[wasm] Implement wasm x64 I16x8 Ops

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 <limits>
 
 #include "src/compilation-info.h"
 #include "src/compiler/code-generator-impl.h"
@@ skipping 2217 matching lines @@
     case kX64I32x4MaxU: {
       CpuFeatureScope sse_scope(masm(), SSE4_1);
       __ pmaxud(i.OutputSimd128Register(), i.InputSimd128Register(1));
       break;
     }
     case kX64S128Zero: {
       XMMRegister dst = i.OutputSimd128Register();
       __ xorps(dst, dst);
       break;
     }
-    case kX64S32x4Select: {
+    case kX64I16x8Splat: {
+      XMMRegister dst = i.OutputSimd128Register();
+      __ movd(dst, i.InputRegister(0));
+      __ pshuflw(dst, dst, 0x0);
+      __ pshufhw(dst, dst, 0x0);
+      __ pshufd(dst, dst, 0x0);
+      break;
+    }
+    case kX64I16x8ExtractLane: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      Register dst = i.OutputRegister();
+      __ xorq(dst, dst);

[bbudge, 2017/04/17 17:34:06]

Is this really needed? It seems like movsx (move word to quad word with sign
extension) should take care of this.

http://www.felixcloutier.com/x86/MOVSX:MOVSXD.html

[gdeepti, 2017/04/17 17:46:22]

Not needed, left behind as a result of merge after assembler changes, here as
well as below. Fixed now. 

+      __ pextrw(dst, i.InputSimd128Register(0), i.InputInt8(1));
+      __ movsxwl(i.OutputRegister(), i.OutputRegister());

[bbudge, 2017/04/17 17:34:06]

nit: use dst here.

[gdeepti, 2017/04/17 17:46:22]

Done.

+      break;
+    }
+    case kX64I16x8ReplaceLane: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      if (instr->InputAt(2)->IsRegister()) {
+        __ pinsrw(i.OutputSimd128Register(), i.InputRegister(2),
+                  i.InputInt8(1));
+      } else {
+        __ pinsrw(i.OutputSimd128Register(), i.InputOperand(2), i.InputInt8(1));
+      }
+      break;
+    }
+    case kX64I16x8Shl: {
+      __ psllw(i.OutputSimd128Register(), i.InputInt8(1));
+      break;
+    }
+    case kX64I16x8ShrS: {
+      __ psraw(i.OutputSimd128Register(), i.InputInt8(1));
+      break;
+    }
+    case kX64I16x8Add: {
+      __ paddw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8AddSaturateS: {
+      __ paddsw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8Sub: {
+      __ psubw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8SubSaturateS: {
+      __ psubsw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8Mul: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      __ pmullw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8MinS: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      __ pminsw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8MaxS: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      __ pmaxsw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8Eq: {
+      __ pcmpeqw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8Ne: {
+      __ pcmpeqw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      __ pcmpeqw(kScratchDoubleReg, kScratchDoubleReg);
+      __ pxor(i.OutputSimd128Register(), kScratchDoubleReg);
+      break;
+    }
+    case kX64I16x8ShrU: {
+      __ psrlw(i.OutputSimd128Register(), i.InputInt8(1));
+      break;
+    }
+    case kX64I16x8AddSaturateU: {
+      __ paddusw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8SubSaturateU: {
+      __ psubusw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8MinU: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      __ pminuw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64I16x8MaxU: {
+      CpuFeatureScope sse_scope(masm(), SSE4_1);
+      __ pmaxuw(i.OutputSimd128Register(), i.InputSimd128Register(1));
+      break;
+    }
+    case kX64S128Select: {
       // Mask used here is stored in dst.
       XMMRegister dst = i.OutputSimd128Register();
       __ movaps(kScratchDoubleReg, i.InputSimd128Register(1));
       __ xorps(kScratchDoubleReg, i.InputSimd128Register(2));
       __ andps(dst, kScratchDoubleReg);
       __ xorps(dst, i.InputSimd128Register(2));
       break;
     }
     case kCheckedLoadInt8:
       ASSEMBLE_CHECKED_LOAD_INTEGER(movsxbl);
@@ skipping 752 matching lines @@
     int padding_size = last_lazy_deopt_pc_ + space_needed - current_pc;
     __ Nop(padding_size);
   }
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8