Skip to content

[Clang] Fix clang crash for fopenmp statement(parallel for) inside lambda function #146772

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[Clang] Fix clang crash for fopenmp statement(parallel for) inside lambda function #146772

wants to merge 1 commit into from

Conversation

xgupta
Copy link
Contributor

@xgupta xgupta commented Jul 2, 2025 •
edited
Loading

Previously, implicit variable like __begin3 used in range-based OpenMP for-loops were not being properly privatized, leading to missing entries in LocalDeclMap and crashes.

This patch ensures such implicit loop counters are handled in EmitOMPPrivateLoopCounters by allocating memory and registering them in the local declaration map before loop codegen.

This fixes #146335

@xgupta
[Clang] Fix clang crash for fopenmp statement(parallel for) inside la…
1794c64 
…mbda function

Previously, implicit variable like `__begin3`  used in range-based
OpenMP for-loops were not being properly privatized, leading to
missing entries in LocalDeclMap and crashes.

This patch ensures such implicit loop counters are handled in
`EmitOMPPrivateLoopCounters` by allocating memory and registering
them in the local declaration map before loop codegen.

This fixes #PR146335
@llvmbot llvmbot added clang Clang issues not falling into any other category clang:codegen IR generation bugs: mangling, exceptions, etc. clang:openmp OpenMP related changes to Clang labels Jul 2, 2025
@llvmbot
Copy link
Member

llvmbot commented Jul 2, 2025 •
edited
Loading

@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-clang

Author: Shivam Gupta (xgupta)

Changes

Previously, implicit variable like __begin3 used in range-based OpenMP for-loops were not being properly privatized, leading to missing entries in LocalDeclMap and crashes.

This patch ensures such implicit loop counters are handled in EmitOMPPrivateLoopCounters by allocating memory and registering them in the local declaration map before loop codegen.

This fixes #PR146335

Patch is 27.18 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/146772.diff

2 Files Affected:

  • (modified) clang/lib/CodeGen/CGStmtOpenMP.cpp (+6)
  • (added) clang/test/OpenMP/for_range_loop_codegen.cpp (+419)
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 5822e0f6db89a..d32f976afa23d 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -2332,6 +2332,12 @@ void CodeGenFunction::EmitOMPPrivateLoopCounters(
   for (const Expr *E : S.counters()) {
     const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
     const auto *PrivateVD = cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl());
+    // Privatize original counter variable (e.g., __beginN, __endN, __rangeN)
+    if (!LocalDeclMap.count(VD)) {
+      Address Addr = CreateMemTemp(VD->getType(), VD->getName());
+      LocalDeclMap.insert({VD, Addr});
+      (void)LoopScope.addPrivate(VD, Addr);
+    }
     // Emit var without initialization.
     AutoVarEmission VarEmission = EmitAutoVarAlloca(*PrivateVD);
     EmitAutoVarCleanups(VarEmission);
diff --git a/clang/test/OpenMP/for_range_loop_codegen.cpp b/clang/test/OpenMP/for_range_loop_codegen.cpp
new file mode 100644
index 0000000000000..bc86c7a477210
--- /dev/null
+++ b/clang/test/OpenMP/for_range_loop_codegen.cpp
@@ -0,0 +1,419 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --clang-args ['-fopenmp', '-std=c++20'] --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
+// RUN: %clang_cc1 -verify -fopenmp -std=c++20 -emit-llvm %s -o - | FileCheck %s
+// expected-no-diagnostics
+#ifndef HEADER
+#define HEADER
+
+enum a {};
+class b {
+public:
+  b(int);
+};
+struct c {
+  typedef b h;
+};
+using e = b;
+using f = b;
+template <typename> class g {};
+template <typename l> class i {
+public:
+  using j = long;
+  l::h aa;
+  i(g<l>) : aa{0} {}
+  bool operator!=(i);
+  void operator++();
+  auto operator*() {
+    int k;
+    return k;
+  }
+  j operator-(i);
+  void operator+=(j);
+};
+class o : public g<c> {
+public:
+  using p = int;
+  o(p, e, f);
+};
+// CHECK1-LABEL: @_Z5begin1gI1cE(
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca [[CLASS_I:%.*]], align 1
+// CHECK1-NEXT:    [[AC:%.*]] = alloca [[CLASS_G:%.*]], align 1
+// CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[CLASS_G]], align 1
+// CHECK1-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[AC]], i32 0, i32 0
+// CHECK1-NEXT:    store i8 [[AC_COERCE:%.*]], ptr [[COERCE_DIVE]], align 1
+// CHECK1-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[AGG_TMP]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i8, ptr [[COERCE_DIVE1]], align 1
+// CHECK1-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1iI1cEC1E1gIS0_E(ptr noundef nonnull align 1 dereferenceable(1) [[RETVAL]], i8 [[TMP0]])
+// CHECK1-NEXT:    [[COERCE_DIVE2:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[RETVAL]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE2]], align 1
+// CHECK1-NEXT:    ret i8 [[TMP1]]
+auto begin(g<c> ac) {
+  i ad(ac);
+  return ad;
+}
+template <typename l> auto end(l ac) {
+  i ad(ac);
+  return ad;
+}
+using m = b;
+using n = b;
+class D {
+protected:
+  D(int, int, n, m, a);
+};
+class H : D {
+  int af;
+  o t;
+
+public:
+  a v;
+  int q;
+// CHECK1-LABEL: @_ZN1HC1E1bS0_(
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[AG:%.*]] = alloca [[CLASS_B:%.*]], align 1
+// CHECK1-NEXT:    [[R:%.*]] = alloca [[CLASS_B]], align 1
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AG]], i32 0, i32 0
+// CHECK1-NEXT:    store i8 [[AG_COERCE:%.*]], ptr [[COERCE_DIVE]], align 1
+// CHECK1-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[R]], i32 0, i32 0
+// CHECK1-NEXT:    store i8 [[R_COERCE:%.*]], ptr [[COERCE_DIVE1]], align 1
+// CHECK1-NEXT:    store ptr [[THIS:%.*]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[THIS2:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[COERCE_DIVE3:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AG]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i8, ptr [[COERCE_DIVE3]], align 1
+// CHECK1-NEXT:    [[COERCE_DIVE4:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[R]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE4]], align 1
+// CHECK1-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1HC2E1bS0_(ptr noundef nonnull align 4 dereferenceable(16) [[THIS2]], i8 [[TMP0]], i8 [[TMP1]])
+// CHECK1-NEXT:    ret ptr [[THIS2]]
+  H(n ag, m r) : D(0, 0, ag, r, a{}), af(0), t{o(0, 0, 0)}, q(0), v() {}
+// CHECK1-LABEL: @_ZN1H1sEv(
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store ptr [[THIS:%.*]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[THIS1]], i64 4
+// CHECK1-NEXT:    ret void
+  auto s() { return t; }
+};
+template <typename> class F { void ak(H &) const; };
+template <typename am> void F<am>::ak(H &an) const {
+  auto ao = an.s();
+  auto ap = [ao](H) {
+#pragma omp for
+    for (auto d : ao)
+      ;
+  };
+  H ar(0, 0);
+  ap(ar);
+}
+// CHECK1-LABEL: @_ZNK1FIdE2akER1H(
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[AN_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[AO:%.*]] = alloca [[CLASS_O:%.*]], align 1
+// CHECK1-NEXT:    [[UNDEF_AGG_TMP:%.*]] = alloca [[CLASS_O]], align 1
+// CHECK1-NEXT:    [[AP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
+// CHECK1-NEXT:    [[AR:%.*]] = alloca [[CLASS_H:%.*]], align 4
+// CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[CLASS_B:%.*]], align 1
+// CHECK1-NEXT:    [[AGG_TMP2:%.*]] = alloca [[CLASS_B]], align 1
+// CHECK1-NEXT:    [[AGG_TMP6:%.*]] = alloca [[CLASS_H]], align 4
+// CHECK1-NEXT:    store ptr [[THIS:%.*]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[AN:%.*]], ptr [[AN_ADDR]], align 8
+// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AN_ADDR]], align 8, !nonnull [[META3:![0-9]+]], !align [[META4:![0-9]+]]
+// CHECK1-NEXT:    call void @_ZN1H1sEv(ptr noundef nonnull align 4 dereferenceable(16) [[TMP0]])
+// CHECK1-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1bC1Ei(ptr noundef nonnull align 1 dereferenceable(1) [[AGG_TMP]], i32 noundef 0)
+// CHECK1-NEXT:    [[CALL3:%.*]] = call noundef ptr @_ZN1bC1Ei(ptr noundef nonnull align 1 dereferenceable(1) [[AGG_TMP2]], i32 noundef 0)
+// CHECK1-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AGG_TMP]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE]], align 1
+// CHECK1-NEXT:    [[COERCE_DIVE4:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AGG_TMP2]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i8, ptr [[COERCE_DIVE4]], align 1
+// CHECK1-NEXT:    [[CALL5:%.*]] = call noundef ptr @_ZN1HC1E1bS0_(ptr noundef nonnull align 4 dereferenceable(16) [[AR]], i8 [[TMP1]], i8 [[TMP2]])
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[AGG_TMP6]], ptr align 4 [[AR]], i64 16, i1 false)
+// CHECK1-NEXT:    [[TMP3:%.*]] = load [2 x i64], ptr [[AGG_TMP6]], align 4
+// CHECK1-NEXT:    call void @_ZZNK1FIdE2akER1HENKUlS1_E_clES1_(ptr noundef nonnull align 1 dereferenceable(1) [[AP]], [2 x i64] [[TMP3]])
+// CHECK1-NEXT:    ret void
+template class F<double>;
+#endif
+// CHECK-LABEL: define {{[^@]+}}@_Z5begin1gI1cE
+// CHECK-SAME: (i8 [[AC_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[CLASS_I:%.*]], align 1
+// CHECK-NEXT:    [[AC:%.*]] = alloca [[CLASS_G:%.*]], align 1
+// CHECK-NEXT:    [[AGG_TMP:%.*]] = alloca [[CLASS_G]], align 1
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[AC]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[AC_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[AGG_TMP]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[COERCE_DIVE1]], align 1
+// CHECK-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1iI1cEC1E1gIS0_E(ptr noundef nonnull align 1 dereferenceable(1) [[RETVAL]], i8 [[TMP0]])
+// CHECK-NEXT:    [[COERCE_DIVE2:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[RETVAL]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE2]], align 1
+// CHECK-NEXT:    ret i8 [[TMP1]]
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZN1iI1cEC1E1gIS0_E
+// CHECK-SAME: (ptr noundef nonnull returned align 1 dereferenceable(1) [[THIS:%.*]], i8 [[DOTCOERCE:%.*]]) unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = alloca [[CLASS_G:%.*]], align 1
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[TMP0]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[DOTCOERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[COERCE_DIVE2:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[TMP0]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE2]], align 1
+// CHECK-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1iI1cEC2E1gIS0_E(ptr noundef nonnull align 1 dereferenceable(1) [[THIS1]], i8 [[TMP1]])
+// CHECK-NEXT:    ret ptr [[THIS1]]
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZN1iI1cEC2E1gIS0_E
+// CHECK-SAME: (ptr noundef nonnull returned align 1 dereferenceable(1) [[THIS:%.*]], i8 [[DOTCOERCE:%.*]]) unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = alloca [[CLASS_G:%.*]], align 1
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[TMP0]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[DOTCOERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1bC1Ei(ptr noundef nonnull align 1 dereferenceable(1) [[THIS1]], i32 noundef 0)
+// CHECK-NEXT:    ret ptr [[THIS1]]
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZNK1FIdE2akER1H
+// CHECK-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(16) [[AN:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[AN_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[AO:%.*]] = alloca [[CLASS_O:%.*]], align 1
+// CHECK-NEXT:    [[UNDEF_AGG_TMP:%.*]] = alloca [[CLASS_O]], align 1
+// CHECK-NEXT:    [[AP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
+// CHECK-NEXT:    [[AR:%.*]] = alloca [[CLASS_H:%.*]], align 4
+// CHECK-NEXT:    [[AGG_TMP:%.*]] = alloca [[CLASS_B:%.*]], align 1
+// CHECK-NEXT:    [[AGG_TMP2:%.*]] = alloca [[CLASS_B]], align 1
+// CHECK-NEXT:    [[AGG_TMP6:%.*]] = alloca [[CLASS_H]], align 4
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[AN]], ptr [[AN_ADDR]], align 8
+// CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AN_ADDR]], align 8, !nonnull [[META3:![0-9]+]], !align [[META4:![0-9]+]]
+// CHECK-NEXT:    call void @_ZN1H1sEv(ptr noundef nonnull align 4 dereferenceable(16) [[TMP0]])
+// CHECK-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1bC1Ei(ptr noundef nonnull align 1 dereferenceable(1) [[AGG_TMP]], i32 noundef 0)
+// CHECK-NEXT:    [[CALL3:%.*]] = call noundef ptr @_ZN1bC1Ei(ptr noundef nonnull align 1 dereferenceable(1) [[AGG_TMP2]], i32 noundef 0)
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AGG_TMP]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    [[COERCE_DIVE4:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AGG_TMP2]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP2:%.*]] = load i8, ptr [[COERCE_DIVE4]], align 1
+// CHECK-NEXT:    [[CALL5:%.*]] = call noundef ptr @_ZN1HC1E1bS0_(ptr noundef nonnull align 4 dereferenceable(16) [[AR]], i8 [[TMP1]], i8 [[TMP2]])
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[AGG_TMP6]], ptr align 4 [[AR]], i64 16, i1 false)
+// CHECK-NEXT:    [[TMP3:%.*]] = load [2 x i64], ptr [[AGG_TMP6]], align 4
+// CHECK-NEXT:    call void @_ZZNK1FIdE2akER1HENKUlS1_E_clES1_(ptr noundef nonnull align 1 dereferenceable(1) [[AP]], [2 x i64] [[TMP3]])
+// CHECK-NEXT:    ret void
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZN1H1sEv
+// CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(16) [[THIS:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[THIS1]], i64 4
+// CHECK-NEXT:    ret void
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZN1HC1E1bS0_
+// CHECK-SAME: (ptr noundef nonnull returned align 4 dereferenceable(16) [[THIS:%.*]], i8 [[AG_COERCE:%.*]], i8 [[R_COERCE:%.*]]) unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[AG:%.*]] = alloca [[CLASS_B:%.*]], align 1
+// CHECK-NEXT:    [[R:%.*]] = alloca [[CLASS_B]], align 1
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AG]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[AG_COERCE]], ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[R]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[R_COERCE]], ptr [[COERCE_DIVE1]], align 1
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[THIS2:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[COERCE_DIVE3:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[AG]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP0:%.*]] = load i8, ptr [[COERCE_DIVE3]], align 1
+// CHECK-NEXT:    [[COERCE_DIVE4:%.*]] = getelementptr inbounds nuw [[CLASS_B]], ptr [[R]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[COERCE_DIVE4]], align 1
+// CHECK-NEXT:    [[CALL:%.*]] = call noundef ptr @_ZN1HC2E1bS0_(ptr noundef nonnull align 4 dereferenceable(16) [[THIS2]], i8 [[TMP0]], i8 [[TMP1]])
+// CHECK-NEXT:    ret ptr [[THIS2]]
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@_ZZNK1FIdE2akER1HENKUlS1_E_clES1_
+// CHECK-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[THIS:%.*]], [2 x i64] [[DOTCOERCE:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = alloca [[CLASS_H:%.*]], align 4
+// CHECK-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[TMP:%.*]] = alloca [[CLASS_I:%.*]], align 1
+// CHECK-NEXT:    [[__RANGE3:%.*]] = alloca ptr, align 8
+// CHECK-NEXT:    [[__END3:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[AGG_TMP:%.*]] = alloca [[CLASS_O:%.*]], align 1
+// CHECK-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[AGG_TMP3:%.*]] = alloca [[CLASS_G:%.*]], align 1
+// CHECK-NEXT:    [[DOTCAPTURE_EXPR_7:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[DOTCAPTURE_EXPR_8:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[AGG_TMP9:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[__BEGIN3:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[__BEGIN313:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[AGG_TMP14:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[__BEGIN318:%.*]] = alloca [[CLASS_I]], align 1
+// CHECK-NEXT:    [[D:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
+// CHECK-NEXT:    store [2 x i64] [[DOTCOERCE]], ptr [[TMP0]], align 4
+// CHECK-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK-NEXT:    store ptr [[THIS1]], ptr [[__RANGE3]], align 8
+// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[__RANGE3]], align 8, !nonnull [[META3]]
+// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[CLASS_O]], ptr [[AGG_TMP]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP3:%.*]] = load i8, ptr [[COERCE_DIVE]], align 1
+// CHECK-NEXT:    [[CALL:%.*]] = call i8 @_Z3endI1oEDaT_(i8 [[TMP3]])
+// CHECK-NEXT:    [[COERCE_DIVE2:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[__END3]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[CALL]], ptr [[COERCE_DIVE2]], align 1
+// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[__RANGE3]], align 8, !nonnull [[META3]]
+// CHECK-NEXT:    [[COERCE_DIVE4:%.*]] = getelementptr inbounds nuw [[CLASS_G]], ptr [[AGG_TMP3]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP5:%.*]] = load i8, ptr [[COERCE_DIVE4]], align 1
+// CHECK-NEXT:    [[CALL5:%.*]] = call i8 @_Z5begin1gI1cE(i8 [[TMP5]])
+// CHECK-NEXT:    [[COERCE_DIVE6:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[DOTCAPTURE_EXPR_]], i32 0, i32 0
+// CHECK-NEXT:    store i8 [[CALL5]], ptr [[COERCE_DIVE6]], align 1
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[DOTCAPTURE_EXPR_7]], ptr align 1 [[__END3]], i64 1, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[AGG_TMP9]], ptr align 1 [[DOTCAPTURE_EXPR_]], i64 1, i1 false)
+// CHECK-NEXT:    [[COERCE_DIVE10:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[AGG_TMP9]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP6:%.*]] = load i8, ptr [[COERCE_DIVE10]], align 1
+// CHECK-NEXT:    [[COERCE_VAL_II:%.*]] = zext i8 [[TMP6]] to i64
+// CHECK-NEXT:    [[CALL11:%.*]] = call noundef i64 @_ZN1iI1cEmiES1_(ptr noundef nonnull align 1 dereferenceable(1) [[DOTCAPTURE_EXPR_7]], i64 [[COERCE_VAL_II]])
+// CHECK-NEXT:    [[SUB:%.*]] = sub nsw i64 [[CALL11]], 1
+// CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
+// CHECK-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
+// CHECK-NEXT:    [[SUB12:%.*]] = sub nsw i64 [[DIV]], 1
+// CHECK-NEXT:    store i64 [[SUB12]], ptr [[DOTCAPTURE_EXPR_8]], align 8
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[__BEGIN3]], ptr align 1 [[DOTCAPTURE_EXPR_]], i64 1, i1 false)
+// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 1 [[AGG_TMP14]], ptr align 1 [[__END3]], i64 1, i1 false)
+// CHECK-NEXT:    [[COERCE_DIVE15:%.*]] = getelementptr inbounds nuw [[CLASS_I]], ptr [[AGG_TMP14]], i32 0, i32 0
+// CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[COERCE_DIVE15]], align 1
+// CHECK-NEXT:    [[COERCE_VAL_II16:%.*]] = zext i8 [[TMP7]] to i64
+// CHECK-NEXT:    [[CALL17:%.*]] = call noundef i1 @_ZN1iI1cEneES1_(ptr noundef nonnull align 1 dereferenceable(1) [[__BEGIN3]], i64 [[COERCE_VAL_II16]])
+// CHECK-NEXT:    br i1 [[CALL17]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK:       omp.precond.then:
+// CHECK-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_8]], align 8
+// CHECK-NEXT:    store i64 [[TMP8]], ptr [[DOTOMP_UB]], align 8
+// CHECK-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// CHECK-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_8]], align 8
+// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP9]], [[TMP10]]
+// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK:       ...
[truncated]

alexey-bataev
alexey-bataev reviewed Jul 2, 2025
View reviewed changes
clang/lib/CodeGen/CGStmtOpenMP.cpp
Comment on lines +2335 to +2340
// Privatize original counter variable (e.g., __beginN, __endN, __rangeN)
if (!LocalDeclMap.count(VD)) {
Address Addr = CreateMemTemp(VD->getType(), VD->getName());
LocalDeclMap.insert({VD, Addr});
(void)LoopScope.addPrivate(VD, Addr);
}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I assume this is too broad. Why the original variables are not captured and privatized? Looks like we're missing some analysis

clang/test/OpenMP/for_range_loop_codegen.cpp
template <typename am> void F<am>::ak(H &an) const {
auto ao = an.s();
auto ap = [ao](H) {
#pragma omp for
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. hmm, the subject says about parallel for, but here it is only for. Why is that?
  2. Need to simplify the test

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@alexey-bataev alexey-bataev alexey-bataev left review comments

@shiltian shiltian Awaiting requested review from shiltian

@zahiraam zahiraam Awaiting requested review from zahiraam

Assignees
No one assigned
Labels
clang:codegen IR generation bugs: mangling, exceptions, etc. clang:openmp OpenMP related changes to Clang clang Clang issues not falling into any other category
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

clang crash for fopenmp statement(parallel for) inside lambda function while generating llvm-ir
3 participants
@xgupta @llvmbot @alexey-bataev