[X86][CostModel] Adjust the costs of ZERO_EXTEND/SIGN_EXTEND with less than 128-bit inputs

[craig.topper]

Summary

Now that we legalize by widening, the element types here won't change. Previously these were modeled as the elements being widened and then the instruction might become an AND or SHL/ASHR pair. But now they'll become something like a ZERO_EXTEND_VECTOR_INREG/SIGN_EXTEND_VECTOR_INREG.

For AVX2, when the destination type is legal its clear the cost should be 1 since we have extend instructions that can produce 256 bit vectors from less than 128 bit vectors. I'm a little less sure about AVX1 costs, but I think the ones I changed were definitely too high, but they might still be too high.

[RKSimon]

LGTM - I agree the AVX1 256-bit result cases might need further tweaking (especially for extension ratio = 2 as we can cheaply use movx+unpckh), but that can wait for now.

llvm/test/Transforms/SLPVectorizer/X86/sext.ll

611 ↗

(On Diff #214918)

Please can you raise a bug about the lower v2i8 sext vectorizing but the upper 2 x i8 not?

llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp

static const TypeConversionCostTblEntry AVX512FConversionTbl[] = {
 
     // v16i1 -> v16i32 - load + broadcast
     { ISD::SIGN_EXTEND, MVT::v16i32, MVT::v16i1,  2 },
     { ISD::ZERO_EXTEND, MVT::v16i32, MVT::v16i1,  2 },
     { ISD::SIGN_EXTEND, MVT::v16i32, MVT::v16i8,  1 },
     { ISD::ZERO_EXTEND, MVT::v16i32, MVT::v16i8,  1 },
     { ISD::SIGN_EXTEND, MVT::v16i32, MVT::v16i16, 1 },
     { ISD::ZERO_EXTEND, MVT::v16i32, MVT::v16i16, 1 },
-    { ISD::ZERO_EXTEND, MVT::v8i64,  MVT::v8i16,  1 },
+    { ISD::SIGN_EXTEND, MVT::v8i64,  MVT::v8i8,   1 },
+    { ISD::ZERO_EXTEND, MVT::v8i64,  MVT::v8i8,   1 },
     { ISD::SIGN_EXTEND, MVT::v8i64,  MVT::v8i16,  1 },
+    { ISD::ZERO_EXTEND, MVT::v8i64,  MVT::v8i16,  1 },
     { ISD::SIGN_EXTEND, MVT::v8i64,  MVT::v8i32,  1 },
     { ISD::ZERO_EXTEND, MVT::v8i64,  MVT::v8i32,  1 },
 
     { ISD::SINT_TO_FP,  MVT::v8f64,  MVT::v8i1,   4 },
     { ISD::SINT_TO_FP,  MVT::v16f32, MVT::v16i1,  3 },
     { ISD::SINT_TO_FP,  MVT::v8f64,  MVT::v8i8,   2 },
     { ISD::SINT_TO_FP,  MVT::v16f32, MVT::v16i8,  2 },
     { ISD::SINT_TO_FP,  MVT::v8f64,  MVT::v8i16,  2 },
static const TypeConversionCostTblEntry AVX512FConversionTbl[] = {
     { ISD::FP_TO_UINT,  MVT::v16i8,  MVT::v16f32, 2 },
   };
 
   static const TypeConversionCostTblEntry AVX2ConversionTbl[] = {
     { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i1,   3 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i1,   3 },
     { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i1,   3 },
     { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i1,   3 },
-    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,   3 },
-    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i8,   3 },
-    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,   3 },
-    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i8,   3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,   1 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i8,   1 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,   1 },
+    { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i8,   1 },
     { ISD::SIGN_EXTEND, MVT::v16i16, MVT::v16i8,  1 },
     { ISD::ZERO_EXTEND, MVT::v16i16, MVT::v16i8,  1 },
-    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16,  3 },
-    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i16,  3 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16,  1 },
+    { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i16,  1 },
     { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i16,  1 },
     { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i16,  1 },
     { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i32,  1 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i32,  1 },
 
     { ISD::TRUNCATE,    MVT::v4i8,   MVT::v4i64,  2 },
     { ISD::TRUNCATE,    MVT::v4i16,  MVT::v4i64,  2 },
     { ISD::TRUNCATE,    MVT::v4i32,  MVT::v4i64,  2 },
     { ISD::TRUNCATE,    MVT::v8i8,   MVT::v8i32,  2 },
     { ISD::TRUNCATE,    MVT::v8i16,  MVT::v8i32,  2 },
     { ISD::TRUNCATE,    MVT::v8i32,  MVT::v8i64,  4 },
 
     { ISD::FP_EXTEND,   MVT::v8f64,  MVT::v8f32,  3 },
     { ISD::FP_ROUND,    MVT::v8f32,  MVT::v8f64,  3 },
 
     { ISD::UINT_TO_FP,  MVT::v8f32,  MVT::v8i32,  8 },
   };
 
   static const TypeConversionCostTblEntry AVXConversionTbl[] = {
     { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i1,  6 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i1,  4 },
     { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i1,  7 },
     { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i1,  4 },
-    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,  6 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i8,  4 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i8,  4 },
-    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,  7 },
+    { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i8,  4 },
     { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i8,  4 },
     { ISD::SIGN_EXTEND, MVT::v16i16, MVT::v16i8, 4 },
     { ISD::ZERO_EXTEND, MVT::v16i16, MVT::v16i8, 4 },
-    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16, 6 },
+    { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i16, 4 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i16, 3 },
     { ISD::SIGN_EXTEND, MVT::v8i32,  MVT::v8i16, 4 },
     { ISD::ZERO_EXTEND, MVT::v8i32,  MVT::v8i16, 4 },
     { ISD::SIGN_EXTEND, MVT::v4i64,  MVT::v4i32, 4 },
     { ISD::ZERO_EXTEND, MVT::v4i64,  MVT::v4i32, 4 },
 
     { ISD::TRUNCATE,    MVT::v16i8, MVT::v16i16, 4 },
     { ISD::TRUNCATE,    MVT::v8i8,  MVT::v8i32,  4 },

llvm/trunk/test/Analysis/CostModel/X86/cast.ll

 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %Z = zext <8 x i1> %in to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %S = sext <8 x i1> %in to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %A1 = zext <16 x i8> undef to <16 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %A2 = sext <16 x i8> undef to <16 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %A = sext <8 x i16> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %B = zext <8 x i16> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C = sext <4 x i32> undef to <4 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C.v8i8.z = zext <8 x i8> undef to <8 x i32>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i16.z = zext <4 x i16> undef to <4 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C.v4i8.z = zext <4 x i8> undef to <4 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D = zext <4 x i32> undef to <4 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %D1 = zext <8 x i32> undef to <8 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %D2 = sext <8 x i32> undef to <8 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D3 = zext <16 x i16> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D4 = zext <16 x i8> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %D5 = zext <16 x i1> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %E = trunc <4 x i64> undef to <4 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %F = trunc <8 x i32> undef to <8 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %F1 = trunc <16 x i16> undef to <16 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %F2 = trunc <8 x i32> undef to <8 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %F3 = trunc <4 x i64> undef to <4 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %G = trunc <8 x i64> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %G1 = trunc <16 x i32> undef to <16 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %G2 = trunc <16 x i32> undef to <16 x i8>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'zext_sext'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %Z = zext <8 x i1> %in to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %S = sext <8 x i1> %in to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A1 = zext <16 x i8> undef to <16 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A2 = sext <16 x i8> undef to <16 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = sext <8 x i16> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = zext <8 x i16> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = sext <4 x i32> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v8i8.z = zext <8 x i8> undef to <8 x i32>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i16.z = zext <4 x i16> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i8.z = zext <4 x i8> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v8i8.z = zext <8 x i8> undef to <8 x i32>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i16.z = zext <4 x i16> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i8.z = zext <4 x i8> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = zext <4 x i32> undef to <4 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %D1 = zext <8 x i32> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %D2 = sext <8 x i32> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D3 = zext <16 x i16> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D4 = zext <16 x i8> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %D5 = zext <16 x i1> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %E = trunc <4 x i64> undef to <4 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F = trunc <8 x i32> undef to <8 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %F1 = trunc <16 x i16> undef to <16 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F2 = trunc <8 x i32> undef to <8 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F3 = trunc <4 x i64> undef to <4 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %G = trunc <8 x i64> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %G1 = trunc <16 x i32> undef to <16 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %G2 = trunc <16 x i32> undef to <16 x i8>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'zext_sext'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %Z = zext <8 x i1> %in to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %S = sext <8 x i1> %in to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A1 = zext <16 x i8> undef to <16 x i16>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A2 = sext <16 x i8> undef to <16 x i16>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = sext <8 x i16> undef to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = zext <8 x i16> undef to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = sext <4 x i32> undef to <4 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v8i8.z = zext <8 x i8> undef to <8 x i32>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i16.z = zext <4 x i16> undef to <4 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i8.z = zext <4 x i8> undef to <4 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v8i8.z = zext <8 x i8> undef to <8 x i32>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v8i8.s = sext <8 x i8> undef to <8 x i32>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i16.z = zext <4 x i16> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i16.s = sext <4 x i16> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i8.z = zext <4 x i8> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C.v4i8.s = sext <4 x i8> undef to <4 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = zext <4 x i32> undef to <4 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D1 = zext <8 x i32> undef to <8 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D2 = sext <8 x i32> undef to <8 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D3 = zext <16 x i16> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D4 = zext <16 x i8> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %D5 = zext <16 x i1> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %E = trunc <4 x i64> undef to <4 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F = trunc <8 x i32> undef to <8 x i16>

llvm/trunk/test/Analysis/CostModel/X86/extend.ll

 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i16> undef to <2 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i16> undef to <4 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = zext <8 x i16> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = zext <16 x i16> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'zext_vXi16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = zext <2 x i16> undef to <2 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = zext <4 x i16> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i64 = zext <8 x i16> undef to <8 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = zext <4 x i16> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = zext <8 x i16> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i16> undef to <2 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i16> undef to <4 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = zext <8 x i16> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = zext <16 x i16> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'zext_vXi16'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = zext <2 x i16> undef to <2 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = zext <4 x i16> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = zext <4 x i16> undef to <4 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = zext <8 x i16> undef to <8 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i16> undef to <2 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i16> undef to <4 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = zext <8 x i16> undef to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = zext <16 x i16> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'zext_vXi16'

 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i16 = zext <4 x i8> undef to <4 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = zext <8 x i8> undef to <8 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = zext <16 x i8> undef to <16 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = zext <32 x i8> undef to <32 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'zext_vXi8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = zext <2 x i8> undef to <2 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i8> undef to <2 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i8> undef to <4 x i32>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = zext <16 x i8> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i16 = zext <2 x i8> undef to <2 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i16 = zext <4 x i8> undef to <4 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = zext <8 x i8> undef to <8 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = zext <16 x i8> undef to <16 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i16 = zext <32 x i8> undef to <32 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'zext_vXi8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = zext <2 x i8> undef to <2 x i64>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i8> undef to <2 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i8> undef to <4 x i32>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = zext <16 x i8> undef to <16 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i16 = zext <2 x i8> undef to <2 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i16 = zext <4 x i8> undef to <4 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = zext <8 x i8> undef to <8 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = zext <16 x i8> undef to <16 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i16 = zext <32 x i8> undef to <32 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'zext_vXi8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i64 = zext <2 x i8> undef to <2 x i64>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = zext <4 x i8> undef to <4 x i64>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = zext <8 x i8> undef to <8 x i64>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i32 = zext <2 x i8> undef to <2 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = zext <4 x i8> undef to <4 x i32>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = zext <8 x i8> undef to <8 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = zext <16 x i8> undef to <16 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2i16 = zext <2 x i8> undef to <2 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i16 = zext <4 x i8> undef to <4 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = zext <8 x i8> undef to <8 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = zext <16 x i8> undef to <16 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = zext <32 x i8> undef to <32 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i16> undef to <2 x i32>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i16> undef to <4 x i32>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8i32 = sext <8 x i16> undef to <8 x i32>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i16> undef to <16 x i32>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'sext_vXi16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i16> undef to <2 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i16> undef to <2 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i16> undef to <4 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = sext <8 x i16> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i16> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'sext_vXi16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i16> undef to <2 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i16> undef to <2 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i16> undef to <4 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = sext <8 x i16> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i16> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512-LABEL: 'sext_vXi16'
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i16> undef to <2 x i64>
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i16> undef to <2 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i16> undef to <4 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = sext <8 x i16> undef to <8 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = sext <16 x i16> undef to <16 x i32>
 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'sext_vXi16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i16> undef to <2 x i64>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = sext <4 x i16> undef to <4 x i64>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = sext <8 x i16> undef to <8 x i64>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i16> undef to <2 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i16> undef to <4 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = sext <8 x i16> undef to <8 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i16> undef to <16 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2i64 = sext <2 x i16> undef to <2 x i64>
   %V4i64 = sext <4 x i16> undef to <4 x i64>

 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'sext_vXi8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i8> undef to <2 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i8> undef to <2 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i8> undef to <4 x i32>
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i8> undef to <16 x i32>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i16 = sext <2 x i8> undef to <2 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'sext_vXi8'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i8> undef to <2 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i8> undef to <2 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i8> undef to <4 x i32>
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i8> undef to <16 x i32>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i16 = sext <2 x i8> undef to <2 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512F-LABEL: 'sext_vXi8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i8> undef to <2 x i64>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i8> undef to <2 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i8> undef to <4 x i32>
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = sext <16 x i8> undef to <16 x i32>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i16 = sext <2 x i8> undef to <2 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX512BW-LABEL: 'sext_vXi8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i8> undef to <2 x i64>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i8> undef to <2 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i8> undef to <4 x i32>
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i32 = sext <16 x i8> undef to <16 x i32>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i16 = sext <2 x i8> undef to <2 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; BTVER2-LABEL: 'sext_vXi8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i64 = sext <2 x i8> undef to <2 x i64>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4i64 = sext <4 x i8> undef to <4 x i64>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8i64 = sext <8 x i8> undef to <8 x i64>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i32 = sext <2 x i8> undef to <2 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4i32 = sext <4 x i8> undef to <4 x i32>
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8i32 = sext <8 x i8> undef to <8 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i32 = sext <16 x i8> undef to <16 x i32>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2i16 = sext <2 x i8> undef to <2 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4i16 = sext <4 x i8> undef to <4 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8i16 = sext <8 x i8> undef to <8 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16i16 = sext <16 x i8> undef to <16 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32i16 = sext <32 x i8> undef to <32 x i16>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;

llvm/trunk/test/Analysis/CostModel/X86/min-legal-vector-width.ll

 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
 ; RUN: opt < %s  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,VEC256,AVX
 ; RUN: opt < %s  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512vl,+avx512bw,+avx512dq,+prefer-256-bit | FileCheck %s --check-prefixes=CHECK,VEC256,SKX256
 ; RUN: opt < %s  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512vl,+avx512bw,+avx512dq,-prefer-256-bit | FileCheck %s --check-prefixes=CHECK,VEC512
 
 define void @zext256() "min-legal-vector-width"="256" {
 ; VEC256-LABEL: 'zext256'
-; VEC256-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %A = zext <8 x i16> undef to <8 x i64>
+; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %A = zext <8 x i16> undef to <8 x i64>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %B = zext <8 x i32> undef to <8 x i64>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C = zext <16 x i8> undef to <16 x i32>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D = zext <16 x i16> undef to <16 x i32>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %E = zext <32 x i8> undef to <32 x i16>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; VEC512-LABEL: 'zext256'
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = zext <8 x i16> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = zext <8 x i32> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = zext <16 x i8> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = zext <16 x i16> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %E = zext <32 x i8> undef to <32 x i16>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %A = zext <8 x i16> undef to <8 x i64>
   %B = zext <8 x i32> undef to <8 x i64>
   %C = zext <16 x i8> undef to <16 x i32>
   %D = zext <16 x i16> undef to <16 x i32>
   %E = zext <32 x i8> undef to <32 x i16>
   ret void
 }
 
 define void @zext512() "min-legal-vector-width"="512" {
 ; AVX-LABEL: 'zext512'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %A = zext <8 x i16> undef to <8 x i64>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %A = zext <8 x i16> undef to <8 x i64>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %B = zext <8 x i32> undef to <8 x i64>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %C = zext <16 x i8> undef to <16 x i32>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D = zext <16 x i16> undef to <16 x i32>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %E = zext <32 x i8> undef to <32 x i16>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SKX256-LABEL: 'zext512'
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = zext <8 x i16> undef to <8 x i64>
;
   %C = zext <16 x i8> undef to <16 x i32>
   %D = zext <16 x i16> undef to <16 x i32>
   %E = zext <32 x i8> undef to <32 x i16>
   ret void
 }
 
 define void @sext256() "min-legal-vector-width"="256" {
 ; VEC256-LABEL: 'sext256'
-; VEC256-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %A = sext <8 x i8> undef to <8 x i64>
-; VEC256-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %B = sext <8 x i16> undef to <8 x i64>
+; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %A = sext <8 x i8> undef to <8 x i64>
+; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %B = sext <8 x i16> undef to <8 x i64>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C = sext <8 x i32> undef to <8 x i64>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D = sext <16 x i8> undef to <16 x i32>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %E = sext <16 x i16> undef to <16 x i32>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %F = sext <32 x i8> undef to <32 x i16>
 ; VEC256-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; VEC512-LABEL: 'sext256'
-; VEC512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %A = sext <8 x i8> undef to <8 x i64>
+; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = sext <8 x i8> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = sext <8 x i16> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = sext <8 x i32> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = sext <16 x i8> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %E = sext <16 x i16> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F = sext <32 x i8> undef to <32 x i16>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %A = sext <8 x i8> undef to <8 x i64>
   %B = sext <8 x i16> undef to <8 x i64>
   %C = sext <8 x i32> undef to <8 x i64>
   %D = sext <16 x i8> undef to <16 x i32>
   %E = sext <16 x i16> undef to <16 x i32>
   %F = sext <32 x i8> undef to <32 x i16>
   ret void
 }
 
 define void @sext512() "min-legal-vector-width"="512" {
 ; AVX-LABEL: 'sext512'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %A = sext <8 x i8> undef to <8 x i64>
-; AVX-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %B = sext <8 x i16> undef to <8 x i64>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %A = sext <8 x i8> undef to <8 x i64>
+; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %B = sext <8 x i16> undef to <8 x i64>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %C = sext <8 x i32> undef to <8 x i64>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %D = sext <16 x i8> undef to <16 x i32>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %E = sext <16 x i16> undef to <16 x i32>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %F = sext <32 x i8> undef to <32 x i16>
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SKX256-LABEL: 'sext512'
-; SKX256-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %A = sext <8 x i8> undef to <8 x i64>
+; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = sext <8 x i8> undef to <8 x i64>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = sext <8 x i16> undef to <8 x i64>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = sext <8 x i32> undef to <8 x i64>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = sext <16 x i8> undef to <16 x i32>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %E = sext <16 x i16> undef to <16 x i32>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F = sext <32 x i8> undef to <32 x i16>
 ; SKX256-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; VEC512-LABEL: 'sext512'
-; VEC512-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %A = sext <8 x i8> undef to <8 x i64>
+; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %A = sext <8 x i8> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %B = sext <8 x i16> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %C = sext <8 x i32> undef to <8 x i64>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %D = sext <16 x i8> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %E = sext <16 x i16> undef to <16 x i32>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F = sext <32 x i8> undef to <32 x i16>
 ; VEC512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %A = sext <8 x i8> undef to <8 x i64>
   %B = sext <8 x i16> undef to <8 x i64>
   %C = sext <8 x i32> undef to <8 x i64>
   %D = sext <16 x i8> undef to <16 x i32>
   %E = sext <16 x i16> undef to <16 x i32>
   %F = sext <32 x i8> undef to <32 x i16>
   ret void
 }

llvm/trunk/test/Transforms/SLPVectorizer/X86/cast.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -basicaa -slp-vectorizer -dce -S | FileCheck %s
-; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -basicaa -slp-vectorizer -dce -S | FileCheck %s
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -basicaa -slp-vectorizer -dce -S | FileCheck %s --check-prefixes=CHECK,SSE42
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -basicaa -slp-vectorizer -dce -S | FileCheck %s --check-prefixes=CHECK,AVX
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 ; int test_sext_4i8_to_4i32(int * restrict A, char * restrict B) {
 ;     A[0] = B[0];
 ;     A[1] = B[1];
 ;     A[2] = B[2];
 ;     A[3] = B[3];
entry:
   %3 = load i16, i16* %arrayidx8, align 1
   %conv9 = zext i16 %3 to i32
   %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 3
   store i32 %conv9, i32* %arrayidx10, align 4
   ret i32 undef
 }
 
 define i64 @test_sext_4i16_to_4i64(i64* noalias nocapture %A, i16* noalias nocapture %B) {
-; CHECK-LABEL: @test_sext_4i16_to_4i64(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <2 x i16>*
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i16>, <2 x i16>* [[TMP0]], align 1
-; CHECK-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i64>
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to <2 x i64>*
-; CHECK-NEXT:    store <2 x i64> [[TMP2]], <2 x i64>* [[TMP3]], align 4
-; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[B]], i64 2
-; CHECK-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 2
-; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i16* [[ARRAYIDX5]] to <2 x i16>*
-; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i16>, <2 x i16>* [[TMP4]], align 1
-; CHECK-NEXT:    [[TMP6:%.*]] = sext <2 x i16> [[TMP5]] to <2 x i64>
-; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i64* [[ARRAYIDX7]] to <2 x i64>*
-; CHECK-NEXT:    store <2 x i64> [[TMP6]], <2 x i64>* [[TMP7]], align 4
-; CHECK-NEXT:    ret i64 undef
+; SSE42-LABEL: @test_sext_4i16_to_4i64(
+; SSE42-NEXT:  entry:
+; SSE42-NEXT:    [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <2 x i16>*
+; SSE42-NEXT:    [[TMP1:%.*]] = load <2 x i16>, <2 x i16>* [[TMP0]], align 1
+; SSE42-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i64>
+; SSE42-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to <2 x i64>*
+; SSE42-NEXT:    store <2 x i64> [[TMP2]], <2 x i64>* [[TMP3]], align 4
+; SSE42-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[B]], i64 2
+; SSE42-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 2
+; SSE42-NEXT:    [[TMP4:%.*]] = bitcast i16* [[ARRAYIDX5]] to <2 x i16>*
+; SSE42-NEXT:    [[TMP5:%.*]] = load <2 x i16>, <2 x i16>* [[TMP4]], align 1
+; SSE42-NEXT:    [[TMP6:%.*]] = sext <2 x i16> [[TMP5]] to <2 x i64>
+; SSE42-NEXT:    [[TMP7:%.*]] = bitcast i64* [[ARRAYIDX7]] to <2 x i64>*
+; SSE42-NEXT:    store <2 x i64> [[TMP6]], <2 x i64>* [[TMP7]], align 4
+; SSE42-NEXT:    ret i64 undef
+;
+; AVX-LABEL: @test_sext_4i16_to_4i64(
+; AVX-NEXT:  entry:
+; AVX-NEXT:    [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <4 x i16>*
+; AVX-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 1
+; AVX-NEXT:    [[TMP2:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i64>
+; AVX-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to <4 x i64>*
+; AVX-NEXT:    store <4 x i64> [[TMP2]], <4 x i64>* [[TMP3]], align 4
+; AVX-NEXT:    ret i64 undef
 ;
 entry:
   %0 = load i16, i16* %B, align 1
   %conv = sext i16 %0 to i64
   store i64 %conv, i64* %A, align 4
   %arrayidx2 = getelementptr inbounds i16, i16* %B, i64 1
   %1 = load i16, i16* %arrayidx2, align 1
   %conv3 = sext i16 %1 to i64

llvm/trunk/test/Transforms/SLPVectorizer/X86/sext.ll

 ; SLM-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
 ; SLM-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
 ; SLM-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
 ; SLM-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
 ; SLM-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
 ; SLM-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
 ; SLM-NEXT:    ret <4 x i64> [[V3]]
 ;
-; AVX-LABEL: @loadext_4i8_to_4i64(
-; AVX-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; AVX-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; AVX-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <2 x i8>*
-; AVX-NEXT:    [[TMP2:%.*]] = load <2 x i8>, <2 x i8>* [[TMP1]], align 1
-; AVX-NEXT:    [[I2:%.*]] = load i8, i8* [[P2]], align 1
-; AVX-NEXT:    [[I3:%.*]] = load i8, i8* [[P3]], align 1
-; AVX-NEXT:    [[TMP3:%.*]] = sext <2 x i8> [[TMP2]] to <2 x i64>
-; AVX-NEXT:    [[X2:%.*]] = sext i8 [[I2]] to i64
-; AVX-NEXT:    [[X3:%.*]] = sext i8 [[I3]] to i64
-; AVX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
-; AVX-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
-; AVX-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
-; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
-; AVX-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
-; AVX-NEXT:    ret <4 x i64> [[V3]]
+; AVX1-LABEL: @loadext_4i8_to_4i64(
+; AVX1-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX1-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX1-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX1-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <2 x i8>*
+; AVX1-NEXT:    [[TMP2:%.*]] = load <2 x i8>, <2 x i8>* [[TMP1]], align 1
+; AVX1-NEXT:    [[I2:%.*]] = load i8, i8* [[P2]], align 1
+; AVX1-NEXT:    [[I3:%.*]] = load i8, i8* [[P3]], align 1
+; AVX1-NEXT:    [[TMP3:%.*]] = sext <2 x i8> [[TMP2]] to <2 x i64>
+; AVX1-NEXT:    [[X2:%.*]] = sext i8 [[I2]] to i64
+; AVX1-NEXT:    [[X3:%.*]] = sext i8 [[I3]] to i64
+; AVX1-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
+; AVX1-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
+; AVX1-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX1-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
+; AVX1-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
+; AVX1-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX2-LABEL: @loadext_4i8_to_4i64(
+; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <4 x i8>*
+; AVX2-NEXT:    [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* [[TMP1]], align 1
+; AVX2-NEXT:    [[TMP3:%.*]] = sext <4 x i8> [[TMP2]] to <4 x i64>
+; AVX2-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX2-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX2-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX2-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX2-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX2-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX2-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX512-LABEL: @loadext_4i8_to_4i64(
+; AVX512-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX512-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX512-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <4 x i8>*
+; AVX512-NEXT:    [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* [[TMP1]], align 1
+; AVX512-NEXT:    [[TMP3:%.*]] = sext <4 x i8> [[TMP2]] to <4 x i64>
+; AVX512-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX512-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX512-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX512-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX512-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX512-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX512-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX512-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX512-NEXT:    ret <4 x i64> [[V3]]
 ;
   %p1 = getelementptr inbounds i8, i8* %p0, i64 1
   %p2 = getelementptr inbounds i8, i8* %p0, i64 2
   %p3 = getelementptr inbounds i8, i8* %p0, i64 3
   %i0 = load i8, i8* %p0, align 1
   %i1 = load i8, i8* %p1, align 1
   %i2 = load i8, i8* %p2, align 1
   %i3 = load i8, i8* %p3, align 1
;
   %v4 = insertelement <8 x i16>   %v3, i16 %x4, i32 4
   %v5 = insertelement <8 x i16>   %v4, i16 %x5, i32 5
   %v6 = insertelement <8 x i16>   %v5, i16 %x6, i32 6
   %v7 = insertelement <8 x i16>   %v6, i16 %x7, i32 7
   ret <8 x i16> %v7
 }
 
 define <8 x i32> @loadext_8i8_to_8i32(i8* %p0) {
-; SSE-LABEL: @loadext_8i8_to_8i32(
-; SSE-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; SSE-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; SSE-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; SSE-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
-; SSE-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
-; SSE-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
-; SSE-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
-; SSE-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <8 x i8>*
-; SSE-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[TMP1]], align 1
-; SSE-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[TMP2]] to <8 x i32>
-; SSE-NEXT:    [[TMP4:%.*]] = extractelement <8 x i32> [[TMP3]], i32 0
-; SSE-NEXT:    [[V0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP4]], i32 0
-; SSE-NEXT:    [[TMP5:%.*]] = extractelement <8 x i32> [[TMP3]], i32 1
-; SSE-NEXT:    [[V1:%.*]] = insertelement <8 x i32> [[V0]], i32 [[TMP5]], i32 1
-; SSE-NEXT:    [[TMP6:%.*]] = extractelement <8 x i32> [[TMP3]], i32 2
-; SSE-NEXT:    [[V2:%.*]] = insertelement <8 x i32> [[V1]], i32 [[TMP6]], i32 2
-; SSE-NEXT:    [[TMP7:%.*]] = extractelement <8 x i32> [[TMP3]], i32 3
-; SSE-NEXT:    [[V3:%.*]] = insertelement <8 x i32> [[V2]], i32 [[TMP7]], i32 3
-; SSE-NEXT:    [[TMP8:%.*]] = extractelement <8 x i32> [[TMP3]], i32 4
-; SSE-NEXT:    [[V4:%.*]] = insertelement <8 x i32> [[V3]], i32 [[TMP8]], i32 4
-; SSE-NEXT:    [[TMP9:%.*]] = extractelement <8 x i32> [[TMP3]], i32 5
-; SSE-NEXT:    [[V5:%.*]] = insertelement <8 x i32> [[V4]], i32 [[TMP9]], i32 5
-; SSE-NEXT:    [[TMP10:%.*]] = extractelement <8 x i32> [[TMP3]], i32 6
-; SSE-NEXT:    [[V6:%.*]] = insertelement <8 x i32> [[V5]], i32 [[TMP10]], i32 6
-; SSE-NEXT:    [[TMP11:%.*]] = extractelement <8 x i32> [[TMP3]], i32 7
-; SSE-NEXT:    [[V7:%.*]] = insertelement <8 x i32> [[V6]], i32 [[TMP11]], i32 7
-; SSE-NEXT:    ret <8 x i32> [[V7]]
-;
-; AVX1-LABEL: @loadext_8i8_to_8i32(
-; AVX1-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; AVX1-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; AVX1-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; AVX1-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
-; AVX1-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
-; AVX1-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
-; AVX1-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
-; AVX1-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <4 x i8>*
-; AVX1-NEXT:    [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* [[TMP1]], align 1
-; AVX1-NEXT:    [[I4:%.*]] = load i8, i8* [[P4]], align 1
-; AVX1-NEXT:    [[I5:%.*]] = load i8, i8* [[P5]], align 1
-; AVX1-NEXT:    [[I6:%.*]] = load i8, i8* [[P6]], align 1
-; AVX1-NEXT:    [[I7:%.*]] = load i8, i8* [[P7]], align 1
-; AVX1-NEXT:    [[TMP3:%.*]] = sext <4 x i8> [[TMP2]] to <4 x i32>
-; AVX1-NEXT:    [[X4:%.*]] = sext i8 [[I4]] to i32
-; AVX1-NEXT:    [[X5:%.*]] = sext i8 [[I5]] to i32
-; AVX1-NEXT:    [[X6:%.*]] = sext i8 [[I6]] to i32
-; AVX1-NEXT:    [[X7:%.*]] = sext i8 [[I7]] to i32
-; AVX1-NEXT:    [[TMP4:%.*]] = extractelement <4 x i32> [[TMP3]], i32 0
-; AVX1-NEXT:    [[V0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP4]], i32 0
-; AVX1-NEXT:    [[TMP5:%.*]] = extractelement <4 x i32> [[TMP3]], i32 1
-; AVX1-NEXT:    [[V1:%.*]] = insertelement <8 x i32> [[V0]], i32 [[TMP5]], i32 1
-; AVX1-NEXT:    [[TMP6:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
-; AVX1-NEXT:    [[V2:%.*]] = insertelement <8 x i32> [[V1]], i32 [[TMP6]], i32 2
-; AVX1-NEXT:    [[TMP7:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
-; AVX1-NEXT:    [[V3:%.*]] = insertelement <8 x i32> [[V2]], i32 [[TMP7]], i32 3
-; AVX1-NEXT:    [[V4:%.*]] = insertelement <8 x i32> [[V3]], i32 [[X4]], i32 4
-; AVX1-NEXT:    [[V5:%.*]] = insertelement <8 x i32> [[V4]], i32 [[X5]], i32 5
-; AVX1-NEXT:    [[V6:%.*]] = insertelement <8 x i32> [[V5]], i32 [[X6]], i32 6
-; AVX1-NEXT:    [[V7:%.*]] = insertelement <8 x i32> [[V6]], i32 [[X7]], i32 7
-; AVX1-NEXT:    ret <8 x i32> [[V7]]
-;
-; AVX2-LABEL: @loadext_8i8_to_8i32(
-; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; AVX2-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
-; AVX2-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
-; AVX2-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
-; AVX2-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
-; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <8 x i8>*
-; AVX2-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[TMP1]], align 1
-; AVX2-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[TMP2]] to <8 x i32>
-; AVX2-NEXT:    [[TMP4:%.*]] = extractelement <8 x i32> [[TMP3]], i32 0
-; AVX2-NEXT:    [[V0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP4]], i32 0
-; AVX2-NEXT:    [[TMP5:%.*]] = extractelement <8 x i32> [[TMP3]], i32 1
-; AVX2-NEXT:    [[V1:%.*]] = insertelement <8 x i32> [[V0]], i32 [[TMP5]], i32 1
-; AVX2-NEXT:    [[TMP6:%.*]] = extractelement <8 x i32> [[TMP3]], i32 2
-; AVX2-NEXT:    [[V2:%.*]] = insertelement <8 x i32> [[V1]], i32 [[TMP6]], i32 2
-; AVX2-NEXT:    [[TMP7:%.*]] = extractelement <8 x i32> [[TMP3]], i32 3
-; AVX2-NEXT:    [[V3:%.*]] = insertelement <8 x i32> [[V2]], i32 [[TMP7]], i32 3
-; AVX2-NEXT:    [[TMP8:%.*]] = extractelement <8 x i32> [[TMP3]], i32 4
-; AVX2-NEXT:    [[V4:%.*]] = insertelement <8 x i32> [[V3]], i32 [[TMP8]], i32 4
-; AVX2-NEXT:    [[TMP9:%.*]] = extractelement <8 x i32> [[TMP3]], i32 5
-; AVX2-NEXT:    [[V5:%.*]] = insertelement <8 x i32> [[V4]], i32 [[TMP9]], i32 5
-; AVX2-NEXT:    [[TMP10:%.*]] = extractelement <8 x i32> [[TMP3]], i32 6
-; AVX2-NEXT:    [[V6:%.*]] = insertelement <8 x i32> [[V5]], i32 [[TMP10]], i32 6
-; AVX2-NEXT:    [[TMP11:%.*]] = extractelement <8 x i32> [[TMP3]], i32 7
-; AVX2-NEXT:    [[V7:%.*]] = insertelement <8 x i32> [[V6]], i32 [[TMP11]], i32 7
-; AVX2-NEXT:    ret <8 x i32> [[V7]]
-;
-; AVX512-LABEL: @loadext_8i8_to_8i32(
-; AVX512-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; AVX512-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; AVX512-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; AVX512-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
-; AVX512-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
-; AVX512-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
-; AVX512-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
-; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <8 x i8>*
-; AVX512-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[TMP1]], align 1
-; AVX512-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[TMP2]] to <8 x i32>
-; AVX512-NEXT:    [[TMP4:%.*]] = extractelement <8 x i32> [[TMP3]], i32 0
-; AVX512-NEXT:    [[V0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP4]], i32 0
-; AVX512-NEXT:    [[TMP5:%.*]] = extractelement <8 x i32> [[TMP3]], i32 1
-; AVX512-NEXT:    [[V1:%.*]] = insertelement <8 x i32> [[V0]], i32 [[TMP5]], i32 1
-; AVX512-NEXT:    [[TMP6:%.*]] = extractelement <8 x i32> [[TMP3]], i32 2
-; AVX512-NEXT:    [[V2:%.*]] = insertelement <8 x i32> [[V1]], i32 [[TMP6]], i32 2
-; AVX512-NEXT:    [[TMP7:%.*]] = extractelement <8 x i32> [[TMP3]], i32 3
-; AVX512-NEXT:    [[V3:%.*]] = insertelement <8 x i32> [[V2]], i32 [[TMP7]], i32 3
-; AVX512-NEXT:    [[TMP8:%.*]] = extractelement <8 x i32> [[TMP3]], i32 4
-; AVX512-NEXT:    [[V4:%.*]] = insertelement <8 x i32> [[V3]], i32 [[TMP8]], i32 4
-; AVX512-NEXT:    [[TMP9:%.*]] = extractelement <8 x i32> [[TMP3]], i32 5
-; AVX512-NEXT:    [[V5:%.*]] = insertelement <8 x i32> [[V4]], i32 [[TMP9]], i32 5
-; AVX512-NEXT:    [[TMP10:%.*]] = extractelement <8 x i32> [[TMP3]], i32 6
-; AVX512-NEXT:    [[V6:%.*]] = insertelement <8 x i32> [[V5]], i32 [[TMP10]], i32 6
-; AVX512-NEXT:    [[TMP11:%.*]] = extractelement <8 x i32> [[TMP3]], i32 7
-; AVX512-NEXT:    [[V7:%.*]] = insertelement <8 x i32> [[V6]], i32 [[TMP11]], i32 7
-; AVX512-NEXT:    ret <8 x i32> [[V7]]
+; CHECK-LABEL: @loadext_8i8_to_8i32(
+; CHECK-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; CHECK-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; CHECK-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; CHECK-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
+; CHECK-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
+; CHECK-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
+; CHECK-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <8 x i8>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[TMP2]] to <8 x i32>
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <8 x i32> [[TMP3]], i32 0
+; CHECK-NEXT:    [[V0:%.*]] = insertelement <8 x i32> undef, i32 [[TMP4]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <8 x i32> [[TMP3]], i32 1
+; CHECK-NEXT:    [[V1:%.*]] = insertelement <8 x i32> [[V0]], i32 [[TMP5]], i32 1
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <8 x i32> [[TMP3]], i32 2
+; CHECK-NEXT:    [[V2:%.*]] = insertelement <8 x i32> [[V1]], i32 [[TMP6]], i32 2
+; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <8 x i32> [[TMP3]], i32 3
+; CHECK-NEXT:    [[V3:%.*]] = insertelement <8 x i32> [[V2]], i32 [[TMP7]], i32 3
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <8 x i32> [[TMP3]], i32 4
+; CHECK-NEXT:    [[V4:%.*]] = insertelement <8 x i32> [[V3]], i32 [[TMP8]], i32 4
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <8 x i32> [[TMP3]], i32 5
+; CHECK-NEXT:    [[V5:%.*]] = insertelement <8 x i32> [[V4]], i32 [[TMP9]], i32 5
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <8 x i32> [[TMP3]], i32 6
+; CHECK-NEXT:    [[V6:%.*]] = insertelement <8 x i32> [[V5]], i32 [[TMP10]], i32 6
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <8 x i32> [[TMP3]], i32 7
+; CHECK-NEXT:    [[V7:%.*]] = insertelement <8 x i32> [[V6]], i32 [[TMP11]], i32 7
+; CHECK-NEXT:    ret <8 x i32> [[V7]]
 ;
   %p1 = getelementptr inbounds i8, i8* %p0, i64 1
   %p2 = getelementptr inbounds i8, i8* %p0, i64 2
   %p3 = getelementptr inbounds i8, i8* %p0, i64 3
   %p4 = getelementptr inbounds i8, i8* %p0, i64 4
   %p5 = getelementptr inbounds i8, i8* %p0, i64 5
   %p6 = getelementptr inbounds i8, i8* %p0, i64 6
   %p7 = getelementptr inbounds i8, i8* %p0, i64 7

 ; SLM-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
 ; SLM-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
 ; SLM-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
 ; SLM-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
 ; SLM-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
 ; SLM-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
 ; SLM-NEXT:    ret <4 x i64> [[V3]]
 ;
-; AVX-LABEL: @loadext_4i16_to_4i64(
-; AVX-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
-; AVX-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
-; AVX-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <2 x i16>*
-; AVX-NEXT:    [[TMP2:%.*]] = load <2 x i16>, <2 x i16>* [[TMP1]], align 1
-; AVX-NEXT:    [[I2:%.*]] = load i16, i16* [[P2]], align 1
-; AVX-NEXT:    [[I3:%.*]] = load i16, i16* [[P3]], align 1
-; AVX-NEXT:    [[TMP3:%.*]] = sext <2 x i16> [[TMP2]] to <2 x i64>
-; AVX-NEXT:    [[X2:%.*]] = sext i16 [[I2]] to i64
-; AVX-NEXT:    [[X3:%.*]] = sext i16 [[I3]] to i64
-; AVX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
-; AVX-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
-; AVX-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
-; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
-; AVX-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
-; AVX-NEXT:    ret <4 x i64> [[V3]]
+; AVX1-LABEL: @loadext_4i16_to_4i64(
+; AVX1-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX1-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX1-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX1-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <2 x i16>*
+; AVX1-NEXT:    [[TMP2:%.*]] = load <2 x i16>, <2 x i16>* [[TMP1]], align 1
+; AVX1-NEXT:    [[I2:%.*]] = load i16, i16* [[P2]], align 1
+; AVX1-NEXT:    [[I3:%.*]] = load i16, i16* [[P3]], align 1
+; AVX1-NEXT:    [[TMP3:%.*]] = sext <2 x i16> [[TMP2]] to <2 x i64>
+; AVX1-NEXT:    [[X2:%.*]] = sext i16 [[I2]] to i64
+; AVX1-NEXT:    [[X3:%.*]] = sext i16 [[I3]] to i64
+; AVX1-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
+; AVX1-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
+; AVX1-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX1-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
+; AVX1-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
+; AVX1-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX2-LABEL: @loadext_4i16_to_4i64(
+; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <4 x i16>*
+; AVX2-NEXT:    [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* [[TMP1]], align 1
+; AVX2-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP2]] to <4 x i64>
+; AVX2-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX2-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX2-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX2-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX2-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX2-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX2-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX512-LABEL: @loadext_4i16_to_4i64(
+; AVX512-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX512-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX512-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <4 x i16>*
+; AVX512-NEXT:    [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* [[TMP1]], align 1
+; AVX512-NEXT:    [[TMP3:%.*]] = sext <4 x i16> [[TMP2]] to <4 x i64>
+; AVX512-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX512-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX512-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX512-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX512-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX512-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX512-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX512-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX512-NEXT:    ret <4 x i64> [[V3]]
 ;
   %p1 = getelementptr inbounds i16, i16* %p0, i64 1
   %p2 = getelementptr inbounds i16, i16* %p0, i64 2
   %p3 = getelementptr inbounds i16, i16* %p0, i64 3
   %i0 = load i16, i16* %p0, align 1
   %i1 = load i16, i16* %p1, align 1
   %i2 = load i16, i16* %p2, align 1
   %i3 = load i16, i16* %p3, align 1

llvm/trunk/test/Transforms/SLPVectorizer/X86/zext.ll

 ; SLM-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
 ; SLM-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
 ; SLM-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
 ; SLM-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
 ; SLM-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
 ; SLM-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
 ; SLM-NEXT:    ret <4 x i64> [[V3]]
 ;
-; AVX-LABEL: @loadext_4i8_to_4i64(
-; AVX-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
-; AVX-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
-; AVX-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <2 x i8>*
-; AVX-NEXT:    [[TMP2:%.*]] = load <2 x i8>, <2 x i8>* [[TMP1]], align 1
-; AVX-NEXT:    [[I2:%.*]] = load i8, i8* [[P2]], align 1
-; AVX-NEXT:    [[I3:%.*]] = load i8, i8* [[P3]], align 1
-; AVX-NEXT:    [[TMP3:%.*]] = zext <2 x i8> [[TMP2]] to <2 x i64>
-; AVX-NEXT:    [[X2:%.*]] = zext i8 [[I2]] to i64
-; AVX-NEXT:    [[X3:%.*]] = zext i8 [[I3]] to i64
-; AVX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
-; AVX-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
-; AVX-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
-; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
-; AVX-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
-; AVX-NEXT:    ret <4 x i64> [[V3]]
+; AVX1-LABEL: @loadext_4i8_to_4i64(
+; AVX1-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX1-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX1-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX1-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <2 x i8>*
+; AVX1-NEXT:    [[TMP2:%.*]] = load <2 x i8>, <2 x i8>* [[TMP1]], align 1
+; AVX1-NEXT:    [[I2:%.*]] = load i8, i8* [[P2]], align 1
+; AVX1-NEXT:    [[I3:%.*]] = load i8, i8* [[P3]], align 1
+; AVX1-NEXT:    [[TMP3:%.*]] = zext <2 x i8> [[TMP2]] to <2 x i64>
+; AVX1-NEXT:    [[X2:%.*]] = zext i8 [[I2]] to i64
+; AVX1-NEXT:    [[X3:%.*]] = zext i8 [[I3]] to i64
+; AVX1-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
+; AVX1-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
+; AVX1-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX1-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
+; AVX1-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
+; AVX1-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX2-LABEL: @loadext_4i8_to_4i64(
+; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <4 x i8>*
+; AVX2-NEXT:    [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* [[TMP1]], align 1
+; AVX2-NEXT:    [[TMP3:%.*]] = zext <4 x i8> [[TMP2]] to <4 x i64>
+; AVX2-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX2-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX2-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX2-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX2-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX2-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX2-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX512-LABEL: @loadext_4i8_to_4i64(
+; AVX512-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
+; AVX512-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
+; AVX512-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
+; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <4 x i8>*
+; AVX512-NEXT:    [[TMP2:%.*]] = load <4 x i8>, <4 x i8>* [[TMP1]], align 1
+; AVX512-NEXT:    [[TMP3:%.*]] = zext <4 x i8> [[TMP2]] to <4 x i64>
+; AVX512-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX512-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX512-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX512-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX512-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX512-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX512-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX512-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX512-NEXT:    ret <4 x i64> [[V3]]
 ;
   %p1 = getelementptr inbounds i8, i8* %p0, i64 1
   %p2 = getelementptr inbounds i8, i8* %p0, i64 2
   %p3 = getelementptr inbounds i8, i8* %p0, i64 3
   %i0 = load i8, i8* %p0, align 1
   %i1 = load i8, i8* %p1, align 1
   %i2 = load i8, i8* %p2, align 1
   %i3 = load i8, i8* %p3, align 1

 ; SLM-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
 ; SLM-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
 ; SLM-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
 ; SLM-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
 ; SLM-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
 ; SLM-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
 ; SLM-NEXT:    ret <4 x i64> [[V3]]
 ;
-; AVX-LABEL: @loadext_4i16_to_4i64(
-; AVX-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
-; AVX-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
-; AVX-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <2 x i16>*
-; AVX-NEXT:    [[TMP2:%.*]] = load <2 x i16>, <2 x i16>* [[TMP1]], align 1
-; AVX-NEXT:    [[I2:%.*]] = load i16, i16* [[P2]], align 1
-; AVX-NEXT:    [[I3:%.*]] = load i16, i16* [[P3]], align 1
-; AVX-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP2]] to <2 x i64>
-; AVX-NEXT:    [[X2:%.*]] = zext i16 [[I2]] to i64
-; AVX-NEXT:    [[X3:%.*]] = zext i16 [[I3]] to i64
-; AVX-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
-; AVX-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
-; AVX-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
-; AVX-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
-; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
-; AVX-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
-; AVX-NEXT:    ret <4 x i64> [[V3]]
+; AVX1-LABEL: @loadext_4i16_to_4i64(
+; AVX1-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX1-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX1-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX1-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <2 x i16>*
+; AVX1-NEXT:    [[TMP2:%.*]] = load <2 x i16>, <2 x i16>* [[TMP1]], align 1
+; AVX1-NEXT:    [[I2:%.*]] = load i16, i16* [[P2]], align 1
+; AVX1-NEXT:    [[I3:%.*]] = load i16, i16* [[P3]], align 1
+; AVX1-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP2]] to <2 x i64>
+; AVX1-NEXT:    [[X2:%.*]] = zext i16 [[I2]] to i64
+; AVX1-NEXT:    [[X3:%.*]] = zext i16 [[I3]] to i64
+; AVX1-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP3]], i32 0
+; AVX1-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP3]], i32 1
+; AVX1-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX1-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[X2]], i32 2
+; AVX1-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[X3]], i32 3
+; AVX1-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX2-LABEL: @loadext_4i16_to_4i64(
+; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <4 x i16>*
+; AVX2-NEXT:    [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* [[TMP1]], align 1
+; AVX2-NEXT:    [[TMP3:%.*]] = zext <4 x i16> [[TMP2]] to <4 x i64>
+; AVX2-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX2-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX2-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX2-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX2-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX2-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX2-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX2-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX2-NEXT:    ret <4 x i64> [[V3]]
+;
+; AVX512-LABEL: @loadext_4i16_to_4i64(
+; AVX512-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
+; AVX512-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
+; AVX512-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
+; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <4 x i16>*
+; AVX512-NEXT:    [[TMP2:%.*]] = load <4 x i16>, <4 x i16>* [[TMP1]], align 1
+; AVX512-NEXT:    [[TMP3:%.*]] = zext <4 x i16> [[TMP2]] to <4 x i64>
+; AVX512-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
+; AVX512-NEXT:    [[V0:%.*]] = insertelement <4 x i64> undef, i64 [[TMP4]], i32 0
+; AVX512-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
+; AVX512-NEXT:    [[V1:%.*]] = insertelement <4 x i64> [[V0]], i64 [[TMP5]], i32 1
+; AVX512-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
+; AVX512-NEXT:    [[V2:%.*]] = insertelement <4 x i64> [[V1]], i64 [[TMP6]], i32 2
+; AVX512-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
+; AVX512-NEXT:    [[V3:%.*]] = insertelement <4 x i64> [[V2]], i64 [[TMP7]], i32 3
+; AVX512-NEXT:    ret <4 x i64> [[V3]]
 ;
   %p1 = getelementptr inbounds i16, i16* %p0, i64 1
   %p2 = getelementptr inbounds i16, i16* %p0, i64 2
   %p3 = getelementptr inbounds i16, i16* %p0, i64 3
   %i0 = load i16, i16* %p0, align 1
   %i1 = load i16, i16* %p1, align 1
   %i2 = load i16, i16* %p2, align 1
   %i3 = load i16, i16* %p3, align 1