[MPIWG Fortran] [mpiwg-tools] Proposal: MPI_SIZEOF not profiled

[N.M. Maclaren]

On May 19 2014, Bill Long wrote:
>
> I'm still concerned about making a distinction between Fortran and C that 
> is entirely artificial. If a Fortran program wants to avoid interception 
> of MPI_WTICK, for example, it is trivial to write a wrapper C function 
> my_wtick_ that contains the macro for MPI_Wtick. Macros in C are just 
> another form of inlining. I think it would be better to merge together 
> text from Martin and Rolf to end up with something simpler and clearer:
>
>
> An implementation is allowed to implement MPI_WTIME, MPI_WTICK, 
> PMPI_WTIME, PMPI_WTICK, MPI_SIZEOF, PMPI_SIZEOF, and the handle- and 
> status-conversion functions (MPI_Group_f2c, etc.) in Sections 17.2.4 and 
> 17.2.5, and no others, such that they are not interceptable.
>
> Advice to implementors. Implementors should document which routines are 
> implemented such that they are not interceptable. (End of advice to 
> implementors.)
>
> Advice to users. If these routines are implemented
> as not interceptable, they will not 
> work with the MPI profiling interface. (End of advice to users.) 

Yes!  That says precisely what it means and leaves any special code
generated for those functions as entirely the implementor's business.


Regards,
Nick Maclaren.