Alan,

thanks for the report, I addressed this issue in
https://github.com/open-mpi/ompi/pull/8116

As a temporary workaround, you can apply the attached patch.

FWIW, f18 (shipped with LLVM 11.0.0) is still in development and uses
gfortran under the hood.

Cheers,

Gilles

On Wed, Oct 21, 2020 at 12:44 AM Alan Wild via users
<users@lists.open-mpi.org> wrote:
>
> More specifically building the new “flang” compiler and compiling openmpi 
> with the combination of clang/flang rather than clang/gfortran.
>
> Configure is passing (including support for 16 byte REAL and COMPLEX types.  
> However there is one file that uses REAL128 and CONPLE(REAL128) and I’m 
> getting type in supported KIND=-1 errors.
>
> (Quick aside I’m surprised that the one file is using the ISO_FORTRAN_ENV 
> type names but configure is only checking for the non-standard REAL*# names.  
> Feels like an oversight to me.)
>
> From what I’ve been able to find (and reading their iso_fortran_env.f90 
> module file the compiler should support the two types. I barely know enough 
> FORTRAN to be dangerous (took one semester of F77 in like 1996) so I’m not 
> really sure what I’m looking at here or what to try next.
>
> I would really like to provide an openmpi build to my users that is a “pure 
> LLVM” build.
>
> -Alan
> --
> a...@madllama.net http://humbleville.blogspot.com

Attachment: configure.diff
Description: Binary data

Reply via email to