e7c6e3557b
Fixes #112538 The problem was that the host associated symbol for the threadprivate variable doesn't have all of the symbol attributes (e.g. POINTER). This caused the lowering code to generate the wrong type, eventually hitting an assertion.
32 lines
1.5 KiB
Fortran
32 lines
1.5 KiB
Fortran
! Simple test for lowering of OpenMP Threadprivate Directive with a pointer var
|
|
! from a common block.
|
|
|
|
!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
|
|
!RUN: bbc -hlfir -emit-hlfir -fopenmp %s -o - | FileCheck %s
|
|
|
|
! Regression test for a compiler crash
|
|
|
|
module mmm
|
|
integer,pointer::nam1
|
|
common /com1/nam1,nam2
|
|
!$omp threadprivate(/com1/)
|
|
end
|
|
use mmm
|
|
!$omp parallel copyin(nam1)
|
|
!$omp end parallel
|
|
end
|
|
|
|
|
|
! CHECK-LABEL: fir.global common @com1_(dense<0> : vector<28xi8>) {alignment = 8 : i64} : !fir.array<28xi8>
|
|
|
|
! CHECK-LABEL: func.func @_QQmain() {
|
|
! CHECK: %[[VAL_0:.*]] = fir.address_of(@com1_) : !fir.ref<!fir.array<28xi8>>
|
|
! CHECK: omp.parallel {
|
|
! CHECK: %[[VAL_17:.*]] = omp.threadprivate %[[VAL_0]] : !fir.ref<!fir.array<28xi8>> -> !fir.ref<!fir.array<28xi8>>
|
|
! CHECK: %[[VAL_18:.*]] = fir.convert %[[VAL_17]] : (!fir.ref<!fir.array<28xi8>>) -> !fir.ref<!fir.array<?xi8>>
|
|
! CHECK: %[[VAL_19:.*]] = arith.constant 0 : index
|
|
! CHECK: %[[VAL_20:.*]] = fir.coordinate_of %[[VAL_18]], %[[VAL_19]] : (!fir.ref<!fir.array<?xi8>>, index) -> !fir.ref<i8>
|
|
! CHECK: %[[VAL_21:.*]] = fir.convert %[[VAL_20]] : (!fir.ref<i8>) -> !fir.ref<!fir.box<!fir.ptr<i32>>>
|
|
! CHECK: %[[VAL_22:.*]]:2 = hlfir.declare %[[VAL_21]] {fortran_attrs = #{{.*}}<pointer>, uniq_name = "_QMmmmEnam1"} : (!fir.ref<!fir.box<!fir.ptr<i32>>>) -> (!fir.ref<!fir.box<!fir.ptr<i32>>>, !fir.ref<!fir.box<!fir.ptr<i32>>>)
|
|
|