AMDGPU: Fix missing and broken check lines in atomic tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@272343 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/AMDGPU/global_atomics.ll b/test/CodeGen/AMDGPU/global_atomics.ll
index 30b197a3604fda7db16e270319e6dea5ff81fc90..743ad7c278be4b7b5ce3977d51949a1615b3baa7 100644 (file)
--- a/test/CodeGen/AMDGPU/global_atomics.ll
+++ b/test/CodeGen/AMDGPU/global_atomics.ll
@@ -1,13 +1,35 @@
-; 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
-
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
 
 ; FUNC-LABEL: {{^}}atomic_add_i32_offset:
 ; GCN: buffer_atomic_add v{{[0-9]+}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
 define void @atomic_add_i32_offset(i32 addrspace(1)* %out, i32 %in) {
 entry:
-  %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
-  %0  = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_add_i32_soffset:
+; GCN: s_mov_b32 [[SREG:s[0-9]+]], 0x8ca0
+; GCN: buffer_atomic_add v{{[0-9]+}}, off, s[{{[0-9]+}}:{{[0-9]+}}], [[SREG]]{{$}}
+define void @atomic_add_i32_soffset(i32 addrspace(1)* %out, i32 %in) {
+entry:
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 9000
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_add_i32_huge_offset:
+; SI-DAG: v_mov_b32_e32 v[[PTRLO:[0-9]+]], 0xdeac
+; SI-DAG: v_mov_b32_e32 v[[PTRHI:[0-9]+]], 0xabcd
+; SI: buffer_atomic_add v{{[0-9]+}}, v{{\[}}[[PTRLO]]:[[PTRHI]]{{\]}}, s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
+; VI: flat_atomic_add
+define void @atomic_add_i32_huge_offset(i32 addrspace(1)* %out, i32 %in) {
+entry:
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 47224239175595
+
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -16,21 +38,20 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, 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{{$}}
 ; VI: flat_atomic_add v[{{[0-9]+:[0-9]+}}], v{{[0-9]+$}}
-
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -41,9 +62,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile add i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -51,7 +72,7 @@ entry:
 ; GCN: buffer_atomic_add v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile add i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -60,8 +81,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile add i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -71,7 +92,7 @@ entry:
 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
+  %val = atomicrmw volatile add i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -82,8 +103,8 @@ entry:
 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
+  %val = atomicrmw volatile add i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -91,8 +112,8 @@ entry:
 ; GCN: buffer_atomic_and v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -101,9 +122,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -113,8 +134,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -125,9 +146,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile and i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -135,7 +156,7 @@ entry:
 ; GCN: buffer_atomic_and v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile and i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -144,8 +165,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile and i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -155,7 +176,7 @@ entry:
 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
+  %val = atomicrmw volatile and i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -166,8 +187,8 @@ entry:
 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
+  %val = atomicrmw volatile and i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -175,8 +196,8 @@ entry:
 ; GCN: buffer_atomic_sub v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -185,9 +206,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -197,8 +218,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -209,9 +230,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile sub i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -219,7 +240,7 @@ entry:
 ; GCN: buffer_atomic_sub v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile sub i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -228,8 +249,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile sub i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -239,7 +260,7 @@ entry:
 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
+  %val = atomicrmw volatile sub i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -250,8 +271,8 @@ entry:
 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
+  %val = atomicrmw volatile sub i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -259,8 +280,8 @@ entry:
 ; GCN: buffer_atomic_smax v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -269,9 +290,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -281,8 +302,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -293,9 +314,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile max i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -303,7 +324,7 @@ entry:
 ; GCN: buffer_atomic_smax v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile max i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -312,8 +333,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile max i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -323,7 +344,7 @@ entry:
 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
+  %val = atomicrmw volatile max i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -334,8 +355,8 @@ entry:
 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
+  %val = atomicrmw volatile max i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -343,8 +364,8 @@ entry:
 ; GCN: buffer_atomic_umax v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -353,9 +374,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -365,8 +386,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -377,9 +398,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile umax i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -387,7 +408,7 @@ entry:
 ; GCN: buffer_atomic_umax v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile umax i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -396,8 +417,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile umax i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -407,7 +428,7 @@ entry:
 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
+  %val = atomicrmw volatile umax i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -418,8 +439,8 @@ entry:
 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
+  %val = atomicrmw volatile umax i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -427,8 +448,8 @@ entry:
 ; GCN: buffer_atomic_smin v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -437,9 +458,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -449,8 +470,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -461,9 +482,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile min i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -471,7 +492,7 @@ entry:
 ; GCN: buffer_atomic_smin v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile min i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -480,8 +501,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile min i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -491,7 +512,7 @@ entry:
 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
+  %val = atomicrmw volatile min i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -502,8 +523,8 @@ entry:
 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
+  %val = atomicrmw volatile min i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -511,8 +532,8 @@ entry:
 ; GCN: buffer_atomic_umin v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -521,9 +542,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -533,8 +554,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -545,9 +566,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile umin i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -555,7 +576,7 @@ entry:
 ; GCN: buffer_atomic_umin v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile umin i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -564,8 +585,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile umin i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -575,7 +596,7 @@ entry:
 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
+  %val = atomicrmw volatile umin i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -586,8 +607,8 @@ entry:
 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
+  %val = atomicrmw volatile umin i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -595,8 +616,8 @@ entry:
 ; GCN: buffer_atomic_or v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -605,9 +626,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -617,8 +638,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -629,9 +650,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile or i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -639,7 +660,7 @@ entry:
 ; GCN: buffer_atomic_or v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile or i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -648,8 +669,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile or i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -659,7 +680,7 @@ entry:
 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
+  %val = atomicrmw volatile or i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -670,8 +691,8 @@ entry:
 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
+  %val = atomicrmw volatile or i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -679,8 +700,8 @@ entry:
 ; GCN: buffer_atomic_swap v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -689,32 +710,35 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, 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{{$}}
+
+; VI: flat_atomic_swap v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}{{$}}
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = 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{{$}}
+
 ; VI: flat_atomic_swap [[RET:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}} glc{{$}}
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -722,7 +746,7 @@ entry:
 ; GCN: buffer_atomic_swap v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -731,8 +755,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -742,7 +766,7 @@ entry:
 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
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -753,19 +777,17 @@ entry:
 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
+  %val = atomicrmw volatile xchg i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
-; CMP_SWAP
-
 ; FUNC-LABEL: {{^}}atomic_cmpxchg_i32_offset:
 ; GCN: buffer_atomic_cmpswap v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
 define void @atomic_cmpxchg_i32_offset(i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
-  %0  = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
 
@@ -774,20 +796,22 @@ entry:
 ; GCN: buffer_store_dword v[[RET]]
 define void @atomic_cmpxchg_i32_ret_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i32 %old) {
 entry:
-  %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
-  %0  = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
-  %1  = extractvalue { i32, i1 } %0, 0
-  store i32 %1, i32 addrspace(1)* %out2
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
+  %extract0 = extractvalue { i32, i1 } %val, 0
+  store i32 %extract0, i32 addrspace(1)* %out2
   ret void
 }
 
 ; FUNC-LABEL: {{^}}atomic_cmpxchg_i32_addr64_offset:
 ; SI: buffer_atomic_cmpswap v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
+
+; VI: flat_atomic_cmpswap v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}]{{$}}
 define void @atomic_cmpxchg_i32_addr64_offset(i32 addrspace(1)* %out, i32 %in, i64 %index, i32 %old) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
-  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
-  %0   = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
 
@@ -798,10 +822,10 @@ entry:
 define void @atomic_cmpxchg_i32_ret_addr64_offset(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index, i32 %old) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
-  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
-  %0   = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
-  %1  = extractvalue { i32, i1 } %0, 0
-  store i32 %1, i32 addrspace(1)* %out2
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = cmpxchg volatile i32 addrspace(1)* %gep, i32 %old, i32 %in seq_cst seq_cst
+  %extract0 = extractvalue { i32, i1 } %val, 0
+  store i32 %extract0, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -809,7 +833,7 @@ entry:
 ; GCN: buffer_atomic_cmpswap v[{{[0-9]+:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_cmpxchg_i32(i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
-  %0  = cmpxchg volatile i32 addrspace(1)* %out, i32 %old, i32 %in seq_cst seq_cst
+  %val = cmpxchg volatile i32 addrspace(1)* %out, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
 
@@ -818,9 +842,9 @@ entry:
 ; GCN: buffer_store_dword v[[RET]]
 define void @atomic_cmpxchg_i32_ret(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i32 %old) {
 entry:
-  %0  = cmpxchg volatile i32 addrspace(1)* %out, i32 %old, i32 %in seq_cst seq_cst
-  %1  = extractvalue { i32, i1 } %0, 0
-  store i32 %1, i32 addrspace(1)* %out2
+  %val = cmpxchg volatile i32 addrspace(1)* %out, i32 %old, i32 %in seq_cst seq_cst
+  %extract0 = extractvalue { i32, i1 } %val, 0
+  store i32 %extract0, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -830,7 +854,7 @@ entry:
 define void @atomic_cmpxchg_i32_addr64(i32 addrspace(1)* %out, i32 %in, i64 %index, i32 %old) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
-  %0  = cmpxchg volatile i32 addrspace(1)* %ptr, i32 %old, i32 %in seq_cst seq_cst
+  %val = cmpxchg volatile i32 addrspace(1)* %ptr, i32 %old, i32 %in seq_cst seq_cst
   ret void
 }
 
@@ -841,9 +865,9 @@ entry:
 define void @atomic_cmpxchg_i32_ret_addr64(i32 addrspace(1)* %out, i32 addrspace(1)* %out2, i32 %in, i64 %index, i32 %old) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
-  %0  = cmpxchg volatile i32 addrspace(1)* %ptr, i32 %old, i32 %in seq_cst seq_cst
-  %1  = extractvalue { i32, i1 } %0, 0
-  store i32 %1, i32 addrspace(1)* %out2
+  %val = cmpxchg volatile i32 addrspace(1)* %ptr, i32 %old, i32 %in seq_cst seq_cst
+  %extract0 = extractvalue { i32, i1 } %val, 0
+  store i32 %extract0, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -851,8 +875,8 @@ entry:
 ; GCN: buffer_atomic_xor v{{[0-9]+}}, off, 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -861,9 +885,9 @@ entry:
 ; 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
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
+  %val = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -873,8 +897,8 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
   ret void
 }
 
@@ -885,9 +909,9 @@ entry:
 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
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = atomicrmw volatile xor i32 addrspace(1)* %gep, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -895,7 +919,7 @@ entry:
 ; GCN: buffer_atomic_xor v{{[0-9]+}}, off, 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
+  %val = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
   ret void
 }
 
@@ -904,8 +928,8 @@ entry:
 ; 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
+  %val = atomicrmw volatile xor i32 addrspace(1)* %out, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
@@ -915,7 +939,7 @@ entry:
 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
+  %val = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
   ret void
 }
 
@@ -926,21 +950,20 @@ entry:
 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
+  %val = atomicrmw volatile xor i32 addrspace(1)* %ptr, i32 %in seq_cst
+  store i32 %val, i32 addrspace(1)* %out2
   ret void
 }
 
-; ATOMIC_LOAD
 ; FUNC-LABEL: {{^}}atomic_load_i32_offset:
 ; SI: buffer_load_dword [[RET:v[0-9]+]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc{{$}}
 ; VI: flat_load_dword [[RET:v[0-9]+]], v[{{[0-9]+}}:{{[0-9]+}}] glc{{$}}
 ; GCN: buffer_store_dword [[RET]]
 define void @atomic_load_i32_offset(i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
-  %gep = getelementptr i32, i32 addrspace(1)* %in, i32 4
-  %0  = load atomic i32, i32 addrspace(1)* %gep  seq_cst, align 4
-  store i32 %0, i32 addrspace(1)* %out
+  %gep = getelementptr i32, i32 addrspace(1)* %in, i64 4
+  %val = load atomic i32, i32 addrspace(1)* %gep  seq_cst, align 4
+  store i32 %val, i32 addrspace(1)* %out
   ret void
 }
 
@@ -950,8 +973,8 @@ entry:
 ; GCN: buffer_store_dword [[RET]]
 define void @atomic_load_i32(i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
-  %0  = load atomic i32, i32 addrspace(1)* %in seq_cst, align 4
-  store i32 %0, i32 addrspace(1)* %out
+  %val = load atomic i32, i32 addrspace(1)* %in seq_cst, align 4
+  store i32 %val, i32 addrspace(1)* %out
   ret void
 }
 
@@ -962,9 +985,9 @@ entry:
 define void @atomic_load_i32_addr64_offset(i32 addrspace(1)* %in, i32 addrspace(1)* %out, i64 %index) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %in, i64 %index
-  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
-  %0  = load atomic i32, i32 addrspace(1)* %gep seq_cst, align 4
-  store i32 %0, i32 addrspace(1)* %out
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
+  %val = load atomic i32, i32 addrspace(1)* %gep seq_cst, align 4
+  store i32 %val, i32 addrspace(1)* %out
   ret void
 }
 
@@ -975,66 +998,17 @@ entry:
 define void @atomic_load_i32_addr64(i32 addrspace(1)* %in, i32 addrspace(1)* %out, i64 %index) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %in, i64 %index
-  %0  = load atomic i32, i32 addrspace(1)* %ptr seq_cst, align 4
-  store i32 %0, i32 addrspace(1)* %out
-  ret void
-}
-
-; FUNC-LABEL: {{^}}atomic_load_i64_offset:
-; SI: 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]]
-define void @atomic_load_i64_offset(i64 addrspace(1)* %in, i64 addrspace(1)* %out) {
-entry:
-  %gep = getelementptr i64, i64 addrspace(1)* %in, i64 4
-  %0  = load atomic i64, i64 addrspace(1)* %gep  seq_cst, align 8
-  store i64 %0, i64 addrspace(1)* %out
-  ret void
-}
-
-; FUNC-LABEL: {{^}}atomic_load_i64:
-; SI: 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]]
-define void @atomic_load_i64(i64 addrspace(1)* %in, i64 addrspace(1)* %out) {
-entry:
-  %0  = load atomic i64, i64 addrspace(1)* %in seq_cst, align 8
-  store i64 %0, i64 addrspace(1)* %out
+  %val = load atomic i32, i32 addrspace(1)* %ptr seq_cst, align 4
+  store i32 %val, i32 addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}atomic_load_i64_addr64_offset:
-; SI: 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]]
-define 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
-  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %0  = load atomic i64, i64 addrspace(1)* %gep seq_cst, align 8
-  store i64 %0, i64 addrspace(1)* %out
-  ret void
-}
-
-; FUNC-LABEL: {{^}}atomic_load_i64_addr64:
-; SI: 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{{$}}
-; GCN: buffer_store_dwordx2 [[RET]]
-define void @atomic_load_i64_addr64(i64 addrspace(1)* %in, i64 addrspace(1)* %out, i64 %index) {
-entry:
-  %ptr = getelementptr i64, i64 addrspace(1)* %in, i64 %index
-  %0  = load atomic i64, i64 addrspace(1)* %ptr seq_cst, align 8
-  store i64 %0, i64 addrspace(1)* %out
-  ret void
-}
-
-; ATOMIC_STORE
 ; FUNC-LABEL: {{^}}atomic_store_i32_offset:
 ; SI: buffer_store_dword {{v[0-9]+}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc{{$}}
 ; VI: flat_store_dword v[{{[0-9]+}}:{{[0-9]+}}], {{v[0-9]+}} glc{{$}}
 define void @atomic_store_i32_offset(i32 %in, i32 addrspace(1)* %out) {
 entry:
-  %gep = getelementptr i32, i32 addrspace(1)* %out, i32 4
+  %gep = getelementptr i32, i32 addrspace(1)* %out, i64 4
   store atomic i32 %in, i32 addrspace(1)* %gep  seq_cst, align 4
   ret void
 }
@@ -1054,7 +1028,7 @@ entry:
 define void @atomic_store_i32_addr64_offset(i32 %in, i32 addrspace(1)* %out, i64 %index) {
 entry:
   %ptr = getelementptr i32, i32 addrspace(1)* %out, i64 %index
-  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i32 4
+  %gep = getelementptr i32, i32 addrspace(1)* %ptr, i64 4
   store atomic i32 %in, i32 addrspace(1)* %gep seq_cst, align 4
   ret void
 }
@@ -1068,43 +1042,3 @@ entry:
   store atomic i32 %in, i32 addrspace(1)* %ptr seq_cst, align 4
   ret void
 }
-
-; FUNC-LABEL: {{^}}atomic_store_i64_offset:
-; SI: buffer_store_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
-; VI: flat_store_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}] glc{{$}}
-define void @atomic_store_i64_offset(i64 %in, i64 addrspace(1)* %out) {
-entry:
-  %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  store atomic i64 %in, i64 addrspace(1)* %gep  seq_cst, align 8
-  ret void
-}
-
-; FUNC-LABEL: {{^}}atomic_store_i64:
-; SI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
-; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}] glc
-define 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:
-; SI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
-; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}] glc{{$}}
-define void @atomic_store_i64_addr64_offset(i64 %in, i64 addrspace(1)* %out, i64 %index) {
-entry:
-  %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  store atomic i64 %in, i64 addrspace(1)* %gep seq_cst, align 8
-  ret void
-}
-
-; FUNC-LABEL: {{^}}atomic_store_i64_addr64:
-; SI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
-; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}] glc{{$}}
-define void @atomic_store_i64_addr64(i64 %in, i64 addrspace(1)* %out, i64 %index) {
-entry:
-  %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  store atomic i64 %in, i64 addrspace(1)* %ptr seq_cst, align 8
-  ret void
-}

diff --git a/test/CodeGen/AMDGPU/global_atomics_i64.ll b/test/CodeGen/AMDGPU/global_atomics_i64.ll
index bc37825e59f0c526d0ae4dbd445dbdfac31d237c..1ef87661ac5ca174d15fa7f3f78d8f2578bb398f 100644 (file)
--- a/test/CodeGen/AMDGPU/global_atomics_i64.ll
+++ b/test/CodeGen/AMDGPU/global_atomics_i64.ll
@@ -1,13 +1,12 @@
 ; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %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{{$}}
 define void @atomic_add_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -17,7 +16,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -29,7 +28,7 @@ define void @atomic_add_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -41,7 +40,7 @@ define void @atomic_add_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -50,7 +49,7 @@ entry:
 ; GCN: buffer_atomic_add_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_add_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -59,7 +58,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -70,7 +69,7 @@ entry:
 define void @atomic_add_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -81,7 +80,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile add i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile add i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -91,7 +90,7 @@ entry:
 define void @atomic_and_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -101,7 +100,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -113,7 +112,7 @@ define void @atomic_and_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -125,7 +124,7 @@ define void @atomic_and_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -134,7 +133,7 @@ entry:
 ; GCN: buffer_atomic_and_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_and_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -143,7 +142,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -154,7 +153,7 @@ entry:
 define void @atomic_and_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -165,7 +164,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile and i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile and i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -175,7 +174,7 @@ entry:
 define void @atomic_sub_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -185,7 +184,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -197,7 +196,7 @@ define void @atomic_sub_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -209,7 +208,7 @@ define void @atomic_sub_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -218,7 +217,7 @@ entry:
 ; GCN: buffer_atomic_sub_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_sub_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -227,7 +226,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -238,7 +237,7 @@ entry:
 define void @atomic_sub_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -249,7 +248,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile sub i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile sub i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -259,7 +258,7 @@ entry:
 define void @atomic_max_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -269,7 +268,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -281,7 +280,7 @@ define void @atomic_max_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -293,7 +292,7 @@ define void @atomic_max_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -302,7 +301,7 @@ entry:
 ; GCN: buffer_atomic_smax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_max_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -311,7 +310,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -322,7 +321,7 @@ entry:
 define void @atomic_max_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -333,7 +332,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile max i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile max i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -343,7 +342,7 @@ entry:
 define void @atomic_umax_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -353,7 +352,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -365,7 +364,7 @@ define void @atomic_umax_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -377,7 +376,7 @@ define void @atomic_umax_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrs
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -386,7 +385,7 @@ entry:
 ; GCN: buffer_atomic_umax_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_umax_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -395,7 +394,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -406,7 +405,7 @@ entry:
 define void @atomic_umax_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -417,7 +416,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile umax i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umax i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -427,7 +426,7 @@ entry:
 define void @atomic_min_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -437,7 +436,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -449,7 +448,7 @@ define void @atomic_min_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -461,7 +460,7 @@ define void @atomic_min_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -470,7 +469,7 @@ entry:
 ; GCN: buffer_atomic_smin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_min_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -479,7 +478,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -490,7 +489,7 @@ entry:
 define void @atomic_min_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -501,7 +500,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile min i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile min i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -511,7 +510,7 @@ entry:
 define void @atomic_umin_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -521,7 +520,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -533,7 +532,7 @@ define void @atomic_umin_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -545,7 +544,7 @@ define void @atomic_umin_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrs
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -554,7 +553,7 @@ entry:
 ; GCN: buffer_atomic_umin_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_umin_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -563,7 +562,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -574,7 +573,7 @@ entry:
 define void @atomic_umin_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -585,7 +584,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile umin i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile umin i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -595,7 +594,7 @@ entry:
 define void @atomic_or_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -605,7 +604,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -617,7 +616,7 @@ define void @atomic_or_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %i
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -629,7 +628,7 @@ define void @atomic_or_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrspa
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -638,7 +637,7 @@ entry:
 ; GCN: buffer_atomic_or_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_or_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -647,7 +646,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -658,7 +657,7 @@ entry:
 define void @atomic_or_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -669,7 +668,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile or i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile or i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -679,7 +678,7 @@ entry:
 define void @atomic_xchg_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -689,18 +688,19 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
 
 ; 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{{\[[0-9]+:[0-9]+\]}}{{$}}
 define void @atomic_xchg_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -712,7 +712,7 @@ define void @atomic_xchg_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrs
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -721,7 +721,7 @@ entry:
 ; GCN: buffer_atomic_swap_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_xchg_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -730,7 +730,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -741,7 +741,7 @@ entry:
 define void @atomic_xchg_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -752,7 +752,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile xchg i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xchg i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -762,7 +762,7 @@ entry:
 define void @atomic_xor_i64_offset(i64 addrspace(1)* %out, i64 %in) {
 entry:
   %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -772,7 +772,7 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -784,7 +784,7 @@ define void @atomic_xor_i64_addr64_offset(i64 addrspace(1)* %out, i64 %in, i64 %
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
   ret void
 }
 
@@ -796,7 +796,7 @@ define void @atomic_xor_i64_ret_addr64_offset(i64 addrspace(1)* %out, i64 addrsp
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
   %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %gep, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -805,7 +805,7 @@ entry:
 ; GCN: buffer_atomic_xor_x2 v{{\[[0-9]+:[0-9]+\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
 define void @atomic_xor_i64(i64 addrspace(1)* %out, i64 %in) {
 entry:
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %out, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %out, i64 %in seq_cst
   ret void
 }
 
@@ -814,7 +814,7 @@ entry:
 ; GCN: buffer_store_dwordx2 [[RET]]
 define 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
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %out, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
@@ -825,7 +825,7 @@ entry:
 define void @atomic_xor_i64_addr64(i64 addrspace(1)* %out, i64 %in, i64 %index) {
 entry:
   %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %ptr, i64 %in seq_cst
   ret void
 }
 
@@ -836,7 +836,191 @@ entry:
 define 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
-  %tmp0  = atomicrmw volatile xor i64 addrspace(1)* %ptr, i64 %in seq_cst
+  %tmp0 = atomicrmw volatile xor i64 addrspace(1)* %ptr, i64 %in seq_cst
   store i64 %tmp0, i64 addrspace(1)* %out2
   ret void
 }
+
+; FUNC-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]]
+define void @atomic_load_i64_offset(i64 addrspace(1)* %in, i64 addrspace(1)* %out) {
+entry:
+  %gep = getelementptr i64, i64 addrspace(1)* %in, i64 4
+  %val = load atomic i64, i64 addrspace(1)* %gep  seq_cst, align 8
+  store i64 %val, i64 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-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]]
+define 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
+  store i64 %val, i64 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-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]]
+define 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
+  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
+  %val = load atomic i64, i64 addrspace(1)* %gep seq_cst, align 8
+  store i64 %val, i64 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-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]]
+define 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
+  %val = load atomic i64, i64 addrspace(1)* %ptr seq_cst, align 8
+  store i64 %val, i64 addrspace(1)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_store_i64_offset:
+; CI: buffer_store_dwordx2 [[RET:v\[[0-9]+:[0-9]+\]]], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:32 glc{{$}}
+; VI: flat_store_dwordx2 [[RET:v\[[0-9]+:[0-9]\]]], v[{{[0-9]+}}:{{[0-9]+}}] glc{{$}}
+define void @atomic_store_i64_offset(i64 %in, i64 addrspace(1)* %out) {
+entry:
+  %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
+  store atomic i64 %in, i64 addrspace(1)* %gep  seq_cst, align 8
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_store_i64:
+; CI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]\]}}, v[{{[0-9]+}}:{{[0-9]+}}] glc
+define 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:
+; CI: buffer_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:32 glc{{$}}
+; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}] glc{{$}}
+define void @atomic_store_i64_addr64_offset(i64 %in, i64 addrspace(1)* %out, i64 %index) {
+entry:
+  %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
+  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
+  store atomic i64 %in, i64 addrspace(1)* %gep seq_cst, align 8
+  ret void
+}
+
+; FUNC-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 glc{{$}}
+; VI: flat_store_dwordx2 {{v\[[0-9]+:[0-9]+\]}}, v[{{[0-9]+:[0-9]+}}] glc{{$}}
+define void @atomic_store_i64_addr64(i64 %in, i64 addrspace(1)* %out, i64 %index) {
+entry:
+  %ptr = getelementptr i64, i64 addrspace(1)* %out, i64 %index
+  store atomic i64 %in, i64 addrspace(1)* %ptr seq_cst, align 8
+  ret void
+}
+
+
+
+
+
+
+
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_offset:
+; GCN: buffer_atomic_cmpswapx2 v[{{[0-9]+}}:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16{{$}}
+define void @atomic_cmpxchg_i64_offset(i64 addrspace(1)* %out, i64 %in, i64 %old) {
+entry:
+  %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
+  %val = cmpxchg volatile i64 addrspace(1)* %gep, i64 %old, i64 %in seq_cst seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_offset:
+; GCN: buffer_atomic_cmpswapx2 v{{\[}}[[RET:[0-9]+]]{{:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 offset:16 glc{{$}}
+; GCN: buffer_store_dwordx2 v[[RET]]
+define void @atomic_cmpxchg_i64_ret_offset(i64 addrspace(1)* %out, i64 addrspace(1)* %out2, i64 %in, i64 %old) {
+entry:
+  %gep = getelementptr i64, i64 addrspace(1)* %out, i64 4
+  %val = cmpxchg volatile i64 addrspace(1)* %gep, i64 %old, i64 %in seq_cst seq_cst
+  %extract0 = extractvalue { i64, i1 } %val, 0
+  store i64 %extract0, i64 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_addr64_offset:
+; SI: buffer_atomic_cmpswapx2 v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16{{$}}
+
+; VI: flat_atomic_cmpswapx2 v[{{[0-9]+\:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}]{{$}}
+define 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
+  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
+  %val  = cmpxchg volatile i64 addrspace(1)* %gep, i64 %old, i64 %in seq_cst seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64_offset:
+; SI: buffer_atomic_cmpswapx2 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 offset:16 glc{{$}}
+; VI: flat_atomic_cmpswapx2 v[[RET:[0-9]+]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}] glc{{$}}
+; GCN: buffer_store_dword v[[RET]]
+define 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
+  %gep = getelementptr i64, i64 addrspace(1)* %ptr, i64 4
+  %val  = cmpxchg volatile i64 addrspace(1)* %gep, i64 %old, i64 %in seq_cst seq_cst
+  %extract0 = extractvalue { i64, i1 } %val, 0
+  store i64 %extract0, i64 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64:
+; GCN: buffer_atomic_cmpswap v[{{[0-9]+:[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0{{$}}
+define 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 v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], off, s[{{[0-9]+}}:{{[0-9]+}}], 0 glc
+; GCN: buffer_store_dword v[[RET]]
+define 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
+  %extract0 = extractvalue { i64, i1 } %val, 0
+  store i64 %extract0, i64 addrspace(1)* %out2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_addr64:
+; SI: buffer_atomic_cmpswap v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64{{$}}
+; VI: flat_atomic_cmpswap v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]{{$}}
+define 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
+  %val = cmpxchg volatile i64 addrspace(1)* %ptr, i64 %old, i64 %in seq_cst seq_cst
+  ret void
+}
+
+; FUNC-LABEL: {{^}}atomic_cmpxchg_i64_ret_addr64:
+; SI: buffer_atomic_cmpswap v{{\[}}[[RET:[0-9]+]]:{{[0-9]+}}], v[{{[0-9]+}}:{{[0-9]+}}], s[{{[0-9]+}}:{{[0-9]+}}], 0 addr64 glc{{$}}
+; VI: flat_atomic_cmpswap v[[RET:[0-9]+]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}] glc{{$}}
+; GCN: buffer_store_dword v[[RET]]
+define 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
+  %val = cmpxchg volatile i64 addrspace(1)* %ptr, i64 %old, i64 %in seq_cst seq_cst
+  %extract0 = extractvalue { i64, i1 } %val, 0
+  store i64 %extract0, i64 addrspace(1)* %out2
+  ret void
+}