itaps-parallel Proposed tag communication functions

Onkar Sahni osahni at scorec.rpi.edu
Sat Mar 8 16:36:12 CST 2008


Here are my initial comments on proposed tag communication functions:

prefix_IExchTags and prefixIExchTagsEnt - seems like a send and there is
no equivalent (or counter-part) recv. I think we agreed on 'to_part' to be
replaceable by 'ALL_PART_NGBRS' (note shared entities will be on different
set of parts contained in 'ALL_PART_NGBRS' of local-part).

prefix_ExchTags and prefix_ExchTagsEnt - if they are synchronous call then
I would recommend to use names prefix_SExchTags and prefix_SExchTagsEnt as
used in MPI (In MPI the communication mode is indicated by a one letter
prefix: B for buffered, S for synchronous, and R for ready, so MPI_Send
may complete before a matching receive is posted). Same as above hold in
this case, there is no equivalent (or counter-part) recv.

prefix_WaitEnt - I am not sure if we need this. I think status object (in
prefix_Wait and prefix_WaitAny) should be enough.

- Onkar

> Hi all,
>    Here they are.  Note that I've added MPI-like request objects to use
> with asynchronous comm'n, and functions to test/wait on these requests,
> much like MPI has.  These functions should also be useful with ghost
> communication.
>
> - tim
>
> --
> ================================================================
> "You will keep in perfect peace him whose mind is
>    steadfast, because he trusts in you."               Isaiah 26:3
>
>               Tim Tautges            Argonne National Laboratory
>           (tautges at mcs.anl.gov)      (telecommuting from UW-Madison)
>           phone: (608) 263-8485      1500 Engineering Dr.
>             fax: (608) 263-4499      Madison, WI 53706
>
>





More information about the itaps-parallel mailing list