llvm-6502/test/CodeGen/AMDGPU/global_atomics.ll
Tom Stellard f799b25cfc AMDGPU/SI: Add VI patterns to select FLAT instructions for global memory ops
Summary:
The MUBUF addr64 bit has been removed on VI, so we must use FLAT
instructions when the pointer is stored in VGPRs.

Reviewers: arsenm

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D11067

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@242673 91177308-0d34-0410-b5e6-96231b3b80d8
2015-07-20 14:28:41 +00:00

922 lines
39 KiB
LLVM

; RUN: llc < %s -march=amdgcn -mcpu=SI -verify-machineinstrs | FileCheck --check-prefix=GCN --check-prefix=SI --check-prefix=FUNC %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck --check-prefix=GCN --check-prefix=VI --check-prefix=FUNC %s
; FUNC-LABEL: {{^}}atomic_add_i32_offset:
; GCN: buffer_atomic_add v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_add_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_ret_offset:
; GCN: buffer_atomic_add [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_add_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_addr64_offset:
; SI: buffer_atomic_add v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_add v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_add_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_ret_addr64_offset:
; SI: buffer_atomic_add [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_add [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_add_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32:
; GCN: buffer_atomic_add v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_add_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile add i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_ret:
; GCN: buffer_atomic_add [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_add_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile add i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_addr64:
; SI: buffer_atomic_add v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_add v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_add_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile add i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_add_i32_ret_addr64:
; SI: buffer_atomic_add [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_add [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_add_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile add i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_offset:
; GCN: buffer_atomic_and v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_and_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_ret_offset:
; GCN: buffer_atomic_and [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_and_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_addr64_offset:
; SI: buffer_atomic_and v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_and v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_and_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_ret_addr64_offset:
; SI: buffer_atomic_and [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_and [[RET:v[0-9]]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_and_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32:
; GCN: buffer_atomic_and v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_and_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile and i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_ret:
; GCN: buffer_atomic_and [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_and_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile and i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_addr64:
; SI: buffer_atomic_and v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_and v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_and_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile and i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_and_i32_ret_addr64:
; SI: buffer_atomic_and [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_and [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_and_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile and i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_offset:
; GCN: buffer_atomic_sub v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_sub_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_ret_offset:
; GCN: buffer_atomic_sub [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_sub_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_addr64_offset:
; SI: buffer_atomic_sub v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_sub v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_sub_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_ret_addr64_offset:
; SI: buffer_atomic_sub [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_sub [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_sub_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32:
; GCN: buffer_atomic_sub v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_sub_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile sub i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_ret:
; GCN: buffer_atomic_sub [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_sub_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile sub i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_addr64:
; SI: buffer_atomic_sub v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_sub v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_sub_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile sub i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_sub_i32_ret_addr64:
; SI: buffer_atomic_sub [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_sub [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_sub_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile sub i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_offset:
; GCN: buffer_atomic_smax v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_max_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_ret_offset:
; GCN: buffer_atomic_smax [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_max_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_addr64_offset:
; SI: buffer_atomic_smax v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smax v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_max_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_ret_addr64_offset:
; SI: buffer_atomic_smax [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smax [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_max_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32:
; GCN: buffer_atomic_smax v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_max_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile max i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_ret:
; GCN: buffer_atomic_smax [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_max_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile max i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_addr64:
; SI: buffer_atomic_smax v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smax v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_max_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile max i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_max_i32_ret_addr64:
; SI: buffer_atomic_smax [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smax [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_max_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile max i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_offset:
; GCN: buffer_atomic_umax v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_umax_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_ret_offset:
; GCN: buffer_atomic_umax [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umax_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_addr64_offset:
; SI: buffer_atomic_umax v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umax v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_umax_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_ret_addr64_offset:
; SI: buffer_atomic_umax [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umax [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umax_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32:
; GCN: buffer_atomic_umax v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_umax_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile umax i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_ret:
; GCN: buffer_atomic_umax [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_umax_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile umax i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_addr64:
; SI: buffer_atomic_umax v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umax v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_umax_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile umax i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umax_i32_ret_addr64:
; SI: buffer_atomic_umax [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umax [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umax_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile umax i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_offset:
; GCN: buffer_atomic_smin v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_min_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_ret_offset:
; GCN: buffer_atomic_smin [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_min_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_addr64_offset:
; SI: buffer_atomic_smin v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smin v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_min_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_ret_addr64_offset:
; SI: buffer_atomic_smin [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smin [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_min_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32:
; GCN: buffer_atomic_smin v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_min_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile min i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_ret:
; GCN: buffer_atomic_smin [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_min_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile min i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_addr64:
; SI: buffer_atomic_smin v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smin v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_min_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile min i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_min_i32_ret_addr64:
; SI: buffer_atomic_smin [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_smin [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_min_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile min i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_offset:
; GCN: buffer_atomic_umin v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_umin_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_ret_offset:
; GCN: buffer_atomic_umin [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umin_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_addr64_offset:
; SI: buffer_atomic_umin v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umin v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_umin_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_ret_addr64_offset:
; SI: buffer_atomic_umin [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umin [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umin_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32:
; GCN: buffer_atomic_umin v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_umin_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile umin i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_ret:
; SI: buffer_atomic_umin [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_umin_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile umin i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_addr64:
; SI: buffer_atomic_umin v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umin v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_umin_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile umin i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_umin_i32_ret_addr64:
; SI: buffer_atomic_umin [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_umin [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_umin_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile umin i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_offset:
; GCN: buffer_atomic_or v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_or_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_ret_offset:
; GCN: buffer_atomic_or [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_or_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_addr64_offset:
; SI: buffer_atomic_or v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_or v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_or_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_ret_addr64_offset:
; SI: buffer_atomic_or [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_or [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_or_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32:
; GCN: buffer_atomic_or v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_or_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile or i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_ret:
; GCN: buffer_atomic_or [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_or_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile or i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_addr64:
; SI: buffer_atomic_or v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_or v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_or_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile or i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_or_i32_ret_addr64:
; SI: buffer_atomic_or [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_or [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_or_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile or i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_offset:
; GCN: buffer_atomic_swap v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_xchg_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_ret_offset:
; GCN: buffer_atomic_swap [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xchg_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_addr64_offset:
; SI: buffer_atomic_swap v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
define void @atomic_xchg_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_ret_addr64_offset:
; SI: buffer_atomic_swap [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_swap [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xchg_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32:
; GCN: buffer_atomic_swap v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_xchg_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_ret:
; GCN: buffer_atomic_swap [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_xchg_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_addr64:
; SI: buffer_atomic_swap v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_swap v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_xchg_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile xchg i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xchg_i32_ret_addr64:
; SI: buffer_atomic_swap [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_swap [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xchg_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile xchg i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_offset:
; GCN: buffer_atomic_xor v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
define void @atomic_xor_i32_offset(i32 addrspace(1)* %out, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_ret_offset:
; GCN: buffer_atomic_xor [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc {{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xor_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
%0 = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_addr64_offset:
; SI: buffer_atomic_xor v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_xor v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_xor_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_ret_addr64_offset:
; SI: buffer_atomic_xor [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_xor [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xor_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
%0 = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32:
; GCN: buffer_atomic_xor v{{[0-9]+}}, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
define void @atomic_xor_i32(i32 addrspace(1)* %out, i32 %in) {
entry:
%0 = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_ret:
; GCN: buffer_atomic_xor [[RET:v[0-9]+]], s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; GCN: buffer_store_dword [[RET]]
define void @atomic_xor_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in) {
entry:
%0 = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_addr64:
; SI: buffer_atomic_xor v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_xor v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
define void @atomic_xor_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
ret void
}
; FUNC-LABEL: {{^}}atomic_xor_i32_ret_addr64:
; SI: buffer_atomic_xor [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: s_movk_i32 flat_scratch_lo, 0x0
; VI: s_movk_i32 flat_scratch_hi, 0x0
; VI: flat_atomic_xor [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
; GCN: buffer_store_dword [[RET]]
define void @atomic_xor_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index) {
entry:
%ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
%0 = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
store i32 %0, i32 addrspace(1)* %out2
ret void
}