[Clang] Fix clang crash for fopenmp statement(parallel for) inside lambda function #146772
+425
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…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
added
clang
|
@llvm/pr-subscribers-clang-codegen @llvm/pr-subscribers-clang Author: Shivam Gupta (xgupta) ChangesPreviously, implicit variable like This patch ensures such implicit loop counters are handled in 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:
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]
|
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); | ||
| } |
There was a problem hiding this comment.
I assume this is too broad. Why the original variables are not captured and privatized? Looks like we're missing some analysis
| template <typename am> void F<am>::ak(H &an) const { | ||
| auto ao = an.s(); | ||
| auto ap = [ao](H) { | ||
| #pragma omp for |
There was a problem hiding this comment.
- hmm, the subject says about parallel for, but here it is only for. Why is that?
- Need to simplify the test
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Previously, implicit variable like
__begin3used 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
EmitOMPPrivateLoopCountersby allocating memory and registering them in the local declaration map before loop codegen.This fixes #146335