Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / 1d6ebdfb66b9d63d34f34ec6ac7ec57eff7cd24b / . / llvm / test / Analysis / CostModel / X86 / arith-fp.ll
blob: d5d47a99e91996f01644f5cd7e3e680ea05254a5 [file] [log] [blame]
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
Simon Pilgrime11195d2020-11-13 17:29:54 +0000[diff] [blame]2; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=-sse2 | FileCheck %s --check-prefixes=SSE1
3; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse2 | FileCheck %s --check-prefixes=SSE2
4; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE42
5; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
6; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
7; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512
8; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512
Simon Pilgrim34b397a2018-04-13 19:30:15 +0000[diff] [blame]9;
Simon Pilgrime11195d2020-11-13 17:29:54 +0000[diff] [blame]10; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=slm | FileCheck %s --check-prefixes=SLM
11; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
12; RUN: opt < %s -enable-no-nans-fp-math -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]13
14target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
15target triple = "x86_64-apple-macosx10.8.0"
16
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]17define i32 @fadd(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]18; SSE1-LABEL: 'fadd'
Simon Pilgrimc2054142019-01-04 16:55:57 +0000[diff] [blame]19; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]20; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fadd <4 x float> undef, undef
21; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fadd <8 x float> undef, undef
22; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fadd <16 x float> undef, undef
23; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fadd double undef, undef
24; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fadd <2 x double> undef, undef
25; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fadd <4 x double> undef, undef
26; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fadd <8 x double> undef, undef
27; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
28;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]29; SSE2-LABEL: 'fadd'
30; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fadd float undef, undef
31; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fadd <4 x float> undef, undef
32; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fadd <8 x float> undef, undef
33; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fadd <16 x float> undef, undef
34; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fadd double undef, undef
35; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> undef, undef
36; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> undef, undef
37; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> undef, undef
38; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
39;
40; SSE42-LABEL: 'fadd'
41; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
42; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
43; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
44; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
45; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
46; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
47; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
48; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
49; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
50;
51; AVX1-LABEL: 'fadd'
52; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
53; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
54; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
55; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
56; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
57; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
58; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
59; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
60; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
61;
62; AVX2-LABEL: 'fadd'
63; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
64; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
65; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fadd <8 x float> undef, undef
66; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fadd <16 x float> undef, undef
67; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
68; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
69; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fadd <4 x double> undef, undef
70; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fadd <8 x double> undef, undef
71; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
72;
73; AVX512-LABEL: 'fadd'
74; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
75; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
76; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fadd <8 x float> undef, undef
77; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fadd <16 x float> undef, undef
78; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
79; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
80; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fadd <4 x double> undef, undef
81; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fadd <8 x double> undef, undef
82; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
83;
84; SLM-LABEL: 'fadd'
85; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
86; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
87; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
88; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
89; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
90; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> undef, undef
91; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> undef, undef
92; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> undef, undef
93; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
94;
95; GLM-LABEL: 'fadd'
96; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
97; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
98; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
99; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
100; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
101; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
102; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
103; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
104; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
105;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]106 %F32 = fadd float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]107 %V4F32 = fadd <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]108 %V8F32 = fadd <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]109 %V16F32 = fadd <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]110
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]111 %F64 = fadd double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]112 %V2F64 = fadd <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]113 %V4F64 = fadd <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]114 %V8F64 = fadd <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]115
116 ret i32 undef
117}
118
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]119define i32 @fsub(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]120; SSE1-LABEL: 'fsub'
Simon Pilgrimc2054142019-01-04 16:55:57 +0000[diff] [blame]121; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]122; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
123; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
124; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
125; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
126; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fsub <2 x double> undef, undef
127; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fsub <4 x double> undef, undef
128; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fsub <8 x double> undef, undef
129; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
130;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]131; SSE2-LABEL: 'fsub'
132; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float undef, undef
133; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
134; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
135; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
136; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
137; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
138; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
139; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
140; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
141;
142; SSE42-LABEL: 'fsub'
143; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
144; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
145; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
146; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
147; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
148; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
149; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
150; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
151; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
152;
153; AVX1-LABEL: 'fsub'
154; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
155; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
156; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
157; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
158; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
159; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
160; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
161; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
162; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
163;
164; AVX2-LABEL: 'fsub'
165; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
166; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
167; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
168; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fsub <16 x float> undef, undef
169; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
170; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
171; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
172; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fsub <8 x double> undef, undef
173; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
174;
175; AVX512-LABEL: 'fsub'
176; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
177; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
178; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
179; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fsub <16 x float> undef, undef
180; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
181; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
182; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
183; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fsub <8 x double> undef, undef
184; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
185;
186; SLM-LABEL: 'fsub'
187; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
188; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
189; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
190; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
191; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
192; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
193; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
194; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
195; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
196;
197; GLM-LABEL: 'fsub'
198; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
199; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
200; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
201; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
202; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
203; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
204; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
205; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
206; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
207;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]208 %F32 = fsub float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]209 %V4F32 = fsub <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]210 %V8F32 = fsub <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]211 %V16F32 = fsub <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]212
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]213 %F64 = fsub double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]214 %V2F64 = fsub <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]215 %V4F64 = fsub <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]216 %V8F64 = fsub <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]217
218 ret i32 undef
219}
220
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]221define i32 @fneg_idiom(i32 %arg) {
222; SSE1-LABEL: 'fneg_idiom'
223; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
224; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
225; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
226; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
227; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double -0.000000e+00, undef
228; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
229; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
230; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
231; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
232;
233; SSE2-LABEL: 'fneg_idiom'
234; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float -0.000000e+00, undef
235; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
236; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
237; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
238; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double -0.000000e+00, undef
239; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
240; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
241; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
242; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
243;
244; SSE42-LABEL: 'fneg_idiom'
245; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
246; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
247; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
248; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
249; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
250; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
251; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
252; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
253; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
254;
255; AVX1-LABEL: 'fneg_idiom'
256; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
257; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
258; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
259; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
260; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
261; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
262; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
263; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
264; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
265;
266; AVX2-LABEL: 'fneg_idiom'
267; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
268; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
269; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
270; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
271; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
272; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
273; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
274; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
275; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
276;
277; AVX512-LABEL: 'fneg_idiom'
278; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
279; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
280; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
281; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
282; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
283; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
284; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
285; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
286; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
287;
288; SLM-LABEL: 'fneg_idiom'
289; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
290; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
291; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
292; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
293; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
294; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
295; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
296; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
297; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
298;
299; GLM-LABEL: 'fneg_idiom'
300; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
301; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
302; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
303; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
304; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double -0.000000e+00, undef
305; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
306; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
307; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
308; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
309;
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]310 %F32 = fsub float -0.0, undef
311 %V4F32 = fsub <4 x float> <float -0.0, float -0.0, float -0.0, float -0.0>, undef
312 %V8F32 = fsub <8 x float> <float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0>, undef
313 %V16F32 = fsub <16 x float> <float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0>, undef
314
315 %F64 = fsub double -0.0, undef
316 %V2F64 = fsub <2 x double> <double -0.0, double -0.0>, undef
317 %V4F64 = fsub <4 x double> <double -0.0, double -0.0, double -0.0, double -0.0>, undef
318 %V8F64 = fsub <8 x double> <double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0>, undef
319
320 ret i32 undef
321}
322
323define i32 @fneg(i32 %arg) {
Craig Topper50d50282019-05-28 04:09:18 +0000[diff] [blame]324; SSE1-LABEL: 'fneg'
325; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
326; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
327; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fneg <8 x float> undef
328; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = fneg <16 x float> undef
329; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fneg double undef
330; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
331; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fneg <4 x double> undef
332; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fneg <8 x double> undef
333; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
334;
335; SSE2-LABEL: 'fneg'
336; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
337; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
338; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fneg <8 x float> undef
339; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = fneg <16 x float> undef
340; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
341; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
342; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fneg <4 x double> undef
343; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fneg <8 x double> undef
344; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
345;
346; SSE42-LABEL: 'fneg'
347; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
348; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
349; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fneg <8 x float> undef
350; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = fneg <16 x float> undef
351; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
352; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
353; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fneg <4 x double> undef
354; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fneg <8 x double> undef
355; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
356;
357; AVX-LABEL: 'fneg'
358; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
359; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
360; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fneg <8 x float> undef
361; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fneg <16 x float> undef
362; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
363; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
364; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fneg <4 x double> undef
365; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fneg <8 x double> undef
366; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
367;
368; AVX512-LABEL: 'fneg'
369; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
370; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
371; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fneg <8 x float> undef
372; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fneg <16 x float> undef
373; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
374; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
375; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fneg <4 x double> undef
376; AVX512-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fneg <8 x double> undef
377; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]378;
379; SLM-LABEL: 'fneg'
Craig Topper50d50282019-05-28 04:09:18 +0000[diff] [blame]380; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
381; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
382; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fneg <8 x float> undef
383; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = fneg <16 x float> undef
384; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
385; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
386; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fneg <4 x double> undef
387; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fneg <8 x double> undef
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]388; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
389;
390; GLM-LABEL: 'fneg'
Craig Topper50d50282019-05-28 04:09:18 +0000[diff] [blame]391; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F32 = fneg float undef
392; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fneg <4 x float> undef
393; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fneg <8 x float> undef
394; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = fneg <16 x float> undef
395; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fneg double undef
396; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fneg <2 x double> undef
397; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fneg <4 x double> undef
398; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fneg <8 x double> undef
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]399; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
400;
Craig Topperba883e92019-05-27 05:27:57 +0000[diff] [blame]401 %F32 = fneg float undef
402 %V4F32 = fneg <4 x float> undef
403 %V8F32 = fneg <8 x float> undef
404 %V16F32 = fneg <16 x float> undef
405
406 %F64 = fneg double undef
407 %V2F64 = fneg <2 x double> undef
408 %V4F64 = fneg <4 x double> undef
409 %V8F64 = fneg <8 x double> undef
410
411 ret i32 undef
412}
413
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]414define i32 @fmul(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]415; SSE1-LABEL: 'fmul'
416; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
417; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
418; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
419; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
420; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
421; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fmul <2 x double> undef, undef
422; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fmul <4 x double> undef, undef
423; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fmul <8 x double> undef, undef
424; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
425;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]426; SSE2-LABEL: 'fmul'
427; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
428; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
429; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
430; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
431; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
432; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fmul <2 x double> undef, undef
433; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fmul <4 x double> undef, undef
434; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fmul <8 x double> undef, undef
435; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
436;
437; SSE42-LABEL: 'fmul'
438; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
439; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
440; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
441; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
442; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
443; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
444; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
445; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
446; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
447;
448; AVX1-LABEL: 'fmul'
449; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
450; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
451; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
452; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
453; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
454; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
455; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
456; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
457; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
458;
459; AVX2-LABEL: 'fmul'
460; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
461; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
462; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
463; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fmul <16 x float> undef, undef
464; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
465; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
466; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
467; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fmul <8 x double> undef, undef
468; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
469;
470; AVX512-LABEL: 'fmul'
471; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
472; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
473; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
474; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fmul <16 x float> undef, undef
475; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
476; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
477; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
478; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fmul <8 x double> undef, undef
479; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
480;
481; SLM-LABEL: 'fmul'
482; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
483; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
484; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
485; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
486; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
487; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fmul <2 x double> undef, undef
488; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fmul <4 x double> undef, undef
489; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fmul <8 x double> undef, undef
490; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
491;
492; GLM-LABEL: 'fmul'
493; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
494; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
495; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
496; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
497; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
498; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
499; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
500; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
501; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
502;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]503 %F32 = fmul float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]504 %V4F32 = fmul <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]505 %V8F32 = fmul <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]506 %V16F32 = fmul <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]507
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]508 %F64 = fmul double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]509 %V2F64 = fmul <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]510 %V4F64 = fmul <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]511 %V8F64 = fmul <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]512
513 ret i32 undef
514}
515
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]516define i32 @fdiv(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]517; SSE1-LABEL: 'fdiv'
518; SSE1-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %F32 = fdiv float undef, undef
519; SSE1-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V4F32 = fdiv <4 x float> undef, undef
520; SSE1-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V8F32 = fdiv <8 x float> undef, undef
521; SSE1-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V16F32 = fdiv <16 x float> undef, undef
522; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fdiv double undef, undef
523; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fdiv <2 x double> undef, undef
524; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fdiv <4 x double> undef, undef
525; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fdiv <8 x double> undef, undef
526; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
527;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]528; SSE2-LABEL: 'fdiv'
529; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %F32 = fdiv float undef, undef
530; SSE2-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
531; SSE2-NEXT: Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
532; SSE2-NEXT: Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
533; SSE2-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %F64 = fdiv double undef, undef
534; SSE2-NEXT: Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
535; SSE2-NEXT: Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
536; SSE2-NEXT: Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
537; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
538;
539; SSE42-LABEL: 'fdiv'
540; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
541; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
542; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
543; SSE42-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
544; SSE42-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
545; SSE42-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
546; SSE42-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
547; SSE42-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
548; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
549;
550; AVX1-LABEL: 'fdiv'
551; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
552; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
553; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
554; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
555; AVX1-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
556; AVX1-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
557; AVX1-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
558; AVX1-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
559; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
560;
561; AVX2-LABEL: 'fdiv'
562; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
563; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
564; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
565; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = fdiv <16 x float> undef, undef
566; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
567; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
568; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
569; AVX2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = fdiv <8 x double> undef, undef
570; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
571;
572; AVX512-LABEL: 'fdiv'
573; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
574; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
575; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
576; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fdiv <16 x float> undef, undef
577; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
578; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
579; AVX512-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
580; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fdiv <8 x double> undef, undef
581; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
582;
583; SLM-LABEL: 'fdiv'
584; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %F32 = fdiv float undef, undef
585; SLM-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
586; SLM-NEXT: Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
587; SLM-NEXT: Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
588; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = fdiv double undef, undef
589; SLM-NEXT: Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
590; SLM-NEXT: Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
591; SLM-NEXT: Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
592; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
593;
594; GLM-LABEL: 'fdiv'
595; GLM-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %F32 = fdiv float undef, undef
596; GLM-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %V4F32 = fdiv <4 x float> undef, undef
597; GLM-NEXT: Cost Model: Found an estimated cost of 70 for instruction: %V8F32 = fdiv <8 x float> undef, undef
598; GLM-NEXT: Cost Model: Found an estimated cost of 140 for instruction: %V16F32 = fdiv <16 x float> undef, undef
599; GLM-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %F64 = fdiv double undef, undef
600; GLM-NEXT: Cost Model: Found an estimated cost of 65 for instruction: %V2F64 = fdiv <2 x double> undef, undef
601; GLM-NEXT: Cost Model: Found an estimated cost of 130 for instruction: %V4F64 = fdiv <4 x double> undef, undef
602; GLM-NEXT: Cost Model: Found an estimated cost of 260 for instruction: %V8F64 = fdiv <8 x double> undef, undef
603; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
604;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]605 %F32 = fdiv float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]606 %V4F32 = fdiv <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]607 %V8F32 = fdiv <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]608 %V16F32 = fdiv <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]609
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]610 %F64 = fdiv double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]611 %V2F64 = fdiv <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]612 %V4F64 = fdiv <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]613 %V8F64 = fdiv <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]614
615 ret i32 undef
616}
617
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]618define i32 @frem(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]619; SSE1-LABEL: 'frem'
620; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]621; SSE1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
622; SSE1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
623; SSE1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]624; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
625; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = frem <2 x double> undef, undef
626; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = frem <4 x double> undef, undef
627; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = frem <8 x double> undef, undef
628; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
629;
630; SSE2-LABEL: 'frem'
631; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]632; SSE2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
633; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
634; SSE2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]635; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
636; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
637; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
638; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
639; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
640;
641; SSE42-LABEL: 'frem'
642; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
643; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
644; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
645; SSE42-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
646; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
647; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
648; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
649; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
650; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]651;
652; AVX-LABEL: 'frem'
653; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
654; AVX-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
Bing1 Yu2c08f1b2020-10-27 10:34:07 +0800[diff] [blame]655; AVX-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V8F32 = frem <8 x float> undef, undef
656; AVX-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]657; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
658; AVX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]659; AVX-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %V4F64 = frem <4 x double> undef, undef
660; AVX-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V8F64 = frem <8 x double> undef, undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]661; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
662;
663; AVX512-LABEL: 'frem'
664; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
665; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
Bing1 Yu2c08f1b2020-10-27 10:34:07 +0800[diff] [blame]666; AVX512-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V8F32 = frem <8 x float> undef, undef
667; AVX512-NEXT: Cost Model: Found an estimated cost of 72 for instruction: %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]668; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
669; AVX512-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]670; AVX512-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %V4F64 = frem <4 x double> undef, undef
671; AVX512-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %V8F64 = frem <8 x double> undef, undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]672; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
673;
674; SLM-LABEL: 'frem'
675; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
676; SLM-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
677; SLM-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
678; SLM-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
679; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
680; SLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
681; SLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
682; SLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
683; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
684;
685; GLM-LABEL: 'frem'
686; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
687; GLM-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
688; GLM-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
689; GLM-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
690; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
691; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
692; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
693; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
694; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
695;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]696 %F32 = frem float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]697 %V4F32 = frem <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]698 %V8F32 = frem <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]699 %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]700
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]701 %F64 = frem double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]702 %V2F64 = frem <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]703 %V4F64 = frem <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]704 %V8F64 = frem <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]705
706 ret i32 undef
707}
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]708
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]709define i32 @fsqrt(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]710; SSE1-LABEL: 'fsqrt'
711; SSE1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
712; SSE1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
713; SSE1-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
714; SSE1-NEXT: Cost Model: Found an estimated cost of 224 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
715; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
716; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
717; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
718; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
719; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
720;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]721; SSE2-LABEL: 'fsqrt'
722; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
723; SSE2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
724; SSE2-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
725; SSE2-NEXT: Cost Model: Found an estimated cost of 224 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
726; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
727; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
728; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
729; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
730; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
731;
732; SSE42-LABEL: 'fsqrt'
733; SSE42-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
734; SSE42-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
735; SSE42-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
736; SSE42-NEXT: Cost Model: Found an estimated cost of 72 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
737; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
738; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
739; SSE42-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
740; SSE42-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
741; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
742;
743; AVX1-LABEL: 'fsqrt'
744; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
745; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
746; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
747; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
748; AVX1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
749; AVX1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
750; AVX1-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
751; AVX1-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
752; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
753;
754; AVX2-LABEL: 'fsqrt'
755; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
756; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
757; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
758; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
759; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
760; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
761; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
762; AVX2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
763; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
764;
765; AVX512-LABEL: 'fsqrt'
766; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
767; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
768; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
769; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
770; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
771; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
772; AVX512-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
773; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
774; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
775;
776; SLM-LABEL: 'fsqrt'
777; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
778; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
779; SLM-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
780; SLM-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
781; SLM-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
782; SLM-NEXT: Cost Model: Found an estimated cost of 70 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
783; SLM-NEXT: Cost Model: Found an estimated cost of 140 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
784; SLM-NEXT: Cost Model: Found an estimated cost of 280 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
785; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
786;
787; GLM-LABEL: 'fsqrt'
788; GLM-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
789; GLM-NEXT: Cost Model: Found an estimated cost of 37 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
790; GLM-NEXT: Cost Model: Found an estimated cost of 74 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
791; GLM-NEXT: Cost Model: Found an estimated cost of 148 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
792; GLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
793; GLM-NEXT: Cost Model: Found an estimated cost of 67 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
794; GLM-NEXT: Cost Model: Found an estimated cost of 134 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
795; GLM-NEXT: Cost Model: Found an estimated cost of 268 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
796; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
797;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]798 %F32 = call float @llvm.sqrt.f32(float undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]799 %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]800 %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]801 %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]802
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]803 %F64 = call double @llvm.sqrt.f64(double undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]804 %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]805 %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]806 %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]807
808 ret i32 undef
809}
810
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]811define i32 @fabs(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]812; SSE1-LABEL: 'fabs'
813; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
814; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
815; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
816; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
817; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
818; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
819; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
820; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
821; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
822;
823; SSE2-LABEL: 'fabs'
824; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
825; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
826; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
827; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
828; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
829; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
830; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
831; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
832; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
833;
834; SSE42-LABEL: 'fabs'
835; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
836; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
837; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
838; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
839; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
840; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
841; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
842; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
843; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]844;
845; AVX-LABEL: 'fabs'
846; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
847; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
848; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
849; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
850; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
851; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
852; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
853; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
854; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
855;
856; AVX512-LABEL: 'fabs'
857; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
858; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
859; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
860; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
861; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
862; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
863; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
864; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
865; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
866;
867; SLM-LABEL: 'fabs'
868; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
869; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
870; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
871; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
872; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
873; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
874; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
875; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
876; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
877;
878; GLM-LABEL: 'fabs'
879; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
880; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
881; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
882; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
883; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
884; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
885; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
886; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
887; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
888;
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]889 %F32 = call float @llvm.fabs.f32(float undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]890 %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]891 %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]892 %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
893
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]894 %F64 = call double @llvm.fabs.f64(double undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]895 %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]896 %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]897 %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
898
899 ret i32 undef
900}
901
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]902define i32 @fcopysign(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]903; SSE1-LABEL: 'fcopysign'
904; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
905; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
906; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
907; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
908; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
909; SSE1-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
910; SSE1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
911; SSE1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
912; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
913;
914; SSE2-LABEL: 'fcopysign'
915; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
916; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
917; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
918; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
919; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
920; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
921; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
922; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
923; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
924;
925; SSE42-LABEL: 'fcopysign'
926; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
927; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
928; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
929; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
930; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
931; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
932; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
933; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
934; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]935;
936; AVX-LABEL: 'fcopysign'
937; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
938; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
939; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
940; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
941; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
942; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
943; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
944; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
945; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
946;
947; AVX512-LABEL: 'fcopysign'
948; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
949; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
950; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
951; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
952; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
953; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
954; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
955; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
956; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
957;
958; SLM-LABEL: 'fcopysign'
959; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
960; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
961; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
962; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
963; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
964; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
965; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
966; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
967; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
968;
969; GLM-LABEL: 'fcopysign'
970; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
971; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
972; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
973; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
974; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
975; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
976; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
977; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
978; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
979;
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]980 %F32 = call float @llvm.copysign.f32(float undef, float undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]981 %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]982 %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]983 %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
984
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]985 %F64 = call double @llvm.copysign.f64(double undef, double undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]986 %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]987 %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]988 %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
989
990 ret i32 undef
991}
992
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]993define i32 @fma(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]994; SSE1-LABEL: 'fma'
995; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]996; SSE1-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
997; SSE1-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
998; SSE1-NEXT: Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]999; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1000; SSE1-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1001; SSE1-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1002; SSE1-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1003; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1004;
1005; SSE2-LABEL: 'fma'
1006; SSE2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]1007; SSE2-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1008; SSE2-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1009; SSE2-NEXT: Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame]1010; SSE2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1011; SSE2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1012; SSE2-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1013; SSE2-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1014; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1015;
1016; SSE42-LABEL: 'fma'
1017; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
1018; SSE42-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1019; SSE42-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1020; SSE42-NEXT: Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1021; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1022; SSE42-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1023; SSE42-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1024; SSE42-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1025; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]1026;
1027; AVX-LABEL: 'fma'
Simon Pilgrim3ede11b2018-04-13 19:12:32 +0000[diff] [blame]1028; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
1029; AVX-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
Bing1 Yu2c08f1b2020-10-27 10:34:07 +0800[diff] [blame]1030; AVX-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1031; AVX-NEXT: Cost Model: Found an estimated cost of 176 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
Simon Pilgrim3ede11b2018-04-13 19:12:32 +0000[diff] [blame]1032; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1033; AVX-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
Simon Pilgrim090cae82020-04-29 11:39:13 +0100[diff] [blame]1034; AVX-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1035; AVX-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]1036; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1037;
1038; AVX512-LABEL: 'fma'
1039; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
1040; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1041; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1042; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1043; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1044; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1045; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1046; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1047; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1048;
1049; SLM-LABEL: 'fma'
1050; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
1051; SLM-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1052; SLM-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1053; SLM-NEXT: Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1054; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1055; SLM-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1056; SLM-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1057; SLM-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1058; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1059;
1060; GLM-LABEL: 'fma'
1061; GLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
1062; GLM-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
1063; GLM-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
1064; GLM-NEXT: Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
1065; GLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
1066; GLM-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
1067; GLM-NEXT: Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
1068; GLM-NEXT: Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
1069; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
1070;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1071 %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1072 %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1073 %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1074 %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1075
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1076 %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1077 %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1078 %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1079 %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1080
1081 ret i32 undef
1082}
1083
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1084declare float @llvm.sqrt.f32(float)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1085declare <4 x float> @llvm.sqrt.v4f32(<4 x float>)
1086declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
1087declare <16 x float> @llvm.sqrt.v16f32(<16 x float>)
1088
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1089declare double @llvm.sqrt.f64(double)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1090declare <2 x double> @llvm.sqrt.v2f64(<2 x double>)
1091declare <4 x double> @llvm.sqrt.v4f64(<4 x double>)
1092declare <8 x double> @llvm.sqrt.v8f64(<8 x double>)
1093
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]1094declare float @llvm.fabs.f32(float)
1095declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
1096declare <8 x float> @llvm.fabs.v8f32(<8 x float>)
1097declare <16 x float> @llvm.fabs.v16f32(<16 x float>)
1098
1099declare double @llvm.fabs.f64(double)
1100declare <2 x double> @llvm.fabs.v2f64(<2 x double>)
1101declare <4 x double> @llvm.fabs.v4f64(<4 x double>)
1102declare <8 x double> @llvm.fabs.v8f64(<8 x double>)
1103
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]1104declare float @llvm.copysign.f32(float, float)
1105declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>)
1106declare <8 x float> @llvm.copysign.v8f32(<8 x float>, <8 x float>)
1107declare <16 x float> @llvm.copysign.v16f32(<16 x float>, <16 x float>)
1108
1109declare double @llvm.copysign.f64(double, double)
1110declare <2 x double> @llvm.copysign.v2f64(<2 x double>, <2 x double>)
1111declare <4 x double> @llvm.copysign.v4f64(<4 x double>, <4 x double>)
1112declare <8 x double> @llvm.copysign.v8f64(<8 x double>, <8 x double>)
1113
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1114declare float @llvm.fma.f32(float, float, float)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1115declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
1116declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
1117declare <16 x float> @llvm.fma.v16f32(<16 x float>, <16 x float>, <16 x float>)
1118
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1119declare double @llvm.fma.f64(double, double, double)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1120declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
1121declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
1122declare <8 x double> @llvm.fma.v8f64(<8 x double>, <8 x double>, <8 x double>)