[mpich2-dev] tests which assume that the implementation supports truncated messages

[Rajeev Thakur]

> Would it make sense to change these tests to be more 
> portable? Could they be made to trigger a different sort of 
> error, perhaps one checked in the HAVE_ERROR_CHECKING 
> portions of the top-level receive code?

It can't be checked in the HAVE_ERROR_CHECKING portions of the top-level
receive code because the size of the matching send is not known at that
time.

Rajeev
 

> -----Original Message-----
> From: owner-mpich2-dev at mcs.anl.gov 
> [mailto:owner-mpich2-dev at mcs.anl.gov] On Behalf Of David Gingold
> Sent: Tuesday, April 01, 2008 3:44 PM
> To: mpich2-dev at mcs.anl.gov
> Subject: [mpich2-dev] tests which assume that the 
> implementation supports truncated messages
> 
> MPICH2 folks -- 
> 
> Some of the tests that come with MPICH2 assume that the MPI 
> implementation supports truncated messages (i.e. 
> point-to-point operations with more data sent than received). 
> It looks like the tests are using this as a way to determine 
> whether errors can be returned in the receive status.
> 
> Our MPI implementation doesn't support truncation. It's not 
> required by the standard, and dispensing with it simplified 
> our code and improved performance.
> 
> Would it make sense to change these tests to be more 
> portable? Could they be made to trigger a different sort of 
> error, perhaps one checked in the HAVE_ERROR_CHECKING 
> portions of the top-level receive code?
> 
> -dg
> 
> --
> David Gingold
> Principal Software Engineer
> SiCortex
> Three Clock Tower Place, Suite 210
> Maynard MA 01754
> (978) 897-0214 x224
> 
> 
> 
>