[Mpiwg-large-counts] Minor improvement to Pythonizing

[Rolf Rabenseifner]

Dear Jeff,

> I made a minor improvement to the Pythonizing script this morning: it now
> automatically adds a \cdeclindex{} entry to each LIS rendering for each "MPI_*"
> C type that is found in a given MPI procedure.

This seems to be the contrary:
The index entries are well chosen according to relevance, 
i.e., not always just all Locations.
And exactly 1-3 defining locations are underlined.

- MPI_Comm, MPI_Datatype: 
  all generation and handling routines, but never the routines, 
  where such a handle is just an Input argument, i.e., just used.

For some others, all locations may be there:
- MPI_File ?
- MPI_Op ?
- MPI_Request ???

Doing this when implementing the index, was many hours of work.

I had not the time to check all.
But having all MPI_Comm or MPI_Datatype in the index makes it unusable!

Best regards
Rolf

  
----- Original Message -----
> From: "mpiwg-large-counts" <mpiwg-large-counts at lists.mpi-forum.org>
> To: "mpiwg-large-counts" <mpiwg-large-counts at lists.mpi-forum.org>
> Cc: "Jeff Squyres" <jsquyres at cisco.com>
> Sent: Saturday, October 26, 2019 5:39:22 PM
> Subject: [Mpiwg-large-counts] Minor improvement to Pythonizing

> I made a minor improvement to the Pythonizing script this morning: it now
> automatically adds a \cdeclindex{} entry to each LIS rendering for each "MPI_*"
> C type that is found in a given MPI procedure.
> 
> Specifically, there are a *lot* of places in the old .tex that manually added a
> \cdeclindex{} entry for some or all of the MPI handles that are used in a given
> binding.  The old LaTeX was not consistent, though:
> 
> - Sometimes only some of the handle types were \cdeclindex{}'ed
> - Sometimes none of the handle types were \cdeclindex{}'ed
> - Sometimes all of the handle types were \cdeclindex{}'ed
> - Sometimes non-handle types were \cdeclindex{}'ed (e.g., MPI_Aint)
> 
> The Pythonizing improvement this morning always lists a \cdeclindex{} for every
> C type that is passed that begins with "MPI_" -- to include MPI_Aint,
> MPI_Count, MPI_Status, MPI_Comm, etc. (because I saw some of these in the
> existing hard-coded LaTeX).
> 
> The end result is that the handle type index is significantly more detailed /
> accurate than it was in MPI-3.1.
> 
> Hopefully this is not controversial.
> 
> --
> Jeff Squyres
> jsquyres at cisco.com
> 
> _______________________________________________
> mpiwg-large-counts mailing list
> mpiwg-large-counts at lists.mpi-forum.org
> https://lists.mpi-forum.org/mailman/listinfo/mpiwg-large-counts

-- 
Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner at hlrs.de .
High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 .
University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 .
Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner .
Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) .