2905 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py2; 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,SSSE33; 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,SSE424; 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=AVX15; 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=AVX26; 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,AVX512F7; 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,AVX512BW8; 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,AVX512DQ9; 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,AVX512VBMI210;11; 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=SLM12; 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=GLM13; 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=XOP14; 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=AVX115; 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,AVX512GFNI16 17target 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"18target triple = "x86_64-apple-macosx10.8.0"19 20;21; Variable Funnel Shifts22;23 24define 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) {25; SSSE3-LABEL: 'var_funnel_i64'26; 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)27; 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)28; 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)29; 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)30; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void31;32; SSE42-LABEL: 'var_funnel_i64'33; 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)34; 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)35; 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)36; 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)37; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void38;39; AVX1-LABEL: 'var_funnel_i64'40; 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)41; 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)42; 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)43; 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)44; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void45;46; AVX2-LABEL: 'var_funnel_i64'47; 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)48; 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)49; 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)50; 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)51; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void52;53; AVX512F-LABEL: 'var_funnel_i64'54; 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)55; 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)56; 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)57; 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)58; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void59;60; AVX512BW-LABEL: 'var_funnel_i64'61; 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)62; 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)63; 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)64; 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)65; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void66;67; AVX512DQ-LABEL: 'var_funnel_i64'68; 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)69; 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)70; 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)71; 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)72; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void73;74; AVX512VBMI2-LABEL: 'var_funnel_i64'75; 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)76; 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)77; 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)78; 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)79; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void80;81; SLM-LABEL: 'var_funnel_i64'82; 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)83; 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)84; 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)85; 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)86; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void87;88; GLM-LABEL: 'var_funnel_i64'89; 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)90; 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)91; 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)92; 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)93; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void94;95; XOP-LABEL: 'var_funnel_i64'96; 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)97; 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)98; 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)99; 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)100; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void101;102; AVX512GFNI-LABEL: 'var_funnel_i64'103; 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)104; 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)105; 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)106; 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)107; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void108;109 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)110 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)111 %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)112 %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)113 ret void114}115 116define 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) {117; SSSE3-LABEL: 'var_funnel_i32'118; 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)119; 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)120; 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)121; 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)122; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void123;124; SSE42-LABEL: 'var_funnel_i32'125; 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)126; 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)127; 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)128; 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)129; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void130;131; AVX1-LABEL: 'var_funnel_i32'132; 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)133; 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)134; 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)135; 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)136; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void137;138; AVX2-LABEL: 'var_funnel_i32'139; 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)140; 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)141; 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)142; 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)143; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void144;145; AVX512F-LABEL: 'var_funnel_i32'146; 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)147; 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)148; 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)149; 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)150; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void151;152; AVX512BW-LABEL: 'var_funnel_i32'153; 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)154; 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)155; 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)156; 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)157; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void158;159; AVX512DQ-LABEL: 'var_funnel_i32'160; 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)161; 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)162; 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)163; 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)164; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void165;166; AVX512VBMI2-LABEL: 'var_funnel_i32'167; 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)168; 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)169; 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)170; 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)171; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void172;173; SLM-LABEL: 'var_funnel_i32'174; 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)175; 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)176; 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)177; 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)178; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void179;180; GLM-LABEL: 'var_funnel_i32'181; 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)182; 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)183; 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)184; 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)185; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void186;187; XOP-LABEL: 'var_funnel_i32'188; 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)189; 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)190; 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)191; 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)192; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void193;194; AVX512GFNI-LABEL: 'var_funnel_i32'195; 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)196; 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)197; 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)198; 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)199; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void200;201 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)202 %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)203 %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)204 %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)205 ret void206}207 208define 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) {209; SSSE3-LABEL: 'var_funnel_i16'210; 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)211; 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)212; 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)213; 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)214; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void215;216; SSE42-LABEL: 'var_funnel_i16'217; 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)218; 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)219; 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)220; 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)221; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void222;223; AVX1-LABEL: 'var_funnel_i16'224; 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)225; 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)226; 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)227; 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)228; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void229;230; AVX2-LABEL: 'var_funnel_i16'231; 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)232; 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)233; 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)234; 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)235; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void236;237; AVX512F-LABEL: 'var_funnel_i16'238; 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)239; 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)240; 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)241; 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)242; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void243;244; AVX512BW-LABEL: 'var_funnel_i16'245; 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)246; 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)247; 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)248; 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)249; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void250;251; AVX512DQ-LABEL: 'var_funnel_i16'252; 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)253; 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)254; 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)255; 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)256; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void257;258; AVX512VBMI2-LABEL: 'var_funnel_i16'259; 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)260; 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)261; 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)262; 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)263; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void264;265; SLM-LABEL: 'var_funnel_i16'266; 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)267; 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)268; 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)269; 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)270; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void271;272; GLM-LABEL: 'var_funnel_i16'273; 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)274; 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)275; 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)276; 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)277; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void278;279; XOP-LABEL: 'var_funnel_i16'280; 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)281; 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)282; 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)283; 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)284; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void285;286; AVX512GFNI-LABEL: 'var_funnel_i16'287; 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)288; 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)289; 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)290; 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)291; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void292;293 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)294 %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)295 %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)296 %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)297 ret void298}299 300define 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) {301; SSSE3-LABEL: 'var_funnel_i8'302; 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)303; 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)304; 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)305; 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)306; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void307;308; SSE42-LABEL: 'var_funnel_i8'309; 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)310; 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)311; 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)312; 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)313; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void314;315; AVX1-LABEL: 'var_funnel_i8'316; 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)317; 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)318; 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)319; 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)320; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void321;322; AVX2-LABEL: 'var_funnel_i8'323; 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)324; 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)325; 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)326; 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)327; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void328;329; AVX512F-LABEL: 'var_funnel_i8'330; 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)331; 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)332; 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)333; 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)334; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void335;336; AVX512BW-LABEL: 'var_funnel_i8'337; 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)338; 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)339; 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)340; 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)341; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void342;343; AVX512DQ-LABEL: 'var_funnel_i8'344; 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)345; 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)346; 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)347; 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)348; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void349;350; AVX512VBMI2-LABEL: 'var_funnel_i8'351; 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)352; 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)353; 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)354; 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)355; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void356;357; SLM-LABEL: 'var_funnel_i8'358; 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)359; 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)360; 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)361; 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)362; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void363;364; GLM-LABEL: 'var_funnel_i8'365; 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)366; 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)367; 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)368; 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)369; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void370;371; XOP-LABEL: 'var_funnel_i8'372; 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)373; 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)374; 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)375; 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)376; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void377;378; AVX512GFNI-LABEL: 'var_funnel_i8'379; 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)380; 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)381; 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)382; 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)383; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void384;385 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)386 %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)387 %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)388 %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)389 ret void390}391 392;393; Uniform Variable Funnel Shifts394;395 396define 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) {397; SSSE3-LABEL: 'splatvar_funnel_i64'398; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer399; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer400; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer401; 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)402; 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)403; 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)404; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void405;406; SSE42-LABEL: 'splatvar_funnel_i64'407; SSE42-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer408; SSE42-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer409; SSE42-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer410; 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)411; 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)412; 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)413; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void414;415; AVX1-LABEL: 'splatvar_funnel_i64'416; AVX1-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer417; 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> zeroinitializer418; 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> zeroinitializer419; 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)420; 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)421; 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)422; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void423;424; AVX2-LABEL: 'splatvar_funnel_i64'425; AVX2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer426; 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> zeroinitializer427; 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> zeroinitializer428; 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)429; 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)430; 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)431; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void432;433; AVX512F-LABEL: 'splatvar_funnel_i64'434; AVX512F-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer435; 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> zeroinitializer436; 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> zeroinitializer437; 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)438; 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)439; 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)440; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void441;442; AVX512BW-LABEL: 'splatvar_funnel_i64'443; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer444; 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> zeroinitializer445; 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> zeroinitializer446; 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)447; 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)448; 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)449; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void450;451; AVX512DQ-LABEL: 'splatvar_funnel_i64'452; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer453; 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> zeroinitializer454; 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> zeroinitializer455; 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)456; 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)457; 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)458; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void459;460; AVX512VBMI2-LABEL: 'splatvar_funnel_i64'461; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer462; 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> zeroinitializer463; 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> zeroinitializer464; 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)465; 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)466; 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)467; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void468;469; SLM-LABEL: 'splatvar_funnel_i64'470; SLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer471; SLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer472; SLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer473; 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)474; 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)475; 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)476; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void477;478; GLM-LABEL: 'splatvar_funnel_i64'479; GLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer480; GLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer481; GLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer482; 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)483; 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)484; 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)485; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void486;487; XOP-LABEL: 'splatvar_funnel_i64'488; XOP-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer489; 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> zeroinitializer490; 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> zeroinitializer491; 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)492; 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)493; 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)494; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void495;496; AVX512GFNI-LABEL: 'splatvar_funnel_i64'497; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer498; 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> zeroinitializer499; 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> zeroinitializer500; 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)501; 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)502; 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)503; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void504;505 %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer506 %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer507 %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer508 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)509 %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)510 %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)511 ret void512}513 514define 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) {515; SSSE3-LABEL: 'splatvar_funnel_i32'516; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer517; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer518; SSSE3-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer519; 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)520; 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)521; 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)522; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void523;524; SSE42-LABEL: 'splatvar_funnel_i32'525; SSE42-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer526; SSE42-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer527; SSE42-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer528; 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)529; 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)530; 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)531; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void532;533; AVX1-LABEL: 'splatvar_funnel_i32'534; AVX1-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer535; 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> zeroinitializer536; 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> zeroinitializer537; 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)538; 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)539; 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)540; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void541;542; AVX2-LABEL: 'splatvar_funnel_i32'543; AVX2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer544; 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> zeroinitializer545; 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> zeroinitializer546; 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)547; 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)548; 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)549; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void550;551; AVX512F-LABEL: 'splatvar_funnel_i32'552; AVX512F-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer553; 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> zeroinitializer554; 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> zeroinitializer555; 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)556; 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)557; 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)558; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void559;560; AVX512BW-LABEL: 'splatvar_funnel_i32'561; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer562; 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> zeroinitializer563; 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> zeroinitializer564; 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)565; 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)566; 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)567; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void568;569; AVX512DQ-LABEL: 'splatvar_funnel_i32'570; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer571; 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> zeroinitializer572; 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> zeroinitializer573; 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)574; 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)575; 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)576; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void577;578; AVX512VBMI2-LABEL: 'splatvar_funnel_i32'579; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer580; 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> zeroinitializer581; 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> zeroinitializer582; 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)583; 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)584; 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)585; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void586;587; SLM-LABEL: 'splatvar_funnel_i32'588; SLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer589; SLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer590; SLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer591; 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)592; 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)593; 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)594; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void595;596; GLM-LABEL: 'splatvar_funnel_i32'597; GLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer598; GLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer599; GLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer600; 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)601; 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)602; 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)603; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void604;605; XOP-LABEL: 'splatvar_funnel_i32'606; XOP-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer607; 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> zeroinitializer608; 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> zeroinitializer609; 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)610; 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)611; 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)612; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void613;614; AVX512GFNI-LABEL: 'splatvar_funnel_i32'615; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer616; 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> zeroinitializer617; 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> zeroinitializer618; 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)619; 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)620; 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)621; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void622;623 %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer624 %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer625 %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer626 %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)627 %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)628 %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)629 ret void630}631 632define 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) {633; SSSE3-LABEL: 'splatvar_funnel_i16'634; 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> zeroinitializer635; 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> zeroinitializer636; 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> zeroinitializer637; 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)638; 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)639; 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)640; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void641;642; SSE42-LABEL: 'splatvar_funnel_i16'643; 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> zeroinitializer644; 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> zeroinitializer645; 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> zeroinitializer646; 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)647; 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)648; 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)649; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void650;651; AVX1-LABEL: 'splatvar_funnel_i16'652; 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> zeroinitializer653; 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> zeroinitializer654; 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> zeroinitializer655; 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)656; 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)657; 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)658; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void659;660; AVX2-LABEL: 'splatvar_funnel_i16'661; 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> zeroinitializer662; 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> zeroinitializer663; 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> zeroinitializer664; 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)665; 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)666; 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)667; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void668;669; AVX512F-LABEL: 'splatvar_funnel_i16'670; 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> zeroinitializer671; 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> zeroinitializer672; 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> zeroinitializer673; 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)674; 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)675; 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)676; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void677;678; AVX512BW-LABEL: 'splatvar_funnel_i16'679; 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> zeroinitializer680; 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> zeroinitializer681; 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> zeroinitializer682; 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)683; 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)684; 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)685; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void686;687; AVX512DQ-LABEL: 'splatvar_funnel_i16'688; 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> zeroinitializer689; 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> zeroinitializer690; 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> zeroinitializer691; 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)692; 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)693; 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)694; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void695;696; AVX512VBMI2-LABEL: 'splatvar_funnel_i16'697; 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> zeroinitializer698; 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> zeroinitializer699; 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> zeroinitializer700; 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)701; 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)702; 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)703; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void704;705; SLM-LABEL: 'splatvar_funnel_i16'706; 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> zeroinitializer707; 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> zeroinitializer708; 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> zeroinitializer709; 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)710; 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)711; 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)712; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void713;714; GLM-LABEL: 'splatvar_funnel_i16'715; 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> zeroinitializer716; 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> zeroinitializer717; 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> zeroinitializer718; 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)719; 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)720; 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)721; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void722;723; XOP-LABEL: 'splatvar_funnel_i16'724; 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> zeroinitializer725; 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> zeroinitializer726; 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> zeroinitializer727; 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)728; 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)729; 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)730; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void731;732; AVX512GFNI-LABEL: 'splatvar_funnel_i16'733; 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> zeroinitializer734; 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> zeroinitializer735; 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> zeroinitializer736; 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)737; 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)738; 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)739; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void740;741 %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer742 %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer743 %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer744 %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)745 %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)746 %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)747 ret void748}749 750define 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) {751; SSSE3-LABEL: 'splatvar_funnel_i8'752; 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> zeroinitializer753; 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> zeroinitializer754; 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> zeroinitializer755; 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)756; 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)757; 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)758; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void759;760; SSE42-LABEL: 'splatvar_funnel_i8'761; 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> zeroinitializer762; 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> zeroinitializer763; 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> zeroinitializer764; 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)765; 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)766; 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)767; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void768;769; AVX1-LABEL: 'splatvar_funnel_i8'770; 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> zeroinitializer771; 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> zeroinitializer772; 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> zeroinitializer773; 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)774; 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)775; 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)776; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void777;778; AVX2-LABEL: 'splatvar_funnel_i8'779; 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> zeroinitializer780; 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> zeroinitializer781; 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> zeroinitializer782; 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)783; 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)784; 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)785; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void786;787; AVX512F-LABEL: 'splatvar_funnel_i8'788; 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> zeroinitializer789; 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> zeroinitializer790; 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> zeroinitializer791; 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)792; 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)793; 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)794; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void795;796; AVX512BW-LABEL: 'splatvar_funnel_i8'797; 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> zeroinitializer798; 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> zeroinitializer799; 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> zeroinitializer800; 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)801; 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)802; 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)803; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void804;805; AVX512DQ-LABEL: 'splatvar_funnel_i8'806; 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> zeroinitializer807; 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> zeroinitializer808; 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> zeroinitializer809; 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)810; 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)811; 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)812; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void813;814; AVX512VBMI2-LABEL: 'splatvar_funnel_i8'815; 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> zeroinitializer816; 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> zeroinitializer817; 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> zeroinitializer818; 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)819; 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)820; 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)821; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void822;823; SLM-LABEL: 'splatvar_funnel_i8'824; 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> zeroinitializer825; 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> zeroinitializer826; 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> zeroinitializer827; 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)828; 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)829; 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)830; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void831;832; GLM-LABEL: 'splatvar_funnel_i8'833; 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> zeroinitializer834; 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> zeroinitializer835; 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> zeroinitializer836; 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)837; 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)838; 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)839; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void840;841; XOP-LABEL: 'splatvar_funnel_i8'842; 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> zeroinitializer843; 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> zeroinitializer844; 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> zeroinitializer845; 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)846; 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)847; 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)848; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void849;850; AVX512GFNI-LABEL: 'splatvar_funnel_i8'851; 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> zeroinitializer852; 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> zeroinitializer853; 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> zeroinitializer854; 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)855; 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)856; 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)857; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void858;859 %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer860 %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer861 %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer862 %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)863 %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)864 %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)865 ret void866}867 868;869; Constant Funnel Shifts870;871 872define 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) {873; SSSE3-LABEL: 'constant_funnel_i64'874; 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)875; 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>)876; 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>)877; 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>)878; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void879;880; SSE42-LABEL: 'constant_funnel_i64'881; 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)882; 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>)883; 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>)884; 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>)885; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void886;887; AVX1-LABEL: 'constant_funnel_i64'888; 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)889; 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>)890; 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>)891; 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>)892; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void893;894; AVX2-LABEL: 'constant_funnel_i64'895; 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)896; 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>)897; 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>)898; 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>)899; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void900;901; AVX512F-LABEL: 'constant_funnel_i64'902; 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)903; 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>)904; 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>)905; 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>)906; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void907;908; AVX512BW-LABEL: 'constant_funnel_i64'909; 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)910; 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>)911; 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>)912; 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>)913; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void914;915; AVX512DQ-LABEL: 'constant_funnel_i64'916; 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)917; 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>)918; 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>)919; 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>)920; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void921;922; AVX512VBMI2-LABEL: 'constant_funnel_i64'923; 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)924; 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>)925; 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>)926; 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>)927; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void928;929; SLM-LABEL: 'constant_funnel_i64'930; 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)931; 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>)932; 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>)933; 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>)934; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void935;936; GLM-LABEL: 'constant_funnel_i64'937; 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)938; 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>)939; 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>)940; 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>)941; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void942;943; XOP-LABEL: 'constant_funnel_i64'944; 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)945; 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>)946; 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>)947; 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>)948; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void949;950; AVX512GFNI-LABEL: 'constant_funnel_i64'951; 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)952; 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>)953; 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>)954; 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>)955; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void956;957 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)958 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)959 %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>)960 %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>)961 ret void962}963 964define 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) {965; SSSE3-LABEL: 'constant_funnel_i32'966; 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)967; 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>)968; 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>)969; 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>)970; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void971;972; SSE42-LABEL: 'constant_funnel_i32'973; 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)974; 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>)975; 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>)976; 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>)977; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void978;979; AVX1-LABEL: 'constant_funnel_i32'980; 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)981; 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>)982; 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>)983; 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>)984; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void985;986; AVX2-LABEL: 'constant_funnel_i32'987; 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)988; 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>)989; 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>)990; 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>)991; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void992;993; AVX512F-LABEL: 'constant_funnel_i32'994; 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)995; 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>)996; 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>)997; 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>)998; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void999;1000; AVX512BW-LABEL: 'constant_funnel_i32'1001; 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)1002; 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>)1003; 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>)1004; 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>)1005; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1006;1007; AVX512DQ-LABEL: 'constant_funnel_i32'1008; 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)1009; 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>)1010; 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>)1011; 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>)1012; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1013;1014; AVX512VBMI2-LABEL: 'constant_funnel_i32'1015; 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)1016; 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>)1017; 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>)1018; 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>)1019; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1020;1021; SLM-LABEL: 'constant_funnel_i32'1022; 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)1023; 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>)1024; 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>)1025; 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>)1026; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1027;1028; GLM-LABEL: 'constant_funnel_i32'1029; 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)1030; 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>)1031; 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>)1032; 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>)1033; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1034;1035; XOP-LABEL: 'constant_funnel_i32'1036; 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)1037; 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>)1038; 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>)1039; 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>)1040; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1041;1042; AVX512GFNI-LABEL: 'constant_funnel_i32'1043; 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)1044; 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>)1045; 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>)1046; 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>)1047; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1048;1049 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)1050 %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>)1051 %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>)1052 %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>)1053 ret void1054}1055 1056define 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) {1057; SSSE3-LABEL: 'constant_funnel_i16'1058; 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)1059; 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>)1060; 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>)1061; 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>)1062; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1063;1064; SSE42-LABEL: 'constant_funnel_i16'1065; 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)1066; 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>)1067; 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>)1068; 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>)1069; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1070;1071; AVX1-LABEL: 'constant_funnel_i16'1072; 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)1073; 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>)1074; 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>)1075; 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>)1076; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1077;1078; AVX2-LABEL: 'constant_funnel_i16'1079; 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)1080; 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>)1081; 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>)1082; 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>)1083; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1084;1085; AVX512F-LABEL: 'constant_funnel_i16'1086; 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)1087; 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>)1088; 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>)1089; 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>)1090; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1091;1092; AVX512BW-LABEL: 'constant_funnel_i16'1093; 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)1094; 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>)1095; 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>)1096; 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>)1097; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1098;1099; AVX512DQ-LABEL: 'constant_funnel_i16'1100; 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)1101; 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>)1102; 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>)1103; 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>)1104; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1105;1106; AVX512VBMI2-LABEL: 'constant_funnel_i16'1107; 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)1108; 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>)1109; 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>)1110; 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>)1111; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1112;1113; SLM-LABEL: 'constant_funnel_i16'1114; 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)1115; 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>)1116; 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>)1117; 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>)1118; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1119;1120; GLM-LABEL: 'constant_funnel_i16'1121; 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)1122; 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>)1123; 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>)1124; 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>)1125; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1126;1127; XOP-LABEL: 'constant_funnel_i16'1128; 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)1129; 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>)1130; 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>)1131; 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>)1132; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1133;1134; AVX512GFNI-LABEL: 'constant_funnel_i16'1135; 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)1136; 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>)1137; 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>)1138; 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>)1139; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1140;1141 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)1142 %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>)1143 %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>)1144 %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>)1145 ret void1146}1147 1148define 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) {1149; SSSE3-LABEL: 'constant_funnel_i8'1150; 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)1151; 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>)1152; 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>)1153; 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>)1154; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1155;1156; SSE42-LABEL: 'constant_funnel_i8'1157; 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)1158; 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>)1159; 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>)1160; 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>)1161; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1162;1163; AVX1-LABEL: 'constant_funnel_i8'1164; 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)1165; 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>)1166; 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>)1167; 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>)1168; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1169;1170; AVX2-LABEL: 'constant_funnel_i8'1171; 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)1172; 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>)1173; 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>)1174; 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>)1175; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1176;1177; AVX512F-LABEL: 'constant_funnel_i8'1178; 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)1179; 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>)1180; 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>)1181; 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>)1182; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1183;1184; AVX512BW-LABEL: 'constant_funnel_i8'1185; 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)1186; 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>)1187; 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>)1188; 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>)1189; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1190;1191; AVX512DQ-LABEL: 'constant_funnel_i8'1192; 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)1193; 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>)1194; 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>)1195; 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>)1196; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1197;1198; AVX512VBMI2-LABEL: 'constant_funnel_i8'1199; 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)1200; 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>)1201; 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>)1202; 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>)1203; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1204;1205; SLM-LABEL: 'constant_funnel_i8'1206; 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)1207; 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>)1208; 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>)1209; 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>)1210; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1211;1212; GLM-LABEL: 'constant_funnel_i8'1213; 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)1214; 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>)1215; 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>)1216; 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>)1217; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1218;1219; XOP-LABEL: 'constant_funnel_i8'1220; 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)1221; 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>)1222; 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>)1223; 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>)1224; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1225;1226; AVX512GFNI-LABEL: 'constant_funnel_i8'1227; 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)1228; 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>)1229; 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>)1230; 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>)1231; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1232;1233 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)1234 %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>)1235 %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>)1236 %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>)1237 ret void1238}1239 1240;1241; Uniform Constant Funnel Shifts1242;1243 1244define 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) {1245; SSSE3-LABEL: 'splatconstant_funnel_i64'1246; 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)1247; 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))1248; 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))1249; 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))1250; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1251;1252; SSE42-LABEL: 'splatconstant_funnel_i64'1253; 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)1254; 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))1255; 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))1256; 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))1257; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1258;1259; AVX1-LABEL: 'splatconstant_funnel_i64'1260; 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)1261; 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))1262; 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))1263; 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))1264; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1265;1266; AVX2-LABEL: 'splatconstant_funnel_i64'1267; 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)1268; 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))1269; 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))1270; 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))1271; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1272;1273; AVX512F-LABEL: 'splatconstant_funnel_i64'1274; 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)1275; 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))1276; 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))1277; 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))1278; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1279;1280; AVX512BW-LABEL: 'splatconstant_funnel_i64'1281; 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)1282; 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))1283; 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))1284; 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))1285; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1286;1287; AVX512DQ-LABEL: 'splatconstant_funnel_i64'1288; 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)1289; 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))1290; 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))1291; 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))1292; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1293;1294; AVX512VBMI2-LABEL: 'splatconstant_funnel_i64'1295; 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)1296; 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))1297; 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))1298; 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))1299; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1300;1301; SLM-LABEL: 'splatconstant_funnel_i64'1302; 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)1303; 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))1304; 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))1305; 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))1306; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1307;1308; GLM-LABEL: 'splatconstant_funnel_i64'1309; 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)1310; 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))1311; 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))1312; 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))1313; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1314;1315; XOP-LABEL: 'splatconstant_funnel_i64'1316; 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)1317; 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))1318; 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))1319; 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))1320; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1321;1322; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'1323; 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)1324; 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))1325; 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))1326; 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))1327; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1328;1329 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)1330 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)1331 %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>)1332 %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>)1333 ret void1334}1335 1336define 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) {1337; SSSE3-LABEL: 'splatconstant_funnel_i32'1338; 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)1339; 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))1340; 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))1341; 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))1342; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1343;1344; SSE42-LABEL: 'splatconstant_funnel_i32'1345; 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)1346; 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))1347; 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))1348; 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))1349; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1350;1351; AVX1-LABEL: 'splatconstant_funnel_i32'1352; 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)1353; 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))1354; 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))1355; 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))1356; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1357;1358; AVX2-LABEL: 'splatconstant_funnel_i32'1359; 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)1360; 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))1361; 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))1362; 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))1363; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1364;1365; AVX512F-LABEL: 'splatconstant_funnel_i32'1366; 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)1367; 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))1368; 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))1369; 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))1370; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1371;1372; AVX512BW-LABEL: 'splatconstant_funnel_i32'1373; 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)1374; 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))1375; 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))1376; 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))1377; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1378;1379; AVX512DQ-LABEL: 'splatconstant_funnel_i32'1380; 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)1381; 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))1382; 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))1383; 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))1384; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1385;1386; AVX512VBMI2-LABEL: 'splatconstant_funnel_i32'1387; 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)1388; 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))1389; 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))1390; 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))1391; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1392;1393; SLM-LABEL: 'splatconstant_funnel_i32'1394; 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)1395; 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))1396; 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))1397; 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))1398; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1399;1400; GLM-LABEL: 'splatconstant_funnel_i32'1401; 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)1402; 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))1403; 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))1404; 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))1405; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1406;1407; XOP-LABEL: 'splatconstant_funnel_i32'1408; 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)1409; 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))1410; 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))1411; 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))1412; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1413;1414; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'1415; 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)1416; 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))1417; 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))1418; 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))1419; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1420;1421 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)1422 %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>)1423 %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>)1424 %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>)1425 ret void1426}1427 1428define 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) {1429; SSSE3-LABEL: 'splatconstant_funnel_i16'1430; 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)1431; 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))1432; 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))1433; 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))1434; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1435;1436; SSE42-LABEL: 'splatconstant_funnel_i16'1437; 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)1438; 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))1439; 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))1440; 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))1441; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1442;1443; AVX1-LABEL: 'splatconstant_funnel_i16'1444; 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)1445; 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))1446; 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))1447; 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))1448; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1449;1450; AVX2-LABEL: 'splatconstant_funnel_i16'1451; 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)1452; 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))1453; 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))1454; 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))1455; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1456;1457; AVX512F-LABEL: 'splatconstant_funnel_i16'1458; 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)1459; 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))1460; 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))1461; 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))1462; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1463;1464; AVX512BW-LABEL: 'splatconstant_funnel_i16'1465; 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)1466; 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))1467; 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))1468; 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))1469; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1470;1471; AVX512DQ-LABEL: 'splatconstant_funnel_i16'1472; 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)1473; 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))1474; 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))1475; 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))1476; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1477;1478; AVX512VBMI2-LABEL: 'splatconstant_funnel_i16'1479; 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)1480; 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))1481; 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))1482; 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))1483; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1484;1485; SLM-LABEL: 'splatconstant_funnel_i16'1486; 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)1487; 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))1488; 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))1489; 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))1490; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1491;1492; GLM-LABEL: 'splatconstant_funnel_i16'1493; 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)1494; 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))1495; 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))1496; 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))1497; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1498;1499; XOP-LABEL: 'splatconstant_funnel_i16'1500; 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)1501; 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))1502; 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))1503; 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))1504; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1505;1506; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'1507; 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)1508; 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))1509; 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))1510; 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))1511; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1512;1513 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)1514 %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>)1515 %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>)1516 %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>)1517 ret void1518}1519 1520define 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) {1521; SSSE3-LABEL: 'splatconstant_funnel_i8'1522; 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)1523; 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))1524; 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))1525; 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))1526; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1527;1528; SSE42-LABEL: 'splatconstant_funnel_i8'1529; 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)1530; 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))1531; 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))1532; 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))1533; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1534;1535; AVX1-LABEL: 'splatconstant_funnel_i8'1536; 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)1537; 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))1538; 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))1539; 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))1540; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1541;1542; AVX2-LABEL: 'splatconstant_funnel_i8'1543; 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)1544; 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))1545; 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))1546; 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))1547; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1548;1549; AVX512F-LABEL: 'splatconstant_funnel_i8'1550; 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)1551; 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))1552; 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))1553; 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))1554; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1555;1556; AVX512BW-LABEL: 'splatconstant_funnel_i8'1557; 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)1558; 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))1559; 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))1560; 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))1561; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1562;1563; AVX512DQ-LABEL: 'splatconstant_funnel_i8'1564; 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)1565; 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))1566; 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))1567; 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))1568; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1569;1570; AVX512VBMI2-LABEL: 'splatconstant_funnel_i8'1571; 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)1572; 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))1573; 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))1574; 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))1575; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1576;1577; SLM-LABEL: 'splatconstant_funnel_i8'1578; 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)1579; 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))1580; 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))1581; 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))1582; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1583;1584; GLM-LABEL: 'splatconstant_funnel_i8'1585; 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)1586; 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))1587; 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))1588; 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))1589; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1590;1591; XOP-LABEL: 'splatconstant_funnel_i8'1592; 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)1593; 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))1594; 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))1595; 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))1596; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1597;1598; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'1599; 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)1600; 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))1601; 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))1602; 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))1603; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1604;1605 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)1606 %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>)1607 %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>)1608 %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>)1609 ret void1610}1611 1612;1613; Variable Unary Funnel Shifts (Rotates)1614;1615 1616define 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) {1617; SSE-LABEL: 'var_rotate_i64'1618; 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)1619; 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)1620; 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)1621; 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)1622; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1623;1624; AVX1-LABEL: 'var_rotate_i64'1625; 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)1626; 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)1627; 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)1628; 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)1629; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1630;1631; AVX2-LABEL: 'var_rotate_i64'1632; 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)1633; 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)1634; 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)1635; 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)1636; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1637;1638; AVX512-LABEL: 'var_rotate_i64'1639; 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)1640; 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)1641; 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)1642; 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)1643; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1644;1645; SLM-LABEL: 'var_rotate_i64'1646; 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)1647; 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)1648; 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)1649; 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)1650; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1651;1652; GLM-LABEL: 'var_rotate_i64'1653; 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)1654; 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)1655; 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)1656; 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)1657; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1658;1659; XOP-LABEL: 'var_rotate_i64'1660; 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)1661; 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)1662; 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)1663; 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)1664; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1665;1666 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)1667 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)1668 %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)1669 %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)1670 ret void1671}1672 1673define 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) {1674; SSSE3-LABEL: 'var_rotate_i32'1675; 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)1676; 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)1677; 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)1678; 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)1679; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1680;1681; SSE42-LABEL: 'var_rotate_i32'1682; 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)1683; 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)1684; 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)1685; 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)1686; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1687;1688; AVX1-LABEL: 'var_rotate_i32'1689; 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)1690; 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)1691; 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)1692; 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)1693; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1694;1695; AVX2-LABEL: 'var_rotate_i32'1696; 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)1697; 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)1698; 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)1699; 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)1700; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1701;1702; AVX512-LABEL: 'var_rotate_i32'1703; 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)1704; 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)1705; 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)1706; 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)1707; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1708;1709; SLM-LABEL: 'var_rotate_i32'1710; 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)1711; 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)1712; 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)1713; 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)1714; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1715;1716; GLM-LABEL: 'var_rotate_i32'1717; 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)1718; 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)1719; 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)1720; 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)1721; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1722;1723; XOP-LABEL: 'var_rotate_i32'1724; 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)1725; 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)1726; 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)1727; 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)1728; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1729;1730 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)1731 %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)1732 %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)1733 %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)1734 ret void1735}1736 1737define 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) {1738; SSSE3-LABEL: 'var_rotate_i16'1739; 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)1740; 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)1741; 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)1742; 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)1743; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1744;1745; SSE42-LABEL: 'var_rotate_i16'1746; 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)1747; 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)1748; 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)1749; 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)1750; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1751;1752; AVX1-LABEL: 'var_rotate_i16'1753; 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)1754; 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)1755; 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)1756; 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)1757; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1758;1759; AVX2-LABEL: 'var_rotate_i16'1760; 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)1761; 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)1762; 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)1763; 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)1764; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1765;1766; AVX512F-LABEL: 'var_rotate_i16'1767; 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)1768; 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)1769; 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)1770; 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)1771; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1772;1773; AVX512BW-LABEL: 'var_rotate_i16'1774; 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)1775; 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)1776; 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)1777; 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)1778; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1779;1780; AVX512DQ-LABEL: 'var_rotate_i16'1781; 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)1782; 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)1783; 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)1784; 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)1785; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1786;1787; AVX512VBMI2-LABEL: 'var_rotate_i16'1788; 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)1789; 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)1790; 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)1791; 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)1792; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1793;1794; SLM-LABEL: 'var_rotate_i16'1795; 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)1796; 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)1797; 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)1798; 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)1799; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1800;1801; GLM-LABEL: 'var_rotate_i16'1802; 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)1803; 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)1804; 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)1805; 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)1806; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1807;1808; XOP-LABEL: 'var_rotate_i16'1809; 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)1810; 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)1811; 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)1812; 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)1813; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1814;1815; AVX512GFNI-LABEL: 'var_rotate_i16'1816; 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)1817; 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)1818; 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)1819; 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)1820; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1821;1822 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)1823 %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)1824 %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)1825 %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)1826 ret void1827}1828 1829define 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) {1830; SSSE3-LABEL: 'var_rotate_i8'1831; 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)1832; 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)1833; 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)1834; 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)1835; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1836;1837; SSE42-LABEL: 'var_rotate_i8'1838; 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)1839; 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)1840; 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)1841; 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)1842; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1843;1844; AVX1-LABEL: 'var_rotate_i8'1845; 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)1846; 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)1847; 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)1848; 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)1849; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1850;1851; AVX2-LABEL: 'var_rotate_i8'1852; 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)1853; 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)1854; 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)1855; 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)1856; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1857;1858; AVX512F-LABEL: 'var_rotate_i8'1859; 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)1860; 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)1861; 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)1862; 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)1863; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1864;1865; AVX512BW-LABEL: 'var_rotate_i8'1866; 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)1867; 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)1868; 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)1869; 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)1870; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1871;1872; AVX512DQ-LABEL: 'var_rotate_i8'1873; 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)1874; 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)1875; 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)1876; 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)1877; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1878;1879; AVX512VBMI2-LABEL: 'var_rotate_i8'1880; 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)1881; 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)1882; 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)1883; 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)1884; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1885;1886; SLM-LABEL: 'var_rotate_i8'1887; 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)1888; 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)1889; 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)1890; 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)1891; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1892;1893; GLM-LABEL: 'var_rotate_i8'1894; 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)1895; 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)1896; 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)1897; 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)1898; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1899;1900; XOP-LABEL: 'var_rotate_i8'1901; 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)1902; 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)1903; 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)1904; 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)1905; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1906;1907; AVX512GFNI-LABEL: 'var_rotate_i8'1908; 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)1909; 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)1910; 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)1911; 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)1912; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1913;1914 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)1915 %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)1916 %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)1917 %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)1918 ret void1919}1920 1921;1922; Uniform Variable Unary Funnel Shifts (Rotates)1923;1924 1925define 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) {1926; SSE-LABEL: 'splatvar_rotate_i64'1927; SSE-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1928; SSE-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer1929; SSE-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer1930; 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)1931; 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)1932; 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)1933; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1934;1935; AVX1-LABEL: 'splatvar_rotate_i64'1936; AVX1-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1937; 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> zeroinitializer1938; 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> zeroinitializer1939; 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)1940; 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)1941; 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)1942; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1943;1944; AVX2-LABEL: 'splatvar_rotate_i64'1945; AVX2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1946; 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> zeroinitializer1947; 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> zeroinitializer1948; 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)1949; 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)1950; 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)1951; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1952;1953; AVX512-LABEL: 'splatvar_rotate_i64'1954; AVX512-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1955; 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> zeroinitializer1956; 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> zeroinitializer1957; 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)1958; 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)1959; 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)1960; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1961;1962; SLM-LABEL: 'splatvar_rotate_i64'1963; SLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1964; SLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer1965; SLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer1966; 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)1967; 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)1968; 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)1969; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1970;1971; GLM-LABEL: 'splatvar_rotate_i64'1972; GLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1973; GLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer1974; GLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer1975; 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)1976; 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)1977; 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)1978; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1979;1980; XOP-LABEL: 'splatvar_rotate_i64'1981; XOP-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1982; 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> zeroinitializer1983; 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> zeroinitializer1984; 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)1985; 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)1986; 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)1987; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void1988;1989 %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer1990 %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer1991 %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer1992 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)1993 %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)1994 %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)1995 ret void1996}1997 1998define 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) {1999; SSE-LABEL: 'splatvar_rotate_i32'2000; SSE-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2001; SSE-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer2002; SSE-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer2003; 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)2004; 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)2005; 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)2006; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2007;2008; AVX1-LABEL: 'splatvar_rotate_i32'2009; AVX1-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2010; 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> zeroinitializer2011; 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> zeroinitializer2012; 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)2013; 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)2014; 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)2015; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2016;2017; AVX2-LABEL: 'splatvar_rotate_i32'2018; AVX2-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2019; 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> zeroinitializer2020; 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> zeroinitializer2021; 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)2022; 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)2023; 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)2024; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2025;2026; AVX512-LABEL: 'splatvar_rotate_i32'2027; AVX512-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2028; 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> zeroinitializer2029; 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> zeroinitializer2030; 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)2031; 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)2032; 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)2033; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2034;2035; SLM-LABEL: 'splatvar_rotate_i32'2036; SLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2037; SLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer2038; SLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer2039; 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)2040; 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)2041; 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)2042; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2043;2044; GLM-LABEL: 'splatvar_rotate_i32'2045; GLM-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2046; GLM-NEXT: Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer2047; GLM-NEXT: Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer2048; 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)2049; 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)2050; 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)2051; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2052;2053; XOP-LABEL: 'splatvar_rotate_i32'2054; XOP-NEXT: Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2055; 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> zeroinitializer2056; 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> zeroinitializer2057; 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)2058; 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)2059; 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)2060; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2061;2062 %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer2063 %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer2064 %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer2065 %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)2066 %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)2067 %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)2068 ret void2069}2070 2071define 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) {2072; SSE-LABEL: 'splatvar_rotate_i16'2073; 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> zeroinitializer2074; 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> zeroinitializer2075; 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> zeroinitializer2076; 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)2077; 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)2078; 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)2079; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2080;2081; AVX1-LABEL: 'splatvar_rotate_i16'2082; 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> zeroinitializer2083; 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> zeroinitializer2084; 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> zeroinitializer2085; 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)2086; 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)2087; 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)2088; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2089;2090; AVX2-LABEL: 'splatvar_rotate_i16'2091; 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> zeroinitializer2092; 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> zeroinitializer2093; 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> zeroinitializer2094; 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)2095; 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)2096; 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)2097; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2098;2099; AVX512F-LABEL: 'splatvar_rotate_i16'2100; 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> zeroinitializer2101; 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> zeroinitializer2102; 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> zeroinitializer2103; 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)2104; 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)2105; 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)2106; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2107;2108; AVX512BW-LABEL: 'splatvar_rotate_i16'2109; 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> zeroinitializer2110; 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> zeroinitializer2111; 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> zeroinitializer2112; 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)2113; 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)2114; 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)2115; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2116;2117; AVX512DQ-LABEL: 'splatvar_rotate_i16'2118; 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> zeroinitializer2119; 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> zeroinitializer2120; 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> zeroinitializer2121; 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)2122; 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)2123; 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)2124; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2125;2126; AVX512VBMI2-LABEL: 'splatvar_rotate_i16'2127; 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> zeroinitializer2128; 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> zeroinitializer2129; 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> zeroinitializer2130; 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)2131; 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)2132; 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)2133; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2134;2135; SLM-LABEL: 'splatvar_rotate_i16'2136; 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> zeroinitializer2137; 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> zeroinitializer2138; 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> zeroinitializer2139; 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)2140; 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)2141; 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)2142; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2143;2144; GLM-LABEL: 'splatvar_rotate_i16'2145; 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> zeroinitializer2146; 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> zeroinitializer2147; 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> zeroinitializer2148; 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)2149; 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)2150; 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)2151; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2152;2153; XOP-LABEL: 'splatvar_rotate_i16'2154; 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> zeroinitializer2155; 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> zeroinitializer2156; 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> zeroinitializer2157; 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)2158; 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)2159; 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)2160; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2161;2162; AVX512GFNI-LABEL: 'splatvar_rotate_i16'2163; 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> zeroinitializer2164; 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> zeroinitializer2165; 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> zeroinitializer2166; 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)2167; 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)2168; 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)2169; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2170;2171 %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer2172 %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer2173 %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer2174 %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)2175 %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)2176 %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)2177 ret void2178}2179 2180define 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) {2181; SSE-LABEL: 'splatvar_rotate_i8'2182; 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> zeroinitializer2183; 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> zeroinitializer2184; 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> zeroinitializer2185; 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)2186; 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)2187; 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)2188; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2189;2190; AVX1-LABEL: 'splatvar_rotate_i8'2191; 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> zeroinitializer2192; 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> zeroinitializer2193; 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> zeroinitializer2194; 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)2195; 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)2196; 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)2197; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2198;2199; AVX2-LABEL: 'splatvar_rotate_i8'2200; 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> zeroinitializer2201; 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> zeroinitializer2202; 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> zeroinitializer2203; 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)2204; 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)2205; 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)2206; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2207;2208; AVX512F-LABEL: 'splatvar_rotate_i8'2209; 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> zeroinitializer2210; 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> zeroinitializer2211; 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> zeroinitializer2212; 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)2213; 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)2214; 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)2215; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2216;2217; AVX512BW-LABEL: 'splatvar_rotate_i8'2218; 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> zeroinitializer2219; 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> zeroinitializer2220; 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> zeroinitializer2221; 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)2222; 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)2223; 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)2224; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2225;2226; AVX512DQ-LABEL: 'splatvar_rotate_i8'2227; 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> zeroinitializer2228; 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> zeroinitializer2229; 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> zeroinitializer2230; 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)2231; 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)2232; 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)2233; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2234;2235; AVX512VBMI2-LABEL: 'splatvar_rotate_i8'2236; 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> zeroinitializer2237; 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> zeroinitializer2238; 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> zeroinitializer2239; 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)2240; 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)2241; 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)2242; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2243;2244; SLM-LABEL: 'splatvar_rotate_i8'2245; 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> zeroinitializer2246; 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> zeroinitializer2247; 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> zeroinitializer2248; 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)2249; 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)2250; 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)2251; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2252;2253; GLM-LABEL: 'splatvar_rotate_i8'2254; 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> zeroinitializer2255; 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> zeroinitializer2256; 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> zeroinitializer2257; 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)2258; 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)2259; 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)2260; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2261;2262; XOP-LABEL: 'splatvar_rotate_i8'2263; 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> zeroinitializer2264; 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> zeroinitializer2265; 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> zeroinitializer2266; 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)2267; 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)2268; 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)2269; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2270;2271; AVX512GFNI-LABEL: 'splatvar_rotate_i8'2272; 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> zeroinitializer2273; 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> zeroinitializer2274; 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> zeroinitializer2275; 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)2276; 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)2277; 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)2278; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2279;2280 %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer2281 %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer2282 %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer2283 %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)2284 %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)2285 %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)2286 ret void2287}2288 2289;2290; Constant Unary Funnel Shifts (Rotates)2291;2292 2293define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {2294; SSE-LABEL: 'constant_rotate_i64'2295; SSE-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2296; 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>)2297; 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>)2298; 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>)2299; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2300;2301; AVX1-LABEL: 'constant_rotate_i64'2302; AVX1-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2303; 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>)2304; 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>)2305; 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>)2306; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2307;2308; AVX2-LABEL: 'constant_rotate_i64'2309; AVX2-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2310; 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>)2311; 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>)2312; 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>)2313; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2314;2315; AVX512-LABEL: 'constant_rotate_i64'2316; AVX512-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2317; 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>)2318; 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>)2319; 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>)2320; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2321;2322; SLM-LABEL: 'constant_rotate_i64'2323; SLM-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2324; 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>)2325; 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>)2326; 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>)2327; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2328;2329; GLM-LABEL: 'constant_rotate_i64'2330; GLM-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2331; 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>)2332; 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>)2333; 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>)2334; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2335;2336; XOP-LABEL: 'constant_rotate_i64'2337; XOP-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2338; 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>)2339; 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>)2340; 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>)2341; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2342;2343 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2344 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)2345 %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>)2346 %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>)2347 ret void2348}2349 2350define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {2351; SSSE3-LABEL: 'constant_rotate_i32'2352; SSSE3-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2353; 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>)2354; 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>)2355; 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>)2356; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2357;2358; SSE42-LABEL: 'constant_rotate_i32'2359; SSE42-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2360; 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>)2361; 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>)2362; 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>)2363; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2364;2365; AVX1-LABEL: 'constant_rotate_i32'2366; AVX1-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2367; 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>)2368; 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>)2369; 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>)2370; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2371;2372; AVX2-LABEL: 'constant_rotate_i32'2373; AVX2-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2374; 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>)2375; 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>)2376; 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>)2377; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2378;2379; AVX512-LABEL: 'constant_rotate_i32'2380; AVX512-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2381; 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>)2382; 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>)2383; 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>)2384; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2385;2386; SLM-LABEL: 'constant_rotate_i32'2387; SLM-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2388; 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>)2389; 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>)2390; 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>)2391; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2392;2393; GLM-LABEL: 'constant_rotate_i32'2394; GLM-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2395; 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>)2396; 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>)2397; 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>)2398; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2399;2400; XOP-LABEL: 'constant_rotate_i32'2401; XOP-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2402; 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>)2403; 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>)2404; 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>)2405; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2406;2407 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)2408 %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>)2409 %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>)2410 %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>)2411 ret void2412}2413 2414define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {2415; SSSE3-LABEL: 'constant_rotate_i16'2416; SSSE3-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2417; 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>)2418; 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>)2419; 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>)2420; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2421;2422; SSE42-LABEL: 'constant_rotate_i16'2423; SSE42-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2424; 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>)2425; 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>)2426; 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>)2427; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2428;2429; AVX1-LABEL: 'constant_rotate_i16'2430; AVX1-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2431; 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>)2432; 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>)2433; 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>)2434; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2435;2436; AVX2-LABEL: 'constant_rotate_i16'2437; AVX2-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2438; 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>)2439; 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>)2440; 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>)2441; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2442;2443; AVX512F-LABEL: 'constant_rotate_i16'2444; AVX512F-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2445; 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>)2446; 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>)2447; 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>)2448; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2449;2450; AVX512BW-LABEL: 'constant_rotate_i16'2451; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2452; 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>)2453; 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>)2454; 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>)2455; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2456;2457; AVX512DQ-LABEL: 'constant_rotate_i16'2458; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2459; 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>)2460; 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>)2461; 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>)2462; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2463;2464; AVX512VBMI2-LABEL: 'constant_rotate_i16'2465; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2466; 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>)2467; 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>)2468; 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>)2469; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2470;2471; SLM-LABEL: 'constant_rotate_i16'2472; SLM-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2473; 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>)2474; 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>)2475; 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>)2476; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2477;2478; GLM-LABEL: 'constant_rotate_i16'2479; GLM-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2480; 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>)2481; 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>)2482; 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>)2483; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2484;2485; XOP-LABEL: 'constant_rotate_i16'2486; XOP-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2487; 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>)2488; 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>)2489; 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>)2490; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2491;2492; AVX512GFNI-LABEL: 'constant_rotate_i16'2493; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2494; 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>)2495; 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>)2496; 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>)2497; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2498;2499 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)2500 %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>)2501 %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>)2502 %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>)2503 ret void2504}2505 2506define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {2507; SSSE3-LABEL: 'constant_rotate_i8'2508; SSSE3-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2509; 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>)2510; 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>)2511; 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>)2512; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2513;2514; SSE42-LABEL: 'constant_rotate_i8'2515; SSE42-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2516; 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>)2517; 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>)2518; 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>)2519; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2520;2521; AVX1-LABEL: 'constant_rotate_i8'2522; AVX1-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2523; 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>)2524; 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>)2525; 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>)2526; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2527;2528; AVX2-LABEL: 'constant_rotate_i8'2529; AVX2-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2530; 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>)2531; 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>)2532; 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>)2533; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2534;2535; AVX512F-LABEL: 'constant_rotate_i8'2536; AVX512F-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2537; 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>)2538; 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>)2539; 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>)2540; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2541;2542; AVX512BW-LABEL: 'constant_rotate_i8'2543; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2544; 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>)2545; 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>)2546; 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>)2547; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2548;2549; AVX512DQ-LABEL: 'constant_rotate_i8'2550; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2551; 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>)2552; 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>)2553; 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>)2554; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2555;2556; AVX512VBMI2-LABEL: 'constant_rotate_i8'2557; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2558; 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>)2559; 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>)2560; 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>)2561; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2562;2563; SLM-LABEL: 'constant_rotate_i8'2564; SLM-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2565; 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>)2566; 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>)2567; 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>)2568; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2569;2570; GLM-LABEL: 'constant_rotate_i8'2571; GLM-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2572; 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>)2573; 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>)2574; 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>)2575; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2576;2577; XOP-LABEL: 'constant_rotate_i8'2578; XOP-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2579; 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>)2580; 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>)2581; 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>)2582; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2583;2584; AVX512GFNI-LABEL: 'constant_rotate_i8'2585; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2586; 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>)2587; 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>)2588; 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>)2589; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2590;2591 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)2592 %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>)2593 %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>)2594 %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>)2595 ret void2596}2597 2598;2599; Uniform Constant Unary Funnel Shifts (Rotates)2600;2601 2602define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64> %a512) {2603; SSE-LABEL: 'splatconstant_rotate_i64'2604; SSE-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2605; 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))2606; 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))2607; 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))2608; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2609;2610; AVX1-LABEL: 'splatconstant_rotate_i64'2611; AVX1-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2612; 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))2613; 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))2614; 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))2615; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2616;2617; AVX2-LABEL: 'splatconstant_rotate_i64'2618; AVX2-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2619; 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))2620; 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))2621; 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))2622; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2623;2624; AVX512-LABEL: 'splatconstant_rotate_i64'2625; AVX512-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2626; 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))2627; 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))2628; 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))2629; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2630;2631; SLM-LABEL: 'splatconstant_rotate_i64'2632; SLM-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2633; 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))2634; 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))2635; 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))2636; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2637;2638; GLM-LABEL: 'splatconstant_rotate_i64'2639; GLM-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2640; 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))2641; 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))2642; 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))2643; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2644;2645; XOP-LABEL: 'splatconstant_rotate_i64'2646; XOP-NEXT: Cost Model: Found costs of 1 for: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2647; 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))2648; 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))2649; 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))2650; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2651;2652 %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)2653 %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)2654 %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>)2655 %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>)2656 ret void2657}2658 2659define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i32> %a512) {2660; SSE-LABEL: 'splatconstant_rotate_i32'2661; SSE-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2662; 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))2663; 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))2664; 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))2665; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2666;2667; AVX1-LABEL: 'splatconstant_rotate_i32'2668; AVX1-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2669; 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))2670; 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))2671; 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))2672; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2673;2674; AVX2-LABEL: 'splatconstant_rotate_i32'2675; AVX2-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2676; 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))2677; 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))2678; 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))2679; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2680;2681; AVX512-LABEL: 'splatconstant_rotate_i32'2682; AVX512-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2683; 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))2684; 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))2685; 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))2686; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2687;2688; SLM-LABEL: 'splatconstant_rotate_i32'2689; SLM-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2690; 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))2691; 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))2692; 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))2693; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2694;2695; GLM-LABEL: 'splatconstant_rotate_i32'2696; GLM-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2697; 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))2698; 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))2699; 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))2700; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2701;2702; XOP-LABEL: 'splatconstant_rotate_i32'2703; XOP-NEXT: Cost Model: Found costs of 1 for: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2704; 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))2705; 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))2706; 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))2707; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2708;2709 %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)2710 %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>)2711 %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>)2712 %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>)2713 ret void2714}2715 2716define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512) {2717; SSE-LABEL: 'splatconstant_rotate_i16'2718; SSE-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2719; 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))2720; 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))2721; 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))2722; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2723;2724; AVX1-LABEL: 'splatconstant_rotate_i16'2725; AVX1-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2726; 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))2727; 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))2728; 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))2729; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2730;2731; AVX2-LABEL: 'splatconstant_rotate_i16'2732; AVX2-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2733; 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))2734; 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))2735; 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))2736; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2737;2738; AVX512F-LABEL: 'splatconstant_rotate_i16'2739; AVX512F-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2740; 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))2741; 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))2742; 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))2743; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2744;2745; AVX512BW-LABEL: 'splatconstant_rotate_i16'2746; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2747; 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))2748; 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))2749; 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))2750; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2751;2752; AVX512DQ-LABEL: 'splatconstant_rotate_i16'2753; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2754; 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))2755; 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))2756; 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))2757; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2758;2759; AVX512VBMI2-LABEL: 'splatconstant_rotate_i16'2760; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2761; 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))2762; 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))2763; 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))2764; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2765;2766; SLM-LABEL: 'splatconstant_rotate_i16'2767; SLM-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2768; 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))2769; 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))2770; 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))2771; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2772;2773; GLM-LABEL: 'splatconstant_rotate_i16'2774; GLM-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2775; 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))2776; 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))2777; 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))2778; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2779;2780; XOP-LABEL: 'splatconstant_rotate_i16'2781; XOP-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2782; 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))2783; 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))2784; 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))2785; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2786;2787; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'2788; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2789; 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))2790; 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))2791; 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))2792; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2793;2794 %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)2795 %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>)2796 %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>)2797 %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>)2798 ret void2799}2800 2801define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512) {2802; SSE-LABEL: 'splatconstant_rotate_i8'2803; SSE-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2804; 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))2805; 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))2806; 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))2807; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2808;2809; AVX1-LABEL: 'splatconstant_rotate_i8'2810; AVX1-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2811; 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))2812; 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))2813; 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))2814; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2815;2816; AVX2-LABEL: 'splatconstant_rotate_i8'2817; AVX2-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2818; 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))2819; 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))2820; 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))2821; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2822;2823; AVX512F-LABEL: 'splatconstant_rotate_i8'2824; AVX512F-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2825; 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))2826; 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))2827; 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))2828; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2829;2830; AVX512BW-LABEL: 'splatconstant_rotate_i8'2831; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2832; 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))2833; 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))2834; 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))2835; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2836;2837; AVX512DQ-LABEL: 'splatconstant_rotate_i8'2838; AVX512DQ-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2839; 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))2840; 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))2841; 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))2842; AVX512DQ-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2843;2844; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'2845; AVX512VBMI2-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2846; 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))2847; 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))2848; 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))2849; AVX512VBMI2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2850;2851; SLM-LABEL: 'splatconstant_rotate_i8'2852; SLM-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2853; 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))2854; 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))2855; 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))2856; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2857;2858; GLM-LABEL: 'splatconstant_rotate_i8'2859; GLM-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2860; 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))2861; 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))2862; 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))2863; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2864;2865; XOP-LABEL: 'splatconstant_rotate_i8'2866; XOP-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2867; 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))2868; 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))2869; 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))2870; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2871;2872; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'2873; AVX512GFNI-NEXT: Cost Model: Found costs of 1 for: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2874; 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))2875; 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))2876; 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))2877; AVX512GFNI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void2878;2879 %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)2880 %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>)2881 %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>)2882 %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>)2883 ret void2884}2885 2886declare i64 @llvm.fshr.i64(i64, i64, i64)2887declare i32 @llvm.fshr.i32(i32, i32, i32)2888declare i16 @llvm.fshr.i16(i16, i16, i16)2889declare i8 @llvm.fshr.i8 (i8, i8, i8)2890 2891declare <2 x i64> @llvm.fshr.v2i64(<2 x i64>, <2 x i64>, <2 x i64>)2892declare <4 x i32> @llvm.fshr.v4i32(<4 x i32>, <4 x i32>, <4 x i32>)2893declare <8 x i16> @llvm.fshr.v8i16(<8 x i16>, <8 x i16>, <8 x i16>)2894declare <16 x i8> @llvm.fshr.v16i8(<16 x i8>, <16 x i8>, <16 x i8>)2895 2896declare <4 x i64> @llvm.fshr.v4i64(<4 x i64>, <4 x i64>, <4 x i64>)2897declare <8 x i32> @llvm.fshr.v8i32(<8 x i32>, <8 x i32>, <8 x i32>)2898declare <16 x i16> @llvm.fshr.v16i16(<16 x i16>, <16 x i16>, <16 x i16>)2899declare <32 x i8> @llvm.fshr.v32i8(<32 x i8>, <32 x i8>, <32 x i8>)2900 2901declare <8 x i64> @llvm.fshr.v8i64(<8 x i64>, <8 x i64>, <8 x i64>)2902declare <16 x i32> @llvm.fshr.v16i32(<16 x i32>, <16 x i32>, <16 x i32>)2903declare <32 x i16> @llvm.fshr.v32i16(<32 x i16>, <32 x i16>, <32 x i16>)2904declare <64 x i8> @llvm.fshr.v64i8(<64 x i8>, <64 x i8>, <64 x i8>)2905