summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll')
-rw-r--r--llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll384
1 files changed, 263 insertions, 121 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
index 8fb47c5eb00..e63b5a9de22 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_i64.ll
@@ -1,8 +1,11 @@
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CI,CIVI %s
+; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,CIVI %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9 %s
; GCN-LABEL: {{^}}atomic_add_i64_offset:
-; GCN: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+
+; GFX9: global_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], off offset:32{{$}}
define amdgpu_kernel void @atomic_add_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -11,8 +14,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_add_i64_ret_offset:
-; GCN: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_add_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -24,6 +29,7 @@ entry:
; GCN-LABEL: {{^}}atomic_add_i64_addr64_offset:
; CI: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_add_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}{{$}}
+; GFX9: global_atomic_add_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_add_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -35,7 +41,9 @@ entry:
; GCN-LABEL: {{^}}atomic_add_i64_ret_addr64_offset:
; CI: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_add_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -46,7 +54,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_add_i64:
-; GCN: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; SIVI: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_add_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_add_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile add i64 addrspace(1)* %out, i64 %in seq_cst
@@ -54,8 +63,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_add_i64_ret:
-; GCN: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_add_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile add i64 addrspace(1)* %out, i64 %in seq_cst
@@ -66,6 +77,7 @@ entry:
; GCN-LABEL: {{^}}atomic_add_i64_addr64:
; CI: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_add_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_add_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_add_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -76,7 +88,9 @@ entry:
; GCN-LABEL: {{^}}atomic_add_i64_ret_addr64:
; CI: buffer_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_add_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_add_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -86,7 +100,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_and_i64_offset:
-; GCN: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_and_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -95,8 +110,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_and_i64_ret_offset:
-; GCN: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_and_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -108,6 +125,7 @@ entry:
; GCN-LABEL: {{^}}atomic_and_i64_addr64_offset:
; CI: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_and_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -119,7 +137,9 @@ entry:
; GCN-LABEL: {{^}}atomic_and_i64_ret_addr64_offset:
; CI: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_and_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -130,7 +150,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_and_i64:
-; GCN: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_and_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile and i64 addrspace(1)* %out, i64 %in seq_cst
@@ -138,8 +159,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_and_i64_ret:
-; GCN: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_and_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile and i64 addrspace(1)* %out, i64 %in seq_cst
@@ -150,6 +173,7 @@ entry:
; GCN-LABEL: {{^}}atomic_and_i64_addr64:
; CI: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_and_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_and_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -160,7 +184,9 @@ entry:
; GCN-LABEL: {{^}}atomic_and_i64_ret_addr64:
; CI: buffer_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_and_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_and_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -170,7 +196,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_sub_i64_offset:
-; GCN: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_sub_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -179,8 +206,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_sub_i64_ret_offset:
-; GCN: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_sub_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -192,6 +221,7 @@ entry:
; GCN-LABEL: {{^}}atomic_sub_i64_addr64_offset:
; CI: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_sub_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -203,7 +233,9 @@ entry:
; GCN-LABEL: {{^}}atomic_sub_i64_ret_addr64_offset:
; CI: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_sub_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -214,7 +246,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_sub_i64:
-; GCN: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_sub_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile sub i64 addrspace(1)* %out, i64 %in seq_cst
@@ -222,8 +255,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_sub_i64_ret:
-; GCN: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_sub_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile sub i64 addrspace(1)* %out, i64 %in seq_cst
@@ -234,6 +269,7 @@ entry:
; GCN-LABEL: {{^}}atomic_sub_i64_addr64:
; CI: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_sub_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_sub_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -244,7 +280,9 @@ entry:
; GCN-LABEL: {{^}}atomic_sub_i64_ret_addr64:
; CI: buffer_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_sub_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_sub_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -254,7 +292,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_max_i64_offset:
-; GCN: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_max_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -263,8 +302,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_max_i64_ret_offset:
-; GCN: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_max_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -276,6 +317,7 @@ entry:
; GCN-LABEL: {{^}}atomic_max_i64_addr64_offset:
; CI: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_max_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -287,7 +329,9 @@ entry:
; GCN-LABEL: {{^}}atomic_max_i64_ret_addr64_offset:
; CI: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -298,7 +342,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_max_i64:
-; GCN: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_max_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile max i64 addrspace(1)* %out, i64 %in seq_cst
@@ -306,8 +351,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_max_i64_ret:
-; GCN: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_max_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile max i64 addrspace(1)* %out, i64 %in seq_cst
@@ -318,6 +365,7 @@ entry:
; GCN-LABEL: {{^}}atomic_max_i64_addr64:
; CI: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_smax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_max_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -328,7 +376,9 @@ entry:
; GCN-LABEL: {{^}}atomic_max_i64_ret_addr64:
; CI: buffer_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_max_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -338,7 +388,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umax_i64_offset:
-; GCN: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_umax_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -347,8 +398,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umax_i64_ret_offset:
-; GCN: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_umax_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -360,6 +413,7 @@ entry:
; GCN-LABEL: {{^}}atomic_umax_i64_addr64_offset:
; CI: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; FX9: global_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} offset:32{{$}}
define amdgpu_kernel void @atomic_umax_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -371,7 +425,9 @@ entry:
; GCN-LABEL: {{^}}atomic_umax_i64_ret_addr64_offset:
; CI: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -382,7 +438,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umax_i64:
-; GCN: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_umax_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile umax i64 addrspace(1)* %out, i64 %in seq_cst
@@ -390,8 +447,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umax_i64_ret:
-; GCN: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_umax_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile umax i64 addrspace(1)* %out, i64 %in seq_cst
@@ -402,6 +461,7 @@ entry:
; GCN-LABEL: {{^}}atomic_umax_i64_addr64:
; CI: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_umax_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_umax_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -412,7 +472,9 @@ entry:
; GCN-LABEL: {{^}}atomic_umax_i64_ret_addr64:
; CI: buffer_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umax_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_umax_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -422,7 +484,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_min_i64_offset:
-; GCN: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_min_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -431,8 +494,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_min_i64_ret_offset:
-; GCN: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_min_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -444,6 +509,7 @@ entry:
; GCN-LABEL: {{^}}atomic_min_i64_addr64_offset:
; CI: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_min_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -455,7 +521,9 @@ entry:
; GCN-LABEL: {{^}}atomic_min_i64_ret_addr64_offset:
; CI: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -466,7 +534,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_min_i64:
-; GCN: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_min_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile min i64 addrspace(1)* %out, i64 %in seq_cst
@@ -474,8 +543,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_min_i64_ret:
-; GCN: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_min_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile min i64 addrspace(1)* %out, i64 %in seq_cst
@@ -486,6 +557,7 @@ entry:
; GCN-LABEL: {{^}}atomic_min_i64_addr64:
; CI: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_smin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_min_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -496,7 +568,9 @@ entry:
; GCN-LABEL: {{^}}atomic_min_i64_ret_addr64:
; CI: buffer_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_smin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_min_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -506,7 +580,9 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umin_i64_offset:
-; GCN: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+
+; GFX9: global_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_umin_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -515,8 +591,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umin_i64_ret_offset:
-; GCN: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_umin_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -528,6 +606,7 @@ entry:
; GCN-LABEL: {{^}}atomic_umin_i64_addr64_offset:
; CI: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_umin_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -539,7 +618,9 @@ entry:
; GCN-LABEL: {{^}}atomic_umin_i64_ret_addr64_offset:
; CI: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_umin_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -550,7 +631,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umin_i64:
-; GCN: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_umin_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile umin i64 addrspace(1)* %out, i64 %in seq_cst
@@ -558,8 +640,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_umin_i64_ret:
-; CI: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_umin_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile umin i64 addrspace(1)* %out, i64 %in seq_cst
@@ -570,6 +654,7 @@ entry:
; GCN-LABEL: {{^}}atomic_umin_i64_addr64:
; CI: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_umin_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_umin_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -580,7 +665,9 @@ entry:
; GCN-LABEL: {{^}}atomic_umin_i64_ret_addr64:
; CI: buffer_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_umin_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_umin_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -590,7 +677,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_or_i64_offset:
-; GCN: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_or_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -599,8 +687,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_or_i64_ret_offset:
-; GCN: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_or_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -612,6 +702,7 @@ entry:
; GCN-LABEL: {{^}}atomic_or_i64_addr64_offset:
; CI: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_or_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -623,7 +714,9 @@ entry:
; GCN-LABEL: {{^}}atomic_or_i64_ret_addr64_offset:
; CI: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_or_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -634,7 +727,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_or_i64:
-; GCN: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_or_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile or i64 addrspace(1)* %out, i64 %in seq_cst
@@ -642,8 +736,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_or_i64_ret:
-; GCN: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_or_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile or i64 addrspace(1)* %out, i64 %in seq_cst
@@ -654,6 +750,7 @@ entry:
; GCN-LABEL: {{^}}atomic_or_i64_addr64:
; CI: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_or_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_or_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -664,7 +761,9 @@ entry:
; GCN-LABEL: {{^}}atomic_or_i64_ret_addr64:
; CI: buffer_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_or_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_or_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -674,7 +773,9 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xchg_i64_offset:
-; GCN: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+
+; GFX9: global_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_xchg_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -683,8 +784,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xchg_i64_ret_offset:
-; GCN: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_xchg_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -696,6 +799,7 @@ entry:
; GCN-LABEL: {{^}}atomic_xchg_i64_addr64_offset:
; CI: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}{{$}}
+; GFX9: global_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_xchg_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -707,7 +811,9 @@ entry:
; GCN-LABEL: {{^}}atomic_xchg_i64_ret_addr64_offset:
; CI: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_xchg_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -718,7 +824,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xchg_i64:
-; GCN: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_xchg_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %out, i64 %in seq_cst
@@ -726,8 +833,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xchg_i64_ret:
-; GCN: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_xchg_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %out, i64 %in seq_cst
@@ -738,6 +847,7 @@ entry:
; GCN-LABEL: {{^}}atomic_xchg_i64_addr64:
; CI: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_swap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_xchg_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -748,7 +858,9 @@ entry:
; GCN-LABEL: {{^}}atomic_xchg_i64_ret_addr64:
; CI: buffer_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_swap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_xchg_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -758,7 +870,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xor_i64_offset:
-; GCN: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; CIVI: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_xor_i64_offset(i64 addrspace(1)* %out, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -767,8 +880,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xor_i64_ret_offset:
-; GCN: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_xor_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -780,6 +895,7 @@ entry:
; GCN-LABEL: {{^}}atomic_xor_i64_addr64_offset:
; CI: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_xor_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -791,7 +907,9 @@ entry:
; GCN-LABEL: {{^}}atomic_xor_i64_ret_addr64_offset:
; CI: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_xor_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -802,7 +920,8 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xor_i64:
-; GCN: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; CIVI: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_xor_i64(i64 addrspace(1)* %out, i64 %in) {
entry:
%tmp0 = atomicrmw volatile xor i64 addrspace(1)* %out, i64 %in seq_cst
@@ -810,8 +929,10 @@ entry:
}
; GCN-LABEL: {{^}}atomic_xor_i64_ret:
-; GCN: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_xor_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in) {
entry:
%tmp0 = atomicrmw volatile xor i64 addrspace(1)* %out, i64 %in seq_cst
@@ -822,6 +943,7 @@ entry:
; GCN-LABEL: {{^}}atomic_xor_i64_addr64:
; CI: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]$}}
+; GFX9: global_atomic_xor_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_xor_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -832,7 +954,9 @@ entry:
; GCN-LABEL: {{^}}atomic_xor_i64_ret_addr64:
; CI: buffer_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}} glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_atomic_xor_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off glc{{$}}
define amdgpu_kernel void @atomic_xor_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -842,15 +966,9 @@ entry:
}
-
-
-
-
-
-
-
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_offset:
-; GCN: buffer_atomic_cmpswap_x2 v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_offset:
+; CIVI: buffer_atomic_cmpswap_x2 v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
+; GFX9: global_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_offset(i64 addrspace(1)* %out, i64 %in, i64 %old) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -858,9 +976,11 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_soffset:
-; GCN: s_mov_b32 [[SREG:s[0-9]+]], 0x11940
-; GCN: buffer_atomic_cmpswap_x2 v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], [[SREG]]{{$}}
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_soffset:
+; CIVI: s_mov_b32 [[SREG:s[0-9]+]], 0x11940
+; CIVI: buffer_atomic_cmpswap_x2 v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], [[SREG]]{{$}}
+
+; GFX9: global_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_soffset(i64 addrspace(1)* %out, i64 %in, i64 %old) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 9000
@@ -868,9 +988,11 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_offset:
-; GCN: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]{{:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; GCN: buffer_store_dwordx2 v{{\[}}[[RET]]:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_ret_offset:
+; CIVI: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]{{:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; CIVI: buffer_store_dwordx2 v{{\[}}[[RET]]:
+
+; GFX9: global_atomic_cmpswap_x2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v{{\[[0-9]+:[0-9]+\]}}, off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %old) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -880,10 +1002,10 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_addr64_offset:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_addr64_offset:
; CI: buffer_atomic_cmpswap_x2 v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
-
; VI: flat_atomic_cmpswap_x2 v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}]{{$}}
+; GFX9: global_atomic_cmpswap_x2 v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], off offset:32{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index, i64 %old) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -892,10 +1014,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64_offset:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64_offset:
; CI: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}] glc{{$}}
-; GCN: buffer_store_dwordx2 v{{\[}}[[RET]]:
+; CIVI: buffer_store_dwordx2 v{{\[}}[[RET]]:
+
+; GFX9: global_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index, i64 %old) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -906,17 +1030,20 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64:
-; GCN: buffer_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64:
+; CIVI: buffer_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+; GFX9: global_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], off{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64(i64 addrspace(1)* %out, i64 %in, i64 %old) {
entry:
%val = cmpxchg volatile i64 addrspace(1)* %out, i64 %old, i64 %in seq_cst seq_cst
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret:
-; GCN: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; GCN: buffer_store_dwordx2 v{{\[}}[[RET]]:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_ret:
+; CIVI: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; CIVI: buffer_store_dwordx2 v{{\[}}[[RET]]:
+
+; GFX9: global_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], off glc{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_ret(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %old) {
entry:
%val = cmpxchg volatile i64 addrspace(1)* %out, i64 %old, i64 %in seq_cst seq_cst
@@ -925,9 +1052,10 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_addr64:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_addr64:
; CI: buffer_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]{{$}}
+; GFX9: global_atomic_cmpswap_x2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], off{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index, i64 %old) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -935,10 +1063,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64:
+; GCN-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64:
; CI: buffer_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}] glc{{$}}
-; GCN: buffer_store_dwordx2 v{{\[}}[[RET]]:
+; CIVI: buffer_store_dwordx2 v{{\[}}[[RET]]:
+
+; GFX9: global_atomic_cmpswap_x2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], off glc{{$}}
define amdgpu_kernel void @atomic_cmpxchg_i64_ret_addr64(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %index, i64 %old) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -948,10 +1078,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_load_i64_offset:
+; GCN-LABEL: {{^}}atomic_load_i64_offset:
; CI: buffer_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
; VI: flat_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}] glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}], off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_load_i64_offset(i64 addrspace(1)* %in, i64 addrspace(1)* %out) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %in, i64 4
@@ -960,10 +1092,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_load_i64:
+; GCN-LABEL: {{^}}atomic_load_i64:
; CI: buffer_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
; VI: flat_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}] glc
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}], off glc{{$}}
define amdgpu_kernel void @atomic_load_i64(i64 addrspace(1)* %in, i64 addrspace(1)* %out) {
entry:
%val = load atomic i64, i64 addrspace(1)* %in seq_cst, align 8
@@ -971,10 +1105,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_load_i64_addr64_offset:
+; GCN-LABEL: {{^}}atomic_load_i64_addr64_offset:
; CI: buffer_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
; VI: flat_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}] glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], off offset:32 glc{{$}}
define amdgpu_kernel void @atomic_load_i64_addr64_offset(i64 addrspace(1)* %in, i64 addrspace(1)* %out, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %in, i64 %index
@@ -984,10 +1120,12 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_load_i64_addr64:
+; GCN-LABEL: {{^}}atomic_load_i64_addr64:
; CI: buffer_load_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
; VI: flat_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}] glc{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
+; CIVI: buffer_store_dwordx2 [[RET]]
+
+; GFX9: global_load_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], off glc{{$}}
define amdgpu_kernel void @atomic_load_i64_addr64(i64 addrspace(1)* %in, i64 addrspace(1)* %out, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %in, i64 %index
@@ -996,9 +1134,10 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_store_i64_offset:
+; GCN-LABEL: {{^}}atomic_store_i64_offset:
; CI: buffer_store_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32{{$}}
; VI: flat_store_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}]{{$}}
+; GFX9: global_store_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}], off offset:32{{$}}
define amdgpu_kernel void @atomic_store_i64_offset(i64 %in, i64 addrspace(1)* %out) {
entry:
%gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
@@ -1006,18 +1145,20 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_store_i64:
+; GCN-LABEL: {{^}}atomic_store_i64:
; CI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}]{{$}}
+; GFX9: global_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}], off{{$}}
define amdgpu_kernel void @atomic_store_i64(i64 %in, i64 addrspace(1)* %out) {
entry:
store atomic i64 %in, i64 addrspace(1)* %out seq_cst, align 8
ret void
}
-; FUNC-LABEL: {{^}}atomic_store_i64_addr64_offset:
+; GCN-LABEL: {{^}}atomic_store_i64_addr64_offset:
; CI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32{{$}}
; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}]{{$}}
+; GFX9: global_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], off offset:32{{$}}
define amdgpu_kernel void @atomic_store_i64_addr64_offset(i64 %in, i64 addrspace(1)* %out, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
@@ -1026,9 +1167,10 @@ entry:
ret void
}
-; FUNC-LABEL: {{^}}atomic_store_i64_addr64:
+; GCN-LABEL: {{^}}atomic_store_i64_addr64:
; CI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}]{{$}}
+; GFX9: global_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}], off{{$}}
define amdgpu_kernel void @atomic_store_i64_addr64(i64 %in, i64 addrspace(1)* %out, i64 %index) {
entry:
%ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
OpenPOWER on IntegriCloud