[llvm] 3dda247 - [CostModel][TTI] Replace BAD_ICMP_PREDICATE with ICMP_EQ for generic funnel shift cost expansion

Wed Oct 6 08:39:26 PDT 2021

Author: Simon Pilgrim
Date: 2021-10-06T16:39:16+01:00
New Revision: 3dda247e180fbe07aea67139e489fa5b3fe0d3e9

URL: https://github.com/llvm/llvm-project/commit/3dda247e180fbe07aea67139e489fa5b3fe0d3e9
DIFF: https://github.com/llvm/llvm-project/commit/3dda247e180fbe07aea67139e489fa5b3fe0d3e9.diff

LOG: [CostModel][TTI] Replace BAD_ICMP_PREDICATE with ICMP_EQ for generic funnel shift cost expansion

The comparison always checks for zero value so know the icmp predicate will be ICMP_EQ

Added: 
    

Modified: 
    llvm/include/llvm/CodeGen/BasicTTIImpl.h
    llvm/test/Analysis/CostModel/X86/fshl.ll
    llvm/test/Analysis/CostModel/X86/fshr.ll
    llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll

Removed: 
    


################################################################################
diff  --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index 7437bbac1a79..cb16051236fd 100644

--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -1465,10 +1465,10 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
         Type *CondTy = RetTy->getWithNewBitWidth(1);
         Cost +=
             thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, RetTy, CondTy,
-                                        CmpInst::BAD_ICMP_PREDICATE, CostKind);
+                                        CmpInst::ICMP_EQ, CostKind);
         Cost +=
             thisT()->getCmpSelInstrCost(BinaryOperator::Select, RetTy, CondTy,
-                                        CmpInst::BAD_ICMP_PREDICATE, CostKind);
+                                        CmpInst::ICMP_EQ, CostKind);
       }
       return Cost;
     }

diff  --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index 42f5dae61698..8ea89ff69d0a 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -22,30 +22,30 @@ target triple = "x86_64-apple-macosx10.8.0"
 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 25 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 50 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 100 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; SSE42-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> %c128)
-; SSE42-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> %c256)
-; SSE42-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> %c512)
+; SSE42-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> %c128)
+; SSE42-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> %c256)
+; SSE42-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> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'var_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; AVX1-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> %c256)
-; AVX1-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> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'var_funnel_i64'
@@ -57,16 +57,16 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ;
 ; SLM-LABEL: 'var_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; SLM-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)
-; SLM-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)
-; SLM-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)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
-; GLM-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> %c128)
-; GLM-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> %c256)
-; GLM-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> %c512)
+; GLM-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> %c128)
+; GLM-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> %c256)
+; GLM-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> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i64'
@@ -86,30 +86,30 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 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 38 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 76 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 152 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 35 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 70 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 140 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
-; SSE42-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> %c128)
-; SSE42-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> %c256)
-; SSE42-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> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'var_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 %c32)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; AVX2-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> %c256)
-; AVX2-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> %c512)
+; AVX2-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> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'var_funnel_i32'
@@ -121,16 +121,16 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ;
 ; SLM-LABEL: 'var_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
-; SLM-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> %c128)
-; SLM-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> %c256)
-; SLM-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> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
-; GLM-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> %c128)
-; GLM-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> %c256)
-; GLM-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> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i32'
@@ -150,37 +150,37 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 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 51 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 102 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 204 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 48 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 96 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 192 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 27 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 57 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 114 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 24 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 54 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 108 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX2-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> %c256)
+; AVX2-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> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'var_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512F-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> %c256)
+; AVX512F-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> %c512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'var_funnel_i16'
@@ -192,23 +192,23 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ;
 ; AVX512DQ-LABEL: 'var_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512DQ-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> %c256)
+; AVX512DQ-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> %c512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'var_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i16'
@@ -228,37 +228,37 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 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 37 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 74 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 148 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 34 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 68 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 136 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 29 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 26 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 58 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 116 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX2-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> %c256)
-; AVX2-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> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'var_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX512F-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> %c256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512F-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> %c512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'var_funnel_i8'
@@ -270,23 +270,23 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ;
 ; AVX512DQ-LABEL: 'var_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX512DQ-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> %c256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512DQ-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> %c512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'var_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i8'
@@ -312,36 +312,36 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; 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 19 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 38 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 76 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'
 ; SSE42-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
 ; SSE42-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
 ; SSE42-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
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i64'
 ; AVX1-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
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i64'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatvar_funnel_i64'
@@ -357,18 +357,18 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-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)
-; SLM-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)
-; SLM-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)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i64'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i64'
@@ -394,36 +394,36 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; 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 12 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 24 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 48 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 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 10 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 20 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 40 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 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
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX1-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)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i32'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX2-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)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatvar_funnel_i32'
@@ -439,18 +439,18 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SLM-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)
+; SLM-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)
+; SLM-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)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i32'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; GLM-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)
+; GLM-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)
+; GLM-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)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i32'
@@ -476,45 +476,45 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; 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 12 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 24 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 48 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 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 10 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 20 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 40 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 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 10 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 24 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 48 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 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 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'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX2-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)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i16'
 ; AVX512F-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
 ; AVX512F-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
 ; AVX512F-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
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512F-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)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i16'
@@ -530,27 +530,27 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; AVX512DQ-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
 ; AVX512DQ-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
 ; AVX512DQ-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
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512DQ-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)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i16'
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; SLM-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> %u256)
-; SLM-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> %u512)
+; SLM-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)
+; SLM-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)
+; SLM-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)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i16'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; GLM-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> %u256)
-; GLM-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> %u512)
+; GLM-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)
+; GLM-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)
+; GLM-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)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i16'
@@ -576,45 +576,45 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; 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 37 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 74 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 148 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 34 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 68 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 136 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'
 ; SSE42-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
 ; SSE42-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
 ; SSE42-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
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i8'
 ; AVX1-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
 ; 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 29 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 26 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 58 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 116 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'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX2-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> %u256)
-; AVX2-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> %u512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i8'
 ; AVX512F-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
 ; AVX512F-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
 ; AVX512F-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
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX512F-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> %u256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512F-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> %u512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i8'
@@ -630,27 +630,27 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; AVX512DQ-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
 ; AVX512DQ-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
 ; AVX512DQ-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
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX512DQ-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> %u256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512DQ-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> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i8'
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i8'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i8'
@@ -678,30 +678,30 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 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 24 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 48 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 96 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SSE42-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 1, i64 7>)
-; SSE42-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 1, i64 7, i64 15, i64 31>)
-; SSE42-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 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 48 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'constant_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'constant_funnel_i64'
@@ -713,16 +713,16 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; SLM-LABEL: 'constant_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SLM-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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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> <i64 1, i64 7>)
+; SLM-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> <i64 1, i64 7, i64 15, i64 31>)
+; SLM-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> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; GLM-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 1, i64 7>)
-; GLM-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 1, i64 7, i64 15, i64 31>)
-; GLM-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 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 24 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 48 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i64'
@@ -742,30 +742,30 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 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 27 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 54 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 108 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 24 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 48 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 96 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'constant_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 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 29 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX2-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 4, i32 5, i32 6, i32 7>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'constant_funnel_i32'
@@ -777,16 +777,16 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; SLM-LABEL: 'constant_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i32'
@@ -806,37 +806,37 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 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 26 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 52 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 104 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 23 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 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>)
+; SSSE3-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>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_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 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 21 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 47 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 94 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 18 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 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 30 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX2-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 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>)
+; AVX2-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 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'constant_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 26 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX512F-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 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 23 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'constant_funnel_i16'
@@ -848,23 +848,23 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ;
 ; AVX512DQ-LABEL: 'constant_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 26 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX512DQ-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 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 23 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'constant_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i16'
@@ -884,37 +884,37 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 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 36 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 72 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 144 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 33 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 66 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 132 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; SSE42-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> <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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 28 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 25 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 57 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 114 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX2-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 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>)
+; AVX2-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 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'constant_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 34 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX512F-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 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 31 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'constant_funnel_i8'
@@ -926,23 +926,23 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ;
 ; AVX512DQ-LABEL: 'constant_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 34 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX512DQ-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 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 31 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'constant_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; SLM-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> <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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
-; GLM-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> <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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i8'
@@ -966,30 +966,30 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 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 18 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 36 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 72 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatconstant_funnel_i64'
@@ -1001,16 +1001,16 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; SLM-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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 20 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 24 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i64'
@@ -1030,30 +1030,30 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 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) {
 ; 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 11 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 22 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 44 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 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 9 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 18 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 36 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 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)
-; AVX1-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> <i32 5, i32 5, i32 5, i32 5>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 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>)
+; AVX1-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>)
+; AVX1-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX1-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> <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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
-; AVX2-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> <i32 5, i32 5, i32 5, i32 5>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatconstant_funnel_i32'
@@ -1065,16 +1065,16 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
-; SLM-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> <i32 5, i32 5, i32 5, i32 5>)
-; SLM-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SLM-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> <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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
-; GLM-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> <i32 5, i32 5, i32 5, i32 5>)
-; GLM-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; GLM-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> <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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i32'
@@ -1094,37 +1094,37 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 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) {
 ; 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 11 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 22 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 44 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 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 9 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 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>)
-; SSE42-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>)
+; 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 9 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 23 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 46 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 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 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; AVX2-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatconstant_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; AVX512F-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 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>)
+; AVX512F-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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatconstant_funnel_i16'
@@ -1136,23 +1136,23 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512DQ-LABEL: 'splatconstant_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; AVX512DQ-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 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>)
+; AVX512DQ-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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; SLM-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
-; GLM-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; GLM-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>)
-; GLM-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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i16'
@@ -1172,37 +1172,37 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 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) {
 ; 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 13 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 26 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 52 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 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 11 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 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>)
-; SSE42-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>)
+; 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 11 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 27 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 54 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 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 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatconstant_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX512F-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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatconstant_funnel_i8'
@@ -1214,23 +1214,23 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512DQ-LABEL: 'splatconstant_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX512DQ-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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; GLM-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>)
-; GLM-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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i8'

diff  --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index 5407fbc45628..9d4f529be062 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -22,30 +22,30 @@ target triple = "x86_64-apple-macosx10.8.0"
 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 25 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 50 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 100 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; SSE42-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> %c128)
-; SSE42-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> %c256)
-; SSE42-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> %c512)
+; SSE42-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> %c128)
+; SSE42-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> %c256)
+; SSE42-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> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'var_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; AVX1-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> %c256)
-; AVX1-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> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'var_funnel_i64'
@@ -57,16 +57,16 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ;
 ; SLM-LABEL: 'var_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; SLM-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)
-; SLM-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)
-; SLM-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)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
-; GLM-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> %c128)
-; GLM-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> %c256)
-; GLM-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> %c512)
+; GLM-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> %c128)
+; GLM-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> %c256)
+; GLM-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> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i64'
@@ -86,30 +86,30 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 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 38 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 76 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 152 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 35 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 70 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 140 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
-; SSE42-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> %c128)
-; SSE42-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> %c256)
-; SSE42-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> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'var_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 %c32)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'var_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
-; AVX2-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> %c256)
-; AVX2-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> %c512)
+; AVX2-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> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'var_funnel_i32'
@@ -121,16 +121,16 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ;
 ; SLM-LABEL: 'var_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
-; SLM-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> %c128)
-; SLM-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> %c256)
-; SLM-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> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
-; GLM-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> %c128)
-; GLM-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> %c256)
-; GLM-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> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 100 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i32'
@@ -150,37 +150,37 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 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 51 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 102 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 204 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 48 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 96 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 192 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 27 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 57 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 114 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 24 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 54 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 108 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX2-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> %c256)
+; AVX2-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> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'var_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512F-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> %c256)
+; AVX512F-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> %c512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'var_funnel_i16'
@@ -192,23 +192,23 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ;
 ; AVX512DQ-LABEL: 'var_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512DQ-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> %c256)
+; AVX512DQ-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> %c512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'var_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i16'
@@ -228,37 +228,37 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 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 37 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 74 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 148 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 34 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 68 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 136 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 29 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 26 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 58 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 116 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX2-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> %c256)
-; AVX2-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> %c512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'var_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX512F-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> %c256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512F-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> %c512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'var_funnel_i8'
@@ -270,23 +270,23 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ;
 ; AVX512DQ-LABEL: 'var_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; AVX512DQ-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> %c256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512DQ-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> %c512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'var_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'var_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'var_funnel_i8'
@@ -312,36 +312,36 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; 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 19 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 38 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 76 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'
 ; SSE42-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
 ; SSE42-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
 ; SSE42-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
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i64'
 ; AVX1-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
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i64'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatvar_funnel_i64'
@@ -357,18 +357,18 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-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)
-; SLM-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)
-; SLM-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)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i64'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i64'
@@ -394,36 +394,36 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; 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 12 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 24 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 48 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 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 10 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 20 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 40 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 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
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX1-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)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i32'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX2-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)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatvar_funnel_i32'
@@ -439,18 +439,18 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; SLM-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)
+; SLM-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)
+; SLM-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)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i32'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; GLM-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)
+; GLM-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)
+; GLM-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)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i32'
@@ -476,45 +476,45 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; 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 12 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 24 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 48 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 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 10 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 20 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 40 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 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 10 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 24 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 48 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 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 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'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX2-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)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i16'
 ; AVX512F-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
 ; AVX512F-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
 ; AVX512F-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
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512F-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)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i16'
@@ -530,27 +530,27 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; AVX512DQ-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
 ; AVX512DQ-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
 ; AVX512DQ-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
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512DQ-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)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i16'
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; SLM-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> %u256)
-; SLM-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> %u512)
+; SLM-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)
+; SLM-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)
+; SLM-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)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i16'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
-; GLM-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> %u256)
-; GLM-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> %u512)
+; GLM-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)
+; GLM-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)
+; GLM-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)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i16'
@@ -576,45 +576,45 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; 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 37 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 74 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 148 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 34 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 68 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 136 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'
 ; SSE42-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
 ; SSE42-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
 ; SSE42-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
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i8'
 ; AVX1-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
 ; 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 29 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 26 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 58 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 116 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'
 ; AVX2-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
 ; AVX2-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
 ; AVX2-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
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX2-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> %u256)
-; AVX2-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> %u512)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i8'
 ; AVX512F-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
 ; AVX512F-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
 ; AVX512F-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
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX512F-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> %u256)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512F-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> %u512)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i8'
@@ -630,27 +630,27 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; AVX512DQ-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
 ; AVX512DQ-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
 ; AVX512DQ-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
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; AVX512DQ-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> %u256)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512DQ-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> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i8'
 ; SLM-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
 ; SLM-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
 ; SLM-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
-; SLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i8'
 ; GLM-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
 ; GLM-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
 ; GLM-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
-; GLM-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 116 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 104 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i8'
@@ -678,30 +678,30 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 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 24 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 48 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 96 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SSE42-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 1, i64 7>)
-; SSE42-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 1, i64 7, i64 15, i64 31>)
-; SSE42-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 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 48 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'constant_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'constant_funnel_i64'
@@ -713,16 +713,16 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; SLM-LABEL: 'constant_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SLM-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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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> <i64 1, i64 7>)
+; SLM-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> <i64 1, i64 7, i64 15, i64 31>)
+; SLM-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> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; GLM-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 1, i64 7>)
-; GLM-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 1, i64 7, i64 15, i64 31>)
-; GLM-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 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 24 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 48 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i64'
@@ -742,30 +742,30 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 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 27 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 54 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 108 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 24 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 48 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 96 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'constant_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 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 32 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 64 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 29 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 58 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX2-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 4, i32 5, i32 6, i32 7>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'constant_funnel_i32'
@@ -777,16 +777,16 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; SLM-LABEL: 'constant_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 44 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 88 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i32'
@@ -806,37 +806,37 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 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 26 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 52 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 104 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 23 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 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>)
+; SSSE3-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>)
 ; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SSE42-LABEL: 'constant_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 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 21 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 47 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 94 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 18 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 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'constant_funnel_i16'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 30 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX2-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 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>)
+; AVX2-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 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'constant_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 26 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX512F-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 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 23 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'constant_funnel_i16'
@@ -848,23 +848,23 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ;
 ; AVX512DQ-LABEL: 'constant_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 13 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 26 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; AVX512DQ-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 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 23 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'constant_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 42 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 84 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 72 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i16'
@@ -884,37 +884,37 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 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 36 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 72 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 144 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 33 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 66 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 132 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; SSE42-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> <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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; 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 28 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 25 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 57 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 114 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 38 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX2-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 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>)
+; AVX2-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 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'constant_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 34 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX512F-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 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 31 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'constant_funnel_i8'
@@ -926,23 +926,23 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ;
 ; AVX512DQ-LABEL: 'constant_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 19 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 34 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 16 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>)
+; AVX512DQ-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 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 31 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'constant_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; SLM-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> <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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'constant_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
-; GLM-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> <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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 56 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 112 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 25 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 50 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 100 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'constant_funnel_i8'
@@ -966,30 +966,30 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 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 18 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 36 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 72 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'
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 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>)
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; SSE42-NEXT:  Cost Model: Found an estimated cost of 24 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>)
 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX1-LABEL: 'splatconstant_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX1-NEXT:  Cost Model: Found an estimated cost of 36 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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatconstant_funnel_i64'
@@ -1001,16 +1001,16 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i64'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; SLM-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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 10 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 20 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>)
+; SLM-NEXT:  Cost Model: Found an estimated cost of 40 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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i64'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 18 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>)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 36 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 12 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>)
+; GLM-NEXT:  Cost Model: Found an estimated cost of 24 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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i64'
@@ -1030,30 +1030,30 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 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) {
 ; 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 11 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 22 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 44 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 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 9 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 18 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 36 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 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)
-; AVX1-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> <i32 5, i32 5, i32 5, i32 5>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 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>)
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 42 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>)
+; AVX1-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>)
+; AVX1-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX1-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> <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>)
 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX2-LABEL: 'splatconstant_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
-; AVX2-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> <i32 5, i32 5, i32 5, i32 5>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512-LABEL: 'splatconstant_funnel_i32'
@@ -1065,16 +1065,16 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i32'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
-; SLM-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> <i32 5, i32 5, i32 5, i32 5>)
-; SLM-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; SLM-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> <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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i32'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
-; GLM-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> <i32 5, i32 5, i32 5, i32 5>)
-; GLM-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> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; GLM-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> <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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i32'
@@ -1094,37 +1094,37 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 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) {
 ; 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 11 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 22 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 44 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 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 9 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 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>)
-; SSE42-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>)
+; 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 9 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 23 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 46 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 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 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; AVX2-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 12 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatconstant_funnel_i16'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; AVX512F-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 14 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>)
+; AVX512F-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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatconstant_funnel_i16'
@@ -1136,23 +1136,23 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512DQ-LABEL: 'splatconstant_funnel_i16'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; AVX512DQ-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 9 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 14 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>)
+; AVX512DQ-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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 6 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; SLM-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i16'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
-; GLM-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> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; GLM-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>)
-; GLM-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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i16'
@@ -1172,37 +1172,37 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 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) {
 ; 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 13 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 26 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 52 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 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 11 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 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>)
-; SSE42-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>)
+; 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 11 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 27 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 54 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 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 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'
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 22 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>)
+; AVX2-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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX2-NEXT:  Cost Model: Found an estimated cost of 16 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>)
 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512F-LABEL: 'splatconstant_funnel_i8'
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512F-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX512F-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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX512F-NEXT:  Cost Model: Found an estimated cost of 15 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>)
 ; AVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512BW-LABEL: 'splatconstant_funnel_i8'
@@ -1214,23 +1214,23 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512DQ-LABEL: 'splatconstant_funnel_i8'
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 18 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>)
+; AVX512DQ-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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 8 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>)
+; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 15 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>)
 ; AVX512DQ-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_funnel_i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; SLM-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; SLM-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>)
-; SLM-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>)
+; SLM-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>)
+; SLM-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>)
+; SLM-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>)
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; GLM-LABEL: 'splatconstant_funnel_i8'
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
-; GLM-NEXT:  Cost Model: Found an estimated cost of 11 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>)
-; GLM-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>)
-; GLM-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>)
+; GLM-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>)
+; GLM-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>)
+; GLM-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>)
 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; XOP-LABEL: 'splatconstant_funnel_i8'

diff  --git a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
index 46ee566eb773..25a0e3830436 100644
--- a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
@@ -256,7 +256,7 @@ define void @ctlz(i32 %a, <16 x i32> %va) {
 define void @fshl(i32 %a, i32 %b, i32 %c, <16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc) {
 ; THRU-LABEL: 'fshl'
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; THRU-NEXT:  Cost Model: Found an estimated cost of 152 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; LATE-LABEL: 'fshl'


        
