; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+ssse3 | FileCheck %s --check-prefixes=SSE,SSSE3
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE,SSE42
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx | FileCheck %s --check-prefixes=AVX1
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512,AVX512F
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512,AVX512BW
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=AVX512,AVX512DQ
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512vbmi2 | FileCheck %s --check-prefixes=AVX512,AVX512VBMI2
;
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mcpu=slm | FileCheck %s --check-prefixes=SLM
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI

target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0"

;
; Variable Funnel Shifts
;

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:21 Lat:22 SizeLat:26 for: %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 costs of RThru:36 CodeSize:42 Lat:44 SizeLat:52 for: %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 costs of RThru:72 CodeSize:84 Lat:88 SizeLat:104 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_funnel_i64'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; SSE42-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:20 SizeLat:22 for: %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 costs of RThru:28 CodeSize:30 Lat:40 SizeLat:44 for: %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 costs of RThru:56 CodeSize:60 Lat:80 SizeLat:88 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_funnel_i64'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX1-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:13 Lat:15 SizeLat:19 for: %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 costs of RThru:25 CodeSize:35 Lat:23 SizeLat:48 for: %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 costs of RThru:50 CodeSize:70 Lat:46 SizeLat:96 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_funnel_i64'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:13 SizeLat:9 for: %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 costs of RThru:14 CodeSize:7 Lat:14 SizeLat:14 for: %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 costs of RThru:28 CodeSize:14 Lat:28 SizeLat:28 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_funnel_i64'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:7 SizeLat:8 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_funnel_i64'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_funnel_i64'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX512DQ-NEXT:  Cost Model: Found costs of 7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:7 SizeLat:8 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; AVX512DQ-NEXT:  Cost Model: Found costs of 7 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_funnel_i64'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_funnel_i64'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; SLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:15 Lat:25 SizeLat:23 for: %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 costs of RThru:40 CodeSize:30 Lat:50 SizeLat:46 for: %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 costs of RThru:80 CodeSize:60 Lat:100 SizeLat:92 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_funnel_i64'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:20 SizeLat:22 for: %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 costs of RThru:28 CodeSize:30 Lat:40 SizeLat:44 for: %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 costs of RThru:56 CodeSize:60 Lat:80 SizeLat:88 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_funnel_i64'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:23 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:46 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_funnel_i64'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
  ret void
}

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:32 Lat:37 SizeLat:38 for: %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 costs of RThru:70 CodeSize:63 Lat:73 SizeLat:75 for: %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 costs of RThru:140 CodeSize:125 Lat:145 SizeLat:149 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_funnel_i32'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:43 SizeLat:35 for: %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 costs of RThru:72 CodeSize:47 Lat:85 SizeLat:69 for: %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 costs of RThru:144 CodeSize:93 Lat:169 SizeLat:137 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_funnel_i32'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX1-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:21 Lat:24 SizeLat:28 for: %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 costs of RThru:36 CodeSize:51 Lat:34 SizeLat:69 for: %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 costs of RThru:72 CodeSize:102 Lat:68 SizeLat:138 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_funnel_i32'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:12 SizeLat:12 for: %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 costs of RThru:14 CodeSize:7 Lat:14 SizeLat:16 for: %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 costs of RThru:28 CodeSize:14 Lat:28 SizeLat:32 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_funnel_i32'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:7 SizeLat:8 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_funnel_i32'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_funnel_i32'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX512DQ-NEXT:  Cost Model: Found costs of 7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:7 SizeLat:8 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; AVX512DQ-NEXT:  Cost Model: Found costs of 7 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_funnel_i32'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_funnel_i32'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:43 SizeLat:35 for: %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 costs of RThru:72 CodeSize:47 Lat:85 SizeLat:69 for: %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 costs of RThru:144 CodeSize:93 Lat:169 SizeLat:137 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_funnel_i32'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; GLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:43 SizeLat:35 for: %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 costs of RThru:72 CodeSize:47 Lat:85 SizeLat:69 for: %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 costs of RThru:144 CodeSize:93 Lat:169 SizeLat:137 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_funnel_i32'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:23 Lat:23 SizeLat:30 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:46 Lat:46 SizeLat:60 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_funnel_i32'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:54 Lat:52 SizeLat:58 for: %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 costs of RThru:92 CodeSize:107 Lat:103 SizeLat:115 for: %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 costs of RThru:184 CodeSize:213 Lat:205 SizeLat:229 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_funnel_i16'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; SSE42-NEXT:  Cost Model: Found costs of RThru:39 CodeSize:39 Lat:46 SizeLat:44 for: %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 costs of RThru:78 CodeSize:77 Lat:91 SizeLat:87 for: %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 costs of RThru:156 CodeSize:153 Lat:181 SizeLat:173 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_funnel_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX1-NEXT:  Cost Model: Found costs of RThru:25 CodeSize:30 Lat:31 SizeLat:39 for: %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 costs of RThru:54 CodeSize:70 Lat:55 SizeLat:92 for: %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 costs of RThru:108 CodeSize:140 Lat:110 SizeLat:184 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_funnel_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX2-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:15 Lat:35 SizeLat:26 for: %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 costs of RThru:22 CodeSize:25 Lat:26 SizeLat:38 for: %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 costs of RThru:44 CodeSize:50 Lat:52 SizeLat:76 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_funnel_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:15 Lat:34 SizeLat:25 for: %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 costs of RThru:21 CodeSize:25 Lat:25 SizeLat:35 for: %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 costs of RThru:32 CodeSize:38 Lat:50 SizeLat:46 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_funnel_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_funnel_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:15 Lat:34 SizeLat:25 for: %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 costs of RThru:21 CodeSize:25 Lat:25 SizeLat:35 for: %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 costs of RThru:32 CodeSize:38 Lat:50 SizeLat:46 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_funnel_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_funnel_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; SLM-NEXT:  Cost Model: Found costs of RThru:41 CodeSize:39 Lat:48 SizeLat:45 for: %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 costs of RThru:82 CodeSize:77 Lat:95 SizeLat:89 for: %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 costs of RThru:164 CodeSize:153 Lat:189 SizeLat:177 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_funnel_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; GLM-NEXT:  Cost Model: Found costs of RThru:39 CodeSize:39 Lat:46 SizeLat:44 for: %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 costs of RThru:78 CodeSize:77 Lat:91 SizeLat:87 for: %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 costs of RThru:156 CodeSize:153 Lat:181 SizeLat:173 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_funnel_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:23 SizeLat:31 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:50 Lat:46 SizeLat:62 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_funnel_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:60 Lat:55 SizeLat:65 for: %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 costs of RThru:66 CodeSize:119 Lat:109 SizeLat:129 for: %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 costs of RThru:132 CodeSize:237 Lat:217 SizeLat:257 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_funnel_i8'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; SSE42-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:40 Lat:57 SizeLat:52 for: %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 costs of RThru:74 CodeSize:79 Lat:113 SizeLat:103 for: %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 costs of RThru:148 CodeSize:157 Lat:225 SizeLat:205 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_funnel_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX1-NEXT:  Cost Model: Found costs of RThru:27 CodeSize:28 Lat:54 SizeLat:41 for: %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 costs of RThru:58 CodeSize:72 Lat:54 SizeLat:102 for: %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 costs of RThru:116 CodeSize:144 Lat:108 SizeLat:204 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_funnel_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:28 Lat:54 SizeLat:40 for: %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 costs of RThru:20 CodeSize:28 Lat:59 SizeLat:56 for: %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 costs of RThru:40 CodeSize:56 Lat:118 SizeLat:112 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_funnel_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:28 Lat:53 SizeLat:39 for: %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 costs of RThru:19 CodeSize:28 Lat:58 SizeLat:53 for: %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 costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_funnel_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:21 SizeLat:15 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:28 Lat:58 SizeLat:39 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:33 Lat:51 SizeLat:38 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_funnel_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:28 Lat:53 SizeLat:39 for: %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 costs of RThru:19 CodeSize:28 Lat:58 SizeLat:53 for: %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 costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_funnel_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:21 SizeLat:15 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:28 Lat:58 SizeLat:39 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:33 Lat:51 SizeLat:38 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_funnel_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; SLM-NEXT:  Cost Model: Found costs of RThru:39 CodeSize:40 Lat:59 SizeLat:53 for: %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 costs of RThru:78 CodeSize:79 Lat:117 SizeLat:105 for: %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 costs of RThru:156 CodeSize:157 Lat:233 SizeLat:209 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_funnel_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; GLM-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:40 Lat:57 SizeLat:52 for: %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 costs of RThru:74 CodeSize:79 Lat:113 SizeLat:103 for: %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 costs of RThru:148 CodeSize:157 Lat:225 SizeLat:205 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_funnel_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:23 SizeLat:31 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:50 Lat:46 SizeLat:62 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_funnel_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:21 SizeLat:15 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:28 Lat:58 SizeLat:39 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:33 Lat:51 SizeLat:38 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
  ret void
}

;
; Uniform Variable Funnel Shifts
;

define void @splatvar_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: 'splatvar_funnel_i64'
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:13 Lat:14 SizeLat:16 for: %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 costs of RThru:28 CodeSize:26 Lat:28 SizeLat:32 for: %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 costs of RThru:56 CodeSize:52 Lat:56 SizeLat:64 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatvar_funnel_i64'
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:12 SizeLat:12 for: %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 costs of RThru:20 CodeSize:14 Lat:24 SizeLat:24 for: %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 costs of RThru:40 CodeSize:28 Lat:48 SizeLat:48 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_funnel_i64'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:13 SizeLat:11 for: %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 costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %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 costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_funnel_i64'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:11 SizeLat:11 for: %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 costs of RThru:10 CodeSize:7 Lat:14 SizeLat:14 for: %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 costs of RThru:20 CodeSize:14 Lat:28 SizeLat:28 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_funnel_i64'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_funnel_i64'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_funnel_i64'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_funnel_i64'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_funnel_i64'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:7 Lat:17 SizeLat:13 for: %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 costs of RThru:32 CodeSize:14 Lat:34 SizeLat:26 for: %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 costs of RThru:64 CodeSize:28 Lat:68 SizeLat:52 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_funnel_i64'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:12 SizeLat:12 for: %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 costs of RThru:20 CodeSize:14 Lat:24 SizeLat:24 for: %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 costs of RThru:40 CodeSize:28 Lat:48 SizeLat:48 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_funnel_i64'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
  %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
  %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
  ret void
}

define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %b32, <4 x i32> %b128, <8 x i32> %b256, <16 x i32> %b512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
; SSSE3-LABEL: 'splatvar_funnel_i32'
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:10 SizeLat:11 for: %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 costs of RThru:20 CodeSize:17 Lat:19 SizeLat:21 for: %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 costs of RThru:40 CodeSize:33 Lat:37 SizeLat:41 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatvar_funnel_i32'
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %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 costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_funnel_i32'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %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 costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %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 costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_funnel_i32'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %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 costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_funnel_i32'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_funnel_i32'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_funnel_i32'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_funnel_i32'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_funnel_i32'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %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 costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_funnel_i32'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %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 costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_funnel_i32'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
  %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
  %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
  ret void
}

define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
; SSSE3-LABEL: 'splatvar_funnel_i16'
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:10 SizeLat:11 for: %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 costs of RThru:20 CodeSize:17 Lat:19 SizeLat:21 for: %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 costs of RThru:40 CodeSize:33 Lat:37 SizeLat:41 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatvar_funnel_i16'
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %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 costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_funnel_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %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 costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %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 costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_funnel_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %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 costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_funnel_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %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 costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %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 costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_funnel_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_funnel_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %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 costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %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 costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_funnel_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_funnel_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:7 Lat:12 SizeLat:11 for: %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 costs of RThru:24 CodeSize:13 Lat:23 SizeLat:21 for: %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 costs of RThru:48 CodeSize:25 Lat:45 SizeLat:41 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_funnel_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %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 costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %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 costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_funnel_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
  %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
  %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
  ret void
}

define void @splatvar_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: 'splatvar_funnel_i8'
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; SSSE3-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:18 Lat:29 SizeLat:25 for: %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 costs of RThru:48 CodeSize:35 Lat:57 SizeLat:49 for: %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 costs of RThru:96 CodeSize:69 Lat:113 SizeLat:97 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatvar_funnel_i8'
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; SSE42-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %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 costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %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 costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_funnel_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:16 Lat:18 SizeLat:22 for: %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 costs of RThru:27 CodeSize:36 Lat:26 SizeLat:47 for: %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 costs of RThru:54 CodeSize:72 Lat:52 SizeLat:94 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_funnel_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:20 SizeLat:21 for: %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 costs of RThru:14 CodeSize:18 Lat:21 SizeLat:27 for: %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 costs of RThru:28 CodeSize:36 Lat:42 SizeLat:54 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_funnel_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %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 costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %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 costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_funnel_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_funnel_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %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 costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %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 costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_funnel_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_funnel_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:16 Lat:31 SizeLat:25 for: %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 costs of RThru:52 CodeSize:31 Lat:61 SizeLat:49 for: %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 costs of RThru:104 CodeSize:61 Lat:121 SizeLat:97 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_funnel_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %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 costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %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 costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_funnel_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:23 SizeLat:31 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:50 Lat:46 SizeLat:62 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
  %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
  %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
  ret void
}

;
; Constant Funnel Shifts
;

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:20 Lat:21 SizeLat:25 for: %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 costs of RThru:34 CodeSize:40 Lat:42 SizeLat:50 for: %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 costs of RThru:68 CodeSize:80 Lat:84 SizeLat:100 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_funnel_i64'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:14 Lat:19 SizeLat:21 for: %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 costs of RThru:26 CodeSize:28 Lat:38 SizeLat:42 for: %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 costs of RThru:52 CodeSize:56 Lat:76 SizeLat:84 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_funnel_i64'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:12 Lat:14 SizeLat:18 for: %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 costs of RThru:24 CodeSize:34 Lat:22 SizeLat:46 for: %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 costs of RThru:48 CodeSize:68 Lat:44 SizeLat:92 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_funnel_i64'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:12 SizeLat:8 for: %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 costs of RThru:13 CodeSize:6 Lat:13 SizeLat:12 for: %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 costs of RThru:26 CodeSize:12 Lat:26 SizeLat:24 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_funnel_i64'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %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>)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_funnel_i64'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_funnel_i64'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %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>)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_funnel_i64'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_funnel_i64'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:14 Lat:24 SizeLat:22 for: %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 costs of RThru:38 CodeSize:28 Lat:48 SizeLat:44 for: %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 costs of RThru:76 CodeSize:56 Lat:96 SizeLat:88 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_funnel_i64'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:14 Lat:19 SizeLat:21 for: %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 costs of RThru:26 CodeSize:28 Lat:38 SizeLat:42 for: %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 costs of RThru:52 CodeSize:56 Lat:76 SizeLat:84 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_funnel_i64'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:11 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; XOP-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:22 Lat:22 SizeLat:28 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:44 Lat:44 SizeLat:56 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_funnel_i64'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:28 Lat:25 SizeLat:32 for: %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 costs of RThru:46 CodeSize:55 Lat:49 SizeLat:63 for: %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 costs of RThru:92 CodeSize:109 Lat:97 SizeLat:125 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_funnel_i32'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:20 Lat:33 SizeLat:25 for: %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 costs of RThru:46 CodeSize:39 Lat:65 SizeLat:49 for: %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 costs of RThru:92 CodeSize:77 Lat:129 SizeLat:97 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_funnel_i32'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:17 Lat:17 SizeLat:24 for: %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 costs of RThru:31 CodeSize:43 Lat:30 SizeLat:60 for: %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 costs of RThru:62 CodeSize:86 Lat:60 SizeLat:120 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_funnel_i32'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:11 SizeLat:11 for: %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 costs of RThru:13 CodeSize:6 Lat:13 SizeLat:14 for: %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 costs of RThru:26 CodeSize:12 Lat:26 SizeLat:28 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_funnel_i32'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %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>)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_funnel_i32'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_funnel_i32'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %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>)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_funnel_i32'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_funnel_i32'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:32 CodeSize:20 Lat:33 SizeLat:31 for: %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 costs of RThru:64 CodeSize:39 Lat:65 SizeLat:61 for: %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 costs of RThru:128 CodeSize:77 Lat:129 SizeLat:121 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_funnel_i32'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:20 Lat:33 SizeLat:25 for: %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 costs of RThru:46 CodeSize:39 Lat:65 SizeLat:49 for: %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 costs of RThru:92 CodeSize:77 Lat:129 SizeLat:97 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_funnel_i32'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:11 SizeLat:7 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:22 Lat:22 SizeLat:28 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:44 Lat:44 SizeLat:56 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_funnel_i32'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
  %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>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:38 Lat:29 SizeLat:38 for: %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 costs of RThru:44 CodeSize:75 Lat:57 SizeLat:75 for: %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 costs of RThru:88 CodeSize:149 Lat:113 SizeLat:149 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_funnel_i16'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:28 Lat:36 SizeLat:33 for: %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 costs of RThru:56 CodeSize:55 Lat:71 SizeLat:65 for: %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 costs of RThru:112 CodeSize:109 Lat:141 SizeLat:129 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_funnel_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:19 Lat:26 SizeLat:28 for: %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 costs of RThru:44 CodeSize:50 Lat:46 SizeLat:71 for: %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 costs of RThru:88 CodeSize:100 Lat:92 SizeLat:142 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_funnel_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:10 Lat:21 SizeLat:16 for: %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 costs of RThru:15 CodeSize:15 Lat:20 SizeLat:24 for: %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 costs of RThru:30 CodeSize:30 Lat:40 SizeLat:48 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_funnel_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:10 Lat:20 SizeLat:15 for: %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 costs of RThru:14 CodeSize:15 Lat:19 SizeLat:22 for: %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 costs of RThru:31 CodeSize:37 Lat:49 SizeLat:45 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_funnel_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_funnel_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:10 Lat:20 SizeLat:15 for: %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 costs of RThru:14 CodeSize:15 Lat:19 SizeLat:22 for: %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 costs of RThru:31 CodeSize:37 Lat:49 SizeLat:45 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_funnel_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_funnel_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:31 CodeSize:28 Lat:38 SizeLat:34 for: %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 costs of RThru:62 CodeSize:55 Lat:75 SizeLat:67 for: %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 costs of RThru:124 CodeSize:109 Lat:149 SizeLat:133 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_funnel_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:28 Lat:36 SizeLat:33 for: %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 costs of RThru:56 CodeSize:55 Lat:71 SizeLat:65 for: %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 costs of RThru:112 CodeSize:109 Lat:141 SizeLat:129 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_funnel_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:11 SizeLat:7 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:22 SizeLat:29 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:48 Lat:44 SizeLat:58 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_funnel_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
  %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>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:32 CodeSize:59 Lat:54 SizeLat:64 for: %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 costs of RThru:64 CodeSize:117 Lat:107 SizeLat:127 for: %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 costs of RThru:128 CodeSize:233 Lat:213 SizeLat:253 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_funnel_i8'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:39 Lat:56 SizeLat:51 for: %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 costs of RThru:72 CodeSize:77 Lat:111 SizeLat:101 for: %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 costs of RThru:144 CodeSize:153 Lat:221 SizeLat:201 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_funnel_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:27 Lat:53 SizeLat:40 for: %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 costs of RThru:57 CodeSize:71 Lat:53 SizeLat:100 for: %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 costs of RThru:114 CodeSize:142 Lat:106 SizeLat:200 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_funnel_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:27 Lat:53 SizeLat:39 for: %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 costs of RThru:19 CodeSize:27 Lat:58 SizeLat:54 for: %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 costs of RThru:38 CodeSize:54 Lat:116 SizeLat:108 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_funnel_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:27 Lat:52 SizeLat:38 for: %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 costs of RThru:18 CodeSize:27 Lat:57 SizeLat:52 for: %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 costs of RThru:39 CodeSize:72 Lat:55 SizeLat:84 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_funnel_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:12 Lat:20 SizeLat:14 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:27 Lat:57 SizeLat:38 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:32 Lat:50 SizeLat:37 for: %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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_funnel_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:27 Lat:52 SizeLat:38 for: %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 costs of RThru:18 CodeSize:27 Lat:57 SizeLat:52 for: %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 costs of RThru:39 CodeSize:72 Lat:55 SizeLat:84 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_funnel_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:12 Lat:20 SizeLat:14 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:27 Lat:57 SizeLat:38 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:32 Lat:50 SizeLat:37 for: %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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_funnel_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:39 Lat:58 SizeLat:52 for: %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 costs of RThru:76 CodeSize:77 Lat:115 SizeLat:103 for: %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 costs of RThru:152 CodeSize:153 Lat:229 SizeLat:205 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_funnel_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:39 Lat:56 SizeLat:51 for: %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 costs of RThru:72 CodeSize:77 Lat:111 SizeLat:101 for: %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 costs of RThru:144 CodeSize:153 Lat:221 SizeLat:201 for: %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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_funnel_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:11 SizeLat:7 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:22 SizeLat:29 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:48 Lat:44 SizeLat:58 for: %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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_funnel_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:12 Lat:20 SizeLat:14 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:27 Lat:57 SizeLat:38 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:32 Lat:50 SizeLat:37 for: %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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
  %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>)
  %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>)
  %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>)
  ret void
}

;
; Uniform Constant Funnel Shifts
;

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:12 Lat:11 SizeLat:13 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:24 Lat:22 SizeLat:26 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:48 Lat:44 SizeLat:52 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatconstant_funnel_i64'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; SSE42-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:12 Lat:18 SizeLat:18 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; SSE42-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:24 Lat:36 SizeLat:36 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_funnel_i64'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:10 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:20 Lat:20 SizeLat:26 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:40 Lat:40 SizeLat:52 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_funnel_i64'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:8 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:9 SizeLat:12 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:18 SizeLat:24 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_funnel_i64'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_funnel_i64'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_funnel_i64'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_funnel_i64'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_funnel_i64'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:6 Lat:14 SizeLat:10 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:12 Lat:28 SizeLat:20 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:52 CodeSize:24 Lat:56 SizeLat:40 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_funnel_i64'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:12 Lat:18 SizeLat:18 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:24 Lat:36 SizeLat:36 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_funnel_i64'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:20 Lat:20 SizeLat:26 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:40 Lat:40 SizeLat:52 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> splat (i64 7))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> splat (i64 7))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> splat (i64 7))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:8 Lat:7 SizeLat:8 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:13 SizeLat:15 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:29 Lat:25 SizeLat:29 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatconstant_funnel_i32'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; SSE42-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; SSE42-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:11 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; SSE42-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:21 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_funnel_i32'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:20 Lat:20 SizeLat:26 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:40 Lat:40 SizeLat:52 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_funnel_i32'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:9 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:18 SizeLat:24 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_funnel_i32'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_funnel_i32'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_funnel_i32'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:6 SizeLat:7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_funnel_i32'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_funnel_i32'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:11 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:21 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_funnel_i32'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:11 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:21 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_funnel_i32'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:20 Lat:20 SizeLat:26 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:40 Lat:40 SizeLat:52 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:4 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> splat (i32 5))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> splat (i32 5))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> splat (i32 5))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
  %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>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:8 Lat:7 SizeLat:8 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:13 SizeLat:15 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:29 Lat:25 SizeLat:29 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatconstant_funnel_i16'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; SSE42-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:11 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:21 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_funnel_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:22 Lat:20 SizeLat:27 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:44 Lat:40 SizeLat:54 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_funnel_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:9 SizeLat:12 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:18 SizeLat:24 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_funnel_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX512F-NEXT:  Cost Model: Found costs of 6 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:18 SizeLat:14 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:37 SizeLat:29 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_funnel_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of 6 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_funnel_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX512DQ-NEXT:  Cost Model: Found costs of 6 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:18 SizeLat:14 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:37 SizeLat:29 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_funnel_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_funnel_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; SLM-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:9 SizeLat:8 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:11 Lat:17 SizeLat:15 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:21 Lat:33 SizeLat:29 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_funnel_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:11 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:21 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_funnel_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:9 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:22 Lat:20 SizeLat:27 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:44 Lat:40 SizeLat:54 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
  %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>)
  %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>)
  %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>)
  ret void
}

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 costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:10 Lat:19 SizeLat:12 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:19 Lat:37 SizeLat:23 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:37 Lat:73 SizeLat:45 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'splatconstant_funnel_i8'
; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; SSE42-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:8 Lat:19 SizeLat:11 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:37 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:29 Lat:73 SizeLat:41 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_funnel_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX1-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:8 Lat:19 SizeLat:11 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:28 Lat:22 SizeLat:33 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:56 Lat:44 SizeLat:66 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_funnel_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:8 Lat:21 SizeLat:11 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:8 Lat:21 SizeLat:16 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:16 Lat:42 SizeLat:32 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_funnel_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:20 SizeLat:14 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:25 Lat:41 SizeLat:27 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_funnel_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:18 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_funnel_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:20 SizeLat:14 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:25 Lat:41 SizeLat:27 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_funnel_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:18 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:20 SizeLat:10 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_funnel_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; SLM-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:8 Lat:21 SizeLat:12 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:15 Lat:41 SizeLat:23 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:29 Lat:81 SizeLat:45 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_funnel_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:8 Lat:19 SizeLat:11 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:15 Lat:37 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:29 Lat:73 SizeLat:41 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_funnel_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; XOP-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:6 Lat:11 SizeLat:7 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:22 SizeLat:29 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:48 Lat:44 SizeLat:58 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:5 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:16 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:16 SizeLat:8 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:16 SizeLat:8 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
  %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>)
  %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>)
  %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>)
  ret void
}

;
; Variable Unary Funnel Shifts (Rotates)
;

define void @var_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %c64, <2 x i64> %c128, <4 x i64> %c256, <8 x i64> %c512) {
; SSE-LABEL: 'var_rotate_i64'
; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; SSE-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:13 Lat:16 SizeLat:18 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; SSE-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:26 Lat:32 SizeLat:36 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; SSE-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:52 Lat:64 SizeLat:72 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_rotate_i64'
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:11 Lat:11 SizeLat:15 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:29 Lat:18 SizeLat:40 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:58 Lat:36 SizeLat:80 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_rotate_i64'
; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:9 SizeLat:5 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:5 Lat:11 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:10 Lat:22 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'var_rotate_i64'
; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_rotate_i64'
; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:13 Lat:16 SizeLat:18 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:26 Lat:32 SizeLat:36 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; SLM-NEXT:  Cost Model: Found costs of RThru:56 CodeSize:52 Lat:64 SizeLat:72 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_rotate_i64'
; GLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; GLM-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:13 Lat:16 SizeLat:18 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; GLM-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:26 Lat:32 SizeLat:36 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; GLM-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:52 Lat:64 SizeLat:72 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_rotate_i64'
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
  ret void
}

define void @var_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
; SSSE3-LABEL: 'var_rotate_i32'
; SSSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:32 CodeSize:28 Lat:34 SizeLat:34 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:64 CodeSize:55 Lat:67 SizeLat:67 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:128 CodeSize:109 Lat:133 SizeLat:133 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_rotate_i32'
; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; SSE42-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:22 Lat:40 SizeLat:32 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; SSE42-NEXT:  Cost Model: Found costs of RThru:66 CodeSize:43 Lat:79 SizeLat:63 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; SSE42-NEXT:  Cost Model: Found costs of RThru:132 CodeSize:85 Lat:157 SizeLat:125 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_rotate_i32'
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:19 Lat:21 SizeLat:25 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:29 CodeSize:45 Lat:29 SizeLat:61 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:58 CodeSize:90 Lat:58 SizeLat:122 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_rotate_i32'
; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:5 Lat:11 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:10 Lat:22 SizeLat:24 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'var_rotate_i32'
; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_rotate_i32'
; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; SLM-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:22 Lat:40 SizeLat:32 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; SLM-NEXT:  Cost Model: Found costs of RThru:66 CodeSize:43 Lat:79 SizeLat:63 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; SLM-NEXT:  Cost Model: Found costs of RThru:132 CodeSize:85 Lat:157 SizeLat:125 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_rotate_i32'
; GLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; GLM-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:22 Lat:40 SizeLat:32 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; GLM-NEXT:  Cost Model: Found costs of RThru:66 CodeSize:43 Lat:79 SizeLat:63 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; GLM-NEXT:  Cost Model: Found costs of RThru:132 CodeSize:85 Lat:157 SizeLat:125 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_rotate_i32'
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
  ret void
}

define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
; SSSE3-LABEL: 'var_rotate_i16'
; SSSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:43 CodeSize:50 Lat:49 SizeLat:54 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:86 CodeSize:99 Lat:97 SizeLat:107 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:172 CodeSize:197 Lat:193 SizeLat:213 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_rotate_i16'
; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:37 Lat:43 SizeLat:41 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; SSE42-NEXT:  Cost Model: Found costs of RThru:72 CodeSize:73 Lat:85 SizeLat:81 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; SSE42-NEXT:  Cost Model: Found costs of RThru:144 CodeSize:145 Lat:169 SizeLat:161 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_rotate_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX1-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:28 Lat:28 SizeLat:36 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:47 CodeSize:62 Lat:50 SizeLat:83 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:94 CodeSize:124 Lat:100 SizeLat:166 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_rotate_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:32 SizeLat:23 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:34 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:68 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_rotate_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:32 SizeLat:23 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:27 CodeSize:29 Lat:41 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_rotate_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:8 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_rotate_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:13 Lat:32 SizeLat:23 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:27 CodeSize:29 Lat:41 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_rotate_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_rotate_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:37 Lat:43 SizeLat:41 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; SLM-NEXT:  Cost Model: Found costs of RThru:72 CodeSize:73 Lat:85 SizeLat:81 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; SLM-NEXT:  Cost Model: Found costs of RThru:144 CodeSize:145 Lat:169 SizeLat:161 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_rotate_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; GLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:37 Lat:43 SizeLat:41 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; GLM-NEXT:  Cost Model: Found costs of RThru:72 CodeSize:73 Lat:85 SizeLat:81 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; GLM-NEXT:  Cost Model: Found costs of RThru:144 CodeSize:145 Lat:169 SizeLat:161 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_rotate_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_rotate_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
  ret void
}

define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
; SSSE3-LABEL: 'var_rotate_i8'
; SSSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:30 CodeSize:56 Lat:52 SizeLat:61 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:60 CodeSize:111 Lat:103 SizeLat:121 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:120 CodeSize:221 Lat:205 SizeLat:241 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'var_rotate_i8'
; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; SSE42-NEXT:  Cost Model: Found costs of RThru:34 CodeSize:38 Lat:54 SizeLat:49 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; SSE42-NEXT:  Cost Model: Found costs of RThru:68 CodeSize:75 Lat:107 SizeLat:97 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; SSE42-NEXT:  Cost Model: Found costs of RThru:136 CodeSize:149 Lat:213 SizeLat:193 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'var_rotate_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:26 Lat:51 SizeLat:38 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:51 CodeSize:64 Lat:49 SizeLat:93 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:102 CodeSize:128 Lat:98 SizeLat:186 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'var_rotate_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX2-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:26 Lat:51 SizeLat:37 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:26 Lat:56 SizeLat:52 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:34 CodeSize:52 Lat:112 SizeLat:104 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'var_rotate_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:26 Lat:51 SizeLat:37 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:26 Lat:56 SizeLat:50 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'var_rotate_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'var_rotate_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:26 Lat:51 SizeLat:37 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:26 Lat:56 SizeLat:50 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'var_rotate_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'var_rotate_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; SLM-NEXT:  Cost Model: Found costs of RThru:34 CodeSize:38 Lat:54 SizeLat:49 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; SLM-NEXT:  Cost Model: Found costs of RThru:68 CodeSize:75 Lat:107 SizeLat:97 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; SLM-NEXT:  Cost Model: Found costs of RThru:136 CodeSize:149 Lat:213 SizeLat:193 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'var_rotate_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; GLM-NEXT:  Cost Model: Found costs of RThru:34 CodeSize:38 Lat:54 SizeLat:49 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; GLM-NEXT:  Cost Model: Found costs of RThru:68 CodeSize:75 Lat:107 SizeLat:97 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; GLM-NEXT:  Cost Model: Found costs of RThru:136 CodeSize:149 Lat:213 SizeLat:193 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'var_rotate_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'var_rotate_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:3 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
  ret void
}

;
; Uniform Variable Unary Funnel Shifts (Rotates)
;

define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512, i64 %c64, <2 x i64> %c128, <4 x i64> %c256, <8 x i64> %c512) {
; SSE-LABEL: 'splatvar_rotate_i64'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:8 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; SSE-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:10 Lat:16 SizeLat:16 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; SSE-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:20 Lat:32 SizeLat:32 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_rotate_i64'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:18 SizeLat:22 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:30 Lat:36 SizeLat:44 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_rotate_i64'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:11 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:10 Lat:22 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'splatvar_rotate_i64'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_rotate_i64'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:5 Lat:8 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; SLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:10 Lat:16 SizeLat:16 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; SLM-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:20 Lat:32 SizeLat:32 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_rotate_i64'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:8 SizeLat:8 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:10 Lat:16 SizeLat:16 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:20 Lat:32 SizeLat:32 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_rotate_i64'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
  %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
  %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
  ret void
}

define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512, i32 %c32, <4 x i32> %c128, <8 x i32> %c256, <16 x i32> %c512) {
; SSE-LABEL: 'splatvar_rotate_i32'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; SSE-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; SSE-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_rotate_i32'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_rotate_i32'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'splatvar_rotate_i32'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_rotate_i32'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_rotate_i32'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_rotate_i32'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
  %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
  %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
  ret void
}

define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
; SSE-LABEL: 'splatvar_rotate_i16'
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; SSE-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; SSE-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_rotate_i16'
; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_rotate_i16'
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_rotate_i16'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_rotate_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:8 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_rotate_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_rotate_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_rotate_i16'
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_rotate_i16'
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_rotate_i16'
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
  %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
  %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
  ret void
}

define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
; SSE-LABEL: 'splatvar_rotate_i8'
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; SSE-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; SSE-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; SSE-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatvar_rotate_i8'
; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:19 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:28 Lat:21 SizeLat:38 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX1-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:56 Lat:42 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatvar_rotate_i8'
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:23 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:32 Lat:36 SizeLat:46 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatvar_rotate_i8'
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatvar_rotate_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatvar_rotate_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatvar_rotate_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatvar_rotate_i8'
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; SLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; SLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; SLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatvar_rotate_i8'
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; GLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; GLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; GLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatvar_rotate_i8'
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
  %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
  %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
  ret void
}

;
; Constant Unary Funnel Shifts (Rotates)
;

define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {
; SSE-LABEL: 'constant_rotate_i64'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; SSE-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:12 Lat:15 SizeLat:17 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; SSE-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:30 SizeLat:34 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; SSE-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:48 Lat:60 SizeLat:68 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_rotate_i64'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:10 SizeLat:14 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:28 Lat:17 SizeLat:38 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:34 CodeSize:56 Lat:34 SizeLat:76 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_rotate_i64'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:4 Lat:8 SizeLat:4 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:4 Lat:10 SizeLat:8 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:8 Lat:20 SizeLat:16 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'constant_rotate_i64'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_rotate_i64'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:12 Lat:15 SizeLat:17 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:24 Lat:30 SizeLat:34 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; SLM-NEXT:  Cost Model: Found costs of RThru:52 CodeSize:48 Lat:60 SizeLat:68 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_rotate_i64'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:12 Lat:15 SizeLat:17 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; GLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:30 SizeLat:34 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; GLM-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:48 Lat:60 SizeLat:68 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_rotate_i64'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
  ret void
}

define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {
; SSSE3-LABEL: 'constant_rotate_i32'
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:24 Lat:22 SizeLat:28 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:47 Lat:43 SizeLat:55 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:80 CodeSize:93 Lat:85 SizeLat:109 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_rotate_i32'
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:30 SizeLat:22 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; SSE42-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:35 Lat:59 SizeLat:43 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SSE42-NEXT:  Cost Model: Found costs of RThru:80 CodeSize:69 Lat:117 SizeLat:85 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_rotate_i32'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:15 Lat:14 SizeLat:21 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:37 Lat:25 SizeLat:52 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:74 Lat:50 SizeLat:104 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_rotate_i32'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:4 Lat:8 SizeLat:8 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:4 Lat:10 SizeLat:10 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:8 Lat:20 SizeLat:20 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'constant_rotate_i32'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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>)
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_rotate_i32'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:29 CodeSize:18 Lat:30 SizeLat:28 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; SLM-NEXT:  Cost Model: Found costs of RThru:58 CodeSize:35 Lat:59 SizeLat:55 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; SLM-NEXT:  Cost Model: Found costs of RThru:116 CodeSize:69 Lat:117 SizeLat:109 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_rotate_i32'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:30 SizeLat:22 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; GLM-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:35 Lat:59 SizeLat:43 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; GLM-NEXT:  Cost Model: Found costs of RThru:80 CodeSize:69 Lat:117 SizeLat:85 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_rotate_i32'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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>)
  ret void
}

define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {
; SSSE3-LABEL: 'constant_rotate_i16'
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:34 Lat:26 SizeLat:34 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:67 Lat:51 SizeLat:67 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:76 CodeSize:133 Lat:101 SizeLat:133 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_rotate_i16'
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:25 CodeSize:26 Lat:33 SizeLat:30 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SSE42-NEXT:  Cost Model: Found costs of RThru:50 CodeSize:51 Lat:65 SizeLat:59 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:100 CodeSize:101 Lat:129 SizeLat:117 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_rotate_i16'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:17 Lat:23 SizeLat:25 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX1-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:42 Lat:41 SizeLat:62 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:74 CodeSize:84 Lat:82 SizeLat:124 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_rotate_i16'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:18 SizeLat:13 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:13 Lat:17 SizeLat:20 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:24 CodeSize:26 Lat:34 SizeLat:40 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_rotate_i16'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:18 SizeLat:13 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:13 Lat:17 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:26 CodeSize:28 Lat:40 SizeLat:36 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_rotate_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:8 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_rotate_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:18 SizeLat:13 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:13 Lat:17 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:26 CodeSize:28 Lat:40 SizeLat:36 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_rotate_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_rotate_i16'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:26 Lat:33 SizeLat:30 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; SLM-NEXT:  Cost Model: Found costs of RThru:52 CodeSize:51 Lat:65 SizeLat:59 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:104 CodeSize:101 Lat:129 SizeLat:117 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_rotate_i16'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:25 CodeSize:26 Lat:33 SizeLat:30 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; GLM-NEXT:  Cost Model: Found costs of RThru:50 CodeSize:51 Lat:65 SizeLat:59 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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 costs of RThru:100 CodeSize:101 Lat:129 SizeLat:117 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_rotate_i16'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_rotate_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
  ret void
}

define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {
; SSSE3-LABEL: 'constant_rotate_i8'
; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; SSSE3-NEXT:  Cost Model: Found costs of RThru:29 CodeSize:55 Lat:51 SizeLat:60 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:58 CodeSize:109 Lat:101 SizeLat:119 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:116 CodeSize:217 Lat:201 SizeLat:237 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'constant_rotate_i8'
; SSE42-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; SSE42-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:37 Lat:53 SizeLat:48 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:66 CodeSize:73 Lat:105 SizeLat:95 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:132 CodeSize:145 Lat:209 SizeLat:189 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'constant_rotate_i8'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:50 SizeLat:37 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:50 CodeSize:63 Lat:48 SizeLat:91 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:100 CodeSize:126 Lat:96 SizeLat:182 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'constant_rotate_i8'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:25 Lat:50 SizeLat:36 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:16 CodeSize:25 Lat:55 SizeLat:50 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:32 CodeSize:50 Lat:110 SizeLat:100 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'constant_rotate_i8'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:25 Lat:50 SizeLat:36 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:16 CodeSize:25 Lat:55 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:34 CodeSize:63 Lat:46 SizeLat:75 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'constant_rotate_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'constant_rotate_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:25 Lat:50 SizeLat:36 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:16 CodeSize:25 Lat:55 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:34 CodeSize:63 Lat:46 SizeLat:75 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'constant_rotate_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'constant_rotate_i8'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:37 Lat:53 SizeLat:48 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:66 CodeSize:73 Lat:105 SizeLat:95 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:132 CodeSize:145 Lat:209 SizeLat:189 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'constant_rotate_i8'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:33 CodeSize:37 Lat:53 SizeLat:48 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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 costs of RThru:66 CodeSize:73 Lat:105 SizeLat:95 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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 costs of RThru:132 CodeSize:145 Lat:209 SizeLat:189 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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 costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'constant_rotate_i8'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'constant_rotate_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
  ret void
}

;
; Uniform Constant Unary Funnel Shifts (Rotates)
;

define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {
; SSE-LABEL: 'splatconstant_rotate_i64'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:5 SizeLat:5 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; SSE-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:10 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; SSE-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:16 Lat:20 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_rotate_i64'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:6 SizeLat:4 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:18 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:28 Lat:30 SizeLat:36 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_rotate_i64'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:4 Lat:6 SizeLat:8 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:8 Lat:12 SizeLat:16 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'splatconstant_rotate_i64'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_rotate_i64'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:4 Lat:5 SizeLat:5 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:8 Lat:10 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:16 Lat:20 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_rotate_i64'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:5 SizeLat:5 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:8 Lat:10 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:16 Lat:20 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_rotate_i64'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> splat (i64 7))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
  %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
  %V4I64  = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
  %V8I64  = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
  ret void
}

define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {
; SSE-LABEL: 'splatconstant_rotate_i32'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; SSE-NEXT:  Cost Model: Found costs of 4 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; SSE-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; SSE-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_rotate_i32'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:6 SizeLat:4 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:18 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:28 Lat:30 SizeLat:36 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_rotate_i32'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:4 Lat:6 SizeLat:8 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:8 Lat:12 SizeLat:16 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'splatconstant_rotate_i32'
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_rotate_i32'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; SLM-NEXT:  Cost Model: Found costs of 4 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_rotate_i32'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; GLM-NEXT:  Cost Model: Found costs of 4 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_rotate_i32'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> splat (i32 5))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
  %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
  %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
  %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <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>)
  ret void
}

define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {
; SSE-LABEL: 'splatconstant_rotate_i16'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; SSE-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_rotate_i16'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:6 SizeLat:4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:18 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:28 Lat:30 SizeLat:36 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_rotate_i16'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:4 Lat:6 SizeLat:8 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:8 Lat:12 SizeLat:16 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_rotate_i16'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX512F-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:16 SizeLat:11 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:16 Lat:28 SizeLat:20 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_rotate_i16'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:5 SizeLat:3 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:5 SizeLat:3 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:5 SizeLat:3 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX512DQ-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:16 SizeLat:11 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:16 Lat:28 SizeLat:20 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_rotate_i16'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_rotate_i16'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; SLM-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_rotate_i16'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; GLM-NEXT:  Cost Model: Found costs of 4 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:7 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:13 Lat:13 SizeLat:13 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_rotate_i16'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> splat (i16 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
  %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
  %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <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>)
  %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <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>)
  ret void
}

define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {
; SSE-LABEL: 'splatconstant_rotate_i8'
; SSE-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:16 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:11 Lat:31 SizeLat:15 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:21 Lat:61 SizeLat:29 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'splatconstant_rotate_i8'
; AVX1-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:16 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:20 Lat:17 SizeLat:24 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:40 Lat:34 SizeLat:48 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'splatconstant_rotate_i8'
; AVX2-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:18 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:18 SizeLat:12 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:12 Lat:36 SizeLat:24 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'splatconstant_rotate_i8'
; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:18 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:18 SizeLat:11 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:32 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'splatconstant_rotate_i8'
; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:8 SizeLat:4 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:9 SizeLat:4 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:9 SizeLat:4 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:18 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:6 Lat:18 SizeLat:11 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:32 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:8 SizeLat:4 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:9 SizeLat:4 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:9 SizeLat:4 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'splatconstant_rotate_i8'
; SLM-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:16 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:11 Lat:31 SizeLat:15 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:21 Lat:61 SizeLat:29 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'splatconstant_rotate_i8'
; GLM-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; GLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:16 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:11 Lat:31 SizeLat:15 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:16 CodeSize:21 Lat:61 SizeLat:29 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'splatconstant_rotate_i8'
; XOP-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> splat (i8 3))
; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
  %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
  %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <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>)
  %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <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>)
  %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <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>)
  ret void
}

declare i64 @llvm.fshr.i64(i64, i64, i64)
declare i32 @llvm.fshr.i32(i32, i32, i32)
declare i16 @llvm.fshr.i16(i16, i16, i16)
declare i8  @llvm.fshr.i8 (i8,  i8,  i8)

declare <2 x i64>  @llvm.fshr.v2i64(<2 x i64>, <2 x i64>, <2 x i64>)
declare <4 x i32>  @llvm.fshr.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)
declare <8 x i16>  @llvm.fshr.v8i16(<8 x i16>, <8 x i16>, <8 x i16>)
declare <16 x i8>  @llvm.fshr.v16i8(<16 x i8>, <16 x i8>, <16 x i8>)

declare <4 x i64>  @llvm.fshr.v4i64(<4 x i64>, <4 x i64>, <4 x i64>)
declare <8 x i32>  @llvm.fshr.v8i32(<8 x i32>, <8 x i32>, <8 x i32>)
declare <16 x i16> @llvm.fshr.v16i16(<16 x i16>, <16 x i16>, <16 x i16>)
declare <32 x i8>  @llvm.fshr.v32i8(<32 x i8>, <32 x i8>, <32 x i8>)

declare <8 x i64>  @llvm.fshr.v8i64(<8 x i64>, <8 x i64>, <8 x i64>)
declare <16 x i32> @llvm.fshr.v16i32(<16 x i32>, <16 x i32>, <16 x i32>)
declare <32 x i16> @llvm.fshr.v32i16(<32 x i16>, <32 x i16>, <32 x i16>)
declare <64 x i8>  @llvm.fshr.v64i8(<64 x i8>, <64 x i8>, <64 x i8>)