1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -mtriple=amdgcn-- -S -structurizecfg -si-annotate-control-flow %s | FileCheck -check-prefix=OPT %s
4 ; This test is designed to check that the backedge from a prior block won't
5 ; reset the variable introduced to record and accumulate the number of threads
6 ; which have already exited the loop.
8 define amdgpu_kernel void @multiple_backedges(i32 %arg, i32* %arg1) {
9 ; OPT-LABEL: @multiple_backedges(
11 ; OPT-NEXT: [[TMP:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
12 ; OPT-NEXT: [[TMP2:%.*]] = shl nsw i32 [[ARG:%.*]], 1
13 ; OPT-NEXT: br label [[LOOP:%.*]]
15 ; OPT-NEXT: [[PHI_BROKEN1:%.*]] = phi i64 [ [[TMP7:%.*]], [[LOOP_END:%.*]] ], [ [[PHI_BROKEN1]], [[LOOP]] ], [ 0, [[ENTRY:%.*]] ]
16 ; OPT-NEXT: [[PHI_BROKEN:%.*]] = phi i64 [ 0, [[LOOP_END]] ], [ [[TMP0:%.*]], [[LOOP]] ], [ 0, [[ENTRY]] ]
17 ; OPT-NEXT: [[TMP4:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP5:%.*]], [[LOOP]] ], [ 0, [[LOOP_END]] ]
18 ; OPT-NEXT: [[TMP5]] = add nsw i32 [[TMP4]], [[TMP]]
19 ; OPT-NEXT: [[TMP6:%.*]] = icmp slt i32 [[ARG]], [[TMP5]]
20 ; OPT-NEXT: [[TMP0]] = call i64 @llvm.amdgcn.if.break.i64.i64(i1 [[TMP6]], i64 [[PHI_BROKEN]])
21 ; OPT-NEXT: [[TMP1:%.*]] = call i1 @llvm.amdgcn.loop.i64(i64 [[TMP0]])
22 ; OPT-NEXT: br i1 [[TMP1]], label [[LOOP_END]], label [[LOOP]]
24 ; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 [[TMP0]])
25 ; OPT-NEXT: [[EXIT:%.*]] = icmp sgt i32 [[TMP5]], [[TMP2]]
26 ; OPT-NEXT: [[TMP7]] = call i64 @llvm.amdgcn.if.break.i64.i64(i1 [[EXIT]], i64 [[PHI_BROKEN1]])
27 ; OPT-NEXT: [[TMP3:%.*]] = call i1 @llvm.amdgcn.loop.i64(i64 [[TMP7]])
28 ; OPT-NEXT: br i1 [[TMP3]], label [[LOOP_EXIT:%.*]], label [[LOOP]]
30 ; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 [[TMP7]])
31 ; OPT-NEXT: [[TMP12:%.*]] = zext i32 [[TMP]] to i64
32 ; OPT-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[ARG1:%.*]], i64 [[TMP12]]
33 ; OPT-NEXT: [[TMP14:%.*]] = addrspacecast i32* [[TMP13]] to i32 addrspace(1)*
34 ; OPT-NEXT: store i32 [[TMP5]], i32 addrspace(1)* [[TMP14]], align 4
38 %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
39 %tmp2 = shl nsw i32 %arg, 1
43 %tmp4 = phi i32 [ 0, %entry ], [ %tmp5, %loop ], [ 0, %loop_end ]
44 %tmp5 = add nsw i32 %tmp4, %tmp
45 %tmp6 = icmp slt i32 %arg, %tmp5
46 br i1 %tmp6, label %loop_end, label %loop
49 %exit = icmp sgt i32 %tmp5, %tmp2
50 br i1 %exit, label %loop_exit, label %loop
53 %tmp12 = zext i32 %tmp to i64
54 %tmp13 = getelementptr inbounds i32, i32* %arg1, i64 %tmp12
55 %tmp14 = addrspacecast i32* %tmp13 to i32 addrspace(1)*
56 store i32 %tmp5, i32 addrspace(1)* %tmp14, align 4
60 ; Function Attrs: nounwind readnone speculatable
61 declare i32 @llvm.amdgcn.workitem.id.x()