Re: comments on datatype encode/decode

Richard Frost (frost@SDSC.EDU)
Mon, 28 Oct 1996 08:17:39 -0800 (PST)

Hi Steve,

Thanks for your comments. I will send updated LaTeX a.s.a.p.

> I have some questions/comments on 7.7.2,7.7.3 for datatype
> encode/decode (comments based on 25 Oct 96 draft):

Some answers:

> - 151/30: I assume the example #'s given are a reference to the
> MPI-1.1 standard? We need the correct reference.

Agreed.

> - 151/30: I find the examples hard to follow. Are they up to date
> relative to the tables? For example, where did the "U" come from.

The examples are wrong and need updating. :-)

> Also, I think it might be useful (for at least one example) to outline
> exactly what would go in each byte since this is a byte encoding.

Agreed.

> - 153/25: What happened to the optional datatypes? I know you sent me
> some mail about resolving it, Richard, but now I don't understand what
> happened.

After listening to the comments of a few vendors, I took author's
priviledge and deleted them. If you want to send architectural
information around, then comment on the MPI_ARCH_ENCODE function
being discussed in the I/O group (but headed for external section 7.7.4).

>
> Minor corrections:
>

> - 149/28: "If comm may be one" wording off

Agreed.

> - 150/1: "MPI_TYPE_ENCODE communicator comm" -> "MPI_TYPE_ENCODE with
> communicator comm"

Agreed.

> - 151+: References to tables non-std (Ricard: I will try to fix this
> once I get a new version from you)

Or please send me a blurb on table reference standard for the MPI document.

> - 151/23: "toplevel" -> "top level"

Agreed.

> - 153/9-24: It would be nice to have all the numbers in a straight
> column.

Agreed.

Again, Thanks for your careful reading.

- Richard