From 940e095f467ef5b9a841de4e6bc8f2ea4557f1fd Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@hotmail.com>
Date: Thu, 15 Feb 2018 23:39:43 +0000
Subject: [PATCH] [OPENMP] Fix PR36399: Crash on C code with ordered doacross
 construct.

Codegen for ordered with doacross construct might produce incorrect code
because of missing cleanup scope for the construct. Without this scope
the final runtime function call could be emitted in the wrong order that
leads to incorrect codegen.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@325304 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/CodeGen/CGStmtOpenMP.cpp           |  2 +
 test/OpenMP/ordered_doacross_codegen.c | 61 ++++++++++++++++++++++++++
 2 files changed, 63 insertions(+)
 create mode 100644 test/OpenMP/ordered_doacross_codegen.c

diff --git a/lib/CodeGen/CGStmtOpenMP.cpp b/lib/CodeGen/CGStmtOpenMP.cpp
index 052ebcaf38..b4cc7ad63c 100644
--- a/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/lib/CodeGen/CGStmtOpenMP.cpp
@@ -2230,6 +2230,7 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
       incrementProfileCounter(&S);
     }
 
+    RunCleanupsScope DoacrossCleanupScope(*this);
     bool Ordered = false;
     if (auto *OrderedClause = S.getSingleClause<OMPOrderedClause>()) {
       if (OrderedClause->getNumForLoops())
@@ -2366,6 +2367,7 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
       return CGF.Builder.CreateIsNotNull(
           CGF.EmitLoadOfScalar(IL, S.getLocStart()));
     });
+    DoacrossCleanupScope.ForceCleanup();
     // We're now done with the loop, so jump to the continuation block.
     if (ContBlock) {
       EmitBranch(ContBlock);
diff --git a/test/OpenMP/ordered_doacross_codegen.c b/test/OpenMP/ordered_doacross_codegen.c
new file mode 100644
index 0000000000..c12df2638f
--- /dev/null
+++ b/test/OpenMP/ordered_doacross_codegen.c
@@ -0,0 +1,61 @@
+// RUN: %clang_cc1 -verify -fopenmp -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -fopenmp -triple x86_64-unknown-unknown -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -triple x86_64-unknown-unknown -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
+// RUN: %clang_cc1 -fopenmp-simd -triple x86_64-unknown-unknown -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -triple x86_64-unknown-unknown -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
+// SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// CHECK: [[KMP_DIM:%.+]] = type { i64, i64, i64 }
+extern int n;
+int a[10], b[10], c[10], d[10];
+void foo();
+
+// CHECK-LABEL: @main()
+int main() {
+  int i;
+// CHECK: [[DIMS:%.+]] = alloca [[KMP_DIM]],
+// CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
+// CHECK: icmp
+// CHECK-NEXT: br i1 %
+// CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIMS]] to i8*
+// CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 24, i1 false)
+// CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIMS]], i32 0, i32 1
+// CHECK: store i64 %{{.+}}, i64* %
+// CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIMS]], i32 0, i32 2
+// CHECK: store i64 1, i64* %
+// CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIMS]] to i8*
+// CHECK: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 1, i8* [[CAST]])
+// CHECK: call void @__kmpc_for_static_init_4(
+#pragma omp for ordered(1)
+  for (i = 0; i < n; ++i) {
+    a[i] = b[i] + 1;
+    foo();
+// CHECK: call void [[FOO:.+]](
+// CHECK: load i32, i32* [[CNT:%.+]],
+// CHECK-NEXT: sext i32 %{{.+}} to i64
+// CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
+// CHECK-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
+#pragma omp ordered depend(source)
+    c[i] = c[i] + 1;
+    foo();
+// CHECK: call void [[FOO]]
+// CHECK: load i32, i32* [[CNT]],
+// CHECK-NEXT: sub nsw i32 %{{.+}}, 2
+// CHECK-NEXT: sext i32 %{{.+}} to i64
+// CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
+// CHECK-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
+#pragma omp ordered depend(sink : i - 2)
+    d[i] = a[i - 2];
+  }
+  // CHECK: call void @__kmpc_for_static_fini(
+  // CHECK: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
+  // CHECK: ret i32 0
+  return 0;
+}
+#endif // HEADER
-- 
2.50.1