151 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py2; RUN: opt -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output -mtriple=thumbv8m.main < %s | FileCheck %s --check-prefix=V8M3; RUN: opt -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output -mtriple=armv8a-linux-gnueabihf < %s | FileCheck %s --check-prefix=NEON4; RUN: opt -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output -mtriple=armv8.1m.main -mattr=+mve < %s | FileCheck %s --check-prefix=MVE5 6target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"7 8define i32 @reduce_i64(i32 %arg) {9; V8M-LABEL: 'reduce_i64'10; V8M-NEXT: Cost Model: Found costs of 2 for: %V1 = call i64 @llvm.vector.reduce.smin.v1i64(<1 x i64> undef)11; V8M-NEXT: Cost Model: Found costs of RThru:13 CodeSize:14 Lat:13 SizeLat:13 for: %V2 = call i64 @llvm.vector.reduce.smin.v2i64(<2 x i64> undef)12; V8M-NEXT: Cost Model: Found costs of RThru:35 CodeSize:38 Lat:35 SizeLat:35 for: %V4 = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> undef)13; V8M-NEXT: Cost Model: Found costs of RThru:79 CodeSize:86 Lat:79 SizeLat:79 for: %V8 = call i64 @llvm.vector.reduce.smin.v8i64(<8 x i64> undef)14; V8M-NEXT: Cost Model: Found costs of RThru:167 CodeSize:182 Lat:167 SizeLat:167 for: %V16 = call i64 @llvm.vector.reduce.smin.v16i64(<16 x i64> undef)15; V8M-NEXT: Cost Model: Found costs of 1 for: ret i32 undef16;17; NEON-LABEL: 'reduce_i64'18; NEON-NEXT: Cost Model: Found costs of 3 for: %V1 = call i64 @llvm.vector.reduce.smin.v1i64(<1 x i64> undef)19; NEON-NEXT: Cost Model: Found costs of 17 for: %V2 = call i64 @llvm.vector.reduce.smin.v2i64(<2 x i64> undef)20; NEON-NEXT: Cost Model: Found costs of 31 for: %V4 = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> undef)21; NEON-NEXT: Cost Model: Found costs of 76 for: %V8 = call i64 @llvm.vector.reduce.smin.v8i64(<8 x i64> undef)22; NEON-NEXT: Cost Model: Found costs of 178 for: %V16 = call i64 @llvm.vector.reduce.smin.v16i64(<16 x i64> undef)23; NEON-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef24;25; MVE-LABEL: 'reduce_i64'26; MVE-NEXT: Cost Model: Found costs of 8 for: %V1 = call i64 @llvm.vector.reduce.smin.v1i64(<1 x i64> undef)27; MVE-NEXT: Cost Model: Found costs of RThru:110 CodeSize:59 Lat:110 SizeLat:110 for: %V2 = call i64 @llvm.vector.reduce.smin.v2i64(<2 x i64> undef)28; MVE-NEXT: Cost Model: Found costs of RThru:212 CodeSize:110 Lat:212 SizeLat:212 for: %V4 = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> undef)29; MVE-NEXT: Cost Model: Found costs of RThru:416 CodeSize:212 Lat:416 SizeLat:416 for: %V8 = call i64 @llvm.vector.reduce.smin.v8i64(<8 x i64> undef)30; MVE-NEXT: Cost Model: Found costs of RThru:824 CodeSize:416 Lat:824 SizeLat:824 for: %V16 = call i64 @llvm.vector.reduce.smin.v16i64(<16 x i64> undef)31; MVE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef32;33 %V1 = call i64 @llvm.vector.reduce.smin.v1i64(<1 x i64> undef)34 %V2 = call i64 @llvm.vector.reduce.smin.v2i64(<2 x i64> undef)35 %V4 = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> undef)36 %V8 = call i64 @llvm.vector.reduce.smin.v8i64(<8 x i64> undef)37 %V16 = call i64 @llvm.vector.reduce.smin.v16i64(<16 x i64> undef)38 ret i32 undef39}40 41define i32 @reduce_i32(i32 %arg) {42; V8M-LABEL: 'reduce_i32'43; V8M-NEXT: Cost Model: Found costs of RThru:7 CodeSize:8 Lat:7 SizeLat:7 for: %V2 = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)44; V8M-NEXT: Cost Model: Found costs of RThru:19 CodeSize:22 Lat:19 SizeLat:19 for: %V4 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)45; V8M-NEXT: Cost Model: Found costs of RThru:43 CodeSize:50 Lat:43 SizeLat:43 for: %V8 = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)46; V8M-NEXT: Cost Model: Found costs of RThru:91 CodeSize:106 Lat:91 SizeLat:91 for: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)47; V8M-NEXT: Cost Model: Found costs of RThru:187 CodeSize:218 Lat:187 SizeLat:187 for: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)48; V8M-NEXT: Cost Model: Found costs of 1 for: ret i32 undef49;50; NEON-LABEL: 'reduce_i32'51; NEON-NEXT: Cost Model: Found costs of 16 for: %V2 = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)52; NEON-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %V4 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)53; NEON-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %V8 = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)54; NEON-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:16 SizeLat:16 for: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)55; NEON-NEXT: Cost Model: Found costs of RThru:32 CodeSize:16 Lat:32 SizeLat:32 for: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)56; NEON-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef57;58; MVE-LABEL: 'reduce_i32'59; MVE-NEXT: Cost Model: Found costs of RThru:58 CodeSize:31 Lat:58 SizeLat:58 for: %V2 = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)60; MVE-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %V4 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)61; MVE-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %V8 = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)62; MVE-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:16 SizeLat:16 for: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)63; MVE-NEXT: Cost Model: Found costs of RThru:32 CodeSize:16 Lat:32 SizeLat:32 for: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)64; MVE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef65;66 %V2 = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)67 %V4 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)68 %V8 = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)69 %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)70 %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)71 ret i32 undef72}73 74define i32 @reduce_i16(i32 %arg) {75; V8M-LABEL: 'reduce_i16'76; V8M-NEXT: Cost Model: Found costs of RThru:7 CodeSize:8 Lat:7 SizeLat:7 for: %V2 = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)77; V8M-NEXT: Cost Model: Found costs of RThru:19 CodeSize:22 Lat:19 SizeLat:19 for: %V4 = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)78; V8M-NEXT: Cost Model: Found costs of RThru:43 CodeSize:50 Lat:43 SizeLat:43 for: %V8 = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)79; V8M-NEXT: Cost Model: Found costs of RThru:91 CodeSize:106 Lat:91 SizeLat:91 for: %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)80; V8M-NEXT: Cost Model: Found costs of RThru:187 CodeSize:218 Lat:187 SizeLat:187 for: %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)81; V8M-NEXT: Cost Model: Found costs of RThru:379 CodeSize:442 Lat:379 SizeLat:379 for: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)82; V8M-NEXT: Cost Model: Found costs of 1 for: ret i32 undef83;84; NEON-LABEL: 'reduce_i16'85; NEON-NEXT: Cost Model: Found costs of 16 for: %V2 = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)86; NEON-NEXT: Cost Model: Found costs of 53 for: %V4 = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)87; NEON-NEXT: Cost Model: Found costs of RThru:6 CodeSize:3 Lat:6 SizeLat:6 for: %V8 = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)88; NEON-NEXT: Cost Model: Found costs of RThru:12 CodeSize:6 Lat:12 SizeLat:12 for: %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)89; NEON-NEXT: Cost Model: Found costs of RThru:24 CodeSize:12 Lat:24 SizeLat:24 for: %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)90; NEON-NEXT: Cost Model: Found costs of RThru:48 CodeSize:24 Lat:48 SizeLat:48 for: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)91; NEON-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef92;93; MVE-LABEL: 'reduce_i16'94; MVE-NEXT: Cost Model: Found costs of RThru:58 CodeSize:31 Lat:58 SizeLat:58 for: %V2 = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)95; MVE-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %V4 = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)96; MVE-NEXT: Cost Model: Found costs of RThru:6 CodeSize:3 Lat:6 SizeLat:6 for: %V8 = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)97; MVE-NEXT: Cost Model: Found costs of RThru:12 CodeSize:6 Lat:12 SizeLat:12 for: %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)98; MVE-NEXT: Cost Model: Found costs of RThru:24 CodeSize:12 Lat:24 SizeLat:24 for: %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)99; MVE-NEXT: Cost Model: Found costs of RThru:48 CodeSize:24 Lat:48 SizeLat:48 for: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)100; MVE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef101;102 %V2 = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)103 %V4 = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)104 %V8 = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)105 %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)106 %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)107 %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)108 ret i32 undef109}110 111define i32 @reduce_i8(i32 %arg) {112; V8M-LABEL: 'reduce_i8'113; V8M-NEXT: Cost Model: Found costs of RThru:7 CodeSize:8 Lat:7 SizeLat:7 for: %V2 = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)114; V8M-NEXT: Cost Model: Found costs of RThru:19 CodeSize:22 Lat:19 SizeLat:19 for: %V4 = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)115; V8M-NEXT: Cost Model: Found costs of RThru:43 CodeSize:50 Lat:43 SizeLat:43 for: %V8 = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)116; V8M-NEXT: Cost Model: Found costs of RThru:91 CodeSize:106 Lat:91 SizeLat:91 for: %V16 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)117; V8M-NEXT: Cost Model: Found costs of RThru:187 CodeSize:218 Lat:187 SizeLat:187 for: %V32 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)118; V8M-NEXT: Cost Model: Found costs of RThru:379 CodeSize:442 Lat:379 SizeLat:379 for: %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)119; V8M-NEXT: Cost Model: Found costs of RThru:763 CodeSize:890 Lat:763 SizeLat:763 for: %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)120; V8M-NEXT: Cost Model: Found costs of 1 for: ret i32 undef121;122; NEON-LABEL: 'reduce_i8'123; NEON-NEXT: Cost Model: Found costs of 16 for: %V2 = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)124; NEON-NEXT: Cost Model: Found costs of 53 for: %V4 = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)125; NEON-NEXT: Cost Model: Found costs of 150 for: %V8 = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)126; NEON-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %V16 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)127; NEON-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:16 SizeLat:16 for: %V32 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)128; NEON-NEXT: Cost Model: Found costs of RThru:32 CodeSize:16 Lat:32 SizeLat:32 for: %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)129; NEON-NEXT: Cost Model: Found costs of RThru:64 CodeSize:32 Lat:64 SizeLat:64 for: %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)130; NEON-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef131;132; MVE-LABEL: 'reduce_i8'133; MVE-NEXT: Cost Model: Found costs of RThru:58 CodeSize:31 Lat:58 SizeLat:58 for: %V2 = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)134; MVE-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %V4 = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)135; MVE-NEXT: Cost Model: Found costs of RThru:6 CodeSize:3 Lat:6 SizeLat:6 for: %V8 = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)136; MVE-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %V16 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)137; MVE-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:16 SizeLat:16 for: %V32 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)138; MVE-NEXT: Cost Model: Found costs of RThru:32 CodeSize:16 Lat:32 SizeLat:32 for: %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)139; MVE-NEXT: Cost Model: Found costs of RThru:64 CodeSize:32 Lat:64 SizeLat:64 for: %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)140; MVE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret i32 undef141;142 %V2 = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)143 %V4 = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)144 %V8 = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)145 %V16 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)146 %V32 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)147 %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)148 %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)149 ret i32 undef150}151