1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
|
; RUN: opt -S -mtriple=amdgcn-- -mcpu=bonaire -loop-reduce < %s | FileCheck -check-prefix=OPT %s
; Make sure the pointer / address space of AtomicRMW is considered
; OPT-LABEL: @test_local_atomicrmw_addressing_loop_uniform_index_max_offset_i32(
; OPT: .lr.ph.preheader:
; OPT: %scevgep2 = getelementptr i8, ptr addrspace(3) %arg1, i32 65532
; OPT: br label %.lr.ph
; OPT: .lr.ph:
; OPT: %lsr.iv3 = phi ptr addrspace(3) [ %scevgep4, %.lr.ph ], [ %scevgep2, %.lr.ph.preheader ]
; OPT: %lsr.iv1 = phi ptr addrspace(3) [ %scevgep, %.lr.ph ], [ %arg0, %.lr.ph.preheader ]
; OPT: %lsr.iv = phi i32 [ %lsr.iv.next, %.lr.ph ], [ %n, %.lr.ph.preheader ]
; OPT: %tmp4 = atomicrmw add ptr addrspace(3) %lsr.iv3, i32 undef seq_cst, align 4
; OPT: %tmp7 = atomicrmw add ptr addrspace(3) %lsr.iv1, i32 undef seq_cst, align 4
; OPT: %0 = atomicrmw add ptr addrspace(3) %lsr.iv1, i32 %tmp8 seq_cst, align 4
; OPT: %scevgep4 = getelementptr i8, ptr addrspace(3) %lsr.iv3, i32 4
; OPT: br i1 %exitcond
define amdgpu_kernel void @test_local_atomicrmw_addressing_loop_uniform_index_max_offset_i32(ptr addrspace(3) noalias nocapture %arg0, ptr addrspace(3) noalias nocapture readonly %arg1, i32 %n) #0 {
bb:
%tmp = icmp sgt i32 %n, 0
br i1 %tmp, label %.lr.ph.preheader, label %._crit_edge
.lr.ph.preheader: ; preds = %bb
br label %.lr.ph
._crit_edge.loopexit: ; preds = %.lr.ph
br label %._crit_edge
._crit_edge: ; preds = %._crit_edge.loopexit, %bb
ret void
.lr.ph: ; preds = %.lr.ph, %.lr.ph.preheader
%indvars.iv = phi i32 [ %indvars.iv.next, %.lr.ph ], [ 0, %.lr.ph.preheader ]
%tmp1 = add nuw nsw i32 %indvars.iv, 16383
%tmp3 = getelementptr inbounds i32, ptr addrspace(3) %arg1, i32 %tmp1
%tmp4 = atomicrmw add ptr addrspace(3) %tmp3, i32 undef seq_cst
%tmp6 = getelementptr inbounds i32, ptr addrspace(3) %arg0, i32 %indvars.iv
%tmp7 = atomicrmw add ptr addrspace(3) %tmp6, i32 undef seq_cst
%tmp8 = add nsw i32 %tmp7, %tmp4
atomicrmw add ptr addrspace(3) %tmp6, i32 %tmp8 seq_cst
%indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
%exitcond = icmp eq i32 %indvars.iv.next, %n
br i1 %exitcond, label %._crit_edge.loopexit, label %.lr.ph
}
; OPT-LABEL: test_local_cmpxchg_addressing_loop_uniform_index_max_offset_i32(
; OPT: .lr.ph.preheader:
; OPT: %scevgep2 = getelementptr i8, ptr addrspace(3) %arg1, i32 65532
; OPT: br label %.lr.ph
; OPT: .lr.ph:
; OPT: %lsr.iv3 = phi ptr addrspace(3) [ %scevgep4, %.lr.ph ], [ %scevgep2, %.lr.ph.preheader ]
; OPT: %lsr.iv1 = phi ptr addrspace(3) [ %scevgep, %.lr.ph ], [ %arg0, %.lr.ph.preheader ]
; OPT: %lsr.iv = phi i32 [ %lsr.iv.next, %.lr.ph ], [ %n, %.lr.ph.preheader ]
; OPT: %tmp4 = cmpxchg ptr addrspace(3) %lsr.iv3, i32 undef, i32 undef seq_cst monotonic, align 4
; OPT: %scevgep4 = getelementptr i8, ptr addrspace(3) %lsr.iv3, i32 4
define amdgpu_kernel void @test_local_cmpxchg_addressing_loop_uniform_index_max_offset_i32(ptr addrspace(3) noalias nocapture %arg0, ptr addrspace(3) noalias nocapture readonly %arg1, i32 %n) #0 {
bb:
%tmp = icmp sgt i32 %n, 0
br i1 %tmp, label %.lr.ph.preheader, label %._crit_edge
.lr.ph.preheader: ; preds = %bb
br label %.lr.ph
._crit_edge.loopexit: ; preds = %.lr.ph
br label %._crit_edge
._crit_edge: ; preds = %._crit_edge.loopexit, %bb
ret void
.lr.ph: ; preds = %.lr.ph, %.lr.ph.preheader
%indvars.iv = phi i32 [ %indvars.iv.next, %.lr.ph ], [ 0, %.lr.ph.preheader ]
%tmp1 = add nuw nsw i32 %indvars.iv, 16383
%tmp3 = getelementptr inbounds i32, ptr addrspace(3) %arg1, i32 %tmp1
%tmp4 = cmpxchg ptr addrspace(3) %tmp3, i32 undef, i32 undef seq_cst monotonic
%tmp4.0 = extractvalue { i32, i1 } %tmp4, 0
%tmp6 = getelementptr inbounds i32, ptr addrspace(3) %arg0, i32 %indvars.iv
%tmp7 = cmpxchg ptr addrspace(3) %tmp6, i32 undef, i32 undef seq_cst monotonic
%tmp7.0 = extractvalue { i32, i1 } %tmp7, 0
%tmp8 = add nsw i32 %tmp7.0, %tmp4.0
atomicrmw add ptr addrspace(3) %tmp6, i32 %tmp8 seq_cst
%indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
%exitcond = icmp eq i32 %indvars.iv.next, %n
br i1 %exitcond, label %._crit_edge.loopexit, label %.lr.ph
}
attributes #0 = { nounwind }
attributes #1 = { nounwind argmemonly }
|
