Lowering x86 adds/addus/subs/subus intrinsics (clang)

[tkrupa]

Summary

This is the patch that lowers x86 intrinsics to native IR
in order to enable optimizations.
Corresponding llvm part: https://reviews.llvm.org/D44785

[craig.topper]

lib/CodeGen/CGBuiltin.cpp

7897

Can we just create the constant in the extended type?

7910

ConstantInt::get should be able to splat on its own if you give it a vector type. Same with getAllOnesValue above

8922

Substraction->Subtraction

8929

Substraction->Subtraction

[tkrupa]

lib/CodeGen/CGBuiltin.cpp

7897

We can't - SignedMaxValue is created in non-extended type and llvm::ConstantInt::get makes an assertion that the input is of the same type. We can perform sext() on SignedMaxValue but the result will be the same.

[craig.topper]

LGTM with that one comment.

lib/CodeGen/CGBuiltin.cpp

7881

You could probably just do ResultType->getScalarSizeInBits() == 8

lib/CodeGen/CGBuiltin.cpp

 
 static Value *EmitX86SExtMask(CodeGenFunction &CGF, Value *Op, 
                               llvm::Type *DstTy) {
   unsigned NumberOfElements = DstTy->getVectorNumElements();
   Value *Mask = getMaskVecValue(CGF, Op, NumberOfElements);
   return CGF.Builder.CreateSExt(Mask, DstTy, "vpmovm2");
 }
 
+// Emit addition with saturation; handles both signed and unsigned intrinsics.
+static Value *EmitX86AddSubSatExpr(CodeGenFunction &CGF, const CallExpr *E,
+                                   SmallVectorImpl<Value *> &Ops,
+                                   bool IsAddition, bool Signed) {
+
+  // Collect vector elements and type data.
+  llvm::Type *ResultType = CGF.ConvertType(E->getType());
+  int NumElements = ResultType->getVectorNumElements();
+
+  llvm::Type *ElementType = ResultType->getVectorElementType();
+  llvm::Type *ExtElementType = ElementType == CGF.Builder.getInt8Ty()
+                                   ? CGF.Builder.getInt16Ty()
+                                   : CGF.Builder.getInt32Ty();
+
+  // Extending vectors to next possible width to make space for possible
+  // overflow.
+  llvm::Type *ExtType = llvm::VectorType::get(ExtElementType, NumElements);
+  Value *VecA = Signed ? CGF.Builder.CreateSExt(Ops[0], ExtType)
+                       : CGF.Builder.CreateZExt(Ops[0], ExtType);
+  Value *VecB = Signed ? CGF.Builder.CreateSExt(Ops[1], ExtType)
+                       : CGF.Builder.CreateZExt(Ops[1], ExtType);
+
+  llvm::Value *ExtProduct = IsAddition ? CGF.Builder.CreateAdd(VecA, VecB)
+                                       : CGF.Builder.CreateSub(VecA, VecB);
+
+  // Create vector of the same type as expected result with max possible values
+  // and extend it to the same type as the product of the addition.
+  APInt SignedMaxValue =
+      llvm::APInt::getSignedMaxValue(ElementType->getPrimitiveSizeInBits());
+  Value *Max = Signed ? llvm::ConstantInt::get(ElementType, SignedMaxValue)
+                      : llvm::Constant::getAllOnesValue(ElementType);
+  Value *MaxVec = CGF.Builder.CreateVectorSplat(NumElements, Max);
+  Value *ExtMaxVec = Signed ? CGF.Builder.CreateSExt(MaxVec, ExtType)
+                            : CGF.Builder.CreateZExt(MaxVec, ExtType);
+  // In Product, replace all overflowed values with max values of non-extended
+  // type.
+  ICmpInst::Predicate Pred = Signed ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE;
+  Value *Cmp = CGF.Builder.CreateICmp(Pred, ExtProduct,
+                                      ExtMaxVec); // 1 if no overflow.
+  Value *SaturatedProduct = CGF.Builder.CreateSelect(
+      Cmp, ExtProduct, ExtMaxVec); // If overflowed, copy from max values.
+
+  if (Signed) {
+    APInt SignedMinValue =
+        llvm::APInt::getSignedMinValue(ElementType->getPrimitiveSizeInBits());
+    Value *Min = llvm::ConstantInt::get(ElementType, SignedMinValue);
+    Value *MinVec = CGF.Builder.CreateVectorSplat(NumElements, Min);
+    Value *ExtMinVec = CGF.Builder.CreateSExt(MinVec, ExtType);
+    Value *IsNegative =
+        CGF.Builder.CreateICmp(ICmpInst::ICMP_SLT, SaturatedProduct, ExtMinVec);
+    SaturatedProduct =
+        CGF.Builder.CreateSelect(IsNegative, ExtMinVec, SaturatedProduct);
+  }
+
+  SaturatedProduct = CGF.Builder.CreateTrunc(SaturatedProduct,
+                                             ResultType); // Trunc to ResultType.
+
+  if (E->getNumArgs() == 4) { // For masked intrinsics.
+    Value *VecSRC = Ops[2];
+    Value *Mask = Ops[3];
+    return EmitX86Select(CGF, Mask, SaturatedProduct, VecSRC);
+  }
+
+  return SaturatedProduct;
+}
+
 Value *CodeGenFunction::EmitX86CpuIs(const CallExpr *E) {
   const Expr *CPUExpr = E->getArg(0)->IgnoreParenCasts();
   StringRef CPUStr = cast<clang::StringLiteral>(CPUExpr)->getString();
   return EmitX86CpuIs(CPUStr);
 }
 
 Value *CodeGenFunction::EmitX86CpuIs(StringRef CPUStr) {
 
case X86::BI__readgsqword: {
     llvm::Type *IntTy = ConvertType(E->getType());
     Value *Ptr = Builder.CreateIntToPtr(EmitScalarExpr(E->getArg(0)),
                                         llvm::PointerType::get(IntTy, 256));
     LoadInst *Load = Builder.CreateAlignedLoad(
         IntTy, Ptr, getContext().getTypeAlignInChars(E->getType()));
     Load->setVolatile(true);
     return Load;
   }
+  case X86::BI__builtin_ia32_paddusb512_mask:
+  case X86::BI__builtin_ia32_paddusw512_mask:
+  case X86::BI__builtin_ia32_paddusb256:
+  case X86::BI__builtin_ia32_paddusw256:
+  case X86::BI__builtin_ia32_paddusb128:
+  case X86::BI__builtin_ia32_paddusw128:
+    return EmitX86AddSubSatExpr(*this, E, Ops, true, false); // Addition, unsigned.
+  case X86::BI__builtin_ia32_paddsb512_mask:
+  case X86::BI__builtin_ia32_paddsw512_mask:
+  case X86::BI__builtin_ia32_paddsb256:
+  case X86::BI__builtin_ia32_paddsw256:
+  case X86::BI__builtin_ia32_paddsb128:
+  case X86::BI__builtin_ia32_paddsw128:
+    return EmitX86AddSubSatExpr(*this, E, Ops, true, true); // Addition, signed.
+  case X86::BI__builtin_ia32_psubusb512_mask:
+  case X86::BI__builtin_ia32_psubusw512_mask:
+  case X86::BI__builtin_ia32_psubusb256:
+  case X86::BI__builtin_ia32_psubusw256:
+  case X86::BI__builtin_ia32_psubusb128:
+  case X86::BI__builtin_ia32_psubusw128:
+    return EmitX86AddSubSatExpr(*this, E, Ops, false, false); // Substraction, unsigned.
+  case X86::BI__builtin_ia32_psubsb512_mask:
+  case X86::BI__builtin_ia32_psubsw512_mask:
+  case X86::BI__builtin_ia32_psubsb256:
+  case X86::BI__builtin_ia32_psubsw256:
+  case X86::BI__builtin_ia32_psubsb128:
+  case X86::BI__builtin_ia32_psubsw128:
+    return EmitX86AddSubSatExpr(*this, E, Ops, false, true); // Substraction, signed.
   }
 }
 
-
 Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
                                            const CallExpr *E) {
   SmallVector<Value*, 4> Ops;
 
   for (unsigned i = 0, e = E->getNumArgs(); i != e; i++)
     Ops.push_back(EmitScalarExpr(E->getArg(i)));
 
   Intrinsic::ID ID = Intrinsic::not_intrinsic;

test/CodeGen/avx2-builtins.c

 __m256i test_mm256_add_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_add_epi64
   // CHECK: add <4 x i64>
   return _mm256_add_epi64(a, b);
 }
 
 __m256i test_mm256_adds_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_adds_epi8
-  // CHECK: call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK-NOT: call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   return _mm256_adds_epi8(a, b);
 }
 
 __m256i test_mm256_adds_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_adds_epi16
-  // CHECK: call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK-NOT: call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   return _mm256_adds_epi16(a, b);
 }
 
 __m256i test_mm256_adds_epu8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_adds_epu8
-  // CHECK: call <32 x i8> @llvm.x86.avx2.paddus.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK-NOT: call <32 x i8> @llvm.x86.avx2.paddus.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   return _mm256_adds_epu8(a, b);
 }
 
 __m256i test_mm256_adds_epu16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_adds_epu16
-  // CHECK: call <16 x i16> @llvm.x86.avx2.paddus.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK-NOT: call <16 x i16> @llvm.x86.avx2.paddus.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   return _mm256_adds_epu16(a, b);
 }
 
 __m256i test_mm256_alignr_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_alignr_epi8
   // CHECK: shufflevector <32 x i8> %{{.*}}, <32 x i8> %{{.*}}, <32 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 33, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48, i32 49>
   return _mm256_alignr_epi8(a, b, 2);
 }

 __m256i test_mm256_sub_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_sub_epi64
   // CHECK: sub <4 x i64>
   return _mm256_sub_epi64(a, b);
 }
 
 __m256i test_mm256_subs_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_subs_epi8
-  // CHECK: call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK-NOT: call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   return _mm256_subs_epi8(a, b);
 }
 
 __m256i test_mm256_subs_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_subs_epi16
-  // CHECK: call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK-NOT: call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   return _mm256_subs_epi16(a, b);
 }
 
 __m256i test_mm256_subs_epu8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_subs_epu8
-  // CHECK: call <32 x i8> @llvm.x86.avx2.psubus.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK-NOT: call <32 x i8> @llvm.x86.avx2.psubus.b(<32 x i8> %{{.*}}, <32 x i8> %{{.*}})
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   return _mm256_subs_epu8(a, b);
 }
 
 __m256i test_mm256_subs_epu16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_subs_epu16
-  // CHECK: call <16 x i16> @llvm.x86.avx2.psubus.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK-NOT: call <16 x i16> @llvm.x86.avx2.psubus.w(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   return _mm256_subs_epu16(a, b);
 }
 
 __m256i test_mm256_unpackhi_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_unpackhi_epi8
   // CHECK: shufflevector <32 x i8> %{{.*}}, <32 x i8> %{{.*}}, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
   return _mm256_unpackhi_epi8(a, b);
 }

test/CodeGen/avx512bw-builtins.c

 __m512i test_mm512_maskz_packus_epi16(__mmask64 __M, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_packus_epi16
   // CHECK: @llvm.x86.avx512.packuswb.512
   // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_maskz_packus_epi16(__M,__A,__B); 
 }
 __m512i test_mm512_adds_epi8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_adds_epi8
-  // CHECK: @llvm.x86.avx512.mask.padds.b.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
   return _mm512_adds_epi8(__A,__B); 
 }
 __m512i test_mm512_mask_adds_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_adds_epi8
-  // CHECK: @llvm.x86.avx512.mask.padds.b.512
-  return _mm512_mask_adds_epi8(__W,__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
 }
 __m512i test_mm512_maskz_adds_epi8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_adds_epi8
-  // CHECK: @llvm.x86.avx512.mask.padds.b.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_maskz_adds_epi8(__U,__A,__B); 
 }
 __m512i test_mm512_adds_epi16(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_adds_epi16
-  // CHECK: @llvm.x86.avx512.mask.padds.w.512
-  return _mm512_adds_epi16(__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
 }
 __m512i test_mm512_mask_adds_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_adds_epi16
-  // CHECK: @llvm.x86.avx512.mask.padds.w.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
   return _mm512_mask_adds_epi16(__W,__U,__A,__B); 
 }
 __m512i test_mm512_maskz_adds_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_adds_epi16
-  // CHECK: @llvm.x86.avx512.mask.padds.w.512
-  return _mm512_maskz_adds_epi16(__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.padds.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
 }
 __m512i test_mm512_adds_epu8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_adds_epu8
-  // CHECK: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
   return _mm512_adds_epu8(__A,__B); 
 }
 __m512i test_mm512_mask_adds_epu8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_adds_epu8
-  // CHECK: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_mask_adds_epu8(__W,__U,__A,__B); 
 }
 __m512i test_mm512_maskz_adds_epu8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_adds_epu8
-  // CHECK: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: add <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_maskz_adds_epu8(__U,__A,__B); 
 }
 __m512i test_mm512_adds_epu16(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_adds_epu16
-  // CHECK: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
   return _mm512_adds_epu16(__A,__B); 
 }
 __m512i test_mm512_mask_adds_epu16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_adds_epu16
-  // CHECK: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
   return _mm512_mask_adds_epu16(__W,__U,__A,__B); 
 }
 __m512i test_mm512_maskz_adds_epu16(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_adds_epu16
-  // CHECK: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK-NOT: @llvm.x86.avx512.mask.paddus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: add <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
   return _mm512_maskz_adds_epu16(__U,__A,__B); 
 }
 __m512i test_mm512_avg_epu8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_avg_epu8
   // CHECK-NOT: @llvm.x86.avx512.mask.pavg.b.512
   // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
   // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
   // CHECK: add <64 x i16> %{{.*}}, %{{.*}}

 __m512i test_mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_shuffle_epi8
   // CHECK: @llvm.x86.avx512.pshuf.b.512
   // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
   return _mm512_maskz_shuffle_epi8(__U,__A,__B); 
 }
 __m512i test_mm512_subs_epi8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_subs_epi8
-  // CHECK: @llvm.x86.avx512.mask.psubs.b.512
-  return _mm512_subs_epi8(__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
 }
 __m512i test_mm512_mask_subs_epi8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_subs_epi8
-  // CHECK: @llvm.x86.avx512.mask.psubs.b.512
-  return _mm512_mask_subs_epi8(__W,__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
 }
 __m512i test_mm512_maskz_subs_epi8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_subs_epi8
-  // CHECK: @llvm.x86.avx512.mask.psubs.b.512
-  return _mm512_maskz_subs_epi8(__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.b.512
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <64 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <64 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <64 x i16> %{{.*}}
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
 }
 __m512i test_mm512_subs_epi16(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_subs_epi16
-  // CHECK: @llvm.x86.avx512.mask.psubs.w.512
-  return _mm512_subs_epi16(__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
 }
 __m512i test_mm512_mask_subs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_subs_epi16
-  // CHECK: @llvm.x86.avx512.mask.psubs.w.512
-  return _mm512_mask_subs_epi16(__W,__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
 }
 __m512i test_mm512_maskz_subs_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_subs_epi16
-  // CHECK: @llvm.x86.avx512.mask.psubs.w.512
-  return _mm512_maskz_subs_epi16(__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubs.w.512
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <32 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <32 x i32> %{{.*}}
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
 }
 __m512i test_mm512_subs_epu8(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_subs_epu8
-  // CHECK: @llvm.x86.avx512.mask.psubus.b.512
-  return _mm512_subs_epu8(__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
 }
 __m512i test_mm512_mask_subs_epu8(__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_subs_epu8
-  // CHECK: @llvm.x86.avx512.mask.psubus.b.512
-  return _mm512_mask_subs_epu8(__W,__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
 }
 __m512i test_mm512_maskz_subs_epu8(__mmask64 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_subs_epu8
-  // CHECK: @llvm.x86.avx512.mask.psubus.b.512
-  return _mm512_maskz_subs_epu8(__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.b.512
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: zext <64 x i8> %{{.*}} to <64 x i16>
+  // CHECK: sub <64 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <64 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i16> %{{.*}}, <64 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <64 x i16> %{{.*}} to <64 x i8>
+  // CHECK: select <64 x i1> %{{.*}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}
 }
 __m512i test_mm512_subs_epu16(__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_subs_epu16
-  // CHECK: @llvm.x86.avx512.mask.psubus.w.512
-  return _mm512_subs_epu16(__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
 }
 __m512i test_mm512_mask_subs_epu16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask_subs_epu16
-  // CHECK: @llvm.x86.avx512.mask.psubus.w.512
-  return _mm512_mask_subs_epu16(__W,__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
 }
 __m512i test_mm512_maskz_subs_epu16(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_maskz_subs_epu16
-  // CHECK: @llvm.x86.avx512.mask.psubus.w.512
-  return _mm512_maskz_subs_epu16(__U,__A,__B); 
+  // CHECK-NOT: @llvm.x86.avx512.mask.psubus.w.512
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: zext <32 x i16> %{{.*}} to <32 x i32>
+  // CHECK: sub <32 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i32> %{{.*}}, <32 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <32 x i32> %{{.*}} to <32 x i16>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
 }
 __m512i test_mm512_mask2_permutex2var_epi16(__m512i __A, __m512i __I, __mmask32 __U, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mask2_permutex2var_epi16
   // CHECK: @llvm.x86.avx512.mask.vpermi2var.hi.512
   return _mm512_mask2_permutex2var_epi16(__A,__I,__U,__B); 
 }
 __m512i test_mm512_permutex2var_epi16(__m512i __A, __m512i __I, __m512i __B) {
   // CHECK-LABEL: @test_mm512_permutex2var_epi16

test/CodeGen/avx512vlbw-builtins.c

__m256i test_mm256_mask_packus_epi16(__m256i __W, __mmask32 __M, __m256i __A,        __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_packus_epi16
   // CHECK: @llvm.x86.avx2.packuswb
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_packus_epi16(__W,__M,__A,__B); 
 }
 
 __m128i test_mm_mask_adds_epi8(__m128i __W, __mmask16 __U, __m128i __A,        __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_adds_epi8
-  // CHECK: @llvm.x86.sse2.padds.b
+  // CHECK-NOT: @llvm.x86.sse2.padds.b
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_mask_adds_epi8(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_adds_epi8(__mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_adds_epi8
-  // CHECK: @llvm.x86.sse2.padds.b
+  // CHECK-NOT: @llvm.x86.sse2.padds.b
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_maskz_adds_epi8(__U,__A,__B); 
 }
 __m256i test_mm256_mask_adds_epi8(__m256i __W, __mmask32 __U, __m256i __A,           __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_adds_epi8
-  // CHECK: @llvm.x86.avx2.padds.b
+  // CHECK-NOT: @llvm.x86.avx2.padds.b
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_adds_epi8(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_adds_epi8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_adds_epi8
-  // CHECK: @llvm.x86.avx2.padds.b
+  // CHECK-NOT: @llvm.x86.avx2.padds.b
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_adds_epi8(__U,__A,__B); 
 }
 __m128i test_mm_mask_adds_epi16(__m128i __W, __mmask8 __U, __m128i __A,         __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_adds_epi16
-  // CHECK: @llvm.x86.sse2.padds.w
+  // CHECK-NOT: @llvm.x86.sse2.padds.w
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_mask_adds_epi16(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_adds_epi16(__mmask8 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_adds_epi16
-  // CHECK: @llvm.x86.sse2.padds.w
+  // CHECK-NOT: @llvm.x86.sse2.padds.w
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_maskz_adds_epi16(__U,__A,__B); 
 }
 __m256i test_mm256_mask_adds_epi16(__m256i __W, __mmask16 __U, __m256i __A,      __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_adds_epi16
-  // CHECK: @llvm.x86.avx2.padds.w
+  // CHECK-NOT: @llvm.x86.avx2.padds.w
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_mask_adds_epi16(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_adds_epi16(__mmask16 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_adds_epi16
-  // CHECK: @llvm.x86.avx2.padds.w
+  // CHECK-NOT: @llvm.x86.avx2.padds.w
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_maskz_adds_epi16(__U,__A,__B); 
 }
-__m128i test_mm_mask_adds_epu8(__m128i __W, __mmask16 __U, __m128i __A,        __m128i __B) {
+__m128i test_mm_mask_adds_epu8(__m128i __W, __mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_adds_epu8
-  // CHECK: @llvm.x86.sse2.paddus.b
+  // CHECK-NOT: @llvm.x86.sse2.paddus.b
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_mask_adds_epu8(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_adds_epu8(__mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_adds_epu8
-  // CHECK: @llvm.x86.sse2.paddus.b
+  // CHECK-NOT: @llvm.x86.sse2.paddus.b
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_maskz_adds_epu8(__U,__A,__B); 
 }
 __m256i test_mm256_mask_adds_epu8(__m256i __W, __mmask32 __U, __m256i __A,           __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_adds_epu8
-  // CHECK: @llvm.x86.avx2.paddus.b
+  // CHECK-NOT: @llvm.x86.avx2.paddus.b
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_adds_epu8(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_adds_epu8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_adds_epu8
-  // CHECK: @llvm.x86.avx2.paddus.b
+  // CHECK-NOT: @llvm.x86.avx2.paddus.b
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: add <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_adds_epu8(__U,__A,__B); 
 }
 __m128i test_mm_mask_adds_epu16(__m128i __W, __mmask8 __U, __m128i __A,         __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_adds_epu16
-  // CHECK: @llvm.x86.sse2.paddus.w
+  // CHECK-NOT: @llvm.x86.sse2.paddus.w
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_mask_adds_epu16(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_adds_epu16(__mmask8 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_adds_epu16
-  // CHECK: @llvm.x86.sse2.paddus.w
+  // CHECK-NOT: @llvm.x86.sse2.paddus.w
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_maskz_adds_epu16(__U,__A,__B); 
 }
 __m256i test_mm256_mask_adds_epu16(__m256i __W, __mmask16 __U, __m256i __A,      __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_adds_epu16
-  // CHECK: @llvm.x86.avx2.paddus.w
+  // CHECK-NOT: @llvm.x86.avx2.paddus.w
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_mask_adds_epu16(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_adds_epu16(__mmask16 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_adds_epu16
-  // CHECK: @llvm.x86.avx2.paddus.w
+  // CHECK-NOT: @llvm.x86.avx2.paddus.w
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: add <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_maskz_adds_epu16(__U,__A,__B); 
 }
 __m128i test_mm_mask_avg_epu8(__m128i __W, __mmask16 __U, __m128i __A,       __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_avg_epu8
   // CHECK-NOT: @llvm.x86.sse2.pavg.b
   // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
   // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>

 __m256i test_mm256_maskz_shuffle_epi8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_shuffle_epi8
   // CHECK: @llvm.x86.avx2.pshuf.b
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_shuffle_epi8(__U,__A,__B); 
 }
 __m128i test_mm_mask_subs_epi8(__m128i __W, __mmask16 __U, __m128i __A,        __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_subs_epi8
-  // CHECK: @llvm.x86.sse2.psubs.b
+  // CHECK-NOT: @llvm.x86.sse2.psubs.b
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_mask_subs_epi8(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_subs_epi8(__mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_subs_epi8
-  // CHECK: @llvm.x86.sse2.psubs.b
+  // CHECK-NOT: @llvm.x86.sse2.psubs.b
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_maskz_subs_epi8(__U,__A,__B); 
 }
 __m256i test_mm256_mask_subs_epi8(__m256i __W, __mmask32 __U, __m256i __A,           __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_subs_epi8
-  // CHECK: @llvm.x86.avx2.psubs.b
+  // CHECK-NOT: @llvm.x86.avx2.psubs.b
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_subs_epi8(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_subs_epi8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_subs_epi8
-  // CHECK: @llvm.x86.avx2.psubs.b
+  // CHECK-NOT: @llvm.x86.avx2.psubs.b
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <32 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <32 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <32 x i16> %{{.*}}
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_subs_epi8(__U,__A,__B); 
 }
 __m128i test_mm_mask_subs_epi16(__m128i __W, __mmask8 __U, __m128i __A,         __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_subs_epi16
-  // CHECK: @llvm.x86.sse2.psubs.w
+  // CHECK-NOT: @llvm.x86.sse2.psubs.w
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_mask_subs_epi16(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_subs_epi16(__mmask8 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_subs_epi16
-  // CHECK: @llvm.x86.sse2.psubs.w
+  // CHECK-NOT: @llvm.x86.sse2.psubs.w
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_maskz_subs_epi16(__U,__A,__B); 
 }
 __m256i test_mm256_mask_subs_epi16(__m256i __W, __mmask16 __U, __m256i __A,      __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_subs_epi16
-  // CHECK: @llvm.x86.avx2.psubs.w
+  // CHECK-NOT: @llvm.x86.avx2.psubs.w
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_mask_subs_epi16(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_subs_epi16(__mmask16 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_subs_epi16
-  // CHECK: @llvm.x86.avx2.psubs.w
+  // CHECK-NOT: @llvm.x86.avx2.psubs.w
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <16 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <16 x i32> %{{.*}}
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_maskz_subs_epi16(__U,__A,__B); 
 }
 __m128i test_mm_mask_subs_epu8(__m128i __W, __mmask16 __U, __m128i __A,        __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_subs_epu8
-  // CHECK: @llvm.x86.sse2.psubus.b
+  // CHECK-NOT: @llvm.x86.sse2.psubus.b
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_mask_subs_epu8(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_subs_epu8(__mmask16 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_subs_epu8
-  // CHECK: @llvm.x86.sse2.psubus.b
+  // CHECK-NOT: @llvm.x86.sse2.psubus.b
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}
   return _mm_maskz_subs_epu8(__U,__A,__B); 
 }
 __m256i test_mm256_mask_subs_epu8(__m256i __W, __mmask32 __U, __m256i __A,           __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_subs_epu8
-  // CHECK: @llvm.x86.avx2.psubus.b
+  // CHECK-NOT: @llvm.x86.avx2.psubus.b
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_mask_subs_epu8(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_subs_epu8(__mmask32 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_subs_epu8
-  // CHECK: @llvm.x86.avx2.psubus.b
+  // CHECK-NOT: @llvm.x86.avx2.psubus.b
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: zext <32 x i8> %{{.*}} to <32 x i16>
+  // CHECK: sub <32 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <32 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <32 x i16> %{{.*}} to <32 x i8>
   // CHECK: select <32 x i1> %{{.*}}, <32 x i8> %{{.*}}, <32 x i8> %{{.*}}
   return _mm256_maskz_subs_epu8(__U,__A,__B); 
 }
 __m128i test_mm_mask_subs_epu16(__m128i __W, __mmask8 __U, __m128i __A,         __m128i __B) {
   // CHECK-LABEL: @test_mm_mask_subs_epu16
-  // CHECK: @llvm.x86.sse2.psubus.w
+  // CHECK-NOT: @llvm.x86.sse2.psubus.w
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_mask_subs_epu16(__W,__U,__A,__B); 
 }
 __m128i test_mm_maskz_subs_epu16(__mmask8 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: @test_mm_maskz_subs_epu16
-  // CHECK: @llvm.x86.sse2.psubus.w
+  // CHECK-NOT: @llvm.x86.sse2.psubus.w
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_maskz_subs_epu16(__U,__A,__B); 
 }
 __m256i test_mm256_mask_subs_epu16(__m256i __W, __mmask16 __U, __m256i __A,      __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask_subs_epu16
-  // CHECK: @llvm.x86.avx2.psubus.w
+  // CHECK-NOT: @llvm.x86.avx2.psubus.w
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_mask_subs_epu16(__W,__U,__A,__B); 
 }
 __m256i test_mm256_maskz_subs_epu16(__mmask16 __U, __m256i __A, __m256i __B) {
   // CHECK-LABEL: @test_mm256_maskz_subs_epu16
-  // CHECK: @llvm.x86.avx2.psubus.w
+  // CHECK-NOT: @llvm.x86.avx2.psubus.w
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: zext <16 x i16> %{{.*}} to <16 x i32>
+  // CHECK: sub <16 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i32> %{{.*}}, <16 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <16 x i32> %{{.*}} to <16 x i16>
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_maskz_subs_epu16(__U,__A,__B); 
 }
 
-
 __m128i test_mm_mask2_permutex2var_epi16(__m128i __A, __m128i __I, __mmask8 __U,            __m128i __B) {
   // CHECK-LABEL: @test_mm_mask2_permutex2var_epi16
   // CHECK: @llvm.x86.avx512.mask.vpermi2var.hi.128
   return _mm_mask2_permutex2var_epi16(__A,__I,__U,__B); 
 }
 __m256i test_mm256_mask2_permutex2var_epi16(__m256i __A, __m256i __I,         __mmask16 __U, __m256i __B) {
   // CHECK-LABEL: @test_mm256_mask2_permutex2var_epi16
   // CHECK: @llvm.x86.avx512.mask.vpermi2var.hi.256

test/CodeGen/sse2-builtins.c

__m128d test_mm_add_sd(__m128d A, __m128d B) {
   // CHECK: extractelement <2 x double> %{{.*}}, i32 0
   // CHECK: fadd double
   // CHECK: insertelement <2 x double> %{{.*}}, double %{{.*}}, i32 0
   return _mm_add_sd(A, B);
 }
 
 __m128i test_mm_adds_epi8(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_adds_epi8
-  // CHECK: call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK-NOT: call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   return _mm_adds_epi8(A, B);
 }
 
 __m128i test_mm_adds_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_adds_epi16
-  // CHECK: call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK-NOT: call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}}, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   return _mm_adds_epi16(A, B);
 }
 
 __m128i test_mm_adds_epu8(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_adds_epu8
-  // CHECK: call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK-NOT: call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: add <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   return _mm_adds_epu8(A, B);
 }
 
 __m128i test_mm_adds_epu16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_adds_epu16
-  // CHECK: call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK-NOT: call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: add <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   return _mm_adds_epu16(A, B);
 }
 
 __m128d test_mm_and_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_and_pd
   // CHECK: and <2 x i64>
   return _mm_and_pd(A, B);
 }
__m128d test_mm_sub_sd(__m128d A, __m128d B) {
   // CHECK: extractelement <2 x double> %{{.*}}, i32 0
   // CHECK: fsub double
   // CHECK: insertelement <2 x double> %{{.*}}, double %{{.*}}, i32 0
   return _mm_sub_sd(A, B);
 }
 
 __m128i test_mm_subs_epi8(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_subs_epi8
-  // CHECK: call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK-NOT: call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <16 x i16> %{{.*}}, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
+  // CHECK: icmp slt <16 x i16> %{{.*}}, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>, <16 x i16> %{{.*}}
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   return _mm_subs_epi8(A, B);
 }
 
 __m128i test_mm_subs_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_subs_epi16
-  // CHECK: call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK-NOT: call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp sle <8 x i32> %{{.*}}, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
+  // CHECK: icmp slt <8 x i32> %{{.*}},  <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32>  <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>, <8 x i32> %{{.*}}
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   return _mm_subs_epi16(A, B);
 }
 
 __m128i test_mm_subs_epu8(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_subs_epu8
-  // CHECK: call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK-NOT: call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %{{.*}}, <16 x i8> %{{.*}})
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: zext <16 x i8> %{{.*}} to <16 x i16>
+  // CHECK: sub <16 x i16> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <16 x i16> %{{.*}}, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+  // CHECK: trunc <16 x i16> %{{.*}} to <16 x i8>
   return _mm_subs_epu8(A, B);
 }
 
 __m128i test_mm_subs_epu16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_subs_epu16
-  // CHECK: call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK-NOT: call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: zext <8 x i16> %{{.*}} to <8 x i32>
+  // CHECK: sub <8 x i32> %{{.*}}, %{{.*}}
+  // CHECK: icmp ule <8 x i32> %{{.*}}, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: select <8 x i1> %{{.*}}, <8 x i32> %{{.*}}, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
+  // CHECK: trunc <8 x i32> %{{.*}} to <8 x i16>
   return _mm_subs_epu16(A, B);
 }
 
 int test_mm_ucomieq_sd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_ucomieq_sd
   // CHECK: call i32 @llvm.x86.sse2.ucomieq.sd(<2 x double> %{{.*}}, <2 x double> %{{.*}})
   return _mm_ucomieq_sd(A, B);
 }