Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] Compile-time MPI_Datatype checking
From: Paul Hargrove (phhargrove_at_[hidden])
Date: 2012-10-31 16:01:04

On Wed, Oct 31, 2012 at 12:47 PM, Ralph Castain <rhc_at_[hidden]> wrote:

> On Oct 31, 2012, at 12:36 PM, Paul Hargrove <phhargrove_at_[hidden]> wrote:
> Ralph,
> I don't think I missed the point about the origin of the concern - we just
> have different view points.
> Jeff indicated that he previously thought the "odd" practice shown in the
> example was uncommon until he learned is was common in codes at Sandia.
> Perhaps you and I have interpreted the "impact" differently:
> + From your responses I gather you are concerned only (or primarily) w/
> folks at Sandia who work on codes that have the "odd" behavior.
>, that isn't accurate. I was only trying to say that the concern
> was raised from that corner for those reasons. And I don't think they would
> consider a code that meets standards as being "odd".

Sorry if I mischaracterized you view - I chose the phrase "I gather" with
the intent of expressing MY understanding/interpretation of your postings.
 I guess I MISunderstood/MISinterpreted.

I chose "odd" for lack of a better term to describe code using a char* for
an array of ints.
I fully agree that the code is standards-compliant - otherwise I'd not care
one iota about warnings or no warnings.

> + On the other hand, I took the Sandia codes as an "existence proof"
> showing that "really smart people" can still write questionable code at
> times. So, I think there could be a non-trivial number of codes developed
> outside of Sandia that would start generating warnings (possibly MANY of
> them).
> Could be - but again, not many are likely to be using clang, and so they
> won't be affected by this change.
> I will now crawl back under my rock - I honestly couldn't care less about
> this, but was trying to explain the concern voiced on the telecon as that
> person is way too busy for this debate.
> [snip - find the full thread in the archives]

> I am crawling back under my rock too, due to a similar lack of true
concern over the outcome.

-Paul [who's only ever used MPI_BYTE]

Paul H. Hargrove                          PHHargrove_at_[hidden]
Future Technologies Group
Computer and Data Sciences Department     Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900