1 ; RUN: llc -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
4 ; FUNC-LABEL: {{^}}test_copy_v4i8:
5 ; SI: buffer_load_dword [[REG:v[0-9]+]]
6 ; SI: buffer_store_dword [[REG]]
8 define void @test_copy_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
9 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
10 store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
14 ; FUNC-LABEL: {{^}}test_copy_v4i8_x2:
15 ; SI: buffer_load_dword [[REG:v[0-9]+]]
16 ; SI: buffer_store_dword [[REG]]
17 ; SI: buffer_store_dword [[REG]]
19 define void @test_copy_v4i8_x2(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
20 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
21 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
22 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
26 ; FUNC-LABEL: {{^}}test_copy_v4i8_x3:
27 ; SI: buffer_load_dword [[REG:v[0-9]+]]
28 ; SI: buffer_store_dword [[REG]]
29 ; SI: buffer_store_dword [[REG]]
30 ; SI: buffer_store_dword [[REG]]
32 define void @test_copy_v4i8_x3(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind {
33 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
34 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
35 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
36 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
40 ; FUNC-LABEL: {{^}}test_copy_v4i8_x4:
41 ; SI: buffer_load_dword [[REG:v[0-9]+]]
42 ; SI: buffer_store_dword [[REG]]
43 ; SI: buffer_store_dword [[REG]]
44 ; SI: buffer_store_dword [[REG]]
45 ; SI: buffer_store_dword [[REG]]
47 define void @test_copy_v4i8_x4(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %out3, <4 x i8> addrspace(1)* %in) nounwind {
48 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
49 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
50 store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
51 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
52 store <4 x i8> %val, <4 x i8> addrspace(1)* %out3, align 4
56 ; FUNC-LABEL: {{^}}test_copy_v4i8_extra_use:
57 ; SI: buffer_load_ubyte
58 ; SI: buffer_load_ubyte
59 ; SI: buffer_load_ubyte
60 ; SI: buffer_load_ubyte
65 ; SI-DAG: buffer_store_byte
66 ; SI-DAG: buffer_store_byte
67 ; SI-DAG: buffer_store_byte
68 ; SI-DAG: buffer_store_byte
69 ; SI-DAG: buffer_store_byte
70 ; SI-DAG: buffer_store_byte
71 ; SI-DAG: buffer_store_byte
72 ; SI_DAG: buffer_store_byte
74 ; After scalarizing v4i8 loads is fixed.
75 ; XSI: buffer_load_dword
80 ; XSI: buffer_store_dword
81 ; XSI: buffer_store_dword
84 define void @test_copy_v4i8_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
85 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
86 %add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9>
87 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
88 store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4
92 ; FUNC-LABEL: {{^}}test_copy_v4i8_x2_extra_use:
93 ; SI: buffer_load_ubyte
94 ; SI: buffer_load_ubyte
95 ; SI: buffer_load_ubyte
96 ; SI: buffer_load_ubyte
101 ; SI-DAG: buffer_store_byte
102 ; SI-DAG: buffer_store_byte
103 ; SI-DAG: buffer_store_byte
104 ; SI-DAG: buffer_store_byte
105 ; SI-DAG: buffer_store_byte
106 ; SI-DAG: buffer_store_byte
107 ; SI-DAG: buffer_store_byte
108 ; SI_DAG: buffer_store_byte
109 ; SI-DAG: buffer_store_byte
110 ; SI-DAG: buffer_store_byte
111 ; SI-DAG: buffer_store_byte
112 ; SI_DAG: buffer_store_byte
114 ; XSI: buffer_load_dword
116 ; XSI: buffer_store_dword
118 ; XSI: buffer_store_dword
119 ; XSI-NEXT: buffer_store_dword
122 define void @test_copy_v4i8_x2_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind {
123 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
124 %add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9>
125 store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
126 store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4
127 store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
131 ; FUNC-LABEL: {{^}}test_copy_v3i8:
135 define void @test_copy_v3i8(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind {
136 %val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 4
137 store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 4
141 ; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_load:
142 ; SI: buffer_load_ubyte
143 ; SI: buffer_load_ubyte
144 ; SI: buffer_load_ubyte
145 ; SI: buffer_load_ubyte
147 define void @test_copy_v4i8_volatile_load(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
148 %val = load volatile <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
149 store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
153 ; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_store:
154 ; SI: buffer_load_ubyte
155 ; SI: buffer_load_ubyte
156 ; SI: buffer_load_ubyte
157 ; SI: buffer_load_ubyte
158 ; SI: buffer_store_byte
159 ; SI: buffer_store_byte
160 ; SI: buffer_store_byte
161 ; SI: buffer_store_byte
163 define void @test_copy_v4i8_volatile_store(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
164 %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
165 store volatile <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4