1; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-4,-unaligned-scratch-access  -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT4,ELT4-ALIGNED,ALIGNED,ALL %s
2; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-8,-unaligned-scratch-access  -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT8,ELT8-ALIGNED,ALIGNED,ALL %s
3; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-16,-unaligned-scratch-access -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT16,ELT16-ALIGNED,ALIGNED,ALL %s
4; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-4,+unaligned-scratch-access  -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT4,ELT4-UNALIGNED,UNALIGNED,ALL %s
5; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-8,+unaligned-scratch-access  -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT8,ELT8-UNALIGNED,UNALIGNED,ALL %s
6; RUN: opt -mtriple=amdgcn-amd-amdhsa -mattr=+max-private-element-size-16,+unaligned-scratch-access -load-store-vectorizer -S -o - %s | FileCheck -check-prefixes=ELT16,ELT16-UNALIGNED,UNALIGNED,ALL %s
7
8target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
9
10; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v4i32
11; ALIGNED: store i32
12; ALIGNED: store i32
13; ALIGNED: store i32
14; ALIGNED: store i32
15
16; ELT8-UNALIGNED: store <2 x i32>
17; ELT8-UNALIGNED: store <2 x i32>
18
19; ELT16-UNALIGNED: store <4 x i32>
20define void @merge_private_store_4_vector_elts_loads_v4i32(i32* %out) #0 {
21  %out.gep.1 = getelementptr i32, i32* %out, i32 1
22  %out.gep.2 = getelementptr i32, i32* %out, i32 2
23  %out.gep.3 = getelementptr i32, i32* %out, i32 3
24
25  store i32 9, i32* %out
26  store i32 1, i32* %out.gep.1
27  store i32 23, i32* %out.gep.2
28  store i32 19, i32* %out.gep.3
29  ret void
30}
31
32; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v4i32_align1(
33; ALIGNED: store i32 9, i32* %out, align 1
34; ALIGNED: store i32 1, i32* %out.gep.1, align 1
35; ALIGNED: store i32 23, i32* %out.gep.2, align 1
36; ALIGNED: store i32 19, i32* %out.gep.3, align 1
37
38; ELT16-UNALIGNED: store <4 x i32> <i32 9, i32 1, i32 23, i32 19>, <4 x i32>* %1, align 1
39
40; ELT8-UNALIGNED: store <2 x i32> <i32 9, i32 1>, <2 x i32>* %1, align 1
41; ELT8-UNALIGNED: store <2 x i32> <i32 23, i32 19>, <2 x i32>* %2, align 1
42
43; ELT4-UNALIGNED: store i32
44; ELT4-UNALIGNED: store i32
45; ELT4-UNALIGNED: store i32
46; ELT4-UNALIGNED: store i32
47define void @merge_private_store_4_vector_elts_loads_v4i32_align1(i32* %out) #0 {
48  %out.gep.1 = getelementptr i32, i32* %out, i32 1
49  %out.gep.2 = getelementptr i32, i32* %out, i32 2
50  %out.gep.3 = getelementptr i32, i32* %out, i32 3
51
52  store i32 9, i32* %out, align 1
53  store i32 1, i32* %out.gep.1, align 1
54  store i32 23, i32* %out.gep.2, align 1
55  store i32 19, i32* %out.gep.3, align 1
56  ret void
57}
58
59; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v4i32_align2(
60; ALIGNED: store i32 9, i32* %out, align 2
61; ALIGNED: store i32 1, i32* %out.gep.1, align 2
62; ALIGNED: store i32 23, i32* %out.gep.2, align 2
63; ALIGNED: store i32 19, i32* %out.gep.3, align 2
64
65; ELT16-UNALIGNED: store <4 x i32> <i32 9, i32 1, i32 23, i32 19>, <4 x i32>* %1, align 2
66
67; ELT8-UNALIGNED: store <2 x i32>
68; ELT8-UNALIGNED: store <2 x i32>
69
70; ELT4-UNALIGNED: store i32
71; ELT4-UNALIGNED: store i32
72; ELT4-UNALIGNED: store i32
73; ELT4-UNALIGNED: store i32
74define void @merge_private_store_4_vector_elts_loads_v4i32_align2(i32* %out) #0 {
75  %out.gep.1 = getelementptr i32, i32* %out, i32 1
76  %out.gep.2 = getelementptr i32, i32* %out, i32 2
77  %out.gep.3 = getelementptr i32, i32* %out, i32 3
78
79  store i32 9, i32* %out, align 2
80  store i32 1, i32* %out.gep.1, align 2
81  store i32 23, i32* %out.gep.2, align 2
82  store i32 19, i32* %out.gep.3, align 2
83  ret void
84}
85
86; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v4i8(
87; ALL: store <4 x i8>
88define void @merge_private_store_4_vector_elts_loads_v4i8(i8* %out) #0 {
89  %out.gep.1 = getelementptr i8, i8* %out, i32 1
90  %out.gep.2 = getelementptr i8, i8* %out, i32 2
91  %out.gep.3 = getelementptr i8, i8* %out, i32 3
92
93  store i8 9, i8* %out, align 4
94  store i8 1, i8* %out.gep.1
95  store i8 23, i8* %out.gep.2
96  store i8 19, i8* %out.gep.3
97  ret void
98}
99
100; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v4i8_align1(
101; ALIGNED: store i8
102; ALIGNED: store i8
103; ALIGNED: store i8
104; ALIGNED: store i8
105
106; UNALIGNED: store <4 x i8> <i8 9, i8 1, i8 23, i8 19>, <4 x i8>* %1, align 1
107define void @merge_private_store_4_vector_elts_loads_v4i8_align1(i8* %out) #0 {
108  %out.gep.1 = getelementptr i8, i8* %out, i32 1
109  %out.gep.2 = getelementptr i8, i8* %out, i32 2
110  %out.gep.3 = getelementptr i8, i8* %out, i32 3
111
112  store i8 9, i8* %out, align 1
113  store i8 1, i8* %out.gep.1, align 1
114  store i8 23, i8* %out.gep.2, align 1
115  store i8 19, i8* %out.gep.3, align 1
116  ret void
117}
118
119; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v2i16(
120; ALL: store <2 x i16>
121define void @merge_private_store_4_vector_elts_loads_v2i16(i16* %out) #0 {
122  %out.gep.1 = getelementptr i16, i16* %out, i32 1
123
124  store i16 9, i16* %out, align 4
125  store i16 12, i16* %out.gep.1
126  ret void
127}
128
129; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v2i16_align2(
130; ALIGNED: store i16
131; ALIGNED: store i16
132
133; UNALIGNED: store <2 x i16> <i16 9, i16 12>, <2 x i16>* %1, align 2
134define void @merge_private_store_4_vector_elts_loads_v2i16_align2(i16* %out) #0 {
135  %out.gep.1 = getelementptr i16, i16* %out, i32 1
136
137  store i16 9, i16* %out, align 2
138  store i16 12, i16* %out.gep.1, align 2
139  ret void
140}
141
142; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v2i16_align1(
143; ALIGNED: store i16
144; ALIGNED: store i16
145
146; UNALIGNED: store <2 x i16> <i16 9, i16 12>, <2 x i16>* %1, align 1
147define void @merge_private_store_4_vector_elts_loads_v2i16_align1(i16* %out) #0 {
148  %out.gep.1 = getelementptr i16, i16* %out, i32 1
149
150  store i16 9, i16* %out, align 1
151  store i16 12, i16* %out.gep.1, align 1
152  ret void
153}
154
155; ALL-LABEL: @merge_private_store_4_vector_elts_loads_v2i16_align8(
156; ALL: store <2 x i16> <i16 9, i16 12>, <2 x i16>* %1, align 8
157define void @merge_private_store_4_vector_elts_loads_v2i16_align8(i16* %out) #0 {
158  %out.gep.1 = getelementptr i16, i16* %out, i32 1
159
160  store i16 9, i16* %out, align 8
161  store i16 12, i16* %out.gep.1, align 2
162  ret void
163}
164
165; ALL-LABEL: @merge_private_store_3_vector_elts_loads_v4i32
166; ELT4: store i32
167; ELT4: store i32
168; ELT4: store i32
169
170; ELT8-ALIGNED: store i32
171; ELT8-ALIGNED: store i32
172; ELT8-ALIGNED: store i32
173
174; ELT8-UNALIGNED: store <2 x i32>
175; ELT8-UNALIGNED: store i32
176
177; ELT16-ALIGNED: store i32
178; ELT16-ALIGNED: store i32
179; ELT16-ALIGNED: store i32
180
181; ELT16-UNALIGNED: store <3 x i32>
182define void @merge_private_store_3_vector_elts_loads_v4i32(i32* %out) #0 {
183  %out.gep.1 = getelementptr i32, i32* %out, i32 1
184  %out.gep.2 = getelementptr i32, i32* %out, i32 2
185
186  store i32 9, i32* %out
187  store i32 1, i32* %out.gep.1
188  store i32 23, i32* %out.gep.2
189  ret void
190}
191
192; ALL-LABEL: @merge_private_store_3_vector_elts_loads_v4i32_align1(
193; ALIGNED: store i32
194; ALIGNED: store i32
195; ALIGNED: store i32
196
197; ELT4-UNALIGNED: store i32
198; ELT4-UNALIGNED: store i32
199; ELT4-UNALIGNED: store i32
200
201; ELT8-UNALIGNED: store <2 x i32>
202; ELT8-UNALIGNED: store i32
203
204; ELT16-UNALIGNED: store <3 x i32>
205define void @merge_private_store_3_vector_elts_loads_v4i32_align1(i32* %out) #0 {
206  %out.gep.1 = getelementptr i32, i32* %out, i32 1
207  %out.gep.2 = getelementptr i32, i32* %out, i32 2
208
209  store i32 9, i32* %out, align 1
210  store i32 1, i32* %out.gep.1, align 1
211  store i32 23, i32* %out.gep.2, align 1
212  ret void
213}
214
215; ALL-LABEL: @merge_private_store_3_vector_elts_loads_v4i8_align1(
216; ALIGNED: store i8
217; ALIGNED: store i8
218; ALIGNED: store i8
219
220; UNALIGNED: store <3 x i8>
221define void @merge_private_store_3_vector_elts_loads_v4i8_align1(i8* %out) #0 {
222  %out.gep.1 = getelementptr i8, i8* %out, i8 1
223  %out.gep.2 = getelementptr i8, i8* %out, i8 2
224
225  store i8 9, i8* %out, align 1
226  store i8 1, i8* %out.gep.1, align 1
227  store i8 23, i8* %out.gep.2, align 1
228  ret void
229}
230
231attributes #0 = { nounwind }
232