[wasm-simd] Fix f32x4.sqrt codegen

It was assuming that the input is always a register, but it isn't.

Bug: chromium:1078399
Change-Id: If14abb8ea34f9febfc04a67a8da260a7e66af7f1
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2182176Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67578}

src/compiler/backend/ia32/code-generator-ia32.cc

@@ -2231,12 +2231,12 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       break;
     }
     case kSSEF32x4Sqrt: {
-      __ sqrtps(i.OutputSimd128Register(), i.InputSimd128Register(0));
+      __ sqrtps(i.OutputSimd128Register(), i.InputOperand(0));
       break;
     }
     case kAVXF32x4Sqrt: {
       CpuFeatureScope avx_scope(tasm(), AVX);
-      __ vsqrtps(i.OutputSimd128Register(), i.InputSimd128Register(0));
+      __ vsqrtps(i.OutputSimd128Register(), i.InputOperand(0));
       break;
     }
     case kIA32F32x4RecipApprox: {