1; RUN: opt -S -passes='openmp-opt' < %s | FileCheck %s
2; RUN: opt -passes=openmp-opt -pass-remarks=openmp-opt -disable-output < %s 2>&1 | FileCheck %s -check-prefix=CHECK-REMARKS
3target datalayout = "e-i64:64-i128:128-v16:16-v32:32-n16:32:64"
4target triple = "nvptx64"
5
6@S = external local_unnamed_addr global i8*
7
8; CHECK-REMARKS: remark: replace_globalization.c:5:7: Replaced globalized variable with 16 bytes of shared memory
9; CHECK-REMARKS: remark: replace_globalization.c:5:14: Replaced globalized variable with 4 bytes of shared memory
10; CHECK: [[SHARED_X:@.+]] = internal addrspace(3) global [16 x i8] undef
11; CHECK: [[SHARED_Y:@.+]] = internal addrspace(3) global [4 x i8] undef
12
13; CHECK: %{{.*}} = call i8* @__kmpc_alloc_shared({{.*}})
14; CHECK: call void @__kmpc_free_shared({{.*}})
15define dso_local void @foo() {
16entry:
17  %x = call i8* @__kmpc_alloc_shared(i64 4)
18  %x_on_stack = bitcast i8* %x to i32*
19  %0 = bitcast i32* %x_on_stack to i8*
20  call void @use(i8* %0)
21  call void @__kmpc_free_shared(i8* %x)
22  ret void
23}
24
25define void @bar() {
26  call void @baz()
27  call void @qux()
28  ret void
29}
30
31; CHECK: %{{.*}} = bitcast i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([16 x i8], [16 x i8] addrspace(3)* [[SHARED_X]], i32 0, i32 0) to i8*) to [4 x i32]*
32define internal void @baz() {
33entry:
34  %tid = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
35  %cmp = icmp eq i32 %tid, 0
36  br i1 %cmp, label %master, label %exit
37master:
38  %x = call i8* @__kmpc_alloc_shared(i64 16), !dbg !11
39  %x_on_stack = bitcast i8* %x to [4 x i32]*
40  %0 = bitcast [4 x i32]* %x_on_stack to i8*
41  call void @use(i8* %0)
42  call void @__kmpc_free_shared(i8* %x)
43  br label %exit
44exit:
45  ret void
46}
47
48; CHECK: %{{.*}} = bitcast i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([4 x i8], [4 x i8] addrspace(3)* [[SHARED_Y]], i32 0, i32 0) to i8*) to [4 x i32]*
49define internal void @qux() {
50entry:
51  %tid = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
52  %ntid = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
53  %warpsize = call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
54  %0 = sub nuw i32 %warpsize, 1
55  %1 = sub nuw i32 %ntid, 1
56  %2 = xor i32 %0, -1
57  %master_tid = and i32 %1, %2
58  %3 = icmp eq i32 %tid, %master_tid
59  br i1 %3, label %master, label %exit
60master:
61  %y = call i8* @__kmpc_alloc_shared(i64 4), !dbg !12
62  %y_on_stack = bitcast i8* %y to [4 x i32]*
63  %4 = bitcast [4 x i32]* %y_on_stack to i8*
64  call void @use(i8* %4)
65  call void @__kmpc_free_shared(i8* %y)
66  br label %exit
67exit:
68  ret void
69}
70
71
72define void @use(i8* %x) {
73entry:
74  store i8* %x, i8** @S
75  ret void
76}
77
78declare i8* @__kmpc_alloc_shared(i64)
79
80declare void @__kmpc_free_shared(i8*)
81
82declare i32 @llvm.nvvm.read.ptx.sreg.tid.x()
83
84declare i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
85
86declare i32 @llvm.nvvm.read.ptx.sreg.warpsize()
87
88
89!llvm.dbg.cu = !{!0}
90!llvm.module.flags = !{!3, !4, !5, !6}
91!nvvm.annotations = !{!7, !8}
92
93!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 12.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
94!1 = !DIFile(filename: "replace_globalization.c", directory: "/tmp/replace_globalization.c")
95!2 = !{}
96!3 = !{i32 2, !"Debug Info Version", i32 3}
97!4 = !{i32 1, !"wchar_size", i32 4}
98!5 = !{i32 7, !"openmp", i32 50}
99!6 = !{i32 7, !"openmp-device", i32 50}
100!7 = !{void ()* @foo, !"kernel", i32 1}
101!8 = !{void ()* @bar, !"kernel", i32 1}
102!9 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 1, type: !10, scopeLine: 1, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
103!10 = !DISubroutineType(types: !2)
104!11 = !DILocation(line: 5, column: 7, scope: !9)
105!12 = !DILocation(line: 5, column: 14, scope: !9)
106