Charm++ parallel programming system

[Edgar Solomonik <solomon AT eecs.berkeley.edu>]

Thanks for the hint Phil, I always forget xlc needs to be a separate
identifier.  I committed the change.  The pami version with xlc built
pretty quickly. The BG/Q installations I've worked with so far all
have much faster login nodes than Intrepid/Surveyor did and all
compile times are way down.

Edgar

On Mon, May 14, 2012 at 9:56 PM, Jeff Hammond 
<jhammond AT alcf.anl.gov>
 wrote:
> if compilation times are empirically bad, give me the data and i will
> throw more tirades regarding the unacceptable nature of a leadership
> computing facility with a login-node bottleneck.  if it was not
> grounds for termination, i would have yanked those PPC970 BGP logins
> out of the racks and burned them in the parking lot a long time ago.
> it's incredibly frustrating that some of my colleagues sabotaged my
> attempts to get better logins by arguing the bottleneck was the
> filesystem, which is provably false by comparison to JSC, at least for
> the codes i care about.
>
> jeff
>
> On Tue, May 15, 2012 at 12:48 AM, Jim Phillips 
> <jim AT ks.uiuc.edu>
>  wrote:
>>
>> The front of of BG/Q is a Power 7 so compilation times are reasonable.
>>
>> -Jim
>>
>>
>>
>> On Mon, 14 May 2012, Phil Miller wrote:
>>
>>> On Mon, May 14, 2012 at 8:02 PM, Edgar Solomonik
>>> <solomon AT eecs.berkeley.edu>
>>>  wrote:
>>>>
>>>> Is there a build of Charm++ on Cetus (BG/Q at ALCF)?  Looks like
>>>> mpi-bluegeneq does a g++ build, but I also see a xlc config file in
>>>> arch.  Is there a faster xlc build I can use?
>>>
>>>
>>> The pami-bluegeneq build should be somewhat better :-). In either
>>> case, the xlc modifier on the build will switch you from g++ to xlC++
>>> - better code, but much slower compilation.
>>>
>>>> Also this little problem in the Charm++ version of git that I pulled
>>>> an hour ago.
>>>>
>>>> TopoManager.C:302
>>>>
>>>> CmiAssert( a>=0 && x<dimNA && b>=0 && b<dimNB && c>=0 && c<dimNC &&
>>>> d>=0 && d<dimND && e>=0 && e<dimNE && t>=0 && t<dimNT );
>>>>
>>>> should be
>>>>
>>>> CmiAssert( a>=0 && a<dimNA && b>=0 && b<dimNB && c>=0 && c<dimNC &&
>>>> d>=0 && d<dimND && e>=0 && e<dimNE && t>=0 && t<dimNT );
>>>
>>>
>>> You should still have commit privileges - go ahead and fix it.
>>>
>>> Phil
>>>
>>> _______________________________________________
>>> charm mailing list
>>> charm AT cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/charm
>>>
>>> _______________________________________________
>>> ppl mailing list
>>> ppl AT cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/ppl
>
>
>
> --
> Jeff Hammond
> Argonne Leadership Computing Facility
> University of Chicago Computation Institute
> jhammond AT alcf.anl.gov
>  / (630) 252-5381
> http://www.linkedin.com/in/jeffhammond
> https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond (in-progress)
> https://wiki.alcf.anl.gov/old/index.php/User:Jhammond (deprecated)
> https://wiki-old.alcf.anl.gov/index.php/User:Jhammond(deprecated)