To: mpi-1sided @ mcs.anl.gov
cc: (bcc: Marc Snir/Watson/IBM Research)
Subject: points of chapter
--0__=tgkSMUTkl7ttkXavDcN5CMQqOYYzLfvVzzEk1o779IEinEB8iAumUKHH
Here are a few comments on the one-sided draft Marc sent out. Some
are small points but we are trying to clean things up.
4:16 - sizeof(type) should be MPI_TYPE_SIZE(MPI_Datatype). Doing
sizeof(MPI_Datatype) is a no no since it returns the size of the
datatype object. Doing sizeof(type) where type is the C equivalent of
MPI_Datatype is ok but less precise.
14:2 - I would spell out w.r.t. Mathematicians would know this but
may not all CS people :-).
22:36 - Should it be:
All processes must provide the same flag value.
====
22:44,23:4 - I think dest[i] should be toranks[i]
5.5.2 - I thought that the text was going to explicitly say it was the
same as the MPI-1 case for pt-2-pt. Did I miss it somewhere? I am
not saying the current text is wrong but I thought we voted to say
this to make it very clear (and not).
Steve
*****
All corrections accepted
*****
--0__=tgkSMUTkl7ttkXavDcN5CMQqOYYzLfvVzzEk1o779IEinEB8iAumUKHH--