRRFloat64(InstructionSelector* selector, ArchOpcode opcode,
+                           Node* node) {
+  X64OperandGenerator g(selector);
+  selector->Emit(opcode, g.DefineAsRegister(node),
+                 g.UseRegister(node->InputAt(0)));
+}
+
+
 void InstructionSelector::VisitLoad(Node* node) {
   MachineType rep = RepresentationOf(OpParameter<LoadRepresentation>(node));
   MachineType typ = TypeOf(OpParameter<LoadRepresentation>(node));
   Node* base = node->InputAt(0);
   Node* index = node->InputAt(1);
 
   ArchOpcode opcode;
   // TODO(titzer): signed/unsigned small loads
   switch (rep) {
     case kRepFloat32:
@@ skipping 583 matching lines @@
        temps);
 }
 
 
 void InstructionSelector::VisitFloat64Sqrt(Node* node) {
   X64OperandGenerator g(this);
   Emit(kSSEFloat64Sqrt, g.DefineAsRegister(node), g.Use(node->InputAt(0)));
 }
 
 
+void InstructionSelector::VisitFloat64Floor(Node* node) {
+  DCHECK(CpuFeatures::IsSupported(SSE4_1));
+  VisitRRFloat64(this, kSSEFloat64Floor, node);
+}
+
+
+void InstructionSelector::VisitFloat64Ceil(Node* node) {
+  DCHECK(CpuFeatures::IsSupported(SSE4_1));
+  VisitRRFloat64(this, kSSEFloat64Ceil, node);
+}
+
+
+void InstructionSelector::VisitFloat64RoundTruncate(Node* node) {
+  DCHECK(CpuFeatures::IsSupported(SSE4_1));
+  VisitRRFloat64(this, kSSEFloat64RoundTruncate, node);
+}
+
+
+void InstructionSelector::VisitFloat64RoundTiesAway(Node* node) {
+  UNREACHABLE();
+}
+
+
 void InstructionSelector::VisitCall(Node* node) {
   X64OperandGenerator g(this);
   CallDescriptor* descriptor = OpParameter<CallDescriptor*>(node);
 
   FrameStateDescriptor* frame_state_descriptor = NULL;
   if (descriptor->NeedsFrameState()) {
     frame_state_descriptor = GetFrameStateDescriptor(
         node->InputAt(static_cast<int>(descriptor->InputCount())));
   }
 
@@ skipping 369 matching lines @@
 
 void InstructionSelector::VisitFloat64LessThanOrEqual(Node* node) {
   FlagsContinuation cont(kUnorderedLessThanOrEqual, node);
   VisitFloat64Compare(this, node, &cont);
 }
 
 
 // static
 MachineOperatorBuilder::Flags
 InstructionSelector::SupportedMachineOperatorFlags() {
+  if (CpuFeatures::IsSupported(SSE4_1)) {
+    return MachineOperatorBuilder::kFloat64Floor |
+           MachineOperatorBuilder::kFloat64Ceil |
+           MachineOperatorBuilder::kFloat64RoundTruncate;
+  }
   return MachineOperatorBuilder::kNoFlags;
 }
-
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8