[J3] [SC22WG5.6536] Did any ASYNCHRONOUS implementations change for TS 29113

[Jeff Hammond]

TS 29113 extends ASYNCHRONOUS to protect MPI nonblocking operations, not just I/O.

Did any Fortran compilers change their implementation of ASYNCHRONOUS in response to this, or was their behavior already sufficient?

I want to know if it is safe in a practical sense for the MPI standard to assume that ASYNCHRONOUS behaves as TS 29113 ie Fortran 2018 prescribes.

For any implementations that did change, can you describe how your compiler detects that procedures like MPI_Irecv have not implemented their inter-process communication using Fortran asynchronous I/O statements and therefore can be treated differently?

Thanks,

Jeff
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20240131/2b2e3e59/attachment.htm>