|
150
|
1 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
|
|
|
2 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,CI %s
|
|
|
3
|
|
|
4 %struct.ByValStruct = type { [4 x i32] }
|
|
|
5 ; Make sure the offset is folded and function's frame register is used
|
|
|
6 ; rather than the global scratch wave offset.
|
|
|
7 ; GCN-LABEL: {{^}}void_func_byval_struct_use_outside_entry_block:
|
|
|
8 ; GCN-NOT: v_lshrrev_b32
|
|
|
9 ; GCN-NOT: s_sub_u32
|
|
|
10
|
|
|
11 ; GCN: s_and_saveexec_b64
|
|
|
12 ; GCN: s_cbranch_execz [[BB1:BB[0-9]+_[0-9]+]]
|
|
|
13
|
|
|
14 ; GCN: buffer_load_dword [[LOAD0:v[0-9]+]], off, s[0:3], s32{{$}}
|
|
|
15 ; GCN-NOT: s32
|
|
|
16 ; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], s32{{$}}
|
|
|
17 ; GCN-NOT: s32
|
|
|
18
|
|
|
19 ; GCN: buffer_load_dword [[LOAD1:v[0-9]+]], off, s[0:3], s32 offset:16{{$}}
|
|
|
20 ; GCN-NOT: s32
|
|
|
21 ; GCN: buffer_store_dword [[LOAD1]], off, s[0:3], s32 offset:16{{$}}
|
|
|
22 ; GCN-NOT: s32
|
|
|
23
|
|
|
24 ; GCN: [[BB1]]
|
|
|
25 ; GCN: s_or_b64 exec, exec
|
|
|
26 define hidden void @void_func_byval_struct_use_outside_entry_block(%struct.ByValStruct addrspace(5)* byval noalias nocapture align 4 %arg0, %struct.ByValStruct addrspace(5)* byval noalias nocapture align 4 %arg1, i1 %cond) #1 {
|
|
|
27 entry:
|
|
|
28 br i1 %cond, label %bb0, label %bb1
|
|
|
29
|
|
|
30 bb0:
|
|
|
31 %arrayidx = getelementptr inbounds %struct.ByValStruct, %struct.ByValStruct addrspace(5)* %arg0, i32 0, i32 0, i32 0
|
|
|
32 %tmp = load volatile i32, i32 addrspace(5)* %arrayidx, align 4
|
|
|
33 %add = add nsw i32 %tmp, 1
|
|
|
34 store volatile i32 %add, i32 addrspace(5)* %arrayidx, align 4
|
|
|
35 %arrayidx2 = getelementptr inbounds %struct.ByValStruct, %struct.ByValStruct addrspace(5)* %arg1, i32 0, i32 0, i32 0
|
|
|
36 %tmp1 = load volatile i32, i32 addrspace(5)* %arrayidx2, align 4
|
|
|
37 %add3 = add nsw i32 %tmp1, 2
|
|
|
38 store volatile i32 %add3, i32 addrspace(5)* %arrayidx2, align 4
|
|
|
39 store volatile i32 9, i32 addrspace(1)* null, align 4
|
|
|
40 br label %bb1
|
|
|
41
|
|
|
42 bb1:
|
|
|
43 ret void
|
|
|
44 }
|
|
|
45 declare hidden void @external_void_func_void() #0
|
|
|
46
|
|
|
47 declare void @llvm.lifetime.start.p5i8(i64, i8 addrspace(5)* nocapture) #3
|
|
|
48 declare void @llvm.lifetime.end.p5i8(i64, i8 addrspace(5)* nocapture) #3
|
|
|
49
|
|
|
50 attributes #0 = { nounwind }
|
|
|
51 attributes #1 = { noinline norecurse nounwind }
|
|
|
52 attributes #2 = { nounwind norecurse "frame-pointer"="all" }
|
