1 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
2 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
3 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE
4 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
5 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
6 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512
7 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512
9 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mcpu=slm | FileCheck %s --check-prefixes=SSE
10 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mcpu=goldmont | FileCheck %s --check-prefixes=SSE
11 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-- -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
13 define i32 @f32(i32 %arg) {
15 ; SSE-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
16 ; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
17 ; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
18 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
19 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
20 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
23 ; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
24 ; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
25 ; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
26 ; AVX1-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
27 ; AVX1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
28 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
31 ; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
32 ; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
33 ; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
34 ; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
35 ; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
36 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
39 ; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
40 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
41 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
42 ; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
43 ; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
44 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
46 %F32 = call float @llvm.maxnum.f32(float undef, float undef)
47 %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
48 %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
49 %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
50 %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
54 define i32 @f64(i32 %arg) {
56 ; SSE-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
57 ; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
58 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
59 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
60 ; SSE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
61 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
64 ; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
65 ; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
66 ; AVX1-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
67 ; AVX1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
68 ; AVX1-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
69 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
72 ; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
73 ; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
74 ; AVX2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
75 ; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
76 ; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
77 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
80 ; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
81 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
82 ; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
83 ; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
84 ; AVX512-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
85 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
87 %f64 = call double @llvm.maxnum.f64(double undef, double undef)
88 %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
89 %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
90 %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
91 %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
95 define i32 @f32_nnan(i32 %arg) {
96 ; SSE-LABEL: 'f32_nnan'
97 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
98 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call nnan <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
99 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call nnan <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
100 ; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call nnan <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
101 ; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call nnan <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
102 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
104 ; AVX-LABEL: 'f32_nnan'
105 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
106 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call nnan <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
107 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call nnan <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
108 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = call nnan <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
109 ; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call nnan <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
110 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
112 ; AVX512-LABEL: 'f32_nnan'
113 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
114 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call nnan <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
115 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call nnan <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
116 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = call nnan <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
117 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call nnan <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
118 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
120 %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
121 %V2F32 = call nnan <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
122 %V4F32 = call nnan <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
123 %V8F32 = call nnan <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
124 %V16F32 = call nnan <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
128 define i32 @f64_nnan(i32 %arg) {
129 ; SSE-LABEL: 'f64_nnan'
130 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
131 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2f64 = call nnan <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
132 ; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4f64 = call nnan <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
133 ; SSE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8f64 = call nnan <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
134 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16f64 = call nnan <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
135 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
137 ; AVX-LABEL: 'f64_nnan'
138 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
139 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2f64 = call nnan <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
140 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4f64 = call nnan <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
141 ; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8f64 = call nnan <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
142 ; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16f64 = call nnan <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
143 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
145 ; AVX512-LABEL: 'f64_nnan'
146 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
147 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2f64 = call nnan <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
148 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4f64 = call nnan <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
149 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8f64 = call nnan <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
150 ; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16f64 = call nnan <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
151 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
153 %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
154 %V2f64 = call nnan <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
155 %V4f64 = call nnan <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
156 %V8f64 = call nnan <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
157 %V16f64 = call nnan <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
161 declare float @llvm.maxnum.f32(float, float)
162 declare <2 x float> @llvm.maxnum.v2f32(<2 x float>, <2 x float>)
163 declare <4 x float> @llvm.maxnum.v4f32(<4 x float>, <4 x float>)
164 declare <8 x float> @llvm.maxnum.v8f32(<8 x float>, <8 x float>)
165 declare <16 x float> @llvm.maxnum.v16f32(<16 x float>, <16 x float>)
167 declare double @llvm.maxnum.f64(double, double)
168 declare <2 x double> @llvm.maxnum.v2f64(<2 x double>, <2 x double>)
169 declare <4 x double> @llvm.maxnum.v4f64(<4 x double>, <4 x double>)
170 declare <8 x double> @llvm.maxnum.v8f64(<8 x double>, <8 x double>)
171 declare <16 x double> @llvm.maxnum.v16f64(<16 x double>, <16 x double>)