[mvapich-discuss] Bug fixes for lazy deregister code

[Olga Shern]

Hi, Jimmy
 
No, these bugs are not related to pathscale compiler,
these are logical bugs, one is related to 32-64 bit and the second one
is related to wrong page calculation.
We and our customers are using pathscale with -02 
and as far as we know there was no problem with this code.
 
Olga 

________________________________

From: jcftang at gmail.com [mailto:jcftang at gmail.com] On Behalf Of Jimmy
Tang
Sent: Thursday, June 29, 2006 12:27 PM
To: Olga Shern; mvapich-discuss at cse.ohio-state.edu
Subject: Re: [mvapich-discuss] Bug fixes for lazy deregister code


Hi Olga,

Regarding these bugs would these bugs manifest themselves as a
consquence of using pathscale compilers compiled with -O2 or higher
optimisations ? one of our users fortran90 codes broke quite badly with
lazy deregister and pathscale with -02 or higher. 


Jimmy.


On 6/29/06, Olga Shern <olgas at voltaire.com> wrote: 

	Hello, Dr. Panda.
	 
	We are going through MVAPICH 0.9.8 code and have noticed that 
	you inserted our improvements for lazy deregister that we sent
you more then a year ago.
	Meantime we found few bugs and fixed them, patch with the
bug-fix's is attached. 
	We will be glad if you can merge these  fixes in the upcoming
MVAPICH version .
	 
	Thanks,
	Olga S.
	Voltaire HPC Solution


	_______________________________________________
	mvapich-discuss mailing list
	mvapich-discuss at cse.ohio-state.edu
	http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss 
	
	
	
	




-- 
Jimmy Tang
Trinity Centre for High Performance Computing,
Lloyd Building, Trinity College Dublin.
http://www.tchpc.tcd.ie/ 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20060629/64d8680c/attachment.html