Hi,
We want feedback on the two ways to implement the async Fortran APIs. Generally, the async APIs have one additional argument, an integer event set id. The options:
(1) This implementation would continue to use the sync API names but would add the event set id as an optional argument. The C async version will be called if the optional argument is present.
(2) The other option is to create separate APIs with the _async_f appended as the naming convention; these APIs would have, for the most part, the extra event set id as part of the signature.
Feedback,
No. (1) is the easier of the two regarding coding efficiency and maintainability (it is just an extra optional parameter added). But the idea strays from the past convention of matching the C naming.
No. (2) Introduces a lot of duplicate code but matches the C naming.
I’m leaning toward the No. (1) option.