[WebAssembly] Custom combines for f64x2.promote_low_f32x4

[tlively]

Summary

Replace the clang builtin function and LLVM intrinsic previously used to select
the f64x2.promote_low_f32x4 instruction with custom combines from standard
SelectionDAG nodes. Implement the new combines to share code with the similar
combines for f64x2.convert_low_i32x4_{s,u}. Resolves PR50232.

clang/include/clang/Basic/BuiltinsWebAssembly.def

 TARGET_BUILTIN(__builtin_wasm_narrow_s_i8x16_i16x8, "V16ScV8sV8s", "nc", "simd128")
 TARGET_BUILTIN(__builtin_wasm_narrow_u_i8x16_i16x8, "V16UcV8sV8s", "nc", "simd128")
 TARGET_BUILTIN(__builtin_wasm_narrow_s_i16x8_i32x4, "V8sV4iV4i", "nc", "simd128")
 TARGET_BUILTIN(__builtin_wasm_narrow_u_i16x8_i32x4, "V8UsV4iV4i", "nc", "simd128")
 
 TARGET_BUILTIN(__builtin_wasm_trunc_sat_zero_s_f64x2_i32x4, "V4iV2d", "nc", "simd128")
 TARGET_BUILTIN(__builtin_wasm_trunc_sat_zero_u_f64x2_i32x4, "V4UiV2d", "nc", "simd128")
 TARGET_BUILTIN(__builtin_wasm_demote_zero_f64x2_f32x4, "V4fV2d", "nc", "simd128")
-TARGET_BUILTIN(__builtin_wasm_promote_low_f32x4_f64x2, "V2dV4f", "nc", "simd128")
 
 TARGET_BUILTIN(__builtin_wasm_load32_zero, "V4iiC*", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_load64_zero, "V2LLiLLiC*", "n", "simd128")
 
 TARGET_BUILTIN(__builtin_wasm_load8_lane, "V16ScScC*V16ScIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_load16_lane, "V8ssC*V8sIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_load32_lane, "V4iiC*V4iIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_load64_lane, "V2LLiLLiC*V2LLiIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_store8_lane, "vSc*V16ScIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_store16_lane, "vs*V8sIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_store32_lane, "vi*V4iIi", "n", "simd128")
 TARGET_BUILTIN(__builtin_wasm_store64_lane, "vLLi*V2LLiIi", "n", "simd128")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

clang/lib/CodeGen/CGBuiltin.cpp

Value *ConcatMask =
                                    Builder.getInt32(2), Builder.getInt32(3)});
     return Builder.CreateShuffleVector(Trunc, Splat, ConcatMask);
   }
   case WebAssembly::BI__builtin_wasm_demote_zero_f64x2_f32x4: {
     Value *Vec = EmitScalarExpr(E->getArg(0));
     Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_demote_zero);
     return Builder.CreateCall(Callee, Vec);
   }
-  case WebAssembly::BI__builtin_wasm_promote_low_f32x4_f64x2: {
-    Value *Vec = EmitScalarExpr(E->getArg(0));
-    Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_promote_low);
-    return Builder.CreateCall(Callee, Vec);
-  }
   case WebAssembly::BI__builtin_wasm_load32_zero: {
     Value *Ptr = EmitScalarExpr(E->getArg(0));
     Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_load32_zero);
     return Builder.CreateCall(Callee, {Ptr});
   }
   case WebAssembly::BI__builtin_wasm_load64_zero: {
     Value *Ptr = EmitScalarExpr(E->getArg(0));
     Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_load64_zero);

clang/lib/Headers/wasm_simd128.h

 typedef unsigned char __u8x8
     __attribute__((__vector_size__(8), __aligned__(8)));
 typedef short __i16x4 __attribute__((__vector_size__(8), __aligned__(8)));
 typedef unsigned short __u16x4
     __attribute__((__vector_size__(8), __aligned__(8)));
 typedef int __i32x2 __attribute__((__vector_size__(8), __aligned__(8)));
 typedef unsigned int __u32x2
     __attribute__((__vector_size__(8), __aligned__(8)));
+typedef float __f32x2 __attribute__((__vector_size__(8), __aligned__(8)));
 
 #define __DEFAULT_FN_ATTRS                                                     \
   __attribute__((__always_inline__, __nodebug__, __target__("simd128"),        \
                  __min_vector_width__(128)))
 
 #define __REQUIRE_CONSTANT(c)                                                  \
   __attribute__((__diagnose_if__(!__builtin_constant_p(c),                     \
                                  #c " must be constant", "error")))

 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_v128_andnot(v128_t __a,
                                                              v128_t __b) {
   return __a & ~__b;
 }
 
 static __inline__ bool __DEFAULT_FN_ATTRS wasm_v128_any_true(v128_t __a) {
   return __builtin_wasm_any_true_v128((__i8x16)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_v128_bitselect(v128_t __a,
                                                                 v128_t __b,
                                                                 v128_t __mask) {
   return (v128_t)__builtin_wasm_bitselect((__i32x4)__a, (__i32x4)__b,
                                           (__i32x4)__mask);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_abs(v128_t __a) {
   return (v128_t)__builtin_wasm_abs_i8x16((__i8x16)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_neg(v128_t __a) {
   return (v128_t)(-(__u8x16)__a);
 }
 
 static __inline__ bool __DEFAULT_FN_ATTRS wasm_i8x16_all_true(v128_t __a) {
   return __builtin_wasm_all_true_i8x16((__i8x16)__a);
 }
 
 static __inline__ int32_t __DEFAULT_FN_ATTRS wasm_i8x16_bitmask(v128_t __a) {
   return __builtin_wasm_bitmask_i8x16((__i8x16)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_popcnt(v128_t __a) {
   return (v128_t)__builtin_wasm_popcnt_i8x16((__i8x16)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_shl(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__i8x16)__a << __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_shr(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__i8x16)__a >> __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_shr(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__u8x16)__a >> __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_add(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)((__u8x16)__a + (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_add_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_add_sat_s_i8x16((__i8x16)__a, (__i8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_add_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_add_sat_u_i8x16((__u8x16)__a, (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_sub(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)((__u8x16)__a - (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_sub_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_sub_sat_s_i8x16((__i8x16)__a, (__i8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_sub_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_sub_sat_u_i8x16((__u8x16)__a, (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_min(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)__builtin_wasm_min_s_i8x16((__i8x16)__a, (__i8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_min(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)__builtin_wasm_min_u_i8x16((__u8x16)__a, (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i8x16_max(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)__builtin_wasm_max_s_i8x16((__i8x16)__a, (__i8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_max(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)__builtin_wasm_max_u_i8x16((__u8x16)__a, (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u8x16_avgr(v128_t __a,
                                                             v128_t __b) {
   return (v128_t)__builtin_wasm_avgr_u_i8x16((__u8x16)__a, (__u8x16)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_abs(v128_t __a) {
   return (v128_t)__builtin_wasm_abs_i16x8((__i16x8)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_neg(v128_t __a) {
   return (v128_t)(-(__u16x8)__a);
 }
 
 static __inline__ bool __DEFAULT_FN_ATTRS wasm_i16x8_all_true(v128_t __a) {
   return __builtin_wasm_all_true_i16x8((__i16x8)__a);
 }
 
 static __inline__ int32_t __DEFAULT_FN_ATTRS wasm_i16x8_bitmask(v128_t __a) {
   return __builtin_wasm_bitmask_i16x8((__i16x8)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_shl(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__i16x8)__a << __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_shr(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__i16x8)__a >> __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u16x8_shr(v128_t __a,
                                                            int32_t __b) {
   return (v128_t)((__u16x8)__a >> __b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_add(v128_t __a,
                                                            v128_t __b) {
   return (v128_t)((__u16x8)__a + (__u16x8)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_add_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_add_sat_s_i16x8((__i16x8)__a, (__i16x8)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_u16x8_add_sat(v128_t __a,
                                                                v128_t __b) {
   return (v128_t)__builtin_wasm_add_sat_u_i16x8((__u16x8)__a, (__u16x8)__b);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS wasm_i16x8_sub(v128_t __a,

 
 static __inline__ v128_t __DEFAULT_FN_ATTRS
 wasm_f32x4_demote_f64x2_zero(v128_t __a) {
   return (v128_t)__builtin_wasm_demote_zero_f64x2_f32x4((__f64x2)__a);
 }
 
 static __inline__ v128_t __DEFAULT_FN_ATTRS
 wasm_f64x2_promote_low_f32x4(v128_t __a) {
-  return (v128_t)__builtin_wasm_promote_low_f32x4_f64x2((__f32x4)__a);
+  return (v128_t) __builtin_convertvector(
+      (__f32x2){((__f32x4)__a)[0], ((__f32x4)__a)[1]}, __f64x2);
 }
 
 #define wasm_i8x16_shuffle(__a, __b, __c0, __c1, __c2, __c3, __c4, __c5, __c6, \
                            __c7, __c8, __c9, __c10, __c11, __c12, __c13,       \
                            __c14, __c15)                                       \
   ((v128_t)__builtin_wasm_shuffle_i8x16(                                       \
       (__i8x16)(__a), (__i8x16)(__b), __c0, __c1, __c2, __c3, __c4, __c5,      \
       __c6, __c7, __c8, __c9, __c10, __c11, __c12, __c13, __c14, __c15))

clang/test/CodeGen/builtins-wasm.c

 }
 
 f32x4 wasm_demote_zero_f64x2_f32x4(f64x2 x) {
   return __builtin_wasm_demote_zero_f64x2_f32x4(x);
   // WEBASSEMBLY: call <4 x float> @llvm.wasm.demote.zero(<2 x double> %x)
   // WEBASSEMBLY: ret
 }
 
-f64x2 wasm_promote_low_f32x4_f64x2(f32x4 x) {
-  return __builtin_wasm_promote_low_f32x4_f64x2(x);
-  // WEBASSEMBLY: call <2 x double> @llvm.wasm.promote.low(<4 x float> %x)
-  // WEBASSEMBLY: ret
-}
-
 i32x4 load32_zero(const int *p) {
   return __builtin_wasm_load32_zero(p);
   // WEBASSEMBLY: call <4 x i32> @llvm.wasm.load32.zero(i32* %p)
   // WEBASSEMBLY: ret
 }
 
 i64x2 load64_zero(const long long *p) {
   return __builtin_wasm_load64_zero(p);

clang/test/Headers/wasm.c

 //
 v128_t test_v128_andnot(v128_t a, v128_t b) {
   return wasm_v128_andnot(a, b);
 }
 
 // CHECK-LABEL: @test_v128_any_true(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.anytrue.v16i8(<16 x i8> [[TMP0]]) #[[ATTR11:[0-9]+]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.anytrue.v16i8(<16 x i8> [[TMP0]]) #[[ATTR10:[0-9]+]]
 // CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-NEXT:    ret i1 [[TOBOOL_I]]
 //
 bool test_v128_any_true(v128_t a) {
   return wasm_v128_any_true(a);
 }
 
 // CHECK-LABEL: @test_v128_bitselect(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <4 x i32> @llvm.wasm.bitselect.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[MASK:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <4 x i32> @llvm.wasm.bitselect.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[MASK:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP0]]
 //
 v128_t test_v128_bitselect(v128_t a, v128_t b, v128_t mask) {
   return wasm_v128_bitselect(a, b, mask);
 }
 
 // CHECK-LABEL: @test_i8x16_abs(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i8> @llvm.abs.v16i8(<16 x i8> [[TMP0]], i1 false) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i8> @llvm.abs.v16i8(<16 x i8> [[TMP0]], i1 false) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i8x16_abs(v128_t a) {
   return wasm_i8x16_abs(a);
 }
 
 // CHECK-LABEL: @test_i8x16_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <16 x i8> zeroinitializer, [[TMP0]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[SUB_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i8x16_neg(v128_t a) {
   return wasm_i8x16_neg(a);
 }
 
 // CHECK-LABEL: @test_i8x16_all_true(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v16i8(<16 x i8> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v16i8(<16 x i8> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-NEXT:    ret i1 [[TOBOOL_I]]
 //
 bool test_i8x16_all_true(v128_t a) {
   return wasm_i8x16_all_true(a);
 }
 
 // CHECK-LABEL: @test_i8x16_bitmask(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v16i8(<16 x i8> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v16i8(<16 x i8> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret i32 [[TMP1]]
 //
 int32_t test_i8x16_bitmask(v128_t a) {
   return wasm_i8x16_bitmask(a);
 }
 
 // CHECK-LABEL: @test_i8x16_popcnt(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i8> @llvm.wasm.popcnt(<16 x i8> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <16 x i8> @llvm.wasm.popcnt(<16 x i8> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i8x16_popcnt(v128_t a) {
   return wasm_i8x16_popcnt(a);
 }
 
 // CHECK-LABEL: @test_i8x16_shl(

 v128_t test_i8x16_add(v128_t a, v128_t b) {
   return wasm_i8x16_add(a, b);
 }
 
 // CHECK-LABEL: @test_i8x16_add_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.sadd.sat.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.sadd.sat.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i8x16_add_sat(v128_t a, v128_t b) {
   return wasm_i8x16_add_sat(a, b);
 }
 
 // CHECK-LABEL: @test_u8x16_add_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.uadd.sat.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.uadd.sat.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u8x16_add_sat(v128_t a, v128_t b) {
   return wasm_u8x16_add_sat(a, b);
 }
 
 // CHECK-LABEL: @test_i8x16_sub(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <16 x i8> [[TMP0]], [[TMP1]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[SUB_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i8x16_sub(v128_t a, v128_t b) {
   return wasm_i8x16_sub(a, b);
 }
 
 // CHECK-LABEL: @test_i8x16_sub_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.sub.sat.signed.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.sub.sat.signed.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i8x16_sub_sat(v128_t a, v128_t b) {
   return wasm_i8x16_sub_sat(a, b);
 }
 
 // CHECK-LABEL: @test_u8x16_sub_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.sub.sat.unsigned.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.sub.sat.unsigned.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u8x16_sub_sat(v128_t a, v128_t b) {
   return wasm_u8x16_sub_sat(a, b);
 }
 
 // CHECK-LABEL: @test_i8x16_min(

 v128_t test_u8x16_max(v128_t a, v128_t b) {
   return wasm_u8x16_max(a, b);
 }
 
 // CHECK-LABEL: @test_u8x16_avgr(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.avgr.unsigned.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.avgr.unsigned.v16i8(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u8x16_avgr(v128_t a, v128_t b) {
   return wasm_u8x16_avgr(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_abs(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.abs.v8i16(<8 x i16> [[TMP0]], i1 false) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.abs.v8i16(<8 x i16> [[TMP0]], i1 false) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i16> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i16x8_abs(v128_t a) {
   return wasm_i16x8_abs(a);
 }
 
 // CHECK-LABEL: @test_i16x8_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <8 x i16> zeroinitializer, [[TMP0]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[SUB_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i16x8_neg(v128_t a) {
   return wasm_i16x8_neg(a);
 }
 
 // CHECK-LABEL: @test_i16x8_all_true(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v8i16(<8 x i16> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v8i16(<8 x i16> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-NEXT:    ret i1 [[TOBOOL_I]]
 //
 bool test_i16x8_all_true(v128_t a) {
   return wasm_i16x8_all_true(a);
 }
 
 // CHECK-LABEL: @test_i16x8_bitmask(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v8i16(<8 x i16> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v8i16(<8 x i16> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret i32 [[TMP1]]
 //
 int32_t test_i16x8_bitmask(v128_t a) {
   return wasm_i16x8_bitmask(a);
 }
 
 // CHECK-LABEL: @test_i16x8_shl(
 // CHECK-NEXT:  entry:

 v128_t test_i16x8_add(v128_t a, v128_t b) {
   return wasm_i16x8_add(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_add_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i16x8_add_sat(v128_t a, v128_t b) {
   return wasm_i16x8_add_sat(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_add_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.uadd.sat.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.uadd.sat.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u16x8_add_sat(v128_t a, v128_t b) {
   return wasm_u16x8_add_sat(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_sub(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <8 x i16> [[TMP0]], [[TMP1]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i16> [[SUB_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i16x8_sub(v128_t a, v128_t b) {
   return wasm_i16x8_sub(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_sub_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.sub.sat.signed.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.sub.sat.signed.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i16x8_sub_sat(v128_t a, v128_t b) {
   return wasm_i16x8_sub_sat(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_sub_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.sub.sat.unsigned.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.sub.sat.unsigned.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u16x8_sub_sat(v128_t a, v128_t b) {
   return wasm_u16x8_sub_sat(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_mul(

 v128_t test_u16x8_max(v128_t a, v128_t b) {
   return wasm_u16x8_max(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_avgr(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.avgr.unsigned.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.avgr.unsigned.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u16x8_avgr(v128_t a, v128_t b) {
   return wasm_u16x8_avgr(a, b);
 }
 
 // CHECK-LABEL: @test_i32x4_abs(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[A:%.*]], i1 false) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[A:%.*]], i1 false) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP0]]
 //
 v128_t test_i32x4_abs(v128_t a) {
   return wasm_i32x4_abs(a);
 }
 
 // CHECK-LABEL: @test_i32x4_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <4 x i32> zeroinitializer, [[A:%.*]]
 // CHECK-NEXT:    ret <4 x i32> [[SUB_I]]
 //
 v128_t test_i32x4_neg(v128_t a) {
   return wasm_i32x4_neg(a);
 }
 
 // CHECK-LABEL: @test_i32x4_all_true(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.wasm.alltrue.v4i32(<4 x i32> [[A:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.wasm.alltrue.v4i32(<4 x i32> [[A:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp ne i32 [[TMP0]], 0
 // CHECK-NEXT:    ret i1 [[TOBOOL_I]]
 //
 bool test_i32x4_all_true(v128_t a) {
   return wasm_i32x4_all_true(a);
 }
 
 // CHECK-LABEL: @test_i32x4_bitmask(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.wasm.bitmask.v4i32(<4 x i32> [[A:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call i32 @llvm.wasm.bitmask.v4i32(<4 x i32> [[A:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    ret i32 [[TMP0]]
 //
 int32_t test_i32x4_bitmask(v128_t a) {
   return wasm_i32x4_bitmask(a);
 }
 
 // CHECK-LABEL: @test_i32x4_shl(
 // CHECK-NEXT:  entry:

 v128_t test_u32x4_max(v128_t a, v128_t b) {
   return wasm_u32x4_max(a, b);
 }
 
 // CHECK-LABEL: @test_i32x4_dot_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.dot(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.dot(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i32x4_dot_i16x8(v128_t a, v128_t b) {
   return wasm_i32x4_dot_i16x8(a, b);
 }
 
 // CHECK-LABEL: @test_i64x2_abs(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[TMP0]], i1 false) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[TMP0]], i1 false) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i64x2_abs(v128_t a) {
   return wasm_i64x2_abs(a);
 }
 
 // CHECK-LABEL: @test_i64x2_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x i64>
 // CHECK-NEXT:    [[SUB_I:%.*]] = sub <2 x i64> zeroinitializer, [[TMP0]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[SUB_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i64x2_neg(v128_t a) {
   return wasm_i64x2_neg(a);
 }
 
 // CHECK-LABEL: @test_i64x2_all_true(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v2i64(<2 x i64> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.alltrue.v2i64(<2 x i64> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK-NEXT:    ret i1 [[TOBOOL_I]]
 //
 bool test_i64x2_all_true(v128_t a) {
   return wasm_i64x2_all_true(a);
 }
 
 // CHECK-LABEL: @test_i64x2_bitmask(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v2i64(<2 x i64> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call i32 @llvm.wasm.bitmask.v2i64(<2 x i64> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret i32 [[TMP1]]
 //
 int32_t test_i64x2_bitmask(v128_t a) {
   return wasm_i64x2_bitmask(a);
 }
 
 // CHECK-LABEL: @test_i64x2_shl(
 // CHECK-NEXT:  entry:

 //
 v128_t test_i64x2_mul(v128_t a, v128_t b) {
   return wasm_i64x2_mul(a, b);
 }
 
 // CHECK-LABEL: @test_f32x4_abs(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_abs(v128_t a) {
   return wasm_f32x4_abs(a);
 }
 
 // CHECK-LABEL: @test_f32x4_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
 // CHECK-NEXT:    [[FNEG_I:%.*]] = fneg <4 x float> [[TMP0]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[FNEG_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_f32x4_neg(v128_t a) {
   return wasm_f32x4_neg(a);
 }
 
 // CHECK-LABEL: @test_f32x4_sqrt(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_sqrt(v128_t a) {
   return wasm_f32x4_sqrt(a);
 }
 
 // CHECK-LABEL: @test_f32x4_ceil(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_ceil(v128_t a) {
   return wasm_f32x4_ceil(a);
 }
 
 // CHECK-LABEL: @test_f32x4_floor(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_floor(v128_t a) {
   return wasm_f32x4_floor(a);
 }
 
 // CHECK-LABEL: @test_f32x4_trunc(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_trunc(v128_t a) {
   return wasm_f32x4_trunc(a);
 }
 
 // CHECK-LABEL: @test_f32x4_nearest(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_nearest(v128_t a) {
   return wasm_f32x4_nearest(a);
 }
 
 // CHECK-LABEL: @test_f32x4_add(

 v128_t test_f32x4_div(v128_t a, v128_t b) {
   return wasm_f32x4_div(a, b);
 }
 
 // CHECK-LABEL: @test_f32x4_min(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.minimum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.minimum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f32x4_min(v128_t a, v128_t b) {
   return wasm_f32x4_min(a, b);
 }
 
 // CHECK-LABEL: @test_f32x4_max(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.maximum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.maximum.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f32x4_max(v128_t a, v128_t b) {
   return wasm_f32x4_max(a, b);
 }
 
 // CHECK-LABEL: @test_f32x4_pmin(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.wasm.pmin.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.wasm.pmin.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f32x4_pmin(v128_t a, v128_t b) {
   return wasm_f32x4_pmin(a, b);
 }
 
 // CHECK-LABEL: @test_f32x4_pmax(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.wasm.pmax.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x float> @llvm.wasm.pmax.v4f32(<4 x float> [[TMP0]], <4 x float> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x float> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f32x4_pmax(v128_t a, v128_t b) {
   return wasm_f32x4_pmax(a, b);
 }
 
 // CHECK-LABEL: @test_f64x2_abs(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_abs(v128_t a) {
   return wasm_f64x2_abs(a);
 }
 
 // CHECK-LABEL: @test_f64x2_neg(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
 // CHECK-NEXT:    [[FNEG_I:%.*]] = fneg <2 x double> [[TMP0]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[FNEG_I]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_f64x2_neg(v128_t a) {
   return wasm_f64x2_neg(a);
 }
 
 // CHECK-LABEL: @test_f64x2_sqrt(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_sqrt(v128_t a) {
   return wasm_f64x2_sqrt(a);
 }
 
 // CHECK-LABEL: @test_f64x2_ceil(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_ceil(v128_t a) {
   return wasm_f64x2_ceil(a);
 }
 
 // CHECK-LABEL: @test_f64x2_floor(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_floor(v128_t a) {
   return wasm_f64x2_floor(a);
 }
 
 // CHECK-LABEL: @test_f64x2_trunc(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_trunc(v128_t a) {
   return wasm_f64x2_trunc(a);
 }
 
 // CHECK-LABEL: @test_f64x2_nearest(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f64x2_nearest(v128_t a) {
   return wasm_f64x2_nearest(a);
 }
 
 // CHECK-LABEL: @test_f64x2_add(

 v128_t test_f64x2_div(v128_t a, v128_t b) {
   return wasm_f64x2_div(a, b);
 }
 
 // CHECK-LABEL: @test_f64x2_min(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.minimum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.minimum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f64x2_min(v128_t a, v128_t b) {
   return wasm_f64x2_min(a, b);
 }
 
 // CHECK-LABEL: @test_f64x2_max(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.maximum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.maximum.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f64x2_max(v128_t a, v128_t b) {
   return wasm_f64x2_max(a, b);
 }
 
 // CHECK-LABEL: @test_f64x2_pmin(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.wasm.pmin.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.wasm.pmin.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f64x2_pmin(v128_t a, v128_t b) {
   return wasm_f64x2_pmin(a, b);
 }
 
 // CHECK-LABEL: @test_f64x2_pmax(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.wasm.pmax.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <2 x double> @llvm.wasm.pmax.v2f64(<2 x double> [[TMP0]], <2 x double> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <2 x double> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_f64x2_pmax(v128_t a, v128_t b) {
   return wasm_f64x2_pmax(a, b);
 }
 
 // CHECK-LABEL: @test_i32x4_trunc_sat_f32x4(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.fptosi.sat.v4i32.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.fptosi.sat.v4i32.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i32x4_trunc_sat_f32x4(v128_t a) {
   return wasm_i32x4_trunc_sat_f32x4(a);
 }
 
 // CHECK-LABEL: @test_u32x4_trunc_sat_f32x4(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_u32x4_trunc_sat_f32x4(v128_t a) {
   return wasm_u32x4_trunc_sat_f32x4(a);
 }
 
 // CHECK-LABEL: @test_f32x4_convert_i32x4(
 // CHECK-NEXT:  entry:

 //
 v128_t test_f64x2_convert_low_u32x4(v128_t a) {
   return wasm_f64x2_convert_low_u32x4(a);
 }
 
 // CHECK-LABEL: @test_i32x4_trunc_sat_f64x2_zero(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i32> @llvm.fptosi.sat.v2i32.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i32> @llvm.fptosi.sat.v2i32.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i32x4_trunc_sat_f64x2_zero(v128_t a) {
   return wasm_i32x4_trunc_sat_f64x2_zero(a);
 }
 
 // CHECK-LABEL: @test_u32x4_trunc_sat_f64x2_zero(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i32> @llvm.fptoui.sat.v2i32.v2f64(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x i32> @llvm.fptoui.sat.v2i32.v2f64(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_u32x4_trunc_sat_f64x2_zero(v128_t a) {
   return wasm_u32x4_trunc_sat_f64x2_zero(a);
 }
 
 // CHECK-LABEL: @test_f32x4_demote_f64x2_zero(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.wasm.demote.zero(<2 x double> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x float> @llvm.wasm.demote.zero(<2 x double> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <4 x float> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_f32x4_demote_f64x2_zero(v128_t a) {
   return wasm_f32x4_demote_f64x2_zero(a);
 }
 
 // CHECK-LABEL: @test_f64x2_promote_low_f32x4(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <2 x double> @llvm.wasm.promote.low(<4 x float> [[TMP0]]) #[[ATTR11]]
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    ret <4 x i32> [[TMP2]]
+// CHECK-NEXT:    [[VECINIT2_I:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> undef, <2 x i32> <i32 0, i32 1>
+// CHECK-NEXT:    [[CONV_I:%.*]] = fpext <2 x float> [[VECINIT2_I]] to <2 x double>
+// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[CONV_I]] to <4 x i32>
+// CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_f64x2_promote_low_f32x4(v128_t a) {
   return wasm_f64x2_promote_low_f32x4(a);
 }
 
 // CHECK-LABEL: @test_i8x16_shuffle(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>

 v128_t test_i64x2_shuffle(v128_t a, v128_t b) {
   return wasm_i64x2_shuffle(a, b, 1, 0);
 }
 
 // CHECK-LABEL: @test_i8x16_swizzle(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.swizzle(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.swizzle(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i8x16_swizzle(v128_t a, v128_t b) {
   return wasm_i8x16_swizzle(a, b);
 }
 
 // CHECK-LABEL: @test_i8x16_narrow_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.narrow.signed.v16i8.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.narrow.signed.v16i8.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i8x16_narrow_i16x8(v128_t a, v128_t b) {
   return wasm_i8x16_narrow_i16x8(a, b);
 }
 
 // CHECK-LABEL: @test_u8x16_narrow_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.narrow.unsigned.v16i8.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <16 x i8> @llvm.wasm.narrow.unsigned.v16i8.v8i16(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <16 x i8> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u8x16_narrow_i16x8(v128_t a, v128_t b) {
   return wasm_u8x16_narrow_i16x8(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_narrow_i32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <8 x i16> @llvm.wasm.narrow.signed.v8i16.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <8 x i16> @llvm.wasm.narrow.signed.v8i16.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i16x8_narrow_i32x4(v128_t a, v128_t b) {
   return wasm_i16x8_narrow_i32x4(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_narrow_i32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <8 x i16> @llvm.wasm.narrow.unsigned.v8i16.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <8 x i16> @llvm.wasm.narrow.unsigned.v8i16.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_u16x8_narrow_i32x4(v128_t a, v128_t b) {
   return wasm_u16x8_narrow_i32x4(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_extend_low_i8x16(

 //
 v128_t test_u64x2_extend_high_u32x4(v128_t a) {
   return wasm_u64x2_extend_high_u32x4(a);
 }
 
 // CHECK-LABEL: @test_i16x8_extadd_pairwise_i8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.wasm.extadd.pairwise.signed.v8i16(<16 x i8> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.wasm.extadd.pairwise.signed.v8i16(<16 x i8> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i16> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i16x8_extadd_pairwise_i8x16(v128_t a) {
   return wasm_i16x8_extadd_pairwise_i8x16(a);
 }
 
 // CHECK-LABEL: @test_u16x8_extadd_pairwise_u8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.wasm.extadd.pairwise.unsigned.v8i16(<16 x i8> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <8 x i16> @llvm.wasm.extadd.pairwise.unsigned.v8i16(<16 x i8> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i16> [[TMP1]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_u16x8_extadd_pairwise_u8x16(v128_t a) {
   return wasm_u16x8_extadd_pairwise_u8x16(a);
 }
 
 // CHECK-LABEL: @test_i32x4_extadd_pairwise_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.wasm.extadd.pairwise.signed.v4i32(<8 x i16> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.wasm.extadd.pairwise.signed.v4i32(<8 x i16> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i32x4_extadd_pairwise_i16x8(v128_t a) {
   return wasm_i32x4_extadd_pairwise_i16x8(a);
 }
 
 // CHECK-LABEL: @test_u32x4_extadd_pairwise_u16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.wasm.extadd.pairwise.unsigned.v4i32(<8 x i16> [[TMP0]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <4 x i32> @llvm.wasm.extadd.pairwise.unsigned.v4i32(<8 x i16> [[TMP0]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_u32x4_extadd_pairwise_u16x8(v128_t a) {
   return wasm_u32x4_extadd_pairwise_u16x8(a);
 }
 
 // CHECK-LABEL: @test_i16x8_extmul_low_i8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.low.signed.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.low.signed.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i16x8_extmul_low_i8x16(v128_t a, v128_t b) {
   return wasm_i16x8_extmul_low_i8x16(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_extmul_high_i8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.high.signed.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.high.signed.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i16x8_extmul_high_i8x16(v128_t a, v128_t b) {
   return wasm_i16x8_extmul_high_i8x16(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_extmul_low_u8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.low.unsigned.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.low.unsigned.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u16x8_extmul_low_u8x16(v128_t a, v128_t b) {
   return wasm_u16x8_extmul_low_u8x16(a, b);
 }
 
 // CHECK-LABEL: @test_u16x8_extmul_high_u8x16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <16 x i8>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.high.unsigned.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.extmul.high.unsigned.v8i16(<16 x i8> [[TMP0]], <16 x i8> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_u16x8_extmul_high_u8x16(v128_t a, v128_t b) {
   return wasm_u16x8_extmul_high_u8x16(a, b);
 }
 
 // CHECK-LABEL: @test_i32x4_extmul_low_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.low.signed.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.low.signed.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i32x4_extmul_low_i16x8(v128_t a, v128_t b) {
   return wasm_i32x4_extmul_low_i16x8(a, b);
 }
 
 // CHECK-LABEL: @test_i32x4_extmul_high_i16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.high.signed.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.high.signed.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_i32x4_extmul_high_i16x8(v128_t a, v128_t b) {
   return wasm_i32x4_extmul_high_i16x8(a, b);
 }
 
 // CHECK-LABEL: @test_u32x4_extmul_low_u16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.low.unsigned.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.low.unsigned.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_u32x4_extmul_low_u16x8(v128_t a, v128_t b) {
   return wasm_u32x4_extmul_low_u16x8(a, b);
 }
 
 // CHECK-LABEL: @test_u32x4_extmul_high_u16x8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.high.unsigned.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <4 x i32> @llvm.wasm.extmul.high.unsigned.v4i32(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 //
 v128_t test_u32x4_extmul_high_u16x8(v128_t a, v128_t b) {
   return wasm_u32x4_extmul_high_u16x8(a, b);
 }
 
 // CHECK-LABEL: @test_i64x2_extmul_low_i32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.low.signed.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.low.signed.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i64x2_extmul_low_i32x4(v128_t a, v128_t b) {
   return wasm_i64x2_extmul_low_i32x4(a, b);
 }
 
 // CHECK-LABEL: @test_i64x2_extmul_high_i32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.high.signed.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.high.signed.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_i64x2_extmul_high_i32x4(v128_t a, v128_t b) {
   return wasm_i64x2_extmul_high_i32x4(a, b);
 }
 
 // CHECK-LABEL: @test_u64x2_extmul_low_u32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.low.unsigned.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.low.unsigned.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_u64x2_extmul_low_u32x4(v128_t a, v128_t b) {
   return wasm_u64x2_extmul_low_u32x4(a, b);
 }
 
 // CHECK-LABEL: @test_u64x2_extmul_high_u32x4(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.high.unsigned.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <2 x i64> @llvm.wasm.extmul.high.unsigned.v2i64(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[TMP0]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 //
 v128_t test_u64x2_extmul_high_u32x4(v128_t a, v128_t b) {
   return wasm_u64x2_extmul_high_u32x4(a, b);
 }
 
 // CHECK-LABEL: @test_i16x8_q15mulr_sat(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to <8 x i16>
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <8 x i16>
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.q15mulr.sat.signed(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR11]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <8 x i16> @llvm.wasm.q15mulr.sat.signed(<8 x i16> [[TMP0]], <8 x i16> [[TMP1]]) #[[ATTR10]]
 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i16> [[TMP2]] to <4 x i32>
 // CHECK-NEXT:    ret <4 x i32> [[TMP3]]
 //
 v128_t test_i16x8_q15mulr_sat(v128_t a, v128_t b) {
   return wasm_i16x8_q15mulr_sat(a, b);
 }

llvm/include/llvm/IR/IntrinsicsWebAssembly.td

def int_wasm_extadd_pairwise_signed :
   Intrinsic<[llvm_anyvector_ty],
             [LLVMSubdivide2VectorType<0>],
             [IntrNoMem, IntrSpeculatable]>;
 def int_wasm_extadd_pairwise_unsigned :
   Intrinsic<[llvm_anyvector_ty],
             [LLVMSubdivide2VectorType<0>],
             [IntrNoMem, IntrSpeculatable]>;
 
-// TODO: Remove these if possible if they are merged to the spec.
+// TODO: Remove this if possible.
 def int_wasm_demote_zero :
   Intrinsic<[llvm_v4f32_ty], [llvm_v2f64_ty],
             [IntrNoMem, IntrSpeculatable]>;
-def int_wasm_promote_low :
-  Intrinsic<[llvm_v2f64_ty], [llvm_v4f32_ty],
-            [IntrNoMem, IntrSpeculatable]>;
 
 //===----------------------------------------------------------------------===//
 // Thread-local storage intrinsics
 //===----------------------------------------------------------------------===//
 
 def int_wasm_tls_size :
   Intrinsic<[llvm_anyint_ty],
             [],

llvm/lib/Target/WebAssembly/WebAssemblyISD.def

 HANDLE_NODETYPE(VEC_SHR_S)
 HANDLE_NODETYPE(VEC_SHR_U)
 HANDLE_NODETYPE(EXTEND_LOW_S)
 HANDLE_NODETYPE(EXTEND_LOW_U)
 HANDLE_NODETYPE(EXTEND_HIGH_S)
 HANDLE_NODETYPE(EXTEND_HIGH_U)
 HANDLE_NODETYPE(CONVERT_LOW_S)
 HANDLE_NODETYPE(CONVERT_LOW_U)
+HANDLE_NODETYPE(PROMOTE_LOW)
 HANDLE_NODETYPE(TRUNC_SAT_ZERO_S)
 HANDLE_NODETYPE(TRUNC_SAT_ZERO_U)
 HANDLE_NODETYPE(THROW)
 HANDLE_NODETYPE(CATCH)
 HANDLE_NODETYPE(MEMORY_COPY)
 HANDLE_NODETYPE(MEMORY_FILL)
 
 // Memory intrinsics
 HANDLE_MEM_NODETYPE(LOAD_SPLAT)
 HANDLE_MEM_NODETYPE(GLOBAL_GET)
 HANDLE_MEM_NODETYPE(GLOBAL_SET)

llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

WebAssemblyTargetLowering::WebAssemblyTargetLowering(
   if (Subtarget->hasSIMD128()) {
     // Hoist bitcasts out of shuffles
     setTargetDAGCombine(ISD::VECTOR_SHUFFLE);
 
     // Combine extends of extract_subvectors into widening ops
     setTargetDAGCombine(ISD::SIGN_EXTEND);
     setTargetDAGCombine(ISD::ZERO_EXTEND);
 
-    // Combine int_to_fp of extract_vectors and vice versa into conversions ops
+    // Combine int_to_fp or fp_extend of extract_vectors and vice versa into
+    // conversions ops
     setTargetDAGCombine(ISD::SINT_TO_FP);
     setTargetDAGCombine(ISD::UINT_TO_FP);
+    setTargetDAGCombine(ISD::FP_EXTEND);
     setTargetDAGCombine(ISD::EXTRACT_SUBVECTOR);
 
     // Combine concat of {s,u}int_to_fp_sat to i32x4.trunc_sat_f64x2_zero_{s,u}
     setTargetDAGCombine(ISD::CONCAT_VECTORS);
 
     // Support saturating add for i8x16 and i16x8
     for (auto Op : {ISD::SADDSAT, ISD::UADDSAT})
       for (auto T : {MVT::v16i8, MVT::v8i16})

 performVectorConvertLowCombine(SDNode *N,
                                TargetLowering::DAGCombinerInfo &DCI) {
   auto &DAG = DCI.DAG;
 
   EVT ResVT = N->getValueType(0);
   if (ResVT != MVT::v2f64)
     return SDValue();
 
-  if (N->getOpcode() == ISD::SINT_TO_FP || N->getOpcode() == ISD::UINT_TO_FP) {
+  auto GetWasmConversionOp = [](unsigned Op) {
+    switch (Op) {
+    case ISD::SINT_TO_FP:
+      return WebAssemblyISD::CONVERT_LOW_S;
+    case ISD::UINT_TO_FP:
+      return WebAssemblyISD::CONVERT_LOW_U;
+    case ISD::FP_EXTEND:
+      return WebAssemblyISD::PROMOTE_LOW;
+    }
+    llvm_unreachable("unexpected op");
+  };
+
+  if (N->getOpcode() == ISD::EXTRACT_SUBVECTOR) {
     // Combine this:
     //
-    //   (v2f64 ({s,u}int_to_fp
-    //     (v2i32 (extract_subvector (v4i32 $x), 0))))
+    //   (v2f64 (extract_subvector
+    //     (v4f64 ({s,u}int_to_fp (v4i32 $x))), 0))
     //
     // into (f64x2.convert_low_i32x4_{s,u} $x).
-    auto Extract = N->getOperand(0);
-    if (Extract.getOpcode() != ISD::EXTRACT_SUBVECTOR)
+    //
+    // Or this:
+    //
+    //  (v2f64 (extract_subvector
+    //    (v4f64 (fp_extend (v4f32 $x))), 0))
+    //
+    // into (f64x2.promote_low_f32x4 $x).
+    auto Conversion = N->getOperand(0);
+    auto ConversionOp = Conversion.getOpcode();
+    MVT ExpectedSourceType;
+    switch (ConversionOp) {
+    case ISD::SINT_TO_FP:
+    case ISD::UINT_TO_FP:
+      ExpectedSourceType = MVT::v4i32;
+      break;
+    case ISD::FP_EXTEND:
+      ExpectedSourceType = MVT::v4f32;
+      break;
+    default:
       return SDValue();
-    if (Extract.getValueType() != MVT::v2i32)
+    }
+
+    if (Conversion.getValueType() != MVT::v4f64)
       return SDValue();
-    auto Source = Extract.getOperand(0);
-    if (Source.getValueType() != MVT::v4i32)
+
+    auto Source = Conversion.getOperand(0);
+    if (Source.getValueType() != ExpectedSourceType)
       return SDValue();
-    auto *IndexNode = dyn_cast<ConstantSDNode>(Extract.getOperand(1));
+
+    auto IndexNode = dyn_cast<ConstantSDNode>(N->getOperand(1));
     if (IndexNode == nullptr || IndexNode->getZExtValue() != 0)
       return SDValue();
 
-    unsigned Op = N->getOpcode() == ISD::SINT_TO_FP
-                      ? WebAssemblyISD::CONVERT_LOW_S
-                      : WebAssemblyISD::CONVERT_LOW_U;
-
+    auto Op = GetWasmConversionOp(ConversionOp);
     return DAG.getNode(Op, SDLoc(N), ResVT, Source);
+  }
 
-  } else if (N->getOpcode() == ISD::EXTRACT_SUBVECTOR) {
-    // Combine this:
-    //
-    //   (v2f64 (extract_subvector
-    //     (v4f64 ({s,u}int_to_fp (v4i32 $x))), 0))
-    //
-    // into (f64x2.convert_low_i32x4_{s,u} $x).
-    auto IntToFP = N->getOperand(0);
-    if (IntToFP.getOpcode() != ISD::SINT_TO_FP &&
-        IntToFP.getOpcode() != ISD::UINT_TO_FP)
-      return SDValue();
-    if (IntToFP.getValueType() != MVT::v4f64)
-      return SDValue();
-    auto Source = IntToFP.getOperand(0);
-    if (Source.getValueType() != MVT::v4i32)
-      return SDValue();
-    auto IndexNode = dyn_cast<ConstantSDNode>(N->getOperand(1));
-    if (IndexNode == nullptr || IndexNode->getZExtValue() != 0)
-      return SDValue();
+  //   (v2f64 ({s,u}int_to_fp
+  //     (v2i32 (extract_subvector (v4i32 $x), 0))))
+  //
+  // Or this:
+  //
+  //   (v2f64 (fp_extend
+  //     (v2f32 (extract_subvector (v4f32 $x), 0))))
+  //
+  // into (f64x2.promote_low_f32x4 $x).
+  auto ConversionOp = N->getOpcode();
+  MVT ExpectedExtractType;
+  MVT ExpectedSourceType;
+  switch (ConversionOp) {
+  case ISD::SINT_TO_FP:
+  case ISD::UINT_TO_FP:
+    ExpectedExtractType = MVT::v2i32;
+    ExpectedSourceType = MVT::v4i32;
+    break;
+  case ISD::FP_EXTEND:
+    ExpectedExtractType = MVT::v2f32;
+    ExpectedSourceType = MVT::v4f32;
+    break;
+  default:
+    llvm_unreachable("unexpected opcode");
+  }
+
+  auto Extract = N->getOperand(0);
+  if (Extract.getOpcode() != ISD::EXTRACT_SUBVECTOR)
+
+  if (Extract.getValueType() != ExpectedExtractType)
+
+  auto Source = Extract.getOperand(0);
+  if (Source.getValueType() != ExpectedSourceType)
+
+  auto *IndexNode = dyn_cast<ConstantSDNode>(Extract.getOperand(1));
 
-    unsigned Op = IntToFP->getOpcode() == ISD::SINT_TO_FP
-                      ? WebAssemblyISD::CONVERT_LOW_S
-                      : WebAssemblyISD::CONVERT_LOW_U;
-
-    return DAG.getNode(Op, SDLoc(N), ResVT, Source);
-
-  } else {
-    llvm_unreachable("unexpected opcode");
-  }
+  unsigned Op = GetWasmConversionOp(ConversionOp);
 }
 
 static SDValue
 performVectorTruncSatLowCombine(SDNode *N,
                                 TargetLowering::DAGCombinerInfo &DCI) {
   auto &DAG = DCI.DAG;
   assert(N->getOpcode() == ISD::CONCAT_VECTORS);
 
default:
     return SDValue();
   case ISD::VECTOR_SHUFFLE:
     return performVECTOR_SHUFFLECombine(N, DCI);
   case ISD::SIGN_EXTEND:
   case ISD::ZERO_EXTEND:
     return performVectorExtendCombine(N, DCI);
   case ISD::SINT_TO_FP:
   case ISD::UINT_TO_FP:
+  case ISD::FP_EXTEND:
   case ISD::EXTRACT_SUBVECTOR:
     return performVectorConvertLowCombine(N, DCI);
   case ISD::CONCAT_VECTORS:
     return performVectorTruncSatLowCombine(N, DCI);
   }
 }

llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td

defm "" : SIMDConvert<I16x8, I8x16, int_wasm_extadd_pairwise_signed,
                       "extadd_pairwise_i8x16_s", 0x7c>;
 defm "" : SIMDConvert<I16x8, I8x16, int_wasm_extadd_pairwise_unsigned,
                       "extadd_pairwise_i8x16_u", 0x7d>;
 defm "" : SIMDConvert<I32x4, I16x8, int_wasm_extadd_pairwise_signed,
                       "extadd_pairwise_i16x8_s", 0x7e>;
 defm "" : SIMDConvert<I32x4, I16x8, int_wasm_extadd_pairwise_unsigned,
                       "extadd_pairwise_i16x8_u", 0x7f>;
 
-// Prototype f64x2 conversions
+// f64x2 <-> f32x4 conversions
 defm "" : SIMDConvert<F32x4, F64x2, int_wasm_demote_zero,
                       "demote_zero_f64x2", 0x5e>;
-defm "" : SIMDConvert<F64x2, F32x4, int_wasm_promote_low,
-                      "promote_low_f32x4", 0x5f>;
+
+def promote_t : SDTypeProfile<1, 1, [SDTCisVec<0>, SDTCisVec<1>]>;
+def promote_low : SDNode<"WebAssemblyISD::PROMOTE_LOW", promote_t>;
+defm "" : SIMDConvert<F64x2, F32x4, promote_low, "promote_low_f32x4", 0x5f>;
 
 //===----------------------------------------------------------------------===//
 // Saturating Rounding Q-Format Multiplication
 //===----------------------------------------------------------------------===//
 
 defm Q15MULR_SAT_S :
   SIMDBinary<I16x8, int_wasm_q15mulr_sat_signed, "q15mulr_sat_s", 0x82>;

llvm/test/CodeGen/WebAssembly/simd-conversions.ll

 ; SIMD128-NEXT: .functype convert_low_u_v2f64_2 (v128) -> (v128){{$}}
 ; SIMD128-NEXT: f64x2.convert_low_i32x4_u $push[[R:[0-9]+]]=, $0
 ; SIMD128-NEXT: return $pop[[R]]
 define <2 x double> @convert_low_u_v2f64_2(<4 x i32> %x) {
   %v = uitofp <4 x i32> %x to <4 x double>
   %a = shufflevector <4 x double> %v, <4 x double> undef, <2 x i32> <i32 0, i32 1>
   ret <2 x double> %a
 }
+
+; CHECK-LABEL: promote_low_v2f64:
+; NO-SIMD128-NOT: f64x2
+; SIMD128-NEXT: .functype promote_low_v2f64 (v128) -> (v128){{$}}
+; SIMD128-NEXT: f64x2.promote_low_f32x4 $push[[R:[0-9]+]]=, $0
+; SIMD128-NEXT: return $pop[[R]]
+define <2 x double> @promote_low_v2f64(<4 x float> %x) {
+  %v = shufflevector <4 x float> %x, <4 x float> undef, <2 x i32> <i32 0, i32 1>
+  %a = fpext <2 x float> %v to <2 x double>
+  ret <2 x double> %a
+}
+
+; CHECK-LABEL: promote_low_v2f64_2:
+; NO-SIMD128-NOT: f64x2
+; SIMD128-NEXT: .functype promote_low_v2f64_2 (v128) -> (v128){{$}}
+; SIMD128-NEXT: f64x2.promote_low_f32x4 $push[[R:[0-9]+]]=, $0
+; SIMD128-NEXT: return $pop[[R]]
+define <2 x double> @promote_low_v2f64_2(<4 x float> %x) {
+  %v = fpext <4 x float> %x to <4 x double>
+  %a = shufflevector <4 x double> %v, <4 x double> undef, <2 x i32> <i32 0, i32 1>
+  ret <2 x double> %a
+}

llvm/test/CodeGen/WebAssembly/simd-intrinsics.ll

 ; CHECK-NEXT: .functype nearest_v2f64 (v128) -> (v128){{$}}
 ; CHECK-NEXT: f64x2.nearest $push[[R:[0-9]+]]=, $0{{$}}
 ; CHECK-NEXT: return $pop[[R]]{{$}}
 declare <2 x double> @llvm.nearbyint.v2f64(<2 x double>)
 define <2 x double> @nearest_v2f64(<2 x double> %a) {
   %v = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> %a)
   ret <2 x double> %v
 }
-
-; CHECK-LABEL: promote_low_v2f64:
-; CHECK-NEXT: .functype promote_low_v2f64 (v128) -> (v128){{$}}
-; CHECK-NEXT: f64x2.promote_low_f32x4 $push[[R:[0-9]+]]=, $0{{$}}
-; CHECK-NEXT: return $pop[[R]]{{$}}
-declare <2 x double> @llvm.wasm.promote.low(<4 x float>)
-define <2 x double> @promote_low_v2f64(<4 x float> %a) {
-  %v = call <2 x double> @llvm.wasm.promote.low(<4 x float> %a)
-  ret <2 x double> %v
-}