Re: amber-developers: bluegene...

From: Carlos P. Sosa <cpsosa.msi.umn.edu>
Date: Mon, 01 Oct 2007 12:11:22 -0500

Ross Walker wrote:

>Hi Carlos,
>
>Yes it would appear to be a problem with xlf 10.1 on Blue Gene. I have
>created a cut down section of code that causes the compiler to crash and
>Mahidhar Tatineni of SDSC should have created a ticket with IBM for this.
>
>So far I have not heard anything back...
>
>All the best
>Ross
>
>/\
>\/
>|\oss Walker
>
>| HPC Consultant and Staff Scientist |
>| San Diego Supercomputer Center |
>| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
>| http://www.rosswalker.co.uk | PGP Key available on request |
>
>Note: Electronic Mail is not secure, has no guarantee of delivery, may not
>be read every day, and should not be used for urgent or sensitive issues.
>
>
>
>>-----Original Message-----
>>From: owner-amber-developers.scripps.edu
>>[mailto:owner-amber-developers.scripps.edu] On Behalf Of
>>Carlos P. Sosa
>>Sent: Monday, October 01, 2007 06:47
>>To: amber-developers.scripps.edu
>>Subject: Re: amber-developers: bluegene...
>>
>>Kim F Wong wrote:
>>
>>
>>
>>>Ross,
>>>
>>>Your fix is good. I've been recycling routines liberally
>>>
>>>
>>and quickly
>>
>>
>>>and this is what happens. Thanks.
>>>
>>>Kim
>>>
>>>Ross Walker wrote:
>>>
>>>
>>>
>>>>Hi Kim,
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>>Can you tell me why evb_io_debug and evb_gfchk are croaking on
>>>>>bluegene ... that is, are there any programming habits I need to
>>>>>break? The evb_io_debug was for my own edification and can be
>>>>>commented out but the evb_gfchk is integral to Berny Schelgel's
>>>>>distributed Gaussian EVB approach. Sorry the codes are messy. I
>>>>>will clean up the debugging soon.
>>>>>
>>>>>
>>>>>
>>>>The xlf compiler has more rigorous error checking than the Intel
>>>>compilers.
>>>>The initial problem was the call arguments to
>>>>
>>>>
>>evb_io_debug. It looks
>>
>>
>>>>like
>>>>you either updated evb_io_debug at some point and forgot to update
>>>>all the
>>>>calls or you meant to call evb_io_debug2 routine instead. I have
>>>>tweaked the
>>>>CVS tree to address this - take a look and see if you
>>>>
>>>>
>>think it is okay.
>>
>>
>>>>The failure beyond this fix looks like some weird compiler bug - I
>>>>want to
>>>>see if I can reproduce it and then I'll check with IBM to
>>>>
>>>>
>>see what it
>>
>>
>>>>might
>>>>be. I wouldn't worry about your style for the moment since style
>>>>shouldn't
>>>>crash a compiler.
>>>>
>>>>All the best
>>>>Ross
>>>>
>>>>/\
>>>>\/
>>>>|\oss Walker
>>>>
>>>>| HPC Consultant and Staff Scientist |
>>>>| San Diego Supercomputer Center |
>>>>| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
>>>>| http://www.rosswalker.co.uk | PGP Key available on request |
>>>>
>>>>Note: Electronic Mail is not secure, has no guarantee of delivery,
>>>>may not
>>>>be read every day, and should not be used for urgent or sensitive
>>>>issues.
>>>>
>>>>
>>>>
>>>>
>>>>
>>Ross,
>>
>>I just got back from a trip and I am trying to catch up with
>>e-mail. Is
>>this still a potential XL compiler problem?
>>
>>--
>>Carlos P. Sosa
>>Address: University of Minnesota
>> Supercomputing Institute for
>> Digital Simulation and Advanced Computation
>> 599 Walter Library
>> 117 Pleasant St. SE
>> Minneapolis, MN 55455
>>email: cpsosa.msi.umn.edu
>>
>>
>>
>>
>
>
>
Hi Ross,

If you have a kernel that I can run and reproduce the error, I can talk
to the compiler developers. I do not have access to the issue that you
opened.

Thanks,


-- 
Carlos P. Sosa
Address: University of Minnesota                           
         Supercomputing Institute for                      
         Digital Simulation and Advanced Computation      
         599 Walter Library
         117 Pleasant St. SE
         Minneapolis, MN 55455       
email:   cpsosa.msi.umn.edu
Received on Wed Oct 03 2007 - 06:07:22 PDT
Custom Search