[Issue]: missing do concurrent feature in the rock 10.0 amdflang 23

[JorgeG94]

Problem Description

this

module m
  implicit none
  type :: inner_t
     real :: x = 0.0
  end type inner_t
  type :: outer_t
     type(inner_t)     :: c          ! nested record, declared BEFORE the allocatables
     real, allocatable :: a(:), b(:)
  end type outer_t
contains
  subroutine k(s, n)
    type(outer_t), intent(inout) :: s
    integer, intent(in) :: n
    integer :: i
    do concurrent (i = 1:n)
       s%b(i) = s%a(i) * 2.0      ! never touches s%c -- doesn't matter
    end do
  end subroutine k
end module m

program mre
  use m
  implicit none
  type(outer_t) :: s
  allocate(s%a(8), s%b(8))
  s%a = 1.0
  call k(s, 8)
  print *, sum(s%b)      ! 16.0
end program mre

does not compile with:

amdflang -O2 -fopenmp --offload-arch=gfx90a -fdo-concurrent-to-openmp=device mre.f90
error: loc("mre.f90":12:37): /__w/rockrel/rockrel/compiler/amd-llvm/flang/lib/Optimizer/OpenMP/DoConcurrentConversion.cpp:609: not yet implemented: Nested record types are not supported yet.
LLVM ERROR: aborting

Operating System

SLE, SP7

CPU

AMD EPYC 7763 64-Core Processor

GPU

MI250X

ROCm Version

10.0.0

Installation Method

Tarball

Installed ROCm Packages / Versions

Installed ROCm packages / versions

Installed like:

wget https://stable.repo.amd.com/rocm/core/tarball/therock-dist-linux-gfx90a-10.0.0.tar.gz
mkdir install
tar -xf *.tar.gz -C install

ROCm Component

flang

Steps to Reproduce

amdflang -O2 -fopenmp --offload-arch=gfx90a -fdo-concurrent-to-openmp=device mre.f90

(Optional for Linux users) Output of rocminfo --support

rocminfo --support output

Paste output here

Additional Information

No response