mirror of
https://github.com/RPCSX/llvm.git
synced 2026-01-31 01:05:23 +01:00
d019e8638a
This switches to the workaround that HSA defaults to for the mesa path. This should be applied to the 4.0 branch. Patch by Vedran Miletić <vedran@miletic.net> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292982 91177308-0d34-0410-b5e6-96231b3b80d8
40 lines
1.6 KiB
LLVM
40 lines
1.6 KiB
LLVM
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
|
|
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
|
|
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
|
|
|
|
; FUNC-LABEL: {{^}}atomic_sub_local:
|
|
; R600: LDS_SUB *
|
|
; SI: ds_sub_u32
|
|
define void @atomic_sub_local(i32 addrspace(3)* %local) {
|
|
%unused = atomicrmw volatile sub i32 addrspace(3)* %local, i32 5 seq_cst
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}atomic_sub_local_const_offset:
|
|
; R600: LDS_SUB *
|
|
; SI: ds_sub_u32 v{{[0-9]+}}, v{{[0-9]+}} offset:16
|
|
define void @atomic_sub_local_const_offset(i32 addrspace(3)* %local) {
|
|
%gep = getelementptr i32, i32 addrspace(3)* %local, i32 4
|
|
%val = atomicrmw volatile sub i32 addrspace(3)* %gep, i32 5 seq_cst
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}atomic_sub_ret_local:
|
|
; R600: LDS_SUB_RET *
|
|
; SI: ds_sub_rtn_u32
|
|
define void @atomic_sub_ret_local(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
|
|
%val = atomicrmw volatile sub i32 addrspace(3)* %local, i32 5 seq_cst
|
|
store i32 %val, i32 addrspace(1)* %out
|
|
ret void
|
|
}
|
|
|
|
; FUNC-LABEL: {{^}}atomic_sub_ret_local_const_offset:
|
|
; R600: LDS_SUB_RET *
|
|
; SI: ds_sub_rtn_u32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:20
|
|
define void @atomic_sub_ret_local_const_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %local) {
|
|
%gep = getelementptr i32, i32 addrspace(3)* %local, i32 5
|
|
%val = atomicrmw volatile sub i32 addrspace(3)* %gep, i32 5 seq_cst
|
|
store i32 %val, i32 addrspace(1)* %out
|
|
ret void
|
|
}
|