|
121
|
1 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
|
|
|
2 ; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
|
|
|
3 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
|
|
120
|
4
|
|
|
5 ; FUNC-LABEL: {{^}}global_load_f64:
|
|
|
6 ; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
|
|
|
7 ; GCN-NOHSA: buffer_store_dwordx2 [[VAL]]
|
|
|
8
|
|
|
9 ; GCN-HSA: flat_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
|
|
|
10 ; GCN-HSA: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, [[VAL]]
|
|
121
|
11 define amdgpu_kernel void @global_load_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
|
|
120
|
12 %ld = load double, double addrspace(1)* %in
|
|
|
13 store double %ld, double addrspace(1)* %out
|
|
|
14 ret void
|
|
|
15 }
|
|
|
16
|
|
|
17 ; FUNC-LABEL: {{^}}global_load_v2f64:
|
|
|
18 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
19 ; GCN-HSA: flat_load_dwordx4
|
|
121
|
20 define amdgpu_kernel void @global_load_v2f64(<2 x double> addrspace(1)* %out, <2 x double> addrspace(1)* %in) #0 {
|
|
120
|
21 entry:
|
|
|
22 %ld = load <2 x double>, <2 x double> addrspace(1)* %in
|
|
|
23 store <2 x double> %ld, <2 x double> addrspace(1)* %out
|
|
|
24 ret void
|
|
|
25 }
|
|
|
26
|
|
|
27 ; FUNC-LABEL: {{^}}global_load_v3f64:
|
|
|
28 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
29 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
30 ; GCN-HSA: flat_load_dwordx4
|
|
|
31 ; GCN-HSA: flat_load_dwordx4
|
|
121
|
32 define amdgpu_kernel void @global_load_v3f64(<3 x double> addrspace(1)* %out, <3 x double> addrspace(1)* %in) #0 {
|
|
120
|
33 entry:
|
|
|
34 %ld = load <3 x double>, <3 x double> addrspace(1)* %in
|
|
|
35 store <3 x double> %ld, <3 x double> addrspace(1)* %out
|
|
|
36 ret void
|
|
|
37 }
|
|
|
38
|
|
|
39 ; FUNC-LABEL: {{^}}global_load_v4f64:
|
|
|
40 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
41 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
42
|
|
|
43 ; GCN-HSA: flat_load_dwordx4
|
|
|
44 ; GCN-HSA: flat_load_dwordx4
|
|
121
|
45 define amdgpu_kernel void @global_load_v4f64(<4 x double> addrspace(1)* %out, <4 x double> addrspace(1)* %in) #0 {
|
|
120
|
46 entry:
|
|
|
47 %ld = load <4 x double>, <4 x double> addrspace(1)* %in
|
|
|
48 store <4 x double> %ld, <4 x double> addrspace(1)* %out
|
|
|
49 ret void
|
|
|
50 }
|
|
|
51
|
|
|
52 ; FUNC-LABEL: {{^}}global_load_v8f64:
|
|
|
53 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
54 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
55 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
56 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
57
|
|
|
58 ; GCN-HSA: flat_load_dwordx4
|
|
|
59 ; GCN-HSA: flat_load_dwordx4
|
|
|
60 ; GCN-HSA: flat_load_dwordx4
|
|
|
61 ; GCN-HSA: flat_load_dwordx4
|
|
121
|
62 define amdgpu_kernel void @global_load_v8f64(<8 x double> addrspace(1)* %out, <8 x double> addrspace(1)* %in) #0 {
|
|
120
|
63 entry:
|
|
|
64 %ld = load <8 x double>, <8 x double> addrspace(1)* %in
|
|
|
65 store <8 x double> %ld, <8 x double> addrspace(1)* %out
|
|
|
66 ret void
|
|
|
67 }
|
|
|
68
|
|
|
69 ; FUNC-LABEL: {{^}}global_load_v16f64:
|
|
|
70 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
71 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
72 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
73 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
74 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
75 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
76 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
77 ; GCN-NOHSA: buffer_load_dwordx4
|
|
|
78
|
|
|
79 ; GCN-HSA: flat_load_dwordx4
|
|
|
80 ; GCN-HSA: flat_load_dwordx4
|
|
|
81 ; GCN-HSA: flat_load_dwordx4
|
|
|
82 ; GCN-HSA: flat_load_dwordx4
|
|
|
83 ; GCN-HSA: flat_load_dwordx4
|
|
|
84 ; GCN-HSA: flat_load_dwordx4
|
|
|
85 ; GCN-HSA: flat_load_dwordx4
|
|
|
86 ; GCN-HSA: flat_load_dwordx4
|
|
121
|
87 define amdgpu_kernel void @global_load_v16f64(<16 x double> addrspace(1)* %out, <16 x double> addrspace(1)* %in) #0 {
|
|
120
|
88 entry:
|
|
|
89 %ld = load <16 x double>, <16 x double> addrspace(1)* %in
|
|
|
90 store <16 x double> %ld, <16 x double> addrspace(1)* %out
|
|
|
91 ret void
|
|
|
92 }
|
|
|
93
|
|
|
94 attributes #0 = { nounwind }
|
