diff --git a/lib/Target/X86/X86TargetTransformInfo.cpp b/lib/Target/X86/X86TargetTransformInfo.cpp
index 1d94eed..ec6dc72 100644
--- a/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -1653,6 +1653,9 @@
   static const CostTblEntry AVX512BWCostTbl[] = {
     { ISD::SETCC,   MVT::v32i16,  1 },
     { ISD::SETCC,   MVT::v64i8,   1 },
+
+    { ISD::SELECT,  MVT::v32i16,  1 },
+    { ISD::SELECT,  MVT::v64i8,   1 },
   };
 
   static const CostTblEntry AVX512CostTbl[] = {
@@ -1660,6 +1663,11 @@
     { ISD::SETCC,   MVT::v16i32,  1 },
     { ISD::SETCC,   MVT::v8f64,   1 },
     { ISD::SETCC,   MVT::v16f32,  1 },
+
+    { ISD::SELECT,  MVT::v8i64,   1 },
+    { ISD::SELECT,  MVT::v16i32,  1 },
+    { ISD::SELECT,  MVT::v8f64,   1 },
+    { ISD::SELECT,  MVT::v16f32,  1 },
   };
 
   static const CostTblEntry AVX2CostTbl[] = {
@@ -1667,6 +1675,11 @@
     { ISD::SETCC,   MVT::v8i32,   1 },
     { ISD::SETCC,   MVT::v16i16,  1 },
     { ISD::SETCC,   MVT::v32i8,   1 },
+
+    { ISD::SELECT,  MVT::v4i64,   1 }, // pblendvb
+    { ISD::SELECT,  MVT::v8i32,   1 }, // pblendvb
+    { ISD::SELECT,  MVT::v16i16,  1 }, // pblendvb
+    { ISD::SELECT,  MVT::v32i8,   1 }, // pblendvb
   };
 
   static const CostTblEntry AVX1CostTbl[] = {
@@ -1677,6 +1690,13 @@
     { ISD::SETCC,   MVT::v8i32,   4 },
     { ISD::SETCC,   MVT::v16i16,  4 },
     { ISD::SETCC,   MVT::v32i8,   4 },
+
+    { ISD::SELECT,  MVT::v4f64,   1 }, // vblendvpd
+    { ISD::SELECT,  MVT::v8f32,   1 }, // vblendvps
+    { ISD::SELECT,  MVT::v4i64,   1 }, // vblendvpd
+    { ISD::SELECT,  MVT::v8i32,   1 }, // vblendvps
+    { ISD::SELECT,  MVT::v16i16,  3 }, // vandps + vandnps + vorps
+    { ISD::SELECT,  MVT::v32i8,   3 }, // vandps + vandnps + vorps
   };
 
   static const CostTblEntry SSE42CostTbl[] = {
@@ -1685,6 +1705,15 @@
     { ISD::SETCC,   MVT::v2i64,   1 },
   };
 
+  static const CostTblEntry SSE41CostTbl[] = {
+    { ISD::SELECT,  MVT::v2f64,   1 }, // blendvpd
+    { ISD::SELECT,  MVT::v4f32,   1 }, // blendvps
+    { ISD::SELECT,  MVT::v2i64,   1 }, // pblendvb
+    { ISD::SELECT,  MVT::v4i32,   1 }, // pblendvb
+    { ISD::SELECT,  MVT::v8i16,   1 }, // pblendvb
+    { ISD::SELECT,  MVT::v16i8,   1 }, // pblendvb
+  };
+
   static const CostTblEntry SSE2CostTbl[] = {
     { ISD::SETCC,   MVT::v2f64,   2 },
     { ISD::SETCC,   MVT::f64,     1 },
@@ -1692,11 +1721,19 @@
     { ISD::SETCC,   MVT::v4i32,   1 },
     { ISD::SETCC,   MVT::v8i16,   1 },
     { ISD::SETCC,   MVT::v16i8,   1 },
+
+    { ISD::SELECT,  MVT::v2f64,   3 }, // andpd + andnpd + orpd
+    { ISD::SELECT,  MVT::v2i64,   3 }, // pand + pandn + por
+    { ISD::SELECT,  MVT::v4i32,   3 }, // pand + pandn + por
+    { ISD::SELECT,  MVT::v8i16,   3 }, // pand + pandn + por
+    { ISD::SELECT,  MVT::v16i8,   3 }, // pand + pandn + por
   };
 
   static const CostTblEntry SSE1CostTbl[] = {
     { ISD::SETCC,   MVT::v4f32,   2 },
     { ISD::SETCC,   MVT::f32,     1 },
+
+    { ISD::SELECT,  MVT::v4f32,   3 }, // andps + andnps + orps
   };
 
   if (ST->hasBWI())
@@ -1719,6 +1756,10 @@
     if (const auto *Entry = CostTableLookup(SSE42CostTbl, ISD, MTy))
       return LT.first * Entry->Cost;
 
+  if (ST->hasSSE41())
+    if (const auto *Entry = CostTableLookup(SSE41CostTbl, ISD, MTy))
+      return LT.first * Entry->Cost;
+
   if (ST->hasSSE2())
     if (const auto *Entry = CostTableLookup(SSE2CostTbl, ISD, MTy))
       return LT.first * Entry->Cost;
diff --git a/test/Analysis/CostModel/X86/fshl.ll b/test/Analysis/CostModel/X86/fshl.ll
index d5da8e7..38621c3 100644
--- a/test/Analysis/CostModel/X86/fshl.ll
+++ b/test/Analysis/CostModel/X86/fshl.ll
@@ -22,9 +22,9 @@
 define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512, i64 %c64, <2 x i64> %c128, <4 x i64> %c256, <8 x i64> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i64'
@@ -93,9 +93,9 @@
 define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i32'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 132 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i32'
@@ -164,9 +164,9 @@
 define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 276 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 284 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i16'
@@ -179,8 +179,8 @@
 ; AVX1-LABEL: 'var_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 146 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i16'
@@ -228,15 +228,15 @@
 ; BDVER2-LABEL: 'var_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'var_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 146 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
@@ -249,9 +249,9 @@
 define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 228 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 236 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i8'
@@ -264,8 +264,8 @@
 ; AVX1-LABEL: 'var_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i8'
@@ -313,15 +313,15 @@
 ; BDVER2-LABEL: 'var_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'var_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
@@ -340,9 +340,9 @@
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i64'
@@ -427,14 +427,23 @@
 }
 
 define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
-; SSE-LABEL: 'splatvar_funnel_i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatvar_funnel_i32'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatvar_funnel_i32'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -509,22 +518,31 @@
 }
 
 define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
-; SSE-LABEL: 'splatvar_funnel_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatvar_funnel_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatvar_funnel_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i16'
@@ -586,8 +604,8 @@
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatvar_funnel_i16'
@@ -595,8 +613,8 @@
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
@@ -613,9 +631,9 @@
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 228 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 236 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i8'
@@ -632,8 +650,8 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i8'
@@ -695,8 +713,8 @@
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatvar_funnel_i8'
@@ -704,8 +722,8 @@
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -724,9 +742,9 @@
 define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i64'
@@ -795,9 +813,9 @@
 define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i32'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i32'
@@ -866,9 +884,9 @@
 define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 78 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i16'
@@ -881,8 +899,8 @@
 ; AVX1-LABEL: 'constant_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i16'
@@ -930,15 +948,15 @@
 ; BDVER2-LABEL: 'constant_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'constant_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
@@ -951,9 +969,9 @@
 define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 224 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 232 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i8'
@@ -966,8 +984,8 @@
 ; AVX1-LABEL: 'constant_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i8'
@@ -1015,15 +1033,15 @@
 ; BDVER2-LABEL: 'constant_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'constant_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
@@ -1040,9 +1058,9 @@
 define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512) {
 ; SSSE3-LABEL: 'splatconstant_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatconstant_funnel_i64'
@@ -1109,12 +1127,19 @@
 }
 
 define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i32'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i32'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
@@ -1173,18 +1198,25 @@
 }
 
 define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i16'
@@ -1232,15 +1264,15 @@
 ; BDVER2-LABEL: 'splatconstant_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatconstant_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
@@ -1251,18 +1283,25 @@
 }
 
 define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i8'
@@ -1310,15 +1349,15 @@
 ; BDVER2-LABEL: 'splatconstant_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatconstant_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
diff --git a/test/Analysis/CostModel/X86/fshr.ll b/test/Analysis/CostModel/X86/fshr.ll
index 7450fc7..31e409c 100644
--- a/test/Analysis/CostModel/X86/fshr.ll
+++ b/test/Analysis/CostModel/X86/fshr.ll
@@ -22,9 +22,9 @@
 define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512, i64 %c64, <2 x i64> %c128, <4 x i64> %c256, <8 x i64> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i64'
@@ -93,9 +93,9 @@
 define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i32'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 132 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i32'
@@ -164,9 +164,9 @@
 define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 276 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 284 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i16'
@@ -179,8 +179,8 @@
 ; AVX1-LABEL: 'var_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 146 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i16'
@@ -228,15 +228,15 @@
 ; BDVER2-LABEL: 'var_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'var_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 142 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 146 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
@@ -249,9 +249,9 @@
 define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSSE3-LABEL: 'var_funnel_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 228 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 236 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'var_funnel_i8'
@@ -264,8 +264,8 @@
 ; AVX1-LABEL: 'var_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i8'
@@ -313,15 +313,15 @@
 ; BDVER2-LABEL: 'var_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'var_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
@@ -340,9 +340,9 @@
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i64'
@@ -427,14 +427,23 @@
 }
 
 define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
-; SSE-LABEL: 'splatvar_funnel_i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatvar_funnel_i32'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatvar_funnel_i32'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -509,22 +518,31 @@
 }
 
 define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
-; SSE-LABEL: 'splatvar_funnel_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatvar_funnel_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatvar_funnel_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i16'
@@ -586,8 +604,8 @@
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatvar_funnel_i16'
@@ -595,8 +613,8 @@
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
@@ -613,9 +631,9 @@
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 228 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 59 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 236 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i8'
@@ -632,8 +650,8 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i8'
@@ -695,8 +713,8 @@
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatvar_funnel_i8'
@@ -704,8 +722,8 @@
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 122 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 126 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -724,9 +742,9 @@
 define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i64'
@@ -795,9 +813,9 @@
 define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i32'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i32'
@@ -866,9 +884,9 @@
 define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i16'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 78 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i16'
@@ -881,8 +899,8 @@
 ; AVX1-LABEL: 'constant_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i16'
@@ -930,15 +948,15 @@
 ; BDVER2-LABEL: 'constant_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'constant_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
@@ -951,9 +969,9 @@
 define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512) {
 ; SSSE3-LABEL: 'constant_funnel_i8'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 224 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 232 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_funnel_i8'
@@ -966,8 +984,8 @@
 ; AVX1-LABEL: 'constant_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i8'
@@ -1015,15 +1033,15 @@
 ; BDVER2-LABEL: 'constant_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'constant_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 124 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
@@ -1040,9 +1058,9 @@
 define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %b64, <2 x i64> %b128, <4 x i64> %b256, <8 x i64> %b512) {
 ; SSSE3-LABEL: 'splatconstant_funnel_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'splatconstant_funnel_i64'
@@ -1109,12 +1127,19 @@
 }
 
 define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i32'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i32'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
@@ -1173,18 +1198,25 @@
 }
 
 define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i16'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i16'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i16'
@@ -1232,15 +1264,15 @@
 ; BDVER2-LABEL: 'splatconstant_funnel_i16'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatconstant_funnel_i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
@@ -1251,18 +1283,25 @@
 }
 
 define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512) {
-; SSE-LABEL: 'splatconstant_funnel_i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; SSSE3-LABEL: 'splatconstant_funnel_i8'
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'splatconstant_funnel_i8'
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i8'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i8'
@@ -1310,15 +1349,15 @@
 ; BDVER2-LABEL: 'splatconstant_funnel_i8'
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BDVER2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BDVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; BDVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; BTVER2-LABEL: 'splatconstant_funnel_i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; BTVER2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; BTVER2-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
diff --git a/test/Analysis/CostModel/X86/reduce-smax-widen.ll b/test/Analysis/CostModel/X86/reduce-smax-widen.ll
index 44b01d3..6eba5c3 100644
--- a/test/Analysis/CostModel/X86/reduce-smax-widen.ll
+++ b/test/Analysis/CostModel/X86/reduce-smax-widen.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-smax.ll b/test/Analysis/CostModel/X86/reduce-smax.ll
index b4b9a3b..4ea2a70 100644
--- a/test/Analysis/CostModel/X86/reduce-smax.ll
+++ b/test/Analysis/CostModel/X86/reduce-smax.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smax.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smax.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smax.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smax.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smax.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smax.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smax.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smax.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smax.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smax.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smax.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smax.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smax.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-smin-widen.ll b/test/Analysis/CostModel/X86/reduce-smin-widen.ll
index 18dd54c..afcf9b2 100644
--- a/test/Analysis/CostModel/X86/reduce-smin-widen.ll
+++ b/test/Analysis/CostModel/X86/reduce-smin-widen.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,19 +124,19 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -150,9 +150,9 @@
 ; AVX1-LABEL: 'reduce_i16'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -197,23 +197,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -231,9 +231,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-smin.ll b/test/Analysis/CostModel/X86/reduce-smin.ll
index f420ad7..0d9adef 100644
--- a/test/Analysis/CostModel/X86/reduce-smin.ll
+++ b/test/Analysis/CostModel/X86/reduce-smin.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.smin.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.smin.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.smin.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.smin.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.smin.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.smin.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.smin.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.smin.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.smin.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.smin.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smin.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smin.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smin.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smin.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.smin.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.smin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.smin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.smin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.smin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.smin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.smin.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.smin.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.smin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.smin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.smin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.smin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.smin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-umax-widen.ll b/test/Analysis/CostModel/X86/reduce-umax-widen.ll
index c48cfdf..c3ef8fd 100644
--- a/test/Analysis/CostModel/X86/reduce-umax-widen.ll
+++ b/test/Analysis/CostModel/X86/reduce-umax-widen.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-umax.ll b/test/Analysis/CostModel/X86/reduce-umax.ll
index 490b0e4..e0a5ab1 100644
--- a/test/Analysis/CostModel/X86/reduce-umax.ll
+++ b/test/Analysis/CostModel/X86/reduce-umax.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umax.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umax.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umax.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umax.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umax.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umax.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umax.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umax.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umax.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umax.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umax.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umax.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umax.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umax.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umax.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umax.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umax.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umax.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umax.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umax.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umax.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umax.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umax.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-umin-widen.ll b/test/Analysis/CostModel/X86/reduce-umin-widen.ll
index b59e2f3..71674ad 100644
--- a/test/Analysis/CostModel/X86/reduce-umin-widen.ll
+++ b/test/Analysis/CostModel/X86/reduce-umin-widen.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/reduce-umin.ll b/test/Analysis/CostModel/X86/reduce-umin.ll
index f2f519c..4a08215 100644
--- a/test/Analysis/CostModel/X86/reduce-umin.ll
+++ b/test/Analysis/CostModel/X86/reduce-umin.ll
@@ -11,18 +11,18 @@
 define i32 @reduce_i64(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i64'
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i64'
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %V1 = call i64 @llvm.experimental.vector.reduce.umin.i64.v1i64(<1 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i64 @llvm.experimental.vector.reduce.umin.i64.v2i64(<2 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V4 = call i64 @llvm.experimental.vector.reduce.umin.i64.v4i64(<4 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8 = call i64 @llvm.experimental.vector.reduce.umin.i64.v8i64(<8 x i64> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V16 = call i64 @llvm.experimental.vector.reduce.umin.i64.v16i64(<16 x i64> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i64'
@@ -67,19 +67,19 @@
 
 define i32 @reduce_i32(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i32'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i32'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i32 @llvm.experimental.vector.reduce.umin.i32.v2i32(<2 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i32 @llvm.experimental.vector.reduce.umin.i32.v4i32(<4 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8 = call i32 @llvm.experimental.vector.reduce.umin.i32.v8i32(<8 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16 = call i32 @llvm.experimental.vector.reduce.umin.i32.v16i32(<16 x i32> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32 = call i32 @llvm.experimental.vector.reduce.umin.i32.v32i32(<32 x i32> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i32'
@@ -124,21 +124,21 @@
 
 define i32 @reduce_i16(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i16'
@@ -154,9 +154,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i16 @llvm.experimental.vector.reduce.umin.i16.v2i16(<2 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i16 @llvm.experimental.vector.reduce.umin.i16.v4i16(<4 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i16 @llvm.experimental.vector.reduce.umin.i16.v8i16(<8 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V16 = call i16 @llvm.experimental.vector.reduce.umin.i16.v16i16(<16 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V32 = call i16 @llvm.experimental.vector.reduce.umin.i16.v32i16(<32 x i16> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 82 for instruction: %V64 = call i16 @llvm.experimental.vector.reduce.umin.i16.v64i16(<64 x i16> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i16'
@@ -206,23 +206,23 @@
 
 define i32 @reduce_i8(i32 %arg) {
 ; SSE2-LABEL: 'reduce_i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 55 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 85 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSSE3-LABEL: 'reduce_i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2 = call i8 @llvm.experimental.vector.reduce.umin.i8.v2i8(<2 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; SSSE3-NEXT:  Cost Model: Found an estimated cost of 49 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; SSE42-LABEL: 'reduce_i8'
@@ -240,9 +240,9 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4 = call i8 @llvm.experimental.vector.reduce.umin.i8.v4i8(<4 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8 = call i8 @llvm.experimental.vector.reduce.umin.i8.v8i8(<8 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16 = call i8 @llvm.experimental.vector.reduce.umin.i8.v16i8(<16 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 81 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %V32 = call i8 @llvm.experimental.vector.reduce.umin.i8.v32i8(<32 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 83 for instruction: %V64 = call i8 @llvm.experimental.vector.reduce.umin.i8.v64i8(<64 x i8> undef)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 97 for instruction: %V128 = call i8 @llvm.experimental.vector.reduce.umin.i8.v128i8(<128 x i8> undef)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'reduce_i8'
diff --git a/test/Analysis/CostModel/X86/vselect-cost.ll b/test/Analysis/CostModel/X86/vselect-cost.ll
index a77d8fa..5829423 100644
--- a/test/Analysis/CostModel/X86/vselect-cost.ll
+++ b/test/Analysis/CostModel/X86/vselect-cost.ll
@@ -9,24 +9,43 @@
 ; Verify the cost of vector select instructions.
 
 define i32 @test_select() {
-; SSE-LABEL: 'test_select'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = select i1 undef, i64 undef, i64 undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = select <2 x i1> undef, <2 x i64> undef, <2 x i64> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = select <4 x i1> undef, <4 x i64> undef, <4 x i64> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = select <8 x i1> undef, <8 x i64> undef, <8 x i64> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = select i1 undef, i32 undef, i32 undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = select <4 x i1> undef, <4 x i32> undef, <4 x i32> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = select <8 x i1> undef, <8 x i32> undef, <8 x i32> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I32 = select <16 x i1> undef, <16 x i32> undef, <16 x i32> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = select i1 undef, i16 undef, i16 undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = select <8 x i1> undef, <8 x i16> undef, <8 x i16> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = select i1 undef, i8 undef, i8 undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = select <16 x i1> undef, <16 x i8> undef, <16 x i8> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+; SSE2-LABEL: 'test_select'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = select i1 undef, i64 undef, i64 undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = select <2 x i1> undef, <2 x i64> undef, <2 x i64> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = select <4 x i1> undef, <4 x i64> undef, <4 x i64> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = select <8 x i1> undef, <8 x i64> undef, <8 x i64> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = select i1 undef, i32 undef, i32 undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4I32 = select <4 x i1> undef, <4 x i32> undef, <4 x i32> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I32 = select <8 x i1> undef, <8 x i32> undef, <8 x i32> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I32 = select <16 x i1> undef, <16 x i32> undef, <16 x i32> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = select i1 undef, i16 undef, i16 undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = select <8 x i1> undef, <8 x i16> undef, <8 x i16> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = select i1 undef, i8 undef, i8 undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = select <16 x i1> undef, <16 x i8> undef, <16 x i8> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
+;
+; SSE41-LABEL: 'test_select'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = select i1 undef, i64 undef, i64 undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = select <2 x i1> undef, <2 x i64> undef, <2 x i64> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I64 = select <4 x i1> undef, <4 x i64> undef, <4 x i64> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I64 = select <8 x i1> undef, <8 x i64> undef, <8 x i64> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = select i1 undef, i32 undef, i32 undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = select <4 x i1> undef, <4 x i32> undef, <4 x i32> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I32 = select <8 x i1> undef, <8 x i32> undef, <8 x i32> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I32 = select <16 x i1> undef, <16 x i32> undef, <16 x i32> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = select i1 undef, i16 undef, i16 undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = select <8 x i1> undef, <8 x i16> undef, <8 x i16> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = select i1 undef, i8 undef, i8 undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = select <16 x i1> undef, <16 x i8> undef, <16 x i8> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX1-LABEL: 'test_select'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = select i1 undef, i64 undef, i64 undef
@@ -39,12 +58,12 @@
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I32 = select <16 x i1> undef, <16 x i32> undef, <16 x i32> undef
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = select i1 undef, i16 undef, i16 undef
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = select <8 x i1> undef, <8 x i16> undef, <8 x i16> undef
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = select i1 undef, i8 undef, i8 undef
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = select <16 x i1> undef, <16 x i8> undef, <16 x i8> undef
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; AVX2-LABEL: 'test_select'
@@ -131,45 +150,85 @@
 ; <4 x float>. Integers of the same size should also use those instructions.
 
 define <2 x i64> @test_2i64(<2 x i64> %a, <2 x i64> %b) {
-; CHECK-LABEL: 'test_2i64'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x i64> %a, <2 x i64> %b
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel
+; SSE2-LABEL: 'test_2i64'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x i64> %a, <2 x i64> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel
+;
+; SSE41-LABEL: 'test_2i64'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x i64> %a, <2 x i64> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel
+;
+; AVX-LABEL: 'test_2i64'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x i64> %a, <2 x i64> %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel
 ;
   %sel = select <2 x i1> <i1 true, i1 false>, <2 x i64> %a, <2 x i64> %b
   ret <2 x i64> %sel
 }
 
 define <2 x double> @test_2double(<2 x double> %a, <2 x double> %b) {
-; CHECK-LABEL: 'test_2double'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x double> %a, <2 x double> %b
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel
+; SSE2-LABEL: 'test_2double'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x double> %a, <2 x double> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel
+;
+; SSE41-LABEL: 'test_2double'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x double> %a, <2 x double> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel
+;
+; AVX-LABEL: 'test_2double'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> <i1 true, i1 false>, <2 x double> %a, <2 x double> %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel
 ;
   %sel = select <2 x i1> <i1 true, i1 false>, <2 x double> %a, <2 x double> %b
   ret <2 x double> %sel
 }
 
 define <4 x i32> @test_4i32(<4 x i32> %a, <4 x i32> %b) {
-; CHECK-LABEL: 'test_4i32'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x i32> %a, <4 x i32> %b
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel
+; SSE2-LABEL: 'test_4i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x i32> %a, <4 x i32> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel
+;
+; SSE41-LABEL: 'test_4i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x i32> %a, <4 x i32> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel
+;
+; AVX-LABEL: 'test_4i32'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x i32> %a, <4 x i32> %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel
 ;
   %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x i32> %a, <4 x i32> %b
   ret <4 x i32> %sel
 }
 
 define <4 x float> @test_4float(<4 x float> %a, <4 x float> %b) {
-; CHECK-LABEL: 'test_4float'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 true>, <4 x float> %a, <4 x float> %b
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel
+; SSE2-LABEL: 'test_4float'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 true>, <4 x float> %a, <4 x float> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel
+;
+; SSE41-LABEL: 'test_4float'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 true>, <4 x float> %a, <4 x float> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel
+;
+; AVX-LABEL: 'test_4float'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 true>, <4 x float> %a, <4 x float> %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel
 ;
   %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 true>, <4 x float> %a, <4 x float> %b
   ret <4 x float> %sel
 }
 
 define <16 x i8> @test_16i8(<16 x i8> %a, <16 x i8> %b) {
-; CHECK-LABEL: 'test_16i8'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <16 x i8> %a, <16 x i8> %b
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel
+; SSE2-LABEL: 'test_16i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <16 x i8> %a, <16 x i8> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel
+;
+; SSE41-LABEL: 'test_16i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <16 x i8> %a, <16 x i8> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel
+;
+; AVX-LABEL: 'test_16i8'
+; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <16 x i8> %a, <16 x i8> %b
+; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel
 ;
   %sel = select <16 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <16 x i8> %a, <16 x i8> %b
   ret <16 x i8> %sel
@@ -178,9 +237,13 @@
 ; AVX added blend instructions with an immediate for <4 x double> and
 ; <8 x float>. Integers of the same size should also use those instructions.
 define <4 x i64> @test_4i64(<4 x i64> %a, <4 x i64> %b) {
-; SSE-LABEL: 'test_4i64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 false, i1 true>, <4 x i64> %a, <4 x i64> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %sel
+; SSE2-LABEL: 'test_4i64'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 false, i1 true>, <4 x i64> %a, <4 x i64> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %sel
+;
+; SSE41-LABEL: 'test_4i64'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 false, i1 true>, <4 x i64> %a, <4 x i64> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %sel
 ;
 ; AVX-LABEL: 'test_4i64'
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 false, i1 true>, <4 x i64> %a, <4 x i64> %b
@@ -191,9 +254,13 @@
 }
 
 define <4 x double> @test_4double(<4 x double> %a, <4 x double> %b) {
-; SSE-LABEL: 'test_4double'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x double> %a, <4 x double> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %sel
+; SSE2-LABEL: 'test_4double'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x double> %a, <4 x double> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %sel
+;
+; SSE41-LABEL: 'test_4double'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x double> %a, <4 x double> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %sel
 ;
 ; AVX-LABEL: 'test_4double'
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> <i1 true, i1 false, i1 true, i1 false>, <4 x double> %a, <4 x double> %b
@@ -204,9 +271,13 @@
 }
 
 define <8 x i32> @test_8i32(<8 x i32> %a, <8 x i32> %b) {
-; SSE-LABEL: 'test_8i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 false>, <8 x i32> %a, <8 x i32> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %sel
+; SSE2-LABEL: 'test_8i32'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 false>, <8 x i32> %a, <8 x i32> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %sel
+;
+; SSE41-LABEL: 'test_8i32'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 false>, <8 x i32> %a, <8 x i32> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %sel
 ;
 ; AVX-LABEL: 'test_8i32'
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 false>, <8 x i32> %a, <8 x i32> %b
@@ -217,9 +288,13 @@
 }
 
 define <8 x float> @test_8float(<8 x float> %a, <8 x float> %b) {
-; SSE-LABEL: 'test_8float'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <8 x float> %a, <8 x float> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %sel
+; SSE2-LABEL: 'test_8float'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <8 x float> %a, <8 x float> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %sel
+;
+; SSE41-LABEL: 'test_8float'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <8 x float> %a, <8 x float> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %sel
 ;
 ; AVX-LABEL: 'test_8float'
 ; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <8 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <8 x float> %a, <8 x float> %b
@@ -231,26 +306,50 @@
 
 ; AVX2
 define <16 x i16> @test_16i16(<16 x i16> %a, <16 x i16> %b) {
-; SSE-LABEL: 'test_16i16'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
+; SSE2-LABEL: 'test_16i16'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
 ;
-; AVX-LABEL: 'test_16i16'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
+; SSE41-LABEL: 'test_16i16'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
+;
+; AVX1-LABEL: 'test_16i16'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
+;
+; AVX2-LABEL: 'test_16i16'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
+;
+; AVX512-LABEL: 'test_16i16'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel
 ;
   %sel = select <16 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <16 x i16> %a, <16 x i16> %b
   ret <16 x i16> %sel
 }
 
 define <32 x i8> @test_32i8(<32 x i8> %a, <32 x i8> %b) {
-; SSE-LABEL: 'test_32i8'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
+; SSE2-LABEL: 'test_32i8'
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
+; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
 ;
-; AVX-LABEL: 'test_32i8'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
+; SSE41-LABEL: 'test_32i8'
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
+; SSE41-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
+;
+; AVX1-LABEL: 'test_32i8'
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
+;
+; AVX2-LABEL: 'test_32i8'
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
+;
+; AVX512-LABEL: 'test_32i8'
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
+; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel
 ;
   %sel = select <32 x i1> <i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true, i1 true, i1 false, i1 true, i1 true>, <32 x i8> %a, <32 x i8> %b
   ret <32 x i8> %sel
diff --git a/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll b/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
index 25e164f..f9ccbf1 100644
--- a/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
+++ b/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -S -loop-vectorize -mtriple=x86_64-apple-darwin %s | FileCheck %s
+; RUN: opt -S -loop-vectorize -mtriple=x86_64-apple-darwin %s | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt -S -loop-vectorize -mtriple=x86_64-apple-darwin -mattr=+avx %s | FileCheck %s --check-prefixes=CHECK,AVX
 
 ; Two mostly identical functions. The only difference is the presence of
 ; fast-math flags on the second. The loop is a pretty simple reduction:
@@ -64,59 +65,85 @@
 }
 
 define double @sumIfVector(double* nocapture readonly %arr) {
-; CHECK-LABEL: @sumIfVector(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> undef, i32 [[INDEX]], i32 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> undef, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1>
-; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr double, double* [[ARR:%.*]], i32 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr double, double* [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[TMP2]] to <2 x double>*
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, <2 x double>* [[TMP3]], align 8
-; CHECK-NEXT:    [[TMP4:%.*]] = fcmp fast une <2 x double> [[WIDE_LOAD]], <double 4.200000e+01, double 4.200000e+01>
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd fast <2 x double> [[VEC_PHI]], [[WIDE_LOAD]]
-; CHECK-NEXT:    [[TMP6:%.*]] = xor <2 x i1> [[TMP4]], <i1 true, i1 true>
-; CHECK-NEXT:    [[PREDPHI]] = select <2 x i1> [[TMP4]], <2 x double> [[TMP5]], <2 x double> [[VEC_PHI]]
-; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
-; CHECK:       middle.block:
-; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <2 x double> [[PREDPHI]], <2 x double> undef, <2 x i32> <i32 1, i32 undef>
-; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x double> [[PREDPHI]], [[RDX_SHUF]]
-; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x double> [[BIN_RDX]], i32 0
-; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 32, 32
-; CHECK-NEXT:    br i1 [[CMP_N]], label [[DONE:%.*]], label [[SCALAR_PH]]
-; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    br label [[LOOP:%.*]]
-; CHECK:       loop:
-; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
-; CHECK-NEXT:    [[TOT:%.*]] = phi double [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
-; CHECK-NEXT:    [[ADDR:%.*]] = getelementptr double, double* [[ARR]], i32 [[I]]
-; CHECK-NEXT:    [[NEXTVAL:%.*]] = load double, double* [[ADDR]]
-; CHECK-NEXT:    [[TST:%.*]] = fcmp fast une double [[NEXTVAL]], 4.200000e+01
-; CHECK-NEXT:    br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
-; CHECK:       do.add:
-; CHECK-NEXT:    [[TOT_NEW:%.*]] = fadd fast double [[TOT]], [[NEXTVAL]]
-; CHECK-NEXT:    br label [[NEXT_ITER]]
-; CHECK:       no.add:
-; CHECK-NEXT:    br label [[NEXT_ITER]]
-; CHECK:       next.iter:
-; CHECK-NEXT:    [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
-; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
-; CHECK-NEXT:    [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
-; CHECK-NEXT:    br i1 [[AGAIN]], label [[LOOP]], label [[DONE]], !llvm.loop !2
-; CHECK:       done:
-; CHECK-NEXT:    [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret double [[TOT_NEXT_LCSSA]]
+; SSE-LABEL: @sumIfVector(
+; SSE-NEXT:  entry:
+; SSE-NEXT:    br label [[LOOP:%.*]]
+; SSE:       loop:
+; SSE-NEXT:    [[I:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
+; SSE-NEXT:    [[TOT:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
+; SSE-NEXT:    [[ADDR:%.*]] = getelementptr double, double* [[ARR:%.*]], i32 [[I]]
+; SSE-NEXT:    [[NEXTVAL:%.*]] = load double, double* [[ADDR]]
+; SSE-NEXT:    [[TST:%.*]] = fcmp fast une double [[NEXTVAL]], 4.200000e+01
+; SSE-NEXT:    br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
+; SSE:       do.add:
+; SSE-NEXT:    [[TOT_NEW:%.*]] = fadd fast double [[TOT]], [[NEXTVAL]]
+; SSE-NEXT:    br label [[NEXT_ITER]]
+; SSE:       no.add:
+; SSE-NEXT:    br label [[NEXT_ITER]]
+; SSE:       next.iter:
+; SSE-NEXT:    [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
+; SSE-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; SSE-NEXT:    [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
+; SSE-NEXT:    br i1 [[AGAIN]], label [[LOOP]], label [[DONE:%.*]]
+; SSE:       done:
+; SSE-NEXT:    [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ]
+; SSE-NEXT:    ret double [[TOT_NEXT_LCSSA]]
+;
+; AVX-LABEL: @sumIfVector(
+; AVX-NEXT:  entry:
+; AVX-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; AVX:       vector.ph:
+; AVX-NEXT:    br label [[VECTOR_BODY:%.*]]
+; AVX:       vector.body:
+; AVX-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; AVX-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
+; AVX-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> undef, i32 [[INDEX]], i32 0
+; AVX-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3>
+; AVX-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; AVX-NEXT:    [[TMP1:%.*]] = getelementptr double, double* [[ARR:%.*]], i32 [[TMP0]]
+; AVX-NEXT:    [[TMP2:%.*]] = getelementptr double, double* [[TMP1]], i32 0
+; AVX-NEXT:    [[TMP3:%.*]] = bitcast double* [[TMP2]] to <4 x double>*
+; AVX-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, <4 x double>* [[TMP3]], align 8
+; AVX-NEXT:    [[TMP4:%.*]] = fcmp fast une <4 x double> [[WIDE_LOAD]], <double 4.200000e+01, double 4.200000e+01, double 4.200000e+01, double 4.200000e+01>
+; AVX-NEXT:    [[TMP5:%.*]] = fadd fast <4 x double> [[VEC_PHI]], [[WIDE_LOAD]]
+; AVX-NEXT:    [[TMP6:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
+; AVX-NEXT:    [[PREDPHI]] = select <4 x i1> [[TMP4]], <4 x double> [[TMP5]], <4 x double> [[VEC_PHI]]
+; AVX-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
+; AVX-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
+; AVX-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
+; AVX:       middle.block:
+; AVX-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x double> [[PREDPHI]], <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; AVX-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x double> [[PREDPHI]], [[RDX_SHUF]]
+; AVX-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x double> [[BIN_RDX]], <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; AVX-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x double> [[BIN_RDX]], [[RDX_SHUF1]]
+; AVX-NEXT:    [[TMP8:%.*]] = extractelement <4 x double> [[BIN_RDX2]], i32 0
+; AVX-NEXT:    [[CMP_N:%.*]] = icmp eq i32 32, 32
+; AVX-NEXT:    br i1 [[CMP_N]], label [[DONE:%.*]], label [[SCALAR_PH]]
+; AVX:       scalar.ph:
+; AVX-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; AVX-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; AVX-NEXT:    br label [[LOOP:%.*]]
+; AVX:       loop:
+; AVX-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
+; AVX-NEXT:    [[TOT:%.*]] = phi double [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TOT_NEXT:%.*]], [[NEXT_ITER]] ]
+; AVX-NEXT:    [[ADDR:%.*]] = getelementptr double, double* [[ARR]], i32 [[I]]
+; AVX-NEXT:    [[NEXTVAL:%.*]] = load double, double* [[ADDR]]
+; AVX-NEXT:    [[TST:%.*]] = fcmp fast une double [[NEXTVAL]], 4.200000e+01
+; AVX-NEXT:    br i1 [[TST]], label [[DO_ADD:%.*]], label [[NO_ADD:%.*]]
+; AVX:       do.add:
+; AVX-NEXT:    [[TOT_NEW:%.*]] = fadd fast double [[TOT]], [[NEXTVAL]]
+; AVX-NEXT:    br label [[NEXT_ITER]]
+; AVX:       no.add:
+; AVX-NEXT:    br label [[NEXT_ITER]]
+; AVX:       next.iter:
+; AVX-NEXT:    [[TOT_NEXT]] = phi double [ [[TOT]], [[NO_ADD]] ], [ [[TOT_NEW]], [[DO_ADD]] ]
+; AVX-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
+; AVX-NEXT:    [[AGAIN:%.*]] = icmp ult i32 [[I_NEXT]], 32
+; AVX-NEXT:    br i1 [[AGAIN]], label [[LOOP]], label [[DONE]], !llvm.loop !2
+; AVX:       done:
+; AVX-NEXT:    [[TOT_NEXT_LCSSA:%.*]] = phi double [ [[TOT_NEXT]], [[NEXT_ITER]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; AVX-NEXT:    ret double [[TOT_NEXT_LCSSA]]
 ;
 entry:
   br label %loop
diff --git a/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll b/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll
index 5efabe1..5b3f5c5 100644
--- a/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll
+++ b/test/Transforms/LoopVectorize/X86/vector-scalar-select-cost.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s  -loop-vectorize -mcpu=core2 -debug-only=loop-vectorize 2>&1 -S | FileCheck %s
+; RUN: opt < %s  -loop-vectorize -mattr=+sse4.2 -debug-only=loop-vectorize 2>&1 -S | FileCheck %s
 ; REQUIRES: asserts
 ; Make sure we use the right select kind when querying select costs.
 
diff --git a/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll b/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
index 5cee363..47e89df 100644
--- a/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
+++ b/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
@@ -1,49 +1,93 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -basicaa -slp-vectorizer -S | FileCheck %s
+; RUN: opt < %s -basicaa -slp-vectorizer -S | FileCheck %s -check-prefixes=CHECK,SSE
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mattr=+avx | FileCheck %s -check-prefixes=CHECK,AVX
 
 target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.10.0"
 
 define void @testfunc(float* nocapture %dest, float* nocapture readonly %src) {
-; CHECK-LABEL: @testfunc(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
-; CHECK:       for.body:
-; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP23:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x float> [[TMP0]], i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> undef, float [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x float> [[TMP3]], float [[TMP4]], i32 1
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x float> undef, float [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP8:%.*]] = fadd <2 x float> [[TMP5]], [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = fmul <2 x float> zeroinitializer, [[TMP0]]
-; CHECK-NEXT:    [[TMP10:%.*]] = fadd <2 x float> [[TMP9]], [[TMP8]]
-; CHECK-NEXT:    [[TMP11:%.*]] = fcmp olt <2 x float> [[TMP10]], <float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP12:%.*]] = select <2 x i1> [[TMP11]], <2 x float> [[TMP10]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP13:%.*]] = fcmp olt <2 x float> [[TMP12]], <float -1.000000e+00, float -1.000000e+00>
-; CHECK-NEXT:    [[TMP14:%.*]] = fmul <2 x float> zeroinitializer, [[TMP12]]
-; CHECK-NEXT:    [[TMP15:%.*]] = select <2 x i1> [[TMP13]], <2 x float> <float -0.000000e+00, float -0.000000e+00>, <2 x float> [[TMP14]]
-; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x float> [[TMP15]], i32 0
-; CHECK-NEXT:    [[TMP17:%.*]] = extractelement <2 x float> [[TMP15]], i32 1
-; CHECK-NEXT:    [[ADD13]] = fadd float [[TMP16]], [[TMP17]]
-; CHECK-NEXT:    [[TMP18:%.*]] = insertelement <2 x float> undef, float [[TMP17]], i32 0
-; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <2 x float> [[TMP18]], float [[ADD13]], i32 1
-; CHECK-NEXT:    [[TMP20:%.*]] = fcmp olt <2 x float> [[TMP19]], <float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP21:%.*]] = select <2 x i1> [[TMP20]], <2 x float> [[TMP19]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
-; CHECK-NEXT:    [[TMP22:%.*]] = fcmp olt <2 x float> [[TMP21]], <float -1.000000e+00, float -1.000000e+00>
-; CHECK-NEXT:    [[TMP23]] = select <2 x i1> [[TMP22]], <2 x float> <float -1.000000e+00, float -1.000000e+00>, <2 x float> [[TMP21]]
-; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
-; CHECK:       for.end:
-; CHECK-NEXT:    ret void
+; SSE-LABEL: @testfunc(
+; SSE-NEXT:  entry:
+; SSE-NEXT:    br label [[FOR_BODY:%.*]]
+; SSE:       for.body:
+; SSE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT:    [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT:    [[S1_055:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[COND_I40:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT:    [[S0_054:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[COND_I44:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
+; SSE-NEXT:    [[TMP0:%.*]] = load float, float* [[ARRAYIDX]], align 4
+; SSE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SSE-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
+; SSE-NEXT:    store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
+; SSE-NEXT:    [[ADD:%.*]] = fadd float [[S0_054]], [[TMP0]]
+; SSE-NEXT:    [[ADD3:%.*]] = fadd float [[S1_055]], [[TMP0]]
+; SSE-NEXT:    [[MUL:%.*]] = fmul float [[S0_054]], 0.000000e+00
+; SSE-NEXT:    [[ADD4:%.*]] = fadd float [[MUL]], [[ADD3]]
+; SSE-NEXT:    [[MUL5:%.*]] = fmul float [[S1_055]], 0.000000e+00
+; SSE-NEXT:    [[ADD6:%.*]] = fadd float [[MUL5]], [[ADD]]
+; SSE-NEXT:    [[CMP_I:%.*]] = fcmp olt float [[ADD6]], 1.000000e+00
+; SSE-NEXT:    [[COND_I:%.*]] = select i1 [[CMP_I]], float [[ADD6]], float 1.000000e+00
+; SSE-NEXT:    [[CMP_I51:%.*]] = fcmp olt float [[COND_I]], -1.000000e+00
+; SSE-NEXT:    [[CMP_I49:%.*]] = fcmp olt float [[ADD4]], 1.000000e+00
+; SSE-NEXT:    [[COND_I50:%.*]] = select i1 [[CMP_I49]], float [[ADD4]], float 1.000000e+00
+; SSE-NEXT:    [[CMP_I47:%.*]] = fcmp olt float [[COND_I50]], -1.000000e+00
+; SSE-NEXT:    [[COND_I_OP:%.*]] = fmul float [[COND_I]], 0.000000e+00
+; SSE-NEXT:    [[MUL10:%.*]] = select i1 [[CMP_I51]], float -0.000000e+00, float [[COND_I_OP]]
+; SSE-NEXT:    [[COND_I50_OP:%.*]] = fmul float [[COND_I50]], 0.000000e+00
+; SSE-NEXT:    [[MUL11:%.*]] = select i1 [[CMP_I47]], float -0.000000e+00, float [[COND_I50_OP]]
+; SSE-NEXT:    [[ADD13]] = fadd float [[MUL10]], [[MUL11]]
+; SSE-NEXT:    [[CMP_I45:%.*]] = fcmp olt float [[ADD13]], 1.000000e+00
+; SSE-NEXT:    [[COND_I46:%.*]] = select i1 [[CMP_I45]], float [[ADD13]], float 1.000000e+00
+; SSE-NEXT:    [[CMP_I43:%.*]] = fcmp olt float [[COND_I46]], -1.000000e+00
+; SSE-NEXT:    [[COND_I44]] = select i1 [[CMP_I43]], float -1.000000e+00, float [[COND_I46]]
+; SSE-NEXT:    [[CMP_I41:%.*]] = fcmp olt float [[MUL11]], 1.000000e+00
+; SSE-NEXT:    [[COND_I42:%.*]] = select i1 [[CMP_I41]], float [[MUL11]], float 1.000000e+00
+; SSE-NEXT:    [[CMP_I39:%.*]] = fcmp olt float [[COND_I42]], -1.000000e+00
+; SSE-NEXT:    [[COND_I40]] = select i1 [[CMP_I39]], float -1.000000e+00, float [[COND_I42]]
+; SSE-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
+; SSE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; SSE:       for.end:
+; SSE-NEXT:    ret void
+;
+; AVX-LABEL: @testfunc(
+; AVX-NEXT:  entry:
+; AVX-NEXT:    br label [[FOR_BODY:%.*]]
+; AVX:       for.body:
+; AVX-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT:    [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT:    [[TMP0:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP23:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
+; AVX-NEXT:    [[TMP1:%.*]] = load float, float* [[ARRAYIDX]], align 4
+; AVX-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; AVX-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
+; AVX-NEXT:    store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
+; AVX-NEXT:    [[TMP2:%.*]] = extractelement <2 x float> [[TMP0]], i32 1
+; AVX-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> undef, float [[TMP2]], i32 0
+; AVX-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP0]], i32 0
+; AVX-NEXT:    [[TMP5:%.*]] = insertelement <2 x float> [[TMP3]], float [[TMP4]], i32 1
+; AVX-NEXT:    [[TMP6:%.*]] = insertelement <2 x float> undef, float [[TMP1]], i32 0
+; AVX-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
+; AVX-NEXT:    [[TMP8:%.*]] = fadd <2 x float> [[TMP5]], [[TMP7]]
+; AVX-NEXT:    [[TMP9:%.*]] = fmul <2 x float> zeroinitializer, [[TMP0]]
+; AVX-NEXT:    [[TMP10:%.*]] = fadd <2 x float> [[TMP9]], [[TMP8]]
+; AVX-NEXT:    [[TMP11:%.*]] = fcmp olt <2 x float> [[TMP10]], <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT:    [[TMP12:%.*]] = select <2 x i1> [[TMP11]], <2 x float> [[TMP10]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT:    [[TMP13:%.*]] = fcmp olt <2 x float> [[TMP12]], <float -1.000000e+00, float -1.000000e+00>
+; AVX-NEXT:    [[TMP14:%.*]] = fmul <2 x float> zeroinitializer, [[TMP12]]
+; AVX-NEXT:    [[TMP15:%.*]] = select <2 x i1> [[TMP13]], <2 x float> <float -0.000000e+00, float -0.000000e+00>, <2 x float> [[TMP14]]
+; AVX-NEXT:    [[TMP16:%.*]] = extractelement <2 x float> [[TMP15]], i32 0
+; AVX-NEXT:    [[TMP17:%.*]] = extractelement <2 x float> [[TMP15]], i32 1
+; AVX-NEXT:    [[ADD13]] = fadd float [[TMP16]], [[TMP17]]
+; AVX-NEXT:    [[TMP18:%.*]] = insertelement <2 x float> undef, float [[TMP17]], i32 0
+; AVX-NEXT:    [[TMP19:%.*]] = insertelement <2 x float> [[TMP18]], float [[ADD13]], i32 1
+; AVX-NEXT:    [[TMP20:%.*]] = fcmp olt <2 x float> [[TMP19]], <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT:    [[TMP21:%.*]] = select <2 x i1> [[TMP20]], <2 x float> [[TMP19]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT:    [[TMP22:%.*]] = fcmp olt <2 x float> [[TMP21]], <float -1.000000e+00, float -1.000000e+00>
+; AVX-NEXT:    [[TMP23]] = select <2 x i1> [[TMP22]], <2 x float> <float -1.000000e+00, float -1.000000e+00>, <2 x float> [[TMP21]]
+; AVX-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
+; AVX-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; AVX:       for.end:
+; AVX-NEXT:    ret void
 ;
 entry:
   br label %for.body
diff --git a/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll b/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
index 8432b91..271244e 100644
--- a/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
+++ b/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
@@ -677,35 +677,27 @@
 ; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
 ; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
 ; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; SSE-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], undef
-; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 undef
-; SSE-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], undef
-; SSE-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 undef
-; SSE-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], undef
-; SSE-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 undef
-; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP12]], undef
-; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP12]], i32 undef
-; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
+; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
+; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
+; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
+; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
+; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
+; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
+; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
+; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
 ; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
-; SSE-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SSE-NEXT:    [[TMP17:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SSE-NEXT:    [[TMP18:%.*]] = icmp sgt i32 [[TMP17]], [[TMP15]]
-; SSE-NEXT:    [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[TMP17]], i32 [[TMP15]]
-; SSE-NEXT:    [[TMP20:%.*]] = icmp sgt i32 [[TMP19]], [[TMP5]]
-; SSE-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP20]], i32 [[TMP19]], i32 [[TMP5]]
-; SSE-NEXT:    [[TMP21:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
-; SSE-NEXT:    [[TMP22:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
-; SSE-NEXT:    [[TMP23:%.*]] = icmp sgt i32 [[OP_EXTRA]], [[TMP22]]
-; SSE-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i32 [[OP_EXTRA]], i32 [[TMP22]]
-; SSE-NEXT:    [[TMP25:%.*]] = select i1 [[TMP4]], i32 3, i32 4
-; SSE-NEXT:    store i32 [[TMP25]], i32* @var, align 8
-; SSE-NEXT:    ret i32 [[TMP24]]
+; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
+; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
+; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
+; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
+; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
+; SSE-NEXT:    [[TMP24:%.*]] = select i1 [[TMP4]], i32 3, i32 4
+; SSE-NEXT:    store i32 [[TMP24]], i32* @var, align 8
+; SSE-NEXT:    ret i32 [[TMP23]]
 ;
 ; AVX-LABEL: @maxi8_mutiple_uses(
 ; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
@@ -853,35 +845,25 @@
 ; SSE-NEXT:    br label [[PP:%.*]]
 ; SSE:       pp:
 ; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
-; SSE-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
-; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], undef
-; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 undef
-; SSE-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], undef
-; SSE-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 undef
-; SSE-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP10]], undef
-; SSE-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32 [[TMP10]], i32 undef
-; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP12]], undef
-; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP12]], i32 undef
-; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
+; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
+; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
+; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
+; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
+; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
+; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
+; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
+; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
+; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
 ; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
 ; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
-; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
 ; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
-; SSE-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <4 x i32> [[TMP6]], [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP6]], <4 x i32> [[RDX_SHUF]]
-; SSE-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
-; SSE-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <4 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
-; SSE-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP2]], <4 x i32> [[RDX_MINMAX_SELECT]], <4 x i32> [[RDX_SHUF1]]
-; SSE-NEXT:    [[TMP20:%.*]] = extractelement <4 x i32> [[RDX_MINMAX_SELECT3]], i32 0
-; SSE-NEXT:    [[TMP21:%.*]] = icmp sgt i32 [[TMP20]], [[TMP15]]
-; SSE-NEXT:    [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP20]], i32 [[TMP15]]
-; SSE-NEXT:    [[TMP23:%.*]] = icmp sgt i32 [[TMP22]], [[TMP18]]
-; SSE-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i32 [[TMP22]], i32 [[TMP18]]
-; SSE-NEXT:    [[TMP25:%.*]] = icmp sgt i32 [[TMP24]], [[TMP5]]
-; SSE-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP25]], i32 [[TMP24]], i32 [[TMP5]]
-; SSE-NEXT:    [[TMP26:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
-; SSE-NEXT:    ret i32 [[OP_EXTRA]]
+; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
+; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
+; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
+; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
+; SSE-NEXT:    ret i32 [[TMP23]]
 ;
 ; AVX-LABEL: @maxi8_wrong_parent(
 ; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16