[wasm-simd][x64][ia32] Fix i8x16.popcnt codegen for ATOM

Fixed: v8:11591
Change-Id: I1d1d3ea94c969e17accb9651cd3e2dc29357ce12
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2780218Reviewed-by: Bill Budge <bbudge@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#73591}

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

@@ -961,8 +961,8 @@ void TurboAssembler::I8x16Popcnt(XMMRegister dst, XMMRegister src,
     // PSHUFB instruction, thus use PSHUFB-free divide-and-conquer
     // algorithm on these processors. ATOM CPU feature captures exactly
     // the right set of processors.
-    xorps(tmp1, tmp1);
-    pavgb(tmp1, src);
+    movaps(tmp1, src);
+    psrlw(tmp1, 1);
     if (dst != src) {
       movaps(dst, src);
     }

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

@@ -2435,8 +2435,8 @@ void TurboAssembler::I8x16Popcnt(XMMRegister dst, XMMRegister src,
     // PSHUFB instruction, thus use PSHUFB-free divide-and-conquer
     // algorithm on these processors. ATOM CPU feature captures exactly
     // the right set of processors.
-    xorps(tmp, tmp);
-    pavgb(tmp, src);
+    movaps(tmp, src);
+    psrlw(tmp, 1);
     if (dst != src) {
       movaps(dst, src);
     }