johnmay @ llnl.gov
04/16/97 05:29 PM
To: mpi-io @ mcs.anl.gov
cc: (bcc: Jean-Pierre Prost/Watson/IBM Research)
Subject: Re: tiling issues
Jean-Pierre Prost writes:
> I would rather keep the tiling for files the same as for memory.
> In other words, if the user really wants to achieve:
> hole, int, hole, int, hole, int, ....
> the filetype used should be:
> (lb, 0), (int, 4)
> which leads to an extent of 8.
I agree that the tiling should work the same for filetypes
and buftypes. Rajeev believes that the definition of a
filetype and a view allows the interpretation that the
tiling of filetypes is different from the tiling of
buftypes. How about adding this to the definition of a view:
The pattern described by a filetype is repeated, beginning
at displacement, to define the view. The pattern of
repetition is defined to be the same pattern that
MPI_TYPE_CONTIGUOUS would produce if it were passed the
filetype and an arbitarily large count. Figure 10.1
shows how the tiling works; note that the filetype in this
example must have an explicit lower bound set in order
for the intial hole to be repeated in the view.
John