1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
|
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE,SSE42
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512
;
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-- -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
define i32 @f32(i32 %arg) {
; SSE2-LABEL: 'f32'
; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; SSE42-LABEL: 'f32'
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
; SSE42-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)
; SSE42-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)
; SSE42-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)
; SSE42-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)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX1-LABEL: 'f32'
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX2-LABEL: 'f32'
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX512-LABEL: 'f32'
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.maxnum.f32(float undef, float undef)
; 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)
; 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)
; 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)
; 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)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%F32 = call float @llvm.maxnum.f32(float undef, float undef)
%V2F32 = call <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
%V4F32 = call <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
%V8F32 = call <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
%V16F32 = call <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
ret i32 undef
}
define i32 @f64(i32 %arg) {
; SSE2-LABEL: 'f64'
; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; SSE42-LABEL: 'f64'
; SSE42-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
; SSE42-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)
; SSE42-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)
; SSE42-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)
; SSE42-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)
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX1-LABEL: 'f64'
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX2-LABEL: 'f64'
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX512-LABEL: 'f64'
; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %f64 = call double @llvm.maxnum.f64(double undef, double undef)
; 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)
; 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)
; 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)
; 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)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%f64 = call double @llvm.maxnum.f64(double undef, double undef)
%V2f64 = call <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
%V4f64 = call <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
%V8f64 = call <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
%V16f64 = call <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
ret i32 undef
}
define i32 @f32_nnan(i32 %arg) {
; SSE-LABEL: 'f32_nnan'
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
; 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)
; 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)
; 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)
; 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)
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX-LABEL: 'f32_nnan'
; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
; 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)
; 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)
; 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)
; 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)
; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX512-LABEL: 'f32_nnan'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
; 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)
; 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)
; 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)
; 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)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%F32 = call nnan float @llvm.maxnum.f32(float undef, float undef)
%V2F32 = call nnan <2 x float> @llvm.maxnum.v2f32(<2 x float> undef, <2 x float> undef)
%V4F32 = call nnan <4 x float> @llvm.maxnum.v4f32(<4 x float> undef, <4 x float> undef)
%V8F32 = call nnan <8 x float> @llvm.maxnum.v8f32(<8 x float> undef, <8 x float> undef)
%V16F32 = call nnan <16 x float> @llvm.maxnum.v16f32(<16 x float> undef, <16 x float> undef)
ret i32 undef
}
define i32 @f64_nnan(i32 %arg) {
; SSE-LABEL: 'f64_nnan'
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
; 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)
; 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)
; 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)
; 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)
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX-LABEL: 'f64_nnan'
; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
; 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)
; 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)
; 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)
; 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)
; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
; AVX512-LABEL: 'f64_nnan'
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
; 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)
; 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)
; 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)
; 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)
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%f64 = call nnan double @llvm.maxnum.f64(double undef, double undef)
%V2f64 = call nnan <2 x double> @llvm.maxnum.v2f64(<2 x double> undef, <2 x double> undef)
%V4f64 = call nnan <4 x double> @llvm.maxnum.v4f64(<4 x double> undef, <4 x double> undef)
%V8f64 = call nnan <8 x double> @llvm.maxnum.v8f64(<8 x double> undef, <8 x double> undef)
%V16f64 = call nnan <16 x double> @llvm.maxnum.v16f64(<16 x double> undef, <16 x double> undef)
ret i32 undef
}
declare float @llvm.maxnum.f32(float, float)
declare <2 x float> @llvm.maxnum.v2f32(<2 x float>, <2 x float>)
declare <4 x float> @llvm.maxnum.v4f32(<4 x float>, <4 x float>)
declare <8 x float> @llvm.maxnum.v8f32(<8 x float>, <8 x float>)
declare <16 x float> @llvm.maxnum.v16f32(<16 x float>, <16 x float>)
declare double @llvm.maxnum.f64(double, double)
declare <2 x double> @llvm.maxnum.v2f64(<2 x double>, <2 x double>)
declare <4 x double> @llvm.maxnum.v4f64(<4 x double>, <4 x double>)
declare <8 x double> @llvm.maxnum.v8f64(<8 x double>, <8 x double>)
declare <16 x double> @llvm.maxnum.v16f64(<16 x double>, <16 x double>)
|
