[wasm-simd] Improve f64x2.convert_low_i32x4 tests

Zero top lanes to ensure that we convert the low lanes of i32x4.

Bug: v8:11265
Change-Id: Id53b65bada406c967c692b584210d4168e2c9183
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2776511Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#73587}

test/cctest/wasm/test-run-wasm-simd.cc

@@ -908,12 +908,15 @@ void RunF64x2ConvertLowI32x4Test(TestExecutionTier execution_tier,
                                  LowerSimd lower_simd, WasmOpcode opcode) {
   WasmRunner<int32_t, SrcType> r(execution_tier, lower_simd);
   double* g = r.builder().template AddGlobal<double>(kWasmS128);
-  // TODO(zhin): set top lanes to 0 to assert conversion happens on low lanes.
-  BUILD(
-      r,
-      WASM_GLOBAL_SET(
-          0, WASM_SIMD_UNOP(opcode, WASM_SIMD_I32x4_SPLAT(WASM_LOCAL_GET(0)))),
-      WASM_ONE);
+  BUILD(r,
+        WASM_GLOBAL_SET(
+            0,
+            WASM_SIMD_UNOP(
+                opcode,
+                // Set top lane of i64x2 == set top 2 lanes of i32x4.
+                WASM_SIMD_I64x2_REPLACE_LANE(
+                    1, WASM_SIMD_I32x4_SPLAT(WASM_LOCAL_GET(0)), WASM_ZERO64))),
+        WASM_ONE);
 
   for (SrcType x : compiler::ValueHelper::GetVector<SrcType>()) {
     r.Call(x);