Re: Suggestion for MPI_Print

Charlie Fineman (Charles.Fineman@Eng.Sun.COM)
Mon, 24 Jun 1996 13:45:06 -0700

Actually, you raise an interesting idea. I wonder if the parallel diff
tool actually *should* support this mode (being able to do a diff on
seperate data streams contained in a single file). It may be
preferable for users (from their perspective) to leverage support for
shared files on their platforms rather than create a seperate file for
each data stream than it would be for them to manage the creation of
one file for each node.

Charlie

John M. May writes:
> > Charlie Fineman:
> > Hmmm, why wouldn't one just use the parallel diff tool (another
> > project proposed at the meeting) to solve this problem?
> [of detecting when different nodes produce different outputs]
>
> MPI_Print as I'm proposing it would write all its output to one place;
> there wouldn't be separate files for each node. In that case, diff
> wouldn't help. Unless, I guess, you first grepped lines from different
> nodes into separate files and then diffed them.
>
> John
>