221 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py2; RUN: opt < %s -mtriple=armv8a-linux-gnueabihf -mattr=+fp64 -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-V83; RUN: opt < %s -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve.fp -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-MVEFP4; RUN: opt < %s -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve -passes="print<cost-model>" -cost-kind=all 2>&1 -disable-output | FileCheck %s --check-prefix=CHECK-MVEI5 6target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"7 8define void @fadd_strict() {9; CHECK-V8-LABEL: 'fadd_strict'10; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:6 Lat:10 SizeLat:6 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)11; CHECK-V8-NEXT: Cost Model: Found costs of RThru:16 CodeSize:12 Lat:20 SizeLat:12 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)12; CHECK-V8-NEXT: Cost Model: Found costs of RThru:32 CodeSize:24 Lat:40 SizeLat:24 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)13; CHECK-V8-NEXT: Cost Model: Found costs of RThru:64 CodeSize:48 Lat:80 SizeLat:48 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)14; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)15; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)16; CHECK-V8-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:24 SizeLat:8 for: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)17; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)18; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)19; CHECK-V8-NEXT: Cost Model: Found costs of RThru:48 CodeSize:20 Lat:28 SizeLat:20 for: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)20; CHECK-V8-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void21;22; CHECK-MVEFP-LABEL: 'fadd_strict'23; CHECK-MVEFP-NEXT: Cost Model: Found costs of 3 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)24; CHECK-MVEFP-NEXT: Cost Model: Found costs of 6 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)25; CHECK-MVEFP-NEXT: Cost Model: Found costs of 12 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)26; CHECK-MVEFP-NEXT: Cost Model: Found costs of 24 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)27; CHECK-MVEFP-NEXT: Cost Model: Found costs of 2 for: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)28; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)29; CHECK-MVEFP-NEXT: Cost Model: Found costs of 8 for: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)30; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)31; CHECK-MVEFP-NEXT: Cost Model: Found costs of 8 for: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)32; CHECK-MVEFP-NEXT: Cost Model: Found costs of 32 for: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)33; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void34;35; CHECK-MVEI-LABEL: 'fadd_strict'36; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)37; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)38; CHECK-MVEI-NEXT: Cost Model: Found costs of 16 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)39; CHECK-MVEI-NEXT: Cost Model: Found costs of 32 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)40; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)41; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)42; CHECK-MVEI-NEXT: Cost Model: Found costs of 16 for: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)43; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)44; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)45; CHECK-MVEI-NEXT: Cost Model: Found costs of 32 for: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)46; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void47;48 %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)49 %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)50 %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)51 %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)52 %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)53 %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)54 %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)55 %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)56 %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)57 %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)58 ret void59}60 61 62define void @fadd_unordered() {63; CHECK-V8-LABEL: 'fadd_unordered'64; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:7 Lat:9 SizeLat:7 for: %fadd_v2f16 = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)65; CHECK-V8-NEXT: Cost Model: Found costs of RThru:20 CodeSize:16 Lat:20 SizeLat:16 for: %fadd_v4f16 = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)66; CHECK-V8-NEXT: Cost Model: Found costs of RThru:44 CodeSize:33 Lat:39 SizeLat:33 for: %fadd_v8f16 = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)67; CHECK-V8-NEXT: Cost Model: Found costs of RThru:92 CodeSize:66 Lat:74 SizeLat:66 for: %fadd_v16f16 = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)68; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fadd_v2f32 = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)69; CHECK-V8-NEXT: Cost Model: Found costs of RThru:6 CodeSize:3 Lat:9 SizeLat:3 for: %fadd_v4f32 = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)70; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fadd_v8f32 = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)71; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fadd_v2f64 = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)72; CHECK-V8-NEXT: Cost Model: Found costs of RThru:12 CodeSize:3 Lat:9 SizeLat:3 for: %fadd_v4f64 = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)73; CHECK-V8-NEXT: Cost Model: Found costs of RThru:52 CodeSize:30 Lat:34 SizeLat:30 for: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)74; CHECK-V8-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void75;76; CHECK-MVEFP-LABEL: 'fadd_unordered'77; CHECK-MVEFP-NEXT: Cost Model: Found costs of 3 for: %fadd_v2f16 = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)78; CHECK-MVEFP-NEXT: Cost Model: Found costs of 6 for: %fadd_v4f16 = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)79; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:8 CodeSize:6 Lat:8 SizeLat:8 for: %fadd_v8f16 = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)80; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %fadd_v16f16 = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)81; CHECK-MVEFP-NEXT: Cost Model: Found costs of 2 for: %fadd_v2f32 = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)82; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fadd_v4f32 = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)83; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:6 CodeSize:5 Lat:6 SizeLat:6 for: %fadd_v8f32 = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)84; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fadd_v2f64 = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)85; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:29 CodeSize:21 Lat:29 SizeLat:29 for: %fadd_v4f64 = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)86; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:76 CodeSize:40 Lat:76 SizeLat:76 for: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)87; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void88;89; CHECK-MVEI-LABEL: 'fadd_unordered'90; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fadd_v2f16 = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)91; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:57 CodeSize:41 Lat:57 SizeLat:57 for: %fadd_v4f16 = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)92; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:169 CodeSize:121 Lat:169 SizeLat:169 for: %fadd_v8f16 = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)93; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:225 CodeSize:161 Lat:225 SizeLat:225 for: %fadd_v16f16 = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)94; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fadd_v2f32 = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)95; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:57 CodeSize:41 Lat:57 SizeLat:57 for: %fadd_v4f32 = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)96; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:85 CodeSize:61 Lat:85 SizeLat:85 for: %fadd_v8f32 = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)97; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fadd_v2f64 = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)98; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:29 CodeSize:21 Lat:29 SizeLat:29 for: %fadd_v4f64 = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)99; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:76 CodeSize:40 Lat:76 SizeLat:76 for: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)100; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void101;102 %fadd_v2f16 = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)103 %fadd_v4f16 = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)104 %fadd_v8f16 = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)105 %fadd_v16f16 = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)106 %fadd_v2f32 = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)107 %fadd_v4f32 = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)108 %fadd_v8f32 = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)109 %fadd_v2f64 = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)110 %fadd_v4f64 = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)111 %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)112 ret void113}114 115define void @fmul_strict() {116; CHECK-V8-LABEL: 'fmul_strict'117; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:6 Lat:10 SizeLat:6 for: %fmul_v2f16 = call half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)118; CHECK-V8-NEXT: Cost Model: Found costs of RThru:16 CodeSize:12 Lat:20 SizeLat:12 for: %fmul_v4f16 = call half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)119; CHECK-V8-NEXT: Cost Model: Found costs of RThru:32 CodeSize:24 Lat:40 SizeLat:24 for: %fmul_v8f16 = call half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)120; CHECK-V8-NEXT: Cost Model: Found costs of RThru:64 CodeSize:48 Lat:80 SizeLat:48 for: %fmul_v16f16 = call half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)121; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fmul_v2f32 = call float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)122; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fmul_v4f32 = call float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)123; CHECK-V8-NEXT: Cost Model: Found costs of RThru:16 CodeSize:8 Lat:24 SizeLat:8 for: %fmul_v8f32 = call float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)124; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fmul_v2f64 = call double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)125; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fmul_v4f64 = call double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)126; CHECK-V8-NEXT: Cost Model: Found costs of RThru:48 CodeSize:20 Lat:28 SizeLat:20 for: %fmul_v4f128 = call fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)127; CHECK-V8-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void128;129; CHECK-MVEFP-LABEL: 'fmul_strict'130; CHECK-MVEFP-NEXT: Cost Model: Found costs of 3 for: %fmul_v2f16 = call half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)131; CHECK-MVEFP-NEXT: Cost Model: Found costs of 6 for: %fmul_v4f16 = call half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)132; CHECK-MVEFP-NEXT: Cost Model: Found costs of 12 for: %fmul_v8f16 = call half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)133; CHECK-MVEFP-NEXT: Cost Model: Found costs of 24 for: %fmul_v16f16 = call half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)134; CHECK-MVEFP-NEXT: Cost Model: Found costs of 2 for: %fmul_v2f32 = call float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)135; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fmul_v4f32 = call float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)136; CHECK-MVEFP-NEXT: Cost Model: Found costs of 8 for: %fmul_v8f32 = call float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)137; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fmul_v2f64 = call double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)138; CHECK-MVEFP-NEXT: Cost Model: Found costs of 8 for: %fmul_v4f64 = call double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)139; CHECK-MVEFP-NEXT: Cost Model: Found costs of 32 for: %fmul_v4f128 = call fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)140; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void141;142; CHECK-MVEI-LABEL: 'fmul_strict'143; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fmul_v2f16 = call half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)144; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fmul_v4f16 = call half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)145; CHECK-MVEI-NEXT: Cost Model: Found costs of 16 for: %fmul_v8f16 = call half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)146; CHECK-MVEI-NEXT: Cost Model: Found costs of 32 for: %fmul_v16f16 = call half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)147; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fmul_v2f32 = call float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)148; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fmul_v4f32 = call float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)149; CHECK-MVEI-NEXT: Cost Model: Found costs of 16 for: %fmul_v8f32 = call float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)150; CHECK-MVEI-NEXT: Cost Model: Found costs of 4 for: %fmul_v2f64 = call double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)151; CHECK-MVEI-NEXT: Cost Model: Found costs of 8 for: %fmul_v4f64 = call double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)152; CHECK-MVEI-NEXT: Cost Model: Found costs of 32 for: %fmul_v4f128 = call fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)153; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void154;155 %fmul_v2f16 = call half @llvm.vector.reduce.fmul.v2f16(half 0.0, <2 x half> undef)156 %fmul_v4f16 = call half @llvm.vector.reduce.fmul.v4f16(half 0.0, <4 x half> undef)157 %fmul_v8f16 = call half @llvm.vector.reduce.fmul.v8f16(half 0.0, <8 x half> undef)158 %fmul_v16f16 = call half @llvm.vector.reduce.fmul.v16f16(half 0.0, <16 x half> undef)159 %fmul_v2f32 = call float @llvm.vector.reduce.fmul.v2f32(float 0.0, <2 x float> undef)160 %fmul_v4f32 = call float @llvm.vector.reduce.fmul.v4f32(float 0.0, <4 x float> undef)161 %fmul_v8f32 = call float @llvm.vector.reduce.fmul.v8f32(float 0.0, <8 x float> undef)162 %fmul_v2f64 = call double @llvm.vector.reduce.fmul.v2f64(double 0.0, <2 x double> undef)163 %fmul_v4f64 = call double @llvm.vector.reduce.fmul.v4f64(double 0.0, <4 x double> undef)164 %fmul_v4f128 = call fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)165 ret void166}167 168 169define void @fmul_unordered() {170; CHECK-V8-LABEL: 'fmul_unordered'171; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:7 Lat:9 SizeLat:7 for: %fmul_v2f16 = call reassoc half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)172; CHECK-V8-NEXT: Cost Model: Found costs of RThru:20 CodeSize:16 Lat:20 SizeLat:16 for: %fmul_v4f16 = call reassoc half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)173; CHECK-V8-NEXT: Cost Model: Found costs of RThru:44 CodeSize:33 Lat:39 SizeLat:33 for: %fmul_v8f16 = call reassoc half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)174; CHECK-V8-NEXT: Cost Model: Found costs of RThru:92 CodeSize:66 Lat:74 SizeLat:66 for: %fmul_v16f16 = call reassoc half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)175; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fmul_v2f32 = call reassoc float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)176; CHECK-V8-NEXT: Cost Model: Found costs of RThru:6 CodeSize:3 Lat:9 SizeLat:3 for: %fmul_v4f32 = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)177; CHECK-V8-NEXT: Cost Model: Found costs of RThru:8 CodeSize:4 Lat:12 SizeLat:4 for: %fmul_v8f32 = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)178; CHECK-V8-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %fmul_v2f64 = call reassoc double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)179; CHECK-V8-NEXT: Cost Model: Found costs of RThru:12 CodeSize:3 Lat:9 SizeLat:3 for: %fmul_v4f64 = call reassoc double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)180; CHECK-V8-NEXT: Cost Model: Found costs of RThru:52 CodeSize:30 Lat:34 SizeLat:30 for: %fmul_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)181; CHECK-V8-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void182;183; CHECK-MVEFP-LABEL: 'fmul_unordered'184; CHECK-MVEFP-NEXT: Cost Model: Found costs of 3 for: %fmul_v2f16 = call reassoc half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)185; CHECK-MVEFP-NEXT: Cost Model: Found costs of 6 for: %fmul_v4f16 = call reassoc half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)186; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:8 CodeSize:6 Lat:8 SizeLat:8 for: %fmul_v8f16 = call reassoc half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)187; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %fmul_v16f16 = call reassoc half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)188; CHECK-MVEFP-NEXT: Cost Model: Found costs of 2 for: %fmul_v2f32 = call reassoc float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)189; CHECK-MVEFP-NEXT: Cost Model: Found costs of 4 for: %fmul_v4f32 = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)190; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:6 CodeSize:5 Lat:6 SizeLat:6 for: %fmul_v8f32 = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)191; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fmul_v2f64 = call reassoc double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)192; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:29 CodeSize:21 Lat:29 SizeLat:29 for: %fmul_v4f64 = call reassoc double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)193; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:76 CodeSize:40 Lat:76 SizeLat:76 for: %fmul_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)194; CHECK-MVEFP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void195;196; CHECK-MVEI-LABEL: 'fmul_unordered'197; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fmul_v2f16 = call reassoc half @llvm.vector.reduce.fmul.v2f16(half 0xH0000, <2 x half> undef)198; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:57 CodeSize:41 Lat:57 SizeLat:57 for: %fmul_v4f16 = call reassoc half @llvm.vector.reduce.fmul.v4f16(half 0xH0000, <4 x half> undef)199; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:169 CodeSize:121 Lat:169 SizeLat:169 for: %fmul_v8f16 = call reassoc half @llvm.vector.reduce.fmul.v8f16(half 0xH0000, <8 x half> undef)200; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:225 CodeSize:161 Lat:225 SizeLat:225 for: %fmul_v16f16 = call reassoc half @llvm.vector.reduce.fmul.v16f16(half 0xH0000, <16 x half> undef)201; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fmul_v2f32 = call reassoc float @llvm.vector.reduce.fmul.v2f32(float 0.000000e+00, <2 x float> undef)202; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:57 CodeSize:41 Lat:57 SizeLat:57 for: %fmul_v4f32 = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 0.000000e+00, <4 x float> undef)203; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:85 CodeSize:61 Lat:85 SizeLat:85 for: %fmul_v8f32 = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 0.000000e+00, <8 x float> undef)204; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:15 CodeSize:11 Lat:15 SizeLat:15 for: %fmul_v2f64 = call reassoc double @llvm.vector.reduce.fmul.v2f64(double 0.000000e+00, <2 x double> undef)205; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:29 CodeSize:21 Lat:29 SizeLat:29 for: %fmul_v4f64 = call reassoc double @llvm.vector.reduce.fmul.v4f64(double 0.000000e+00, <4 x double> undef)206; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:76 CodeSize:40 Lat:76 SizeLat:76 for: %fmul_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)207; CHECK-MVEI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void208;209 %fmul_v2f16 = call reassoc half @llvm.vector.reduce.fmul.v2f16(half 0.0, <2 x half> undef)210 %fmul_v4f16 = call reassoc half @llvm.vector.reduce.fmul.v4f16(half 0.0, <4 x half> undef)211 %fmul_v8f16 = call reassoc half @llvm.vector.reduce.fmul.v8f16(half 0.0, <8 x half> undef)212 %fmul_v16f16 = call reassoc half @llvm.vector.reduce.fmul.v16f16(half 0.0, <16 x half> undef)213 %fmul_v2f32 = call reassoc float @llvm.vector.reduce.fmul.v2f32(float 0.0, <2 x float> undef)214 %fmul_v4f32 = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 0.0, <4 x float> undef)215 %fmul_v8f32 = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 0.0, <8 x float> undef)216 %fmul_v2f64 = call reassoc double @llvm.vector.reduce.fmul.v2f64(double 0.0, <2 x double> undef)217 %fmul_v4f64 = call reassoc double @llvm.vector.reduce.fmul.v4f64(double 0.0, <4 x double> undef)218 %fmul_v4f128 = call reassoc fp128 @llvm.vector.reduce.fmul.v4f128(fp128 undef, <4 x fp128> undef)219 ret void220}221