| blob | a59f36b21db17673b03d71551c433d0e86347d23 |
1 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
2 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mattr=+sse2 -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
3 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mattr=+sse4.2 -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
4 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mattr=+avx -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
5 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mattr=+avx2 -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
6 ;
7 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mcpu=skylake -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,AVX,SKL
8 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mcpu=knl -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,AVX512,KNL
9 ; RUN: opt < %s -S -x86-experimental-vector-widening-legalization -mtriple=x86_64-apple-darwin -mcpu=skx -cost-model -analyze | FileCheck %s --check-prefixes=CHECK,AVX512,SKX
12 define <2 x double> @test1(<2 x i64> %trigger, <2 x double>* %addr, <2 x double> %dst) {
13 ; SSE2-LABEL: 'test1'
14 ; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %mask = icmp eq <2 x i64> %trigger, zeroinitializer
15 ; SSE2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1> %mask, <2 x double> %dst)
16 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
17 ;
18 ; SSE42-LABEL: 'test1'
19 ; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i64> %trigger, zeroinitializer
20 ; SSE42-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1> %mask, <2 x double> %dst)
21 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
22 ;
23 ; AVX-LABEL: 'test1'
24 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i64> %trigger, zeroinitializer
25 ; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1> %mask, <2 x double> %dst)
26 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
27 ;
28 ; AVX512-LABEL: 'test1'
29 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i64> %trigger, zeroinitializer
30 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1> %mask, <2 x double> %dst)
31 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
32 ;
33 %mask = icmp eq <2 x i64> %trigger, zeroinitializer
34 %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1>%mask, <2 x double>%dst)
35 ret <2 x double> %res
36 }
38 define <4 x i32> @test2(<4 x i32> %trigger, <4 x i32>* %addr, <4 x i32> %dst) {
39 ; SSE-LABEL: 'test2'
40 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
41 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %addr, i32 4, <4 x i1> %mask, <4 x i32> %dst)
42 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
43 ;
44 ; AVX-LABEL: 'test2'
45 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
46 ; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %addr, i32 4, <4 x i1> %mask, <4 x i32> %dst)
47 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
48 ;
49 ; AVX512-LABEL: 'test2'
50 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
51 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %res = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %addr, i32 4, <4 x i1> %mask, <4 x i32> %dst)
52 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
53 ;
54 %mask = icmp eq <4 x i32> %trigger, zeroinitializer
55 %res = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %addr, i32 4, <4 x i1>%mask, <4 x i32>%dst)
56 ret <4 x i32> %res
57 }
59 define void @test3(<4 x i32> %trigger, <4 x i32>* %addr, <4 x i32> %val) {
60 ; SSE-LABEL: 'test3'
61 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
62 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %val, <4 x i32>* %addr, i32 4, <4 x i1> %mask)
63 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
64 ;
65 ; AVX-LABEL: 'test3'
66 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
67 ; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %val, <4 x i32>* %addr, i32 4, <4 x i1> %mask)
68 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
69 ;
70 ; AVX512-LABEL: 'test3'
71 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <4 x i32> %trigger, zeroinitializer
72 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %val, <4 x i32>* %addr, i32 4, <4 x i1> %mask)
73 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
74 ;
75 %mask = icmp eq <4 x i32> %trigger, zeroinitializer
76 call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32>%val, <4 x i32>* %addr, i32 4, <4 x i1>%mask)
77 ret void
78 }
80 define <8 x float> @test4(<8 x i32> %trigger, <8 x float>* %addr, <8 x float> %dst) {
81 ; SSE-LABEL: 'test4'
82 ; SSE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %mask = icmp eq <8 x i32> %trigger, zeroinitializer
83 ; SSE-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1> %mask, <8 x float> %dst)
84 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %res
85 ;
86 ; AVX1-LABEL: 'test4'
87 ; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %mask = icmp eq <8 x i32> %trigger, zeroinitializer
88 ; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1> %mask, <8 x float> %dst)
89 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %res
90 ;
91 ; AVX2-LABEL: 'test4'
92 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <8 x i32> %trigger, zeroinitializer
93 ; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1> %mask, <8 x float> %dst)
94 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %res
95 ;
96 ; SKL-LABEL: 'test4'
97 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <8 x i32> %trigger, zeroinitializer
98 ; SKL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1> %mask, <8 x float> %dst)
99 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %res
100 ;
101 ; AVX512-LABEL: 'test4'
102 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <8 x i32> %trigger, zeroinitializer
103 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1> %mask, <8 x float> %dst)
104 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %res
105 ;
106 %mask = icmp eq <8 x i32> %trigger, zeroinitializer
107 %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 4, <8 x i1>%mask, <8 x float>%dst)
108 ret <8 x float> %res
109 }
111 define void @test5(<2 x i32> %trigger, <2 x float>* %addr, <2 x float> %val) {
112 ; SSE-LABEL: 'test5'
113 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
114 ; SSE-NEXT: Cost Model: Found an estimated cost of 7 for instruction: call void @llvm.masked.store.v2f32.p0v2f32(<2 x float> %val, <2 x float>* %addr, i32 4, <2 x i1> %mask)
115 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
116 ;
117 ; AVX-LABEL: 'test5'
118 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
119 ; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: call void @llvm.masked.store.v2f32.p0v2f32(<2 x float> %val, <2 x float>* %addr, i32 4, <2 x i1> %mask)
120 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
121 ;
122 ; AVX512-LABEL: 'test5'
123 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
124 ; AVX512-NEXT: Cost Model: Found an estimated cost of 5 for instruction: call void @llvm.masked.store.v2f32.p0v2f32(<2 x float> %val, <2 x float>* %addr, i32 4, <2 x i1> %mask)
125 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
126 ;
127 %mask = icmp eq <2 x i32> %trigger, zeroinitializer
128 call void @llvm.masked.store.v2f32.p0v2f32(<2 x float>%val, <2 x float>* %addr, i32 4, <2 x i1>%mask)
129 ret void
130 }
132 define void @test6(<2 x i32> %trigger, <2 x i32>* %addr, <2 x i32> %val) {
133 ; SSE-LABEL: 'test6'
134 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
135 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: call void @llvm.masked.store.v2i32.p0v2i32(<2 x i32> %val, <2 x i32>* %addr, i32 4, <2 x i1> %mask)
136 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
137 ;
138 ; AVX-LABEL: 'test6'
139 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
140 ; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: call void @llvm.masked.store.v2i32.p0v2i32(<2 x i32> %val, <2 x i32>* %addr, i32 4, <2 x i1> %mask)
141 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
142 ;
143 ; AVX512-LABEL: 'test6'
144 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
145 ; AVX512-NEXT: Cost Model: Found an estimated cost of 5 for instruction: call void @llvm.masked.store.v2i32.p0v2i32(<2 x i32> %val, <2 x i32>* %addr, i32 4, <2 x i1> %mask)
146 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
147 ;
148 %mask = icmp eq <2 x i32> %trigger, zeroinitializer
149 call void @llvm.masked.store.v2i32.p0v2i32(<2 x i32>%val, <2 x i32>* %addr, i32 4, <2 x i1>%mask)
150 ret void
151 }
153 define <2 x float> @test7(<2 x i32> %trigger, <2 x float>* %addr, <2 x float> %dst) {
154 ; SSE-LABEL: 'test7'
155 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
156 ; SSE-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>* %addr, i32 4, <2 x i1> %mask, <2 x float> %dst)
157 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x float> %res
158 ;
159 ; AVX-LABEL: 'test7'
160 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
161 ; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>* %addr, i32 4, <2 x i1> %mask, <2 x float> %dst)
162 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x float> %res
163 ;
164 ; AVX512-LABEL: 'test7'
165 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
166 ; AVX512-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %res = call <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>* %addr, i32 4, <2 x i1> %mask, <2 x float> %dst)
167 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x float> %res
168 ;
169 %mask = icmp eq <2 x i32> %trigger, zeroinitializer
170 %res = call <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>* %addr, i32 4, <2 x i1>%mask, <2 x float>%dst)
171 ret <2 x float> %res
172 }
174 define <2 x i32> @test8(<2 x i32> %trigger, <2 x i32>* %addr, <2 x i32> %dst) {
175 ; SSE-LABEL: 'test8'
176 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
177 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <2 x i32> @llvm.masked.load.v2i32.p0v2i32(<2 x i32>* %addr, i32 4, <2 x i1> %mask, <2 x i32> %dst)
178 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %res
179 ;
180 ; AVX-LABEL: 'test8'
181 ; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
182 ; AVX-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <2 x i32> @llvm.masked.load.v2i32.p0v2i32(<2 x i32>* %addr, i32 4, <2 x i1> %mask, <2 x i32> %dst)
183 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %res
184 ;
185 ; AVX512-LABEL: 'test8'
186 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %mask = icmp eq <2 x i32> %trigger, zeroinitializer
187 ; AVX512-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %res = call <2 x i32> @llvm.masked.load.v2i32.p0v2i32(<2 x i32>* %addr, i32 4, <2 x i1> %mask, <2 x i32> %dst)
188 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i32> %res
189 ;
190 %mask = icmp eq <2 x i32> %trigger, zeroinitializer
191 %res = call <2 x i32> @llvm.masked.load.v2i32.p0v2i32(<2 x i32>* %addr, i32 4, <2 x i1>%mask, <2 x i32>%dst)
192 ret <2 x i32> %res
193 }
195 define <2 x double> @test_gather_2f64(<2 x double*> %ptrs, <2 x i1> %mask, <2 x double> %src0) {
196 ; SSE-LABEL: 'test_gather_2f64'
197 ; SSE-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
198 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
199 ;
200 ; AVX1-LABEL: 'test_gather_2f64'
201 ; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
202 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
203 ;
204 ; AVX2-LABEL: 'test_gather_2f64'
205 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
206 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
207 ;
208 ; SKL-LABEL: 'test_gather_2f64'
209 ; SKL-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
210 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
211 ;
212 ; AVX512-LABEL: 'test_gather_2f64'
213 ; AVX512-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
214 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %res
215 ;
220 %res = call <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32 4, <2 x i1> %mask, <2 x double> %src0)
221 ret <2 x double> %res
222 }
223 declare <2 x double> @llvm.masked.gather.v2f64.v2p0f64(<2 x double*> %ptrs, i32, <2 x i1> %mask, <2 x double> %src0)
225 define <4 x i32> @test_gather_4i32(<4 x i32*> %ptrs, <4 x i1> %mask, <4 x i32> %src0) {
226 ; SSE-LABEL: 'test_gather_4i32'
227 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
228 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
229 ;
230 ; AVX1-LABEL: 'test_gather_4i32'
231 ; AVX1-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
232 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
233 ;
234 ; AVX2-LABEL: 'test_gather_4i32'
235 ; AVX2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
236 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
237 ;
238 ; SKL-LABEL: 'test_gather_4i32'
239 ; SKL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
240 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
241 ;
242 ; KNL-LABEL: 'test_gather_4i32'
243 ; KNL-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
244 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
245 ;
246 ; SKX-LABEL: 'test_gather_4i32'
247 ; SKX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
248 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
249 ;
254 %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> %mask, <4 x i32> %src0)
255 ret <4 x i32> %res
256 }
258 define <4 x i32> @test_gather_4i32_const_mask(<4 x i32*> %ptrs, <4 x i32> %src0) {
259 ; SSE-LABEL: 'test_gather_4i32_const_mask'
260 ; SSE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
261 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
262 ;
263 ; AVX1-LABEL: 'test_gather_4i32_const_mask'
264 ; AVX1-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
265 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
266 ;
267 ; AVX2-LABEL: 'test_gather_4i32_const_mask'
268 ; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
269 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
270 ;
271 ; SKL-LABEL: 'test_gather_4i32_const_mask'
272 ; SKL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
273 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
274 ;
275 ; KNL-LABEL: 'test_gather_4i32_const_mask'
276 ; KNL-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
277 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
278 ;
279 ; SKX-LABEL: 'test_gather_4i32_const_mask'
280 ; SKX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
281 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %res
282 ;
287 %res = call <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x i32> %src0)
288 ret <4 x i32> %res
289 }
290 declare <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*> %ptrs, i32, <4 x i1> %mask, <4 x i32> %src0)
292 define <16 x float> @test_gather_16f32_const_mask(float* %base, <16 x i32> %ind) {
293 ; SSE2-LABEL: 'test_gather_16f32_const_mask'
294 ; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
295 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
296 ; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
297 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
298 ;
299 ; SSE42-LABEL: 'test_gather_16f32_const_mask'
300 ; SSE42-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
301 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
302 ; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
303 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
304 ;
305 ; AVX1-LABEL: 'test_gather_16f32_const_mask'
306 ; AVX1-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
307 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
308 ; AVX1-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
309 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
310 ;
311 ; AVX2-LABEL: 'test_gather_16f32_const_mask'
312 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
313 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
314 ; AVX2-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
315 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
316 ;
317 ; SKL-LABEL: 'test_gather_16f32_const_mask'
318 ; SKL-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
319 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
320 ; SKL-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
321 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
322 ;
323 ; AVX512-LABEL: 'test_gather_16f32_const_mask'
324 ; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
325 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
326 ; AVX512-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
327 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
328 ;
329 %sext_ind = sext <16 x i32> %ind to <16 x i64>
330 %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
332 %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
333 ret <16 x float>%res
334 }
336 define <16 x float> @test_gather_16f32_var_mask(float* %base, <16 x i32> %ind, <16 x i1>%mask) {
337 ; SSE2-LABEL: 'test_gather_16f32_var_mask'
338 ; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
339 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
340 ; SSE2-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
341 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
342 ;
343 ; SSE42-LABEL: 'test_gather_16f32_var_mask'
344 ; SSE42-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
345 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
346 ; SSE42-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
347 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
348 ;
349 ; AVX1-LABEL: 'test_gather_16f32_var_mask'
350 ; AVX1-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
351 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
352 ; AVX1-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
353 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
354 ;
355 ; AVX2-LABEL: 'test_gather_16f32_var_mask'
356 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
357 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
358 ; AVX2-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
359 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
360 ;
361 ; SKL-LABEL: 'test_gather_16f32_var_mask'
362 ; SKL-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
363 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
364 ; SKL-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
365 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
366 ;
367 ; AVX512-LABEL: 'test_gather_16f32_var_mask'
368 ; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
369 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
370 ; AVX512-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
371 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
372 ;
373 %sext_ind = sext <16 x i32> %ind to <16 x i64>
374 %gep.v = getelementptr float, float* %base, <16 x i64> %sext_ind
376 %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
377 ret <16 x float>%res
378 }
380 define <16 x float> @test_gather_16f32_ra_var_mask(<16 x float*> %ptrs, <16 x i32> %ind, <16 x i1>%mask) {
381 ; SSE2-LABEL: 'test_gather_16f32_ra_var_mask'
382 ; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
383 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
384 ; SSE2-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
385 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
386 ;
387 ; SSE42-LABEL: 'test_gather_16f32_ra_var_mask'
388 ; SSE42-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
389 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
390 ; SSE42-NEXT: Cost Model: Found an estimated cost of 60 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
391 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
392 ;
393 ; AVX1-LABEL: 'test_gather_16f32_ra_var_mask'
394 ; AVX1-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
395 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
396 ; AVX1-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
397 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
398 ;
399 ; AVX2-LABEL: 'test_gather_16f32_ra_var_mask'
400 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
401 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
402 ; AVX2-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
403 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
404 ;
405 ; SKL-LABEL: 'test_gather_16f32_ra_var_mask'
406 ; SKL-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
407 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
408 ; SKL-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
409 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
410 ;
411 ; AVX512-LABEL: 'test_gather_16f32_ra_var_mask'
412 ; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
413 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
414 ; AVX512-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
415 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
416 ;
417 %sext_ind = sext <16 x i32> %ind to <16 x i64>
418 %gep.v = getelementptr float, <16 x float*> %ptrs, <16 x i64> %sext_ind
420 %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32 4, <16 x i1> %mask, <16 x float> undef)
421 ret <16 x float>%res
422 }
424 define <16 x float> @test_gather_16f32_const_mask2(float* %base, <16 x i32> %ind) {
425 ; SSE2-LABEL: 'test_gather_16f32_const_mask2'
426 ; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
427 ; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
428 ; SSE2-NEXT: Cost Model: Found an estimated cost of 23 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
429 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
430 ; SSE2-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
431 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
432 ;
433 ; SSE42-LABEL: 'test_gather_16f32_const_mask2'
434 ; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
435 ; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
436 ; SSE42-NEXT: Cost Model: Found an estimated cost of 11 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
437 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
438 ; SSE42-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
439 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
440 ;
441 ; AVX1-LABEL: 'test_gather_16f32_const_mask2'
442 ; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
443 ; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
444 ; AVX1-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
445 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
446 ; AVX1-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
447 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
448 ;
449 ; AVX2-LABEL: 'test_gather_16f32_const_mask2'
450 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
451 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
452 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
453 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
454 ; AVX2-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
455 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
456 ;
457 ; SKL-LABEL: 'test_gather_16f32_const_mask2'
458 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
459 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
460 ; SKL-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
461 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
462 ; SKL-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
463 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
464 ;
465 ; AVX512-LABEL: 'test_gather_16f32_const_mask2'
466 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
467 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
468 ; AVX512-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %sext_ind = sext <16 x i32> %ind to <16 x i64>
469 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
470 ; AVX512-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
471 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %res
472 ;
473 %broadcast.splatinsert = insertelement <16 x float*> undef, float* %base, i32 0
474 %broadcast.splat = shufflevector <16 x float*> %broadcast.splatinsert, <16 x float*> undef, <16 x i32> zeroinitializer
476 %sext_ind = sext <16 x i32> %ind to <16 x i64>
477 %gep.random = getelementptr float, <16 x float*> %broadcast.splat, <16 x i64> %sext_ind
479 %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.random, i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <16 x float> undef)
480 ret <16 x float>%res
481 }
483 define void @test_scatter_16i32(i32* %base, <16 x i32> %ind, i16 %mask, <16 x i32>%val) {
484 ; SSE-LABEL: 'test_scatter_16i32'
485 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
486 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
487 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
488 ; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %imask = bitcast i16 %mask to <16 x i1>
489 ; SSE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
490 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
491 ;
492 ; AVX1-LABEL: 'test_scatter_16i32'
493 ; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
494 ; AVX1-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
495 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
496 ; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %imask = bitcast i16 %mask to <16 x i1>
497 ; AVX1-NEXT: Cost Model: Found an estimated cost of 64 for instruction: call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
498 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
499 ;
500 ; AVX2-LABEL: 'test_scatter_16i32'
501 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
502 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
503 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
504 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %imask = bitcast i16 %mask to <16 x i1>
505 ; AVX2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
506 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
507 ;
508 ; SKL-LABEL: 'test_scatter_16i32'
509 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
510 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
511 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
512 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %imask = bitcast i16 %mask to <16 x i1>
513 ; SKL-NEXT: Cost Model: Found an estimated cost of 64 for instruction: call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
514 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
515 ;
516 ; AVX512-LABEL: 'test_scatter_16i32'
517 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
518 ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
519 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
520 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %imask = bitcast i16 %mask to <16 x i1>
521 ; AVX512-NEXT: Cost Model: Found an estimated cost of 18 for instruction: call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32> %val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
522 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
523 ;
524 %broadcast.splatinsert = insertelement <16 x i32*> undef, i32* %base, i32 0
525 %broadcast.splat = shufflevector <16 x i32*> %broadcast.splatinsert, <16 x i32*> undef, <16 x i32> zeroinitializer
527 %gep.random = getelementptr i32, <16 x i32*> %broadcast.splat, <16 x i32> %ind
528 %imask = bitcast i16 %mask to <16 x i1>
529 call void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32>%val, <16 x i32*> %gep.random, i32 4, <16 x i1> %imask)
530 ret void
531 }
533 define void @test_scatter_8i32(<8 x i32>%a1, <8 x i32*> %ptr, <8 x i1>%mask) {
534 ; SSE-LABEL: 'test_scatter_8i32'
535 ; SSE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: call void @llvm.masked.scatter.v8i32.v8p0i32(<8 x i32> %a1, <8 x i32*> %ptr, i32 4, <8 x i1> %mask)
536 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
537 ;
538 ; AVX-LABEL: 'test_scatter_8i32'
539 ; AVX-NEXT: Cost Model: Found an estimated cost of 32 for instruction: call void @llvm.masked.scatter.v8i32.v8p0i32(<8 x i32> %a1, <8 x i32*> %ptr, i32 4, <8 x i1> %mask)
540 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
541 ;
542 ; AVX512-LABEL: 'test_scatter_8i32'
543 ; AVX512-NEXT: Cost Model: Found an estimated cost of 10 for instruction: call void @llvm.masked.scatter.v8i32.v8p0i32(<8 x i32> %a1, <8 x i32*> %ptr, i32 4, <8 x i1> %mask)
544 ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
545 ;
547 call void @llvm.masked.scatter.v8i32.v8p0i32(<8 x i32> %a1, <8 x i32*> %ptr, i32 4, <8 x i1> %mask)
548 ret void
549 }
551 declare void @llvm.masked.scatter.v8i32.v8p0i32(<8 x i32> %a1, <8 x i32*> %ptr, i32, <8 x i1> %mask)
553 define void @test_scatter_4i32(<4 x i32>%a1, <4 x i32*> %ptr, <4 x i1>%mask) {
554 ; SSE-LABEL: 'test_scatter_4i32'
555 ; SSE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
556 ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
557 ;
558 ; AVX-LABEL: 'test_scatter_4i32'
559 ; AVX-NEXT: Cost Model: Found an estimated cost of 16 for instruction: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
560 ; AVX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
561 ;
562 ; KNL-LABEL: 'test_scatter_4i32'
563 ; KNL-NEXT: Cost Model: Found an estimated cost of 16 for instruction: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
564 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
565 ;
566 ; SKX-LABEL: 'test_scatter_4i32'
567 ; SKX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
568 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
569 ;
570 call void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32 4, <4 x i1> %mask)
571 ret void
572 }
574 define <4 x float> @test_gather_4f32(float* %ptr, <4 x i32> %ind, <4 x i1>%mask) {
575 ; SSE2-LABEL: 'test_gather_4f32'
576 ; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
577 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
578 ; SSE2-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
579 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
580 ;
581 ; SSE42-LABEL: 'test_gather_4f32'
582 ; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
583 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
584 ; SSE42-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
585 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
586 ;
587 ; AVX1-LABEL: 'test_gather_4f32'
588 ; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
589 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
590 ; AVX1-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
591 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
592 ;
593 ; AVX2-LABEL: 'test_gather_4f32'
594 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
595 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
596 ; AVX2-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
597 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
598 ;
599 ; SKL-LABEL: 'test_gather_4f32'
600 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
601 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
602 ; SKL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
603 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
604 ;
605 ; KNL-LABEL: 'test_gather_4f32'
606 ; KNL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
607 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
608 ; KNL-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
609 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
610 ;
611 ; SKX-LABEL: 'test_gather_4f32'
612 ; SKX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
613 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
614 ; SKX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
615 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
616 ;
617 %sext_ind = sext <4 x i32> %ind to <4 x i64>
618 %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
620 %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> %mask, <4 x float> undef)
621 ret <4 x float>%res
622 }
624 define <4 x float> @test_gather_4f32_const_mask(float* %ptr, <4 x i32> %ind) {
625 ; SSE2-LABEL: 'test_gather_4f32_const_mask'
626 ; SSE2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
627 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
628 ; SSE2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
629 ; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
630 ;
631 ; SSE42-LABEL: 'test_gather_4f32_const_mask'
632 ; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
633 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
634 ; SSE42-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
635 ; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
636 ;
637 ; AVX1-LABEL: 'test_gather_4f32_const_mask'
638 ; AVX1-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
639 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
640 ; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
641 ; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
642 ;
643 ; AVX2-LABEL: 'test_gather_4f32_const_mask'
644 ; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
645 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
646 ; AVX2-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
647 ; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
648 ;
649 ; SKL-LABEL: 'test_gather_4f32_const_mask'
650 ; SKL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
651 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
652 ; SKL-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
653 ; SKL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
654 ;
655 ; KNL-LABEL: 'test_gather_4f32_const_mask'
656 ; KNL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
657 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
658 ; KNL-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
659 ; KNL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
660 ;
661 ; SKX-LABEL: 'test_gather_4f32_const_mask'
662 ; SKX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sext_ind = sext <4 x i32> %ind to <4 x i64>
663 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
664 ; SKX-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
665 ; SKX-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %res
666 ;
667 %sext_ind = sext <4 x i32> %ind to <4 x i64>
668 %gep.v = getelementptr float, float* %ptr, <4 x i64> %sext_ind
670 %res = call <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> undef)
671 ret <4 x float>%res
672 }
674 declare <4 x float> @llvm.masked.gather.v4f32.v4p0f32(<4 x float*> %gep.v, i32, <4 x i1> %mask, <4 x float> )
675 declare void @llvm.masked.scatter.v4i32.v4p0i32(<4 x i32> %a1, <4 x i32*> %ptr, i32, <4 x i1> %mask)
676 declare void @llvm.masked.scatter.v16i32.v16p0i32(<16 x i32>%val, <16 x i32*> %gep.random, i32, <16 x i1> %imask)
677 declare <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %gep.v, i32, <16 x i1> %mask, <16 x float>)
679 declare <16 x i32> @llvm.masked.load.v16i32.p0v16i32(<16 x i32>*, i32, <16 x i1>, <16 x i32>)
680 declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32, <4 x i1>, <4 x i32>)
681 declare <2 x i32> @llvm.masked.load.v2i32.p0v2i32(<2 x i32>*, i32, <2 x i1>, <2 x i32>)
682 declare void @llvm.masked.store.v16i32.p0v16i32(<16 x i32>, <16 x i32>*, i32, <16 x i1>)
683 declare void @llvm.masked.store.v8i32.p0v8i32(<8 x i32>, <8 x i32>*, i32, <8 x i1>)
684 declare void @llvm.masked.store.v4i32.p0v4i32(<4 x i32>, <4 x i32>*, i32, <4 x i1>)
685 declare void @llvm.masked.store.v2f32.p0v2f32(<2 x float>, <2 x float>*, i32, <2 x i1>)
686 declare void @llvm.masked.store.v2i32.p0v2i32(<2 x i32>, <2 x i32>*, i32, <2 x i1>)
687 declare void @llvm.masked.store.v16f32.p0v16f32(<16 x float>, <16 x float>*, i32, <16 x i1>)
688 declare <16 x float> @llvm.masked.load.v16f32.p0v16f32(<16 x float>*, i32, <16 x i1>, <16 x float>)
689 declare <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>*, i32, <8 x i1>, <8 x float>)
690 declare <4 x float> @llvm.masked.load.v4f32.p0v4f32(<4 x float>*, i32, <4 x i1>, <4 x float>)
691 declare <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>*, i32, <2 x i1>, <2 x float>)
692 declare <8 x double> @llvm.masked.load.v8f64.p0v8f64(<8 x double>*, i32, <8 x i1>, <8 x double>)
693 declare <4 x double> @llvm.masked.load.v4f64.p0v4f64(<4 x double>*, i32, <4 x i1>, <4 x double>)
694 declare <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>*, i32, <2 x i1>, <2 x double>)
695 declare void @llvm.masked.store.v8f64.p0v8f64(<8 x double>, <8 x double>*, i32, <8 x i1>)
696 declare void @llvm.masked.store.v2f64.p0v2f64(<2 x double>, <2 x double>*, i32, <2 x i1>)
697 declare void @llvm.masked.store.v2i64.p0v2i64(<2 x i64>, <2 x i64>*, i32, <2 x i1>)