213 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py2; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=aarch64 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-NOFP163; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=aarch64 < %s -mattr=+fullfp16 | FileCheck %s --check-prefixes=CHECK,CHECK-F164; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mtriple=aarch64 < %s -mattr=+bf16 | FileCheck %s --check-prefixes=CHECK,CHECK-BF165 6target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"7 8define void @strict_fp_reductions() {9; CHECK-LABEL: 'strict_fp_reductions'10; CHECK-NEXT: Cost Model: Found costs of RThru:6 CodeSize:5 Lat:10 SizeLat:6 for: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)11; CHECK-NEXT: Cost Model: Found costs of RThru:14 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)12; CHECK-NEXT: Cost Model: Found costs of RThru:28 CodeSize:22 Lat:44 SizeLat:28 for: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)13; CHECK-NEXT: Cost Model: Found costs of RThru:6 CodeSize:5 Lat:10 SizeLat:6 for: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)14; CHECK-NEXT: Cost Model: Found costs of RThru:12 CodeSize:10 Lat:20 SizeLat:12 for: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)15; CHECK-NEXT: Cost Model: Found costs of RThru:20 CodeSize:8 Lat:16 SizeLat:8 for: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)16; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void17;18 %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)19 %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)20 %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)21 %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)22 %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)23 %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)24 ret void25}26 27define void @strict_fp_reductions_fp16() {28; CHECK-NOFP16-LABEL: 'strict_fp_reductions_fp16'29; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:12 CodeSize:5 Lat:10 SizeLat:6 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)30; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:26 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)31; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:54 CodeSize:23 Lat:46 SizeLat:30 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)32; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:108 CodeSize:46 Lat:92 SizeLat:60 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)33; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void34;35; CHECK-F16-LABEL: 'strict_fp_reductions_fp16'36; CHECK-F16-NEXT: Cost Model: Found costs of RThru:6 CodeSize:5 Lat:10 SizeLat:6 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)37; CHECK-F16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)38; CHECK-F16-NEXT: Cost Model: Found costs of RThru:30 CodeSize:23 Lat:46 SizeLat:30 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)39; CHECK-F16-NEXT: Cost Model: Found costs of RThru:60 CodeSize:46 Lat:92 SizeLat:60 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)40; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void41;42; CHECK-BF16-LABEL: 'strict_fp_reductions_fp16'43; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:12 CodeSize:5 Lat:10 SizeLat:6 for: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)44; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:26 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)45; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:54 CodeSize:23 Lat:46 SizeLat:30 for: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)46; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:108 CodeSize:46 Lat:92 SizeLat:60 for: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)47; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void48;49 %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)50 %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)51 %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)52 %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)53 ret void54}55 56define void @strict_fp_reductions_bf16() {57; CHECK-NOFP16-LABEL: 'strict_fp_reductions_bf16'58; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:54 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)59; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void60;61; CHECK-F16-LABEL: 'strict_fp_reductions_bf16'62; CHECK-F16-NEXT: Cost Model: Found costs of RThru:54 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)63; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void64;65; CHECK-BF16-LABEL: 'strict_fp_reductions_bf16'66; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:26 CodeSize:11 Lat:22 SizeLat:14 for: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)67; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void68;69 %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4f8(bfloat 0.0, <4 x bfloat> undef)70 ret void71}72 73define void @fast_fp_reductions() {74; CHECK-LABEL: 'fast_fp_reductions'75; CHECK-NEXT: Cost Model: Found costs of 1 for: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)76; CHECK-NEXT: Cost Model: Found costs of 1 for: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)77; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)78; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)79; CHECK-NEXT: Cost Model: Found costs of 3 for: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)80; CHECK-NEXT: Cost Model: Found costs of 3 for: %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)81; CHECK-NEXT: Cost Model: Found costs of 5 for: %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.000000e+00, <13 x float> undef)82; CHECK-NEXT: Cost Model: Found costs of 3 for: %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.000000e+00, <5 x float> undef)83; CHECK-NEXT: Cost Model: Found costs of 1 for: %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)84; CHECK-NEXT: Cost Model: Found costs of 1 for: %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)85; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)86; CHECK-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)87; CHECK-NEXT: Cost Model: Found costs of 4 for: %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.000000e+00, <7 x double> undef)88; CHECK-NEXT: Cost Model: Found costs of 8 for: %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.000000e+00, <9 x double> undef)89; CHECK-NEXT: Cost Model: Found costs of RThru:12 CodeSize:2 Lat:6 SizeLat:2 for: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)90; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void91;92 %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)93 %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)94 95 %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)96 %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)97 98 %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)99 %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)100 101 %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.0, <13 x float> undef)102 %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.0, <5 x float> undef)103 104 %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)105 %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)106 107 %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)108 %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)109 110 %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.0, <7 x double> undef)111 %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.0, <9 x double> undef)112 113 %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)114 115 ret void116}117 118define void @fast_fp_reductions_fp16() {119; CHECK-NOFP16-LABEL: 'fast_fp_reductions_fp16'120; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:4 Lat:6 SizeLat:4 for: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)121; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:4 Lat:6 SizeLat:4 for: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)122; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)123; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)124; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:48 CodeSize:27 Lat:33 SizeLat:27 for: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)125; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:48 CodeSize:27 Lat:33 SizeLat:27 for: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)126; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:72 CodeSize:44 Lat:52 SizeLat:44 for: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)127; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:72 CodeSize:44 Lat:52 SizeLat:44 for: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)128; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:203 CodeSize:35 Lat:41 SizeLat:35 for: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)129; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:248 CodeSize:35 Lat:41 SizeLat:35 for: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)130; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void131;132; CHECK-F16-LABEL: 'fast_fp_reductions_fp16'133; CHECK-F16-NEXT: Cost Model: Found costs of 2 for: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)134; CHECK-F16-NEXT: Cost Model: Found costs of 2 for: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)135; CHECK-F16-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)136; CHECK-F16-NEXT: Cost Model: Found costs of 2 for: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)137; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)138; CHECK-F16-NEXT: Cost Model: Found costs of 3 for: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)139; CHECK-F16-NEXT: Cost Model: Found costs of 4 for: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)140; CHECK-F16-NEXT: Cost Model: Found costs of 4 for: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)141; CHECK-F16-NEXT: Cost Model: Found costs of 4 for: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)142; CHECK-F16-NEXT: Cost Model: Found costs of 4 for: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)143; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void144;145; CHECK-BF16-LABEL: 'fast_fp_reductions_fp16'146; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:4 Lat:6 SizeLat:4 for: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)147; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:4 Lat:6 SizeLat:4 for: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)148; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)149; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)150; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:48 CodeSize:27 Lat:33 SizeLat:27 for: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)151; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:48 CodeSize:27 Lat:33 SizeLat:27 for: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)152; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:72 CodeSize:44 Lat:52 SizeLat:44 for: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)153; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:72 CodeSize:44 Lat:52 SizeLat:44 for: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)154; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:203 CodeSize:35 Lat:41 SizeLat:35 for: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)155; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:248 CodeSize:35 Lat:41 SizeLat:35 for: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)156; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void157;158 %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)159 %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)160 161 %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)162 %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)163 164 %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)165 %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)166 167 %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)168 %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)169 170 %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0.0, <11 x half> undef)171 %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0.0, <13 x half> undef)172 173 ret void174}175 176define void @fast_fp_reductions_bf16() {177; CHECK-NOFP16-LABEL: 'fast_fp_reductions_bf16'178; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:28 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)179; CHECK-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void180;181; CHECK-F16-LABEL: 'fast_fp_reductions_bf16'182; CHECK-F16-NEXT: Cost Model: Found costs of RThru:28 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)183; CHECK-F16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void184;185; CHECK-BF16-LABEL: 'fast_fp_reductions_bf16'186; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:14 CodeSize:8 Lat:12 SizeLat:8 for: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)187; CHECK-BF16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void188;189 %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4f8(bfloat -0.0, <4 x bfloat> undef)190 ret void191}192 193declare bfloat @llvm.vector.reduce.fadd.v4f8(bfloat, <4 x bfloat>)194declare fp128 @llvm.vector.reduce.fadd.v4f128(fp128, <4 x fp128>)195 196declare half @llvm.vector.reduce.fadd.v2f16(half, <2 x half>)197declare half @llvm.vector.reduce.fadd.v4f16(half, <4 x half>)198declare half @llvm.vector.reduce.fadd.v8f16(half, <8 x half>)199declare half @llvm.vector.reduce.fadd.v16f16(half, <16 x half>)200declare half @llvm.vector.reduce.fadd.v11f16(half, <11 x half>)201declare half @llvm.vector.reduce.fadd.v13f16(half, <13 x half>)202 203declare float @llvm.vector.reduce.fadd.v2f32(float, <2 x float>)204declare float @llvm.vector.reduce.fadd.v4f32(float, <4 x float>)205declare float @llvm.vector.reduce.fadd.v8f32(float, <8 x float>)206declare float @llvm.vector.reduce.fadd.v13f32(float, <13 x float>)207declare float @llvm.vector.reduce.fadd.v5f32(float, <5 x float>)208 209declare double @llvm.vector.reduce.fadd.v2f64(double, <2 x double>)210declare double @llvm.vector.reduce.fadd.v4f64(double, <4 x double>)211declare double @llvm.vector.reduce.fadd.v7f64(double, <7 x double>)212declare double @llvm.vector.reduce.fadd.v9f64(double, <9 x double>)213