[wasm] Implement I32x4ReplaceLane, I32x4Add, I32x4Sub.

src/compiler/x64/code-generator-x64.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/compilation-info.h"
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
@@ skipping 2066 matching lines @@
       XMMRegister dst = i.OutputSimd128Register();
       __ Movd(dst, i.InputRegister(0));
       __ shufps(dst, dst, 0x0);
       break;
     }
     case kX64Int32x4ExtractLane: {
       CpuFeatureScope sse_scope(masm(), SSE4_1);
       __ Pextrd(i.OutputRegister(), i.InputSimd128Register(0), i.InputInt8(1));
       break;
     }
+    case kX64Int32x4ReplaceLane: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      XMMRegister dst = i.OutputSimd128Register();
+      if (instr->InputAt(2)->IsRegister()) {
+        __ Pinsrd(i.InputSimd128Register(0), i.InputRegister(2),
+                  i.InputInt8(1));
+      } else {
+        __ Pinsrd(i.InputSimd128Register(0), i.InputOperand(2), i.InputInt8(1));
+      }
+      __ Movups(dst, i.InputSimd128Register(0));
+      break;
+    }
+    case kX64Int32x4Add: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      XMMRegister dst = i.OutputSimd128Register();
+      __ addps(i.InputSimd128Register(0), i.InputSimd128Register(1));
+      __ Movups(dst, i.InputSimd128Register(0));
+      break;
+    }
+    case kX64Int32x4Sub: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      XMMRegister dst = i.OutputSimd128Register();
+      __ subps(i.InputSimd128Register(0), i.InputSimd128Register(1));
+      __ Movups(dst, i.InputSimd128Register(0));
+      break;
+    }
     case kCheckedLoadInt8:
       ASSEMBLE_CHECKED_LOAD_INTEGER(movsxbl);
       break;
     case kCheckedLoadUint8:
       ASSEMBLE_CHECKED_LOAD_INTEGER(movzxbl);
       break;
     case kCheckedLoadInt16:
       ASSEMBLE_CHECKED_LOAD_INTEGER(movsxwl);
       break;
     case kCheckedLoadUint16:
@@ skipping 618 matching lines @@
     int padding_size = last_lazy_deopt_pc_ + space_needed - current_pc;
     __ Nop(padding_size);
   }
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8