[wasm-simd] Enable SIMD by default

Bug: v8:11511,v8:6020
Change-Id: I07b87b2a3a2ea08def9be2d0da808c013dd6fc2a
Cq-Include-Trybots: luci.v8.try:v8_linux_blink_rel
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2728245
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Cr-Commit-Position: refs/heads/master@{#73237}

src/wasm/baseline/liftoff-compiler.cc

@@ -334,13 +334,6 @@ void CheckBailoutAllowed(LiftoffBailoutReason reason, const char* detail,
   // https://v8.dev/docs/wasm-shipping-checklist. Some are not though. They are
   // listed here explicitly, with a bug assigned to each of them.
 
-  // TODO(6020): Fully implement SIMD in Liftoff.
-  STATIC_ASSERT(kStagedFeatures.has_simd());
-  if (reason == kSimd) {
-    DCHECK(env->enabled_features.has_simd());
-    return;
-  }
-
   // TODO(7581): Fully implement reftypes in Liftoff.
   STATIC_ASSERT(kStagedFeatures.has_reftypes());
   if (reason == kRefTypes) {

src/wasm/wasm-feature-flags.h

@@ -57,12 +57,6 @@
   /* Staged in v8.7 * */                                                       \
   V(return_call, "return call opcodes", false)                                 \
                                                                                \
-  /* Fixed-width SIMD operations. */                                           \
-  /* https://github.com/webassembly/simd */                                    \
-  /* V8 side owner: gdeepti, zhin */                                           \
-  /* Staged in v8.7 * */                                                       \
-  V(simd, "SIMD opcodes", false)                                               \
-                                                                               \
   /* Threads proposal. */                                                      \
   /* https://github.com/webassembly/threads */                                 \
   /* NOTE: This is enabled via chromium flag on desktop systems since v7.4  */ \
@@ -86,7 +80,14 @@
   /* V8 side owner: thibaudm */                                                \
   /* Shipped in v8.6. */                                                       \
   /* ITS: https://groups.google.com/g/v8-users/c/pv2E4yFWeF0 */                \
-  V(mv, "multi-value support", true)
+  V(mv, "multi-value support", true)                                           \
+                                                                               \
+  /* Fixed-width SIMD operations. */                                           \
+  /* https://github.com/webassembly/simd */                                    \
+  /* V8 side owner: gdeepti, zhin */                                           \
+  /* Staged in v8.7 * */                                                       \
+  /* Shipped in v9.1 * */                                                      \
+  V(simd, "SIMD opcodes", true)
 
 // Combination of all available wasm feature flags.
 #define FOREACH_WASM_FEATURE_FLAG(V)        \

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

@@ -368,6 +368,26 @@ bool ExpectFused(TestExecutionTier tier) {
   void RunWasm_##name##_Impl(LowerSimd lower_simd,                        \
                              TestExecutionTier execution_tier)
 
+// Use this for post-mvp instructions, for which we only implement interpreter
+// and TurboFan support. Liftoff support is generally not implemented yet, so
+// don't test that, and neither is scalar lowering.
+#define WASM_SIMD_TEST_POST_MVP(name)                                     \
+  void RunWasm_##name##_Impl(LowerSimd lower_simd,                        \
+                             TestExecutionTier execution_tier);           \
+  TEST(RunWasm_##name##_turbofan) {                                       \
+    if (!CpuFeatures::SupportsWasmSimd128()) return;                      \
+    EXPERIMENTAL_FLAG_SCOPE(simd);                                        \
+    FLAG_SCOPE(wasm_simd_post_mvp);                                       \
+    RunWasm_##name##_Impl(kNoLowerSimd, TestExecutionTier::kTurbofan);    \
+  }                                                                       \
+  TEST(RunWasm_##name##_interpreter) {                                    \
+    EXPERIMENTAL_FLAG_SCOPE(simd);                                        \
+    FLAG_SCOPE(wasm_simd_post_mvp);                                       \
+    RunWasm_##name##_Impl(kNoLowerSimd, TestExecutionTier::kInterpreter); \
+  }                                                                       \
+  void RunWasm_##name##_Impl(LowerSimd lower_simd,                        \
+                             TestExecutionTier execution_tier)
+
 // Returns true if the platform can represent the result.
 template <typename T>
 bool PlatformCanRepresent(T x) {
@@ -607,14 +627,12 @@ WASM_SIMD_TEST(F32x4Sqrt) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4Sqrt, std::sqrt);
 }
 
-WASM_SIMD_TEST(F32x4RecipApprox) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F32x4RecipApprox) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4RecipApprox,
                    base::Recip, false /* !exact */);
 }
 
-WASM_SIMD_TEST(F32x4RecipSqrtApprox) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F32x4RecipSqrtApprox) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4RecipSqrtApprox,
                    base::RecipSqrt, false /* !exact */);
 }
@@ -767,8 +785,7 @@ WASM_SIMD_TEST(F32x4Le) {
 }
 
 #if V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_S390X
-WASM_SIMD_TEST_NO_LOWERING(F32x4Qfma) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F32x4Qfma) {
   WasmRunner<int32_t, float, float, float> r(execution_tier, lower_simd);
   // Set up global to hold mask output.
   float* g = r.builder().AddGlobal<float>(kWasmS128);
@@ -792,8 +809,7 @@ WASM_SIMD_TEST_NO_LOWERING(F32x4Qfma) {
   }
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F32x4Qfms) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F32x4Qfms) {
   WasmRunner<int32_t, float, float, float> r(execution_tier, lower_simd);
   // Set up global to hold mask output.
   float* g = r.builder().AddGlobal<float>(kWasmS128);
@@ -1493,8 +1509,7 @@ WASM_SIMD_TEST(I64x2Mul) {
 }
 
 #if V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_S390X
-WASM_SIMD_TEST_NO_LOWERING(F64x2Qfma) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F64x2Qfma) {
   WasmRunner<int32_t, double, double, double> r(execution_tier, lower_simd);
   // Set up global to hold mask output.
   double* g = r.builder().AddGlobal<double>(kWasmS128);
@@ -1518,8 +1533,7 @@ WASM_SIMD_TEST_NO_LOWERING(F64x2Qfma) {
   }
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F64x2Qfms) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
+WASM_SIMD_TEST_POST_MVP(F64x2Qfms) {
   WasmRunner<int32_t, double, double, double> r(execution_tier, lower_simd);
   // Set up global to hold mask output.
   double* g = r.builder().AddGlobal<double>(kWasmS128);
@@ -3615,9 +3629,7 @@ WASM_SIMD_TEST(SimdF32x4SetGlobal) {
 
 #if V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32
 // TODO(v8:11168): Prototyping prefetch.
-WASM_SIMD_TEST(SimdPrefetch) {
-  FLAG_SCOPE(wasm_simd_post_mvp);
-
+WASM_SIMD_TEST_POST_MVP(SimdPrefetch) {
   {
     // Test PrefetchT.
     WasmRunner<int32_t> r(execution_tier, lower_simd);