121
|
1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX8 -check-prefix=NOAUTO %s
|
|
2 ; RUN: llc -march=amdgcn -mattr=+auto-waitcnt-before-barrier -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX8 -check-prefix=AUTO %s
|
|
3 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=NOAUTO %s
|
|
4 ; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=+auto-waitcnt-before-barrier -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=AUTO %s
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
5
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
6 ; GCN-LABEL: {{^}}test_barrier:
|
121
|
7 ; GFX8: buffer_store_dword
|
|
8 ; GFX9: global_store_dword
|
|
9 ; NOAUTO: s_waitcnt
|
|
10 ; AUTO-NOT: s_waitcnt
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
11 ; GCN: s_barrier
|
121
|
12 define amdgpu_kernel void @test_barrier(i32 addrspace(1)* %out, i32 %size) #0 {
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
13 entry:
|
120
|
14 %tmp = call i32 @llvm.amdgcn.workitem.id.x()
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
15 %tmp1 = getelementptr i32, i32 addrspace(1)* %out, i32 %tmp
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
16 store i32 %tmp, i32 addrspace(1)* %tmp1
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
17 call void @llvm.amdgcn.s.barrier()
|
121
|
18 %tmp3 = sub i32 %size, 1
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
19 %tmp4 = sub i32 %tmp3, %tmp
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
20 %tmp5 = getelementptr i32, i32 addrspace(1)* %out, i32 %tmp4
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
21 %tmp6 = load i32, i32 addrspace(1)* %tmp5
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
22 store i32 %tmp6, i32 addrspace(1)* %tmp1
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
23 ret void
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
24 }
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
25
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
26 declare void @llvm.amdgcn.s.barrier() #1
|
120
|
27 declare i32 @llvm.amdgcn.workitem.id.x() #2
|
100
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
28
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
29 attributes #0 = { nounwind }
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
30 attributes #1 = { convergent nounwind }
|
Miyagi Mitsuki <e135756@ie.u-ryukyu.ac.jp>
parents:
diff
changeset
|
31 attributes #2 = { nounwind readnone }
|