Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / 71d61567c059f5e91d623805e00f9b4ecf966186 / . / llvm / test / Analysis / CostModel / X86 / arith-fp.ll
blob: bd62bd9dc18a496f23f06d96dc086696e1980864 [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 Pilgrim71d61562019-01-04 16:37:01 +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=CHECK,SSE,SSE1
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]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=CHECK,SSE,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=CHECK,SSE,SSE42
Simon Pilgrim3ede11b2018-04-13 19:12:32 +0000[diff] [blame]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=CHECK,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=CHECK,AVX,AVX2
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]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=CHECK,AVX512,AVX512F
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=CHECK,AVX512,AVX512BW
Simon Pilgrim34b397a2018-04-13 19:30:15 +0000[diff] [blame]9;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +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=BTVER2
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'
19; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fadd float undef, undef
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 Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]106; BTVER2-LABEL: 'fadd'
107; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
108; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
109; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
110; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
111; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
112; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
113; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
114; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
115; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
116;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]117 %F32 = fadd float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]118 %V4F32 = fadd <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]119 %V8F32 = fadd <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]120 %V16F32 = fadd <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]121
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]122 %F64 = fadd double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]123 %V2F64 = fadd <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]124 %V4F64 = fadd <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]125 %V8F64 = fadd <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]126
127 ret i32 undef
128}
129
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]130define i32 @fsub(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]131; SSE1-LABEL: 'fsub'
132; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float undef, undef
133; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
134; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
135; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
136; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
137; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fsub <2 x double> undef, undef
138; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fsub <4 x double> undef, undef
139; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fsub <8 x double> undef, undef
140; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
141;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]142; SSE2-LABEL: 'fsub'
143; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float undef, undef
144; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
145; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
146; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
147; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
148; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
149; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
150; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
151; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
152;
153; SSE42-LABEL: 'fsub'
154; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
155; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
156; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
157; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
158; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
159; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
160; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
161; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
162; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
163;
164; AVX1-LABEL: 'fsub'
165; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
166; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
167; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
168; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
169; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
170; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
171; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
172; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
173; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
174;
175; AVX2-LABEL: 'fsub'
176; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
177; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
178; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
179; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fsub <16 x float> undef, undef
180; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
181; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
182; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
183; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fsub <8 x double> undef, undef
184; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
185;
186; AVX512-LABEL: 'fsub'
187; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
188; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
189; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
190; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fsub <16 x float> undef, undef
191; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
192; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
193; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
194; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fsub <8 x double> undef, undef
195; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
196;
197; SLM-LABEL: 'fsub'
198; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
199; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
200; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
201; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
202; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
203; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
204; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
205; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
206; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
207;
208; GLM-LABEL: 'fsub'
209; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
210; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
211; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
212; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
213; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
214; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
215; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
216; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
217; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
218;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]219; BTVER2-LABEL: 'fsub'
220; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
221; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
222; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
223; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
224; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
225; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
226; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
227; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
228; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
229;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]230 %F32 = fsub float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]231 %V4F32 = fsub <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]232 %V8F32 = fsub <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]233 %V16F32 = fsub <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]234
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]235 %F64 = fsub double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]236 %V2F64 = fsub <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]237 %V4F64 = fsub <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]238 %V8F64 = fsub <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]239
240 ret i32 undef
241}
242
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]243define i32 @fmul(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]244; SSE1-LABEL: 'fmul'
245; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
246; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
247; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
248; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
249; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
250; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fmul <2 x double> undef, undef
251; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fmul <4 x double> undef, undef
252; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fmul <8 x double> undef, undef
253; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
254;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]255; SSE2-LABEL: 'fmul'
256; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
257; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
258; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
259; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
260; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
261; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fmul <2 x double> undef, undef
262; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fmul <4 x double> undef, undef
263; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fmul <8 x double> undef, undef
264; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
265;
266; SSE42-LABEL: 'fmul'
267; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
268; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
269; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
270; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
271; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
272; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
273; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
274; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
275; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
276;
277; AVX1-LABEL: 'fmul'
278; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
279; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
280; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
281; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
282; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
283; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
284; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
285; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
286; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
287;
288; AVX2-LABEL: 'fmul'
289; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
290; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
291; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
292; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fmul <16 x float> undef, undef
293; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
294; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
295; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
296; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fmul <8 x double> undef, undef
297; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
298;
299; AVX512-LABEL: 'fmul'
300; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
301; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
302; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
303; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fmul <16 x float> undef, undef
304; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
305; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
306; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
307; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fmul <8 x double> undef, undef
308; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
309;
310; SLM-LABEL: 'fmul'
311; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
312; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
313; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
314; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
315; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
316; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fmul <2 x double> undef, undef
317; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fmul <4 x double> undef, undef
318; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fmul <8 x double> undef, undef
319; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
320;
321; GLM-LABEL: 'fmul'
322; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
323; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
324; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
325; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
326; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
327; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
328; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
329; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
330; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
331;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]332; BTVER2-LABEL: 'fmul'
333; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
334; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
335; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
336; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
337; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
338; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
339; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
340; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
341; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
342;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]343 %F32 = fmul float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]344 %V4F32 = fmul <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]345 %V8F32 = fmul <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]346 %V16F32 = fmul <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]347
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]348 %F64 = fmul double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]349 %V2F64 = fmul <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]350 %V4F64 = fmul <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]351 %V8F64 = fmul <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]352
353 ret i32 undef
354}
355
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]356define i32 @fdiv(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]357; SSE1-LABEL: 'fdiv'
358; SSE1-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %F32 = fdiv float undef, undef
359; SSE1-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %V4F32 = fdiv <4 x float> undef, undef
360; SSE1-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V8F32 = fdiv <8 x float> undef, undef
361; SSE1-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V16F32 = fdiv <16 x float> undef, undef
362; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fdiv double undef, undef
363; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fdiv <2 x double> undef, undef
364; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fdiv <4 x double> undef, undef
365; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fdiv <8 x double> undef, undef
366; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
367;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]368; SSE2-LABEL: 'fdiv'
369; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %F32 = fdiv float undef, undef
370; SSE2-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
371; SSE2-NEXT: Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
372; SSE2-NEXT: Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
373; SSE2-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %F64 = fdiv double undef, undef
374; SSE2-NEXT: Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
375; SSE2-NEXT: Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
376; SSE2-NEXT: Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
377; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
378;
379; SSE42-LABEL: 'fdiv'
380; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
381; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
382; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
383; SSE42-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
384; SSE42-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
385; SSE42-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
386; SSE42-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
387; SSE42-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
388; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
389;
390; AVX1-LABEL: 'fdiv'
391; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
392; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
393; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
394; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
395; AVX1-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
396; AVX1-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
397; AVX1-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
398; AVX1-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
399; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
400;
401; AVX2-LABEL: 'fdiv'
402; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
403; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
404; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
405; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = fdiv <16 x float> undef, undef
406; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
407; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
408; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
409; AVX2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = fdiv <8 x double> undef, undef
410; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
411;
412; AVX512-LABEL: 'fdiv'
413; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
414; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
415; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
416; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fdiv <16 x float> undef, undef
417; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
418; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
419; AVX512-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
420; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fdiv <8 x double> undef, undef
421; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
422;
423; SLM-LABEL: 'fdiv'
424; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %F32 = fdiv float undef, undef
425; SLM-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
426; SLM-NEXT: Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
427; SLM-NEXT: Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
428; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = fdiv double undef, undef
429; SLM-NEXT: Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
430; SLM-NEXT: Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
431; SLM-NEXT: Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
432; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
433;
434; GLM-LABEL: 'fdiv'
435; GLM-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %F32 = fdiv float undef, undef
436; GLM-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %V4F32 = fdiv <4 x float> undef, undef
437; GLM-NEXT: Cost Model: Found an estimated cost of 70 for instruction: %V8F32 = fdiv <8 x float> undef, undef
438; GLM-NEXT: Cost Model: Found an estimated cost of 140 for instruction: %V16F32 = fdiv <16 x float> undef, undef
439; GLM-NEXT: Cost Model: Found an estimated cost of 33 for instruction: %F64 = fdiv double undef, undef
440; GLM-NEXT: Cost Model: Found an estimated cost of 65 for instruction: %V2F64 = fdiv <2 x double> undef, undef
441; GLM-NEXT: Cost Model: Found an estimated cost of 130 for instruction: %V4F64 = fdiv <4 x double> undef, undef
442; GLM-NEXT: Cost Model: Found an estimated cost of 260 for instruction: %V8F64 = fdiv <8 x double> undef, undef
443; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
444;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]445; BTVER2-LABEL: 'fdiv'
446; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
447; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
448; BTVER2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
449; BTVER2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
450; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
451; BTVER2-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
452; BTVER2-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
453; BTVER2-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
454; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
455;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]456 %F32 = fdiv float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]457 %V4F32 = fdiv <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]458 %V8F32 = fdiv <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]459 %V16F32 = fdiv <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]460
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]461 %F64 = fdiv double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]462 %V2F64 = fdiv <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]463 %V4F64 = fdiv <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]464 %V8F64 = fdiv <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]465
466 ret i32 undef
467}
468
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]469define i32 @frem(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]470; SSE1-LABEL: 'frem'
471; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
472; SSE1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
473; SSE1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
474; SSE1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
475; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
476; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = frem <2 x double> undef, undef
477; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = frem <4 x double> undef, undef
478; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = frem <8 x double> undef, undef
479; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
480;
481; SSE2-LABEL: 'frem'
482; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
483; SSE2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
484; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
485; SSE2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
486; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
487; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
488; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
489; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
490; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
491;
492; SSE42-LABEL: 'frem'
493; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
494; SSE42-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
495; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
496; SSE42-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
497; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
498; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
499; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
500; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
501; 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]502;
503; AVX-LABEL: 'frem'
504; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
505; AVX-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
506; AVX-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
507; AVX-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %V16F32 = frem <16 x float> undef, undef
508; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
509; AVX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
510; AVX-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
511; AVX-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F64 = frem <8 x double> undef, undef
512; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
513;
514; AVX512-LABEL: 'frem'
515; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
516; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
517; AVX512-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
518; AVX512-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %V16F32 = frem <16 x float> undef, undef
519; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
520; AVX512-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
521; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
522; AVX512-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V8F64 = frem <8 x double> undef, undef
523; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
524;
525; SLM-LABEL: 'frem'
526; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
527; SLM-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
528; SLM-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
529; SLM-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
530; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
531; SLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
532; SLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
533; SLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
534; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
535;
536; GLM-LABEL: 'frem'
537; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
538; GLM-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
539; GLM-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
540; GLM-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
541; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
542; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
543; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
544; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
545; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
546;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]547; BTVER2-LABEL: 'frem'
548; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
549; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
550; BTVER2-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
551; BTVER2-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %V16F32 = frem <16 x float> undef, undef
552; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
553; BTVER2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
554; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
555; BTVER2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F64 = frem <8 x double> undef, undef
556; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
557;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]558 %F32 = frem float undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]559 %V4F32 = frem <4 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]560 %V8F32 = frem <8 x float> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]561 %V16F32 = frem <16 x float> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]562
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]563 %F64 = frem double undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]564 %V2F64 = frem <2 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]565 %V4F64 = frem <4 x double> undef, undef
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]566 %V8F64 = frem <8 x double> undef, undef
Simon Pilgrim3cd61a02016-08-21 18:34:47 +0000[diff] [blame]567
568 ret i32 undef
569}
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]570
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]571define i32 @fsqrt(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]572; SSE1-LABEL: 'fsqrt'
573; SSE1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
574; SSE1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
575; SSE1-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
576; SSE1-NEXT: Cost Model: Found an estimated cost of 224 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
577; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
578; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
579; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
580; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
581; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
582;
Simon Pilgrim5334a2c2018-04-06 12:36:27 +0000[diff] [blame]583; SSE2-LABEL: 'fsqrt'
584; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
585; SSE2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
586; SSE2-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
587; SSE2-NEXT: Cost Model: Found an estimated cost of 224 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
588; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
589; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
590; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
591; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
592; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
593;
594; SSE42-LABEL: 'fsqrt'
595; SSE42-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
596; SSE42-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
597; SSE42-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
598; SSE42-NEXT: Cost Model: Found an estimated cost of 72 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
599; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
600; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
601; SSE42-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
602; SSE42-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
603; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
604;
605; AVX1-LABEL: 'fsqrt'
606; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
607; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
608; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
609; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
610; AVX1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
611; AVX1-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
612; AVX1-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
613; AVX1-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
614; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
615;
616; AVX2-LABEL: 'fsqrt'
617; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
618; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
619; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
620; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
621; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
622; AVX2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
623; AVX2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
624; AVX2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
625; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
626;
627; AVX512-LABEL: 'fsqrt'
628; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
629; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
630; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
631; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
632; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
633; AVX512-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
634; AVX512-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
635; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
636; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
637;
638; SLM-LABEL: 'fsqrt'
639; SLM-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
640; SLM-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
641; SLM-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
642; SLM-NEXT: Cost Model: Found an estimated cost of 160 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
643; SLM-NEXT: Cost Model: Found an estimated cost of 35 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
644; SLM-NEXT: Cost Model: Found an estimated cost of 70 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
645; SLM-NEXT: Cost Model: Found an estimated cost of 140 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
646; SLM-NEXT: Cost Model: Found an estimated cost of 280 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
647; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
648;
649; GLM-LABEL: 'fsqrt'
650; GLM-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
651; GLM-NEXT: Cost Model: Found an estimated cost of 37 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
652; GLM-NEXT: Cost Model: Found an estimated cost of 74 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
653; GLM-NEXT: Cost Model: Found an estimated cost of 148 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
654; GLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
655; GLM-NEXT: Cost Model: Found an estimated cost of 67 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
656; GLM-NEXT: Cost Model: Found an estimated cost of 134 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
657; GLM-NEXT: Cost Model: Found an estimated cost of 268 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
658; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
659;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]660; BTVER2-LABEL: 'fsqrt'
661; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
662; BTVER2-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
663; BTVER2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
664; BTVER2-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
665; BTVER2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
666; BTVER2-NEXT: Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
667; BTVER2-NEXT: Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
668; BTVER2-NEXT: Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
669; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
670;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]671 %F32 = call float @llvm.sqrt.f32(float undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]672 %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]673 %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]674 %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]675
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]676 %F64 = call double @llvm.sqrt.f64(double undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]677 %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]678 %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]679 %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]680
681 ret i32 undef
682}
683
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]684define i32 @fabs(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]685; SSE1-LABEL: 'fabs'
686; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
687; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
688; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
689; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
690; SSE1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
691; SSE1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
692; SSE1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
693; SSE1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
694; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
695;
696; SSE2-LABEL: 'fabs'
697; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
698; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
699; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
700; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
701; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
702; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
703; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
704; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
705; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
706;
707; SSE42-LABEL: 'fabs'
708; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
709; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
710; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
711; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
712; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
713; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
714; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
715; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
716; 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]717;
718; AVX-LABEL: 'fabs'
719; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
720; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
721; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
722; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
723; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
724; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
725; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
726; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
727; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
728;
729; AVX512-LABEL: 'fabs'
730; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
731; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
732; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
733; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
734; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
735; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
736; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
737; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
738; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
739;
740; SLM-LABEL: 'fabs'
741; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
742; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
743; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
744; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
745; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
746; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
747; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
748; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
749; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
750;
751; GLM-LABEL: 'fabs'
752; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
753; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
754; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
755; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
756; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
757; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
758; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
759; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
760; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
761;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]762; BTVER2-LABEL: 'fabs'
763; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
764; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
765; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
766; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
767; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
768; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
769; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
770; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
771; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
772;
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]773 %F32 = call float @llvm.fabs.f32(float undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]774 %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]775 %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]776 %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
777
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]778 %F64 = call double @llvm.fabs.f64(double undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]779 %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]780 %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]781 %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
782
783 ret i32 undef
784}
785
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]786define i32 @fcopysign(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]787; SSE1-LABEL: 'fcopysign'
788; SSE1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
789; 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)
790; 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)
791; 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)
792; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
793; 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)
794; 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)
795; 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)
796; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
797;
798; SSE2-LABEL: 'fcopysign'
799; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
800; 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)
801; 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)
802; 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)
803; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
804; 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)
805; 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)
806; 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)
807; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
808;
809; SSE42-LABEL: 'fcopysign'
810; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
811; 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)
812; 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)
813; 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)
814; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
815; 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)
816; 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)
817; 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)
818; 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]819;
820; AVX-LABEL: 'fcopysign'
821; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
822; 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)
823; 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)
824; 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)
825; AVX-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
826; 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)
827; 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)
828; 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)
829; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
830;
831; AVX512-LABEL: 'fcopysign'
832; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
833; 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)
834; 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)
835; 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)
836; AVX512-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
837; 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)
838; 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)
839; 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)
840; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
841;
842; SLM-LABEL: 'fcopysign'
843; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
844; 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)
845; 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)
846; 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)
847; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
848; 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)
849; 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)
850; 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)
851; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
852;
853; GLM-LABEL: 'fcopysign'
854; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
855; 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)
856; 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)
857; 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)
858; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
859; 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)
860; 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)
861; 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)
862; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
863;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]864; BTVER2-LABEL: 'fcopysign'
865; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
866; BTVER2-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)
867; BTVER2-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)
868; BTVER2-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)
869; BTVER2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
870; BTVER2-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)
871; BTVER2-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)
872; BTVER2-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)
873; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
874;
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]875 %F32 = call float @llvm.copysign.f32(float undef, float undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]876 %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]877 %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]878 %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
879
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]880 %F64 = call double @llvm.copysign.f64(double undef, double undef)
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]881 %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]882 %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]883 %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
884
885 ret i32 undef
886}
887
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]888define i32 @fma(i32 %arg) {
Simon Pilgrim71d61562019-01-04 16:37:01 +0000[diff] [blame^]889; SSE1-LABEL: 'fma'
890; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
891; 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)
892; 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)
893; 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)
894; SSE1-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
895; 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)
896; 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)
897; 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)
898; SSE1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
899;
900; SSE2-LABEL: 'fma'
901; SSE2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
902; 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)
903; 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)
904; 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)
905; SSE2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
906; 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)
907; 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)
908; 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)
909; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
910;
911; SSE42-LABEL: 'fma'
912; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
913; 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)
914; 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)
915; 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)
916; SSE42-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
917; 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)
918; 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)
919; 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)
920; 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]921;
922; AVX-LABEL: 'fma'
Simon Pilgrim3ede11b2018-04-13 19:12:32 +0000[diff] [blame]923; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
924; 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)
925; AVX-NEXT: Cost Model: Found an estimated cost of 87 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
926; AVX-NEXT: Cost Model: Found an estimated cost of 174 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
927; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
928; 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)
929; 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)
930; 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]931; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
932;
933; AVX512-LABEL: 'fma'
934; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
935; 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)
936; 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)
937; 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)
938; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
939; 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)
940; 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)
941; 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)
942; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
943;
944; SLM-LABEL: 'fma'
945; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
946; 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)
947; 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)
948; 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)
949; SLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
950; 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)
951; 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)
952; 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)
953; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
954;
955; GLM-LABEL: 'fma'
956; GLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
957; 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)
958; 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)
959; 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)
960; GLM-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
961; 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)
962; 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)
963; 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)
964; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
965;
Simon Pilgrimcd9ccf82018-04-20 13:50:33 +0000[diff] [blame]966; BTVER2-LABEL: 'fma'
967; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
968; BTVER2-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)
969; BTVER2-NEXT: Cost Model: Found an estimated cost of 87 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
970; BTVER2-NEXT: Cost Model: Found an estimated cost of 174 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
971; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
972; BTVER2-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)
973; BTVER2-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)
974; BTVER2-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)
975; BTVER2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
976;
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]977 %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]978 %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]979 %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]980 %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]981
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]982 %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]983 %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]984 %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]985 %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]986
987 ret i32 undef
988}
989
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]990declare float @llvm.sqrt.f32(float)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]991declare <4 x float> @llvm.sqrt.v4f32(<4 x float>)
992declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
993declare <16 x float> @llvm.sqrt.v16f32(<16 x float>)
994
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]995declare double @llvm.sqrt.f64(double)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]996declare <2 x double> @llvm.sqrt.v2f64(<2 x double>)
997declare <4 x double> @llvm.sqrt.v4f64(<4 x double>)
998declare <8 x double> @llvm.sqrt.v8f64(<8 x double>)
999
Simon Pilgrim8b021c32016-10-01 16:30:13 +0000[diff] [blame]1000declare float @llvm.fabs.f32(float)
1001declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
1002declare <8 x float> @llvm.fabs.v8f32(<8 x float>)
1003declare <16 x float> @llvm.fabs.v16f32(<16 x float>)
1004
1005declare double @llvm.fabs.f64(double)
1006declare <2 x double> @llvm.fabs.v2f64(<2 x double>)
1007declare <4 x double> @llvm.fabs.v4f64(<4 x double>)
1008declare <8 x double> @llvm.fabs.v8f64(<8 x double>)
1009
Simon Pilgrime0ec5c12016-10-01 16:41:52 +0000[diff] [blame]1010declare float @llvm.copysign.f32(float, float)
1011declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>)
1012declare <8 x float> @llvm.copysign.v8f32(<8 x float>, <8 x float>)
1013declare <16 x float> @llvm.copysign.v16f32(<16 x float>, <16 x float>)
1014
1015declare double @llvm.copysign.f64(double, double)
1016declare <2 x double> @llvm.copysign.v2f64(<2 x double>, <2 x double>)
1017declare <4 x double> @llvm.copysign.v4f64(<4 x double>, <4 x double>)
1018declare <8 x double> @llvm.copysign.v8f64(<8 x double>, <8 x double>)
1019
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1020declare float @llvm.fma.f32(float, float, float)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1021declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
1022declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
1023declare <16 x float> @llvm.fma.v16f32(<16 x float>, <16 x float>, <16 x float>)
1024
Simon Pilgrim91780592016-09-18 21:01:20 +0000[diff] [blame]1025declare double @llvm.fma.f64(double, double, double)
Simon Pilgrim07d7a212016-08-21 19:06:25 +0000[diff] [blame]1026declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
1027declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
1028declare <8 x double> @llvm.fma.v8f64(<8 x double>, <8 x double>, <8 x double>)