[wasm-simd][x64] Optimize signed compares codegen

Improve SSE codegen similar to suggestions in
https://crrev.com/c/2698066:

- s/movdqa/movaps/
- s/por/orps
- s/pxor/xorps
- s/pand/andps
- use movshdup (SSSE3) insted of shuffle

Bug: v8:11415
Change-Id: I2e76cbbe16267c055e24f258354b77994aed47b8
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2713131Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#73057}

src/codegen/ia32/macro-assembler-ia32.cc

@@ -1083,7 +1083,7 @@ void TurboAssembler::I64x2GtS(XMMRegister dst, XMMRegister src0,
     DCHECK_EQ(dst, src0);
     pcmpgtq(dst, src1);
   } else {
-    CpuFeatureScope sse_scope(this, SSSE3);
+    CpuFeatureScope sse_scope(this, SSE3);
     DCHECK_NE(dst, src0);
     DCHECK_NE(dst, src1);
     movaps(dst, src1);
@@ -1115,7 +1115,7 @@ void TurboAssembler::I64x2GeS(XMMRegister dst, XMMRegister src0,
     pcmpeqd(scratch, scratch);
     xorps(dst, scratch);
   } else {
-    CpuFeatureScope sse_scope(this, SSSE3);
+    CpuFeatureScope sse_scope(this, SSE3);
     DCHECK_NE(dst, src0);
     DCHECK_NE(dst, src1);
     movaps(dst, src0);

src/codegen/x64/macro-assembler-x64.cc

@@ -2514,17 +2514,18 @@ void TurboAssembler::I64x2GtS(XMMRegister dst, XMMRegister src0,
     DCHECK_EQ(dst, src0);
     pcmpgtq(dst, src1);
   } else {
+    CpuFeatureScope sse_scope(this, SSE3);
     DCHECK_NE(dst, src0);
     DCHECK_NE(dst, src1);
-    movdqa(dst, src1);
-    movdqa(kScratchDoubleReg, src0);
+    movaps(dst, src1);
+    movaps(kScratchDoubleReg, src0);
     psubq(dst, src0);
     pcmpeqd(kScratchDoubleReg, src1);
-    pand(dst, kScratchDoubleReg);
-    movdqa(kScratchDoubleReg, src0);
+    andps(dst, kScratchDoubleReg);
+    movaps(kScratchDoubleReg, src0);
     pcmpgtd(kScratchDoubleReg, src1);
-    por(dst, kScratchDoubleReg);
-    pshufd(dst, dst, 0xF5);
+    orps(dst, kScratchDoubleReg);
+    movshdup(dst, dst);
   }
 }
 
@@ -2539,25 +2540,26 @@ void TurboAssembler::I64x2GeS(XMMRegister dst, XMMRegister src0,
     CpuFeatureScope sse_scope(this, SSE4_2);
     DCHECK_NE(dst, src0);
     if (dst != src1) {
-      movdqa(dst, src1);
+      movaps(dst, src1);
     }
     pcmpgtq(dst, src0);
     pcmpeqd(kScratchDoubleReg, kScratchDoubleReg);
-    pxor(dst, kScratchDoubleReg);
+    xorps(dst, kScratchDoubleReg);
   } else {
+    CpuFeatureScope sse_scope(this, SSE3);
     DCHECK_NE(dst, src0);
     DCHECK_NE(dst, src1);
-    movdqa(dst, src0);
-    movdqa(kScratchDoubleReg, src1);
+    movaps(dst, src0);
+    movaps(kScratchDoubleReg, src1);
     psubq(dst, src1);
     pcmpeqd(kScratchDoubleReg, src0);
-    pand(dst, kScratchDoubleReg);
-    movdqa(kScratchDoubleReg, src1);
+    andps(dst, kScratchDoubleReg);
+    movaps(kScratchDoubleReg, src1);
     pcmpgtd(kScratchDoubleReg, src0);
-    por(dst, kScratchDoubleReg);
-    pshufd(dst, dst, 0xF5);
+    orps(dst, kScratchDoubleReg);
+    movshdup(dst, dst);
     pcmpeqd(kScratchDoubleReg, kScratchDoubleReg);
-    pxor(dst, kScratchDoubleReg);
+    xorps(dst, kScratchDoubleReg);
   }
 }