xref: /llvm-project/llvm/test/Analysis/CostModel/AArch64/reduce-fadd.ll (revision c5672e21ca2a16ff18cdaa83db11d2edb84c5e14)
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
2; RUN: opt -passes='print<cost-model>' 2>&1 -disable-output -mtriple=aarch64--linux-gnu < %s | FileCheck %s
3; RUN: opt -passes='print<cost-model>' 2>&1 -disable-output -mtriple=aarch64--linux-gnu -mattr=+fullfp16 < %s | FileCheck %s --check-prefix=FP16
4; RUN: opt -passes='print<cost-model>' 2>&1 -disable-output -mtriple=aarch64--linux-gnu -mattr=+bf16 < %s | FileCheck %s --check-prefix=BF16
5
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 an estimated cost of 8 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
11; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
12; CHECK-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
13; CHECK-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
14; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
15; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
16; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
17; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
18; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
19; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)
20; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
21; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
22;
23; FP16-LABEL: 'strict_fp_reductions'
24; FP16-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
25; FP16-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
26; FP16-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
27; FP16-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
28; FP16-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
29; FP16-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
30; FP16-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
31; FP16-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
32; FP16-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
33; FP16-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)
34; FP16-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
35; FP16-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
36;
37; BF16-LABEL: 'strict_fp_reductions'
38; BF16-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
39; BF16-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
40; BF16-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
41; BF16-NEXT:  Cost Model: Found an estimated cost of 76 for instruction: %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
42; BF16-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
43; BF16-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
44; BF16-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
45; BF16-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
46; BF16-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
47; BF16-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR0000, <4 x bfloat> undef)
48; BF16-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
49; BF16-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
50;
51  %fadd_v2f16 = call half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
52  %fadd_v4f16 = call half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
53  %fadd_v8f16 = call half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
54  %fadd_v16f16 = call half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
55  %fadd_v2f32 = call float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
56  %fadd_v4f32 = call float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
57  %fadd_v8f32 = call float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)
58  %fadd_v2f64 = call double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)
59  %fadd_v4f64 = call double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)
60  %fadd_v4f8 = call bfloat @llvm.vector.reduce.fadd.v4f8(bfloat 0.0, <4 x bfloat> undef)
61  %fadd_v4f128 = call fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
62
63  ret void
64}
65
66
67define void @fast_fp_reductions() {
68; CHECK-LABEL: 'fast_fp_reductions'
69; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
70; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
71; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
72; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
73; CHECK-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
74; CHECK-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
75; CHECK-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
76; CHECK-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
77; CHECK-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
78; CHECK-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
79; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
80; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
81; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
82; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
83; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
84; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
85; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.000000e+00, <13 x float> undef)
86; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.000000e+00, <5 x float> undef)
87; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
88; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
89; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
90; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
91; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.000000e+00, <7 x double> undef)
92; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.000000e+00, <9 x double> undef)
93; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)
94; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
95; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
96;
97; FP16-LABEL: 'fast_fp_reductions'
98; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
99; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
100; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
101; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
102; FP16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
103; FP16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
104; FP16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
105; FP16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
106; FP16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
107; FP16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
108; FP16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
109; FP16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
110; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
111; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
112; FP16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
113; FP16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
114; FP16-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.000000e+00, <13 x float> undef)
115; FP16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.000000e+00, <5 x float> undef)
116; FP16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
117; FP16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
118; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
119; FP16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
120; FP16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.000000e+00, <7 x double> undef)
121; FP16-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.000000e+00, <9 x double> undef)
122; FP16-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)
123; FP16-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
124; FP16-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
125;
126; BF16-LABEL: 'fast_fp_reductions'
127; BF16-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
128; BF16-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> undef)
129; BF16-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
130; BF16-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0xH0000, <4 x half> undef)
131; BF16-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
132; BF16-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0xH0000, <8 x half> undef)
133; BF16-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
134; BF16-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0xH0000, <16 x half> undef)
135; BF16-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0xH0000, <11 x half> undef)
136; BF16-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0xH0000, <13 x half> undef)
137; BF16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
138; BF16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> undef)
139; BF16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
140; BF16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> undef)
141; BF16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
142; BF16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.000000e+00, <8 x float> undef)
143; BF16-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.000000e+00, <13 x float> undef)
144; BF16-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.000000e+00, <5 x float> undef)
145; BF16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
146; BF16-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> undef)
147; BF16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
148; BF16-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> undef)
149; BF16-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.000000e+00, <7 x double> undef)
150; BF16-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.000000e+00, <9 x double> undef)
151; BF16-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4bf16(bfloat 0xR8000, <4 x bfloat> undef)
152; BF16-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
153; BF16-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
154;
155  %fadd_v2f16_fast = call fast half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
156  %fadd_v2f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v2f16(half 0.0, <2 x half> undef)
157
158  %fadd_v4f16_fast = call fast half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
159  %fadd_v4f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v4f16(half 0.0, <4 x half> undef)
160
161  %fadd_v8f16 = call fast half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
162  %fadd_v8f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v8f16(half 0.0, <8 x half> undef)
163
164  %fadd_v16f16 = call fast half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
165  %fadd_v16f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v16f16(half 0.0, <16 x half> undef)
166
167  %fadd_v11f16 = call fast half @llvm.vector.reduce.fadd.v11f16(half 0.0, <11 x half> undef)
168  %fadd_v13f16_reassoc = call reassoc half @llvm.vector.reduce.fadd.v13f16(half 0.0, <13 x half> undef)
169
170  %fadd_v2f32 = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
171  %fadd_v2f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v2f32(float 0.0, <2 x float> undef)
172
173  %fadd_v4f32 = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
174  %fadd_v4f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v4f32(float 0.0, <4 x float> undef)
175
176  %fadd_v8f32 = call fast float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)
177  %fadd_v8f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v8f32(float 0.0, <8 x float> undef)
178
179  %fadd_v13f32 = call fast float @llvm.vector.reduce.fadd.v13f32(float 0.0, <13 x float> undef)
180  %fadd_v5f32_reassoc = call reassoc float @llvm.vector.reduce.fadd.v5f32(float 0.0, <5 x float> undef)
181
182  %fadd_v2f64 = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)
183  %fadd_v2f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v2f64(double 0.0, <2 x double> undef)
184
185  %fadd_v4f64 = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)
186  %fadd_v4f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v4f64(double 0.0, <4 x double> undef)
187
188  %fadd_v7f64 = call fast double @llvm.vector.reduce.fadd.v7f64(double 0.0, <7 x double> undef)
189  %fadd_v9f64_reassoc = call reassoc double @llvm.vector.reduce.fadd.v9f64(double 0.0, <9 x double> undef)
190
191  %fadd_v4f8 = call reassoc bfloat @llvm.vector.reduce.fadd.v4f8(bfloat -0.0, <4 x bfloat> undef)
192  %fadd_v4f128 = call reassoc fp128 @llvm.vector.reduce.fadd.v4f128(fp128 undef, <4 x fp128> undef)
193
194  ret void
195}
196
197declare bfloat @llvm.vector.reduce.fadd.v4f8(bfloat, <4 x bfloat>)
198declare fp128 @llvm.vector.reduce.fadd.v4f128(fp128, <4 x fp128>)
199
200declare half @llvm.vector.reduce.fadd.v2f16(half, <2 x half>)
201declare half @llvm.vector.reduce.fadd.v4f16(half, <4 x half>)
202declare half @llvm.vector.reduce.fadd.v8f16(half, <8 x half>)
203declare half @llvm.vector.reduce.fadd.v16f16(half, <16 x half>)
204declare half @llvm.vector.reduce.fadd.v11f16(half, <11 x half>)
205declare half @llvm.vector.reduce.fadd.v13f16(half, <13 x half>)
206
207declare float @llvm.vector.reduce.fadd.v2f32(float, <2 x float>)
208declare float @llvm.vector.reduce.fadd.v4f32(float, <4 x float>)
209declare float @llvm.vector.reduce.fadd.v8f32(float, <8 x float>)
210declare float @llvm.vector.reduce.fadd.v13f32(float, <13 x float>)
211declare float @llvm.vector.reduce.fadd.v5f32(float, <5 x float>)
212
213declare double @llvm.vector.reduce.fadd.v2f64(double, <2 x double>)
214declare double @llvm.vector.reduce.fadd.v4f64(double, <4 x double>)
215declare double @llvm.vector.reduce.fadd.v7f64(double, <7 x double>)
216declare double @llvm.vector.reduce.fadd.v9f64(double, <9 x double>)
217