opencode.siil.arti.ee/llvm-tms34010
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
m4
llvm-tms34010/clang/test/OpenMP/irbuilder_for_iterator.cpp
T
Jan André Reuter 0d20c873dd [OMPIRBuilder] Pass work loop type in ident flags (#189347) 
Flang uses the OMPIRBuilder to lower OpenMP constructs to LLVM IR.
When dealing with work sharing constructs, such as DO, DISTRIBUTE or
SECTIONS/SECTION, OMPIRBuilder needs to construct the call to the OpenMP
runtime, typically `__kmpc_for_static_init` or
`__kmpc_dist_for_static_init`.

The first passed flag to these functions is the `ident_t` struct,
defined in `kmp.h`. Most of the arguments are reserved for usage in
Fortran and unused in `openmp`. However, the `flags` argument is used
throughout the code base to identify specific constructs, such as the
type of work sharing construct.

In https://github.com/llvm/llvm-project/issues/112545, it was identified
that Flang does not provide the correct `ident_t` flags when calling
into e.g. `__kmpc_for_static_init`, causing the following runtime
warning to appear when the OpenMP Tools Interface is used:

```
OMP: Warning #189: OMPT: Cannot determine workshare type; using the default (loop) instead. This issue is fixed in an up-to-date compiler.
```

This PR adds a test, verifying that the work sharing constructs provide
the correct flags to the runtime call. The IR check for SECTIONS/SECTION
maps against `KMP_IDENT_WORK_LOOP`, as the OMPIRBuilder converts the
SECTIONS construct into a DO loop with a switch-case for every SECTION
clause.

In addition, `OMPIRBuilder` is modified, so that the work loop type is
passed via `ident_t.flags` to the respective
`__kmpc_for_static_init`/`__kmpc_dist_for_static_init` calls, letting
the test pass and the `libomp` warning disappear.
Wherever possible, provide the mapping based on the passed `LoopType`.
If not available, I implemented the decision based on the
`DistScheduleSchedType`, or in case of `applyDynamicWorkshareLoop`,
hard-coded this to `OMP_IDENT_FLAG_WORK_LOOP`.

---

As this is the first time dealing with `OMPIRBuilder`, I'm not sure if I
missed anything while implementing these changes.
If there's anything that can be improved here, e.g. with the ident flag
mapping, I'm very grateful for any feedback. I tried to implement this
with the best of my understanding based on taking a look at the existing
code and tests.

For example, I was not sure if adding a test to
`flang/test/Integration/OpenMP/` is the correct approach here, or if I
rather should try to modify
`unittests/Frontend/OpenMPIRBuilderTest.cpp`. In the end, the former
looked to me more straight-forward.

---

Fixes https://github.com/llvm/llvm-project/issues/112545

---------

Signed-off-by: Jan André Reuter <j.reuter@fz-juelich.de>
2026-04-14 12:32:17 +00:00

165 lines
11 KiB
C++
Raw Permalink Blame History

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --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 -fopenmp-enable-irbuilder -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
struct MyIterator {
MyIterator(unsigned pos);
MyIterator(const MyIterator &other);
const MyIterator &operator=(const MyIterator &that);
MyIterator &operator++();
int operator-(const MyIterator &that) const;
MyIterator &operator+=(unsigned a);
MyIterator operator+(unsigned a) const;
bool operator==(const MyIterator &that) const;
bool operator!=(const MyIterator &that) const;
unsigned operator*() const;
};
extern "C" void workshareloop_iterator(float *a, float *b, float *c) {
#pragma omp for
for (MyIterator it = MyIterator(7); it != MyIterator(41); ++it) {
unsigned i = *it;
a[i] = b[i] * c[i];
}
}
#endif // HEADER
// CHECK-LABEL: define {{[^@]+}}@workshareloop_iterator
// CHECK-SAME: (ptr noundef [[A:%.*]], ptr noundef [[B:%.*]], ptr noundef [[C:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[IT:%.*]] = alloca [[STRUCT_MYITERATOR:%.*]], align 1
// CHECK-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
// CHECK-NEXT: [[AGG_CAPTURED1:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
// CHECK-NEXT: [[DOTCOUNT_ADDR:%.*]] = alloca i64, align 8
// CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[P_LASTITER:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[P_LOWERBOUND:%.*]] = alloca i64, align 8
// CHECK-NEXT: [[P_UPPERBOUND:%.*]] = alloca i64, align 8
// CHECK-NEXT: [[P_STRIDE:%.*]] = alloca i64, align 8
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK-NEXT: store ptr [[B]], ptr [[B_ADDR]], align 8
// CHECK-NEXT: store ptr [[C]], ptr [[C_ADDR]], align 8
// CHECK-NEXT: call void @_ZN10MyIteratorC1Ej(ptr noundef nonnull align 1 dereferenceable(1) [[IT]], i32 noundef 7)
// CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0
// CHECK-NEXT: store ptr [[IT]], ptr [[TMP0]], align 8
// CHECK-NEXT: call void @_ZN10MyIteratorC1ERKS_(ptr noundef nonnull align 1 dereferenceable(1) [[AGG_CAPTURED1]], ptr noundef nonnull align 1 dereferenceable(1) [[IT]])
// CHECK-NEXT: call void @__captured_stmt(ptr [[DOTCOUNT_ADDR]], ptr [[AGG_CAPTURED]])
// CHECK-NEXT: [[DOTCOUNT:%.*]] = load i64, ptr [[DOTCOUNT_ADDR]], align 8
// CHECK-NEXT: br label [[OMP_LOOP_PREHEADER:%.*]]
// CHECK: omp_loop.preheader:
// CHECK-NEXT: store i64 0, ptr [[P_LOWERBOUND]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = sub i64 [[DOTCOUNT]], 1
// CHECK-NEXT: store i64 [[TMP1]], ptr [[P_UPPERBOUND]], align 8
// CHECK-NEXT: store i64 1, ptr [[P_STRIDE]], align 8
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
// CHECK-NEXT: call void @__kmpc_for_static_init_8u(ptr @[[GLOB1:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, ptr [[P_LASTITER]], ptr [[P_LOWERBOUND]], ptr [[P_UPPERBOUND]], ptr [[P_STRIDE]], i64 1, i64 0)
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr [[P_LOWERBOUND]], align 8
// CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr [[P_UPPERBOUND]], align 8
// CHECK-NEXT: [[TMP4:%.*]] = sub i64 [[TMP3]], [[TMP2]]
// CHECK-NEXT: [[TMP5:%.*]] = add i64 [[TMP4]], 1
// CHECK-NEXT: br label [[OMP_LOOP_HEADER:%.*]]
// CHECK: omp_loop.header:
// CHECK-NEXT: [[OMP_LOOP_IV:%.*]] = phi i64 [ 0, [[OMP_LOOP_PREHEADER]] ], [ [[OMP_LOOP_NEXT:%.*]], [[OMP_LOOP_INC:%.*]] ]
// CHECK-NEXT: br label [[OMP_LOOP_COND:%.*]]
// CHECK: omp_loop.cond:
// CHECK-NEXT: [[OMP_LOOP_CMP:%.*]] = icmp ult i64 [[OMP_LOOP_IV]], [[TMP5]]
// CHECK-NEXT: br i1 [[OMP_LOOP_CMP]], label [[OMP_LOOP_BODY:%.*]], label [[OMP_LOOP_EXIT:%.*]]
// CHECK: omp_loop.body:
// CHECK-NEXT: [[TMP6:%.*]] = add i64 [[OMP_LOOP_IV]], [[TMP2]]
// CHECK-NEXT: call void @__captured_stmt.1(ptr [[IT]], i64 [[TMP6]], ptr [[AGG_CAPTURED1]])
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_ZNK10MyIteratordeEv(ptr noundef nonnull align 1 dereferenceable(1) [[IT]])
// CHECK-NEXT: store i32 [[CALL]], ptr [[I]], align 4
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[B_ADDR]], align 8
// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[I]], align 4
// CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[TMP8]] to i64
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i64 [[IDXPROM]]
// CHECK-NEXT: [[TMP9:%.*]] = load float, ptr [[ARRAYIDX]], align 4
// CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[C_ADDR]], align 8
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[I]], align 4
// CHECK-NEXT: [[IDXPROM2:%.*]] = zext i32 [[TMP11]] to i64
// CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i64 [[IDXPROM2]]
// CHECK-NEXT: [[TMP12:%.*]] = load float, ptr [[ARRAYIDX3]], align 4
// CHECK-NEXT: [[MUL:%.*]] = fmul float [[TMP9]], [[TMP12]]
// CHECK-NEXT: [[TMP13:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[I]], align 4
// CHECK-NEXT: [[IDXPROM4:%.*]] = zext i32 [[TMP14]] to i64
// CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[IDXPROM4]]
// CHECK-NEXT: store float [[MUL]], ptr [[ARRAYIDX5]], align 4
// CHECK-NEXT: br label [[OMP_LOOP_INC]]
// CHECK: omp_loop.inc:
// CHECK-NEXT: [[OMP_LOOP_NEXT]] = add nuw i64 [[OMP_LOOP_IV]], 1
// CHECK-NEXT: br label [[OMP_LOOP_HEADER]]
// CHECK: omp_loop.exit:
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]])
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM6:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
// CHECK-NEXT: call void @__kmpc_barrier(ptr @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM6]])
// CHECK-NEXT: br label [[OMP_LOOP_AFTER:%.*]]
// CHECK: omp_loop.after:
// CHECK-NEXT: ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@__captured_stmt
// CHECK-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[DISTANCE:%.*]], ptr noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[DISTANCE_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[DOTSTART:%.*]] = alloca [[STRUCT_MYITERATOR:%.*]], align 1
// CHECK-NEXT: [[DOTSTOP:%.*]] = alloca [[STRUCT_MYITERATOR]], align 1
// CHECK-NEXT: [[DOTSTEP:%.*]] = alloca i64, align 8
// CHECK-NEXT: store ptr [[DISTANCE]], ptr [[DISTANCE_ADDR]], align 8
// CHECK-NEXT: store ptr [[__CONTEXT]], ptr [[__CONTEXT_ADDR]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[__CONTEXT_ADDR]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON:%.*]], ptr [[TMP0]], i32 0, i32 0
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TMP1]], align 8, !nonnull [[META2:![0-9]+]]
// CHECK-NEXT: call void @_ZN10MyIteratorC1ERKS_(ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTART]], ptr noundef nonnull align 1 dereferenceable(1) [[TMP2]])
// CHECK-NEXT: call void @_ZN10MyIteratorC1Ej(ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTOP]], i32 noundef 41)
// CHECK-NEXT: store i64 1, ptr [[DOTSTEP]], align 8
// CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr [[DOTSTEP]], align 8
// CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[TMP3]], 0
// CHECK-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK: cond.true:
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_ZNK10MyIteratormiERKS_(ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTART]], ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTOP]])
// CHECK-NEXT: [[CONV:%.*]] = sext i32 [[CALL]] to i64
// CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr [[DOTSTEP]], align 8
// CHECK-NEXT: [[SUB:%.*]] = sub i64 0, [[TMP4]]
// CHECK-NEXT: [[DIV:%.*]] = udiv i64 [[CONV]], [[SUB]]
// CHECK-NEXT: br label [[COND_END:%.*]]
// CHECK: cond.false:
// CHECK-NEXT: [[CALL1:%.*]] = call noundef i32 @_ZNK10MyIteratormiERKS_(ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTOP]], ptr noundef nonnull align 1 dereferenceable(1) [[DOTSTART]])
// CHECK-NEXT: [[CONV2:%.*]] = sext i32 [[CALL1]] to i64
// CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr [[DOTSTEP]], align 8
// CHECK-NEXT: [[DIV3:%.*]] = udiv i64 [[CONV2]], [[TMP5]]
// CHECK-NEXT: br label [[COND_END]]
// CHECK: cond.end:
// CHECK-NEXT: [[COND:%.*]] = phi i64 [ [[DIV]], [[COND_TRUE]] ], [ [[DIV3]], [[COND_FALSE]] ]
// CHECK-NEXT: [[TMP6:%.*]] = load ptr, ptr [[DISTANCE_ADDR]], align 8, !nonnull [[META2]], !align [[META3:![0-9]+]]
// CHECK-NEXT: store i64 [[COND]], ptr [[TMP6]], align 8
// CHECK-NEXT: ret void
//
//
// CHECK-LABEL: define {{[^@]+}}@__captured_stmt.1
// CHECK-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[LOOPVAR:%.*]], i64 noundef [[LOGICAL:%.*]], ptr noalias noundef [[__CONTEXT:%.*]]) #[[ATTR2]] {
// CHECK-NEXT: entry:
// CHECK-NEXT: [[LOOPVAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[LOGICAL_ADDR:%.*]] = alloca i64, align 8
// CHECK-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca ptr, align 8
// CHECK-NEXT: [[REF_TMP:%.*]] = alloca [[STRUCT_MYITERATOR:%.*]], align 1
// CHECK-NEXT: store ptr [[LOOPVAR]], ptr [[LOOPVAR_ADDR]], align 8
// CHECK-NEXT: store i64 [[LOGICAL]], ptr [[LOGICAL_ADDR]], align 8
// CHECK-NEXT: store ptr [[__CONTEXT]], ptr [[__CONTEXT_ADDR]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[__CONTEXT_ADDR]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[LOGICAL_ADDR]], align 8
// CHECK-NEXT: [[MUL:%.*]] = mul i64 1, [[TMP1]]
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[MUL]] to i32
// CHECK-NEXT: call void @_ZNK10MyIteratorplEj(ptr dead_on_unwind writable sret([[STRUCT_MYITERATOR]]) align 1 [[REF_TMP]], ptr noundef nonnull align 1 dereferenceable(1) [[TMP0]], i32 noundef [[CONV]])
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[LOOPVAR_ADDR]], align 8, !nonnull [[META2]]
// CHECK-NEXT: [[CALL:%.*]] = call noundef nonnull align 1 dereferenceable(1) ptr @_ZN10MyIteratoraSERKS_(ptr noundef nonnull align 1 dereferenceable(1) [[TMP2]], ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK-NEXT: ret void
//
Reference in new issue View Git Blame Copy Permalink