| Issue |
140737
|
| Summary |
[flang][OpenMP] refine semantic checks for atomic
|
| Labels |
ABI,
flang:openmp
|
| Assignees |
|
| Reporter |
tblah
|
```
module m1
integer::n=2
type w
integer x(2,2,2),v,e
end type
type(w)::g
contains
subroutine s1
!$omp atomic capture
g%x(n,n,n)=g%x(n,n,n)+f(g%e)
g%v=g%x(n,n,n)
!$omp end atomic
end subroutine
end module
```
This is from the deleted Fujitsu test [0455_0029](https://github.com/fujitsu/compiler-test-suite/blob/a0e269d5d279c3a0781bb74940cf5ec16b4ab109/Fortran/0455/0455_0029.f90).
Currently flang doesn't like that `g%e` is used on the rhs of `g%x(n,n,n)=g%x(n,n,n)+f(g%e)`. But as a different component of the derived type and non-overlapping with `g%x` I believe this is valid.
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
