[torquedev] array_changes and NUMA

David Beer dbeer at adaptivecomputing.com
Tue Jun 8 13:56:42 MDT 2010


Garrick,

I've been working a lot on the NUMA branch - its still undergoing massive development and testing so that it can be compatible with Ultra-violet hardware. Let's let the code stabilize and mature before we merge it back to the trunk. Also, it is not going to be released in 2.5 (its going to be in 3.0), so we wouldn't want to merge it back until that is released. 

Today I have testing time on a UV system, and that's why I didn't respond. I will try to figure out which ENABLE_CSA and USEJOBCREATE is correct when I have a chance.

Thanks for doing what you've done.

David

----- Original Message -----
> Well, I've done what I can. There are a lot of changes in the NUMA
> branch that
> will be difficult to merge back into trunk. But maybe someone should
> start moving small chunks of isolated and working code.
> 
> 
> One such example is in src/lib/Libutil/u_tree.c. NUMA's version was
> extensively changed and doesn't even compile cleanly.
> 
> 
> $ svn log -r3508:3510 ../NUMA/src/lib/Libutils/u_tree.c | less
> ------------------------------------------------------------------------
> r3508 | knielson | 2010-03-18 12:16:20 -0700 (Thu, 18 Mar 2010) | 1
> line
> 
> creating a private branch in develop NUMA support
> ------------------------------------------------------------------------
> r3510 | knielson | 2010-03-19 17:09:41 -0700 (Fri, 19 Mar 2010) | 3
> lines
> 
> Integrated multi-mom code. Need to build in backwards compatibility so
> users can more easily upgrade.
> Changed the tinsert, tfind, etc calls with AVL_insert, AVL_find, etc.
> 
> ------------------------------------------------------------------------
> 
> 
> $ diff -pruN ../NUMA/src/lib/Libutils/u_tree.c
> ../../trunk/src/lib/Libutils/u_tree.c | wc -l
> 684
> 
> 
> $ make -C ../NUMA-build/src/lib/Libutils/
> make: Entering directory
> `/usr/local/src/garrickrpm/crisvn/torque/branches/NUMA-build/src/lib/Libutils'
> if gcc -DHAVE_CONFIG_H -I. -I../../../../NUMA/src/lib/Libutils
> -I../../../src/include -I../../../../NUMA/src/include -g -O2 -W -Wall
> -Wno-unused-parameter -Wno-long-long -pedantic -Werror
> -D_LARGEFILE64_SOURCE -MT u_tree.o -MD -MP -MF ".deps/u_tree.Tpo" -c
> -o u_tree.o ../../../../NUMA/src/lib/Libutils/u_tree.c; \
> then mv -f ".deps/u_tree.Tpo" ".deps/u_tree.Po"; else rm -f
> ".deps/u_tree.Tpo"; exit 1; fi
> cc1: warnings being treated as errors
> ../../../../NUMA/src/lib/Libutils/u_tree.c: In function
> ‘AVL_delete_node’: ../../../../NUMA/src/lib/Libutils/u_tree.c:226:
> warning: ‘h2’ may be used uninitialized in this function
> make: *** [u_tree.o] Error 1
> 
> 
> 
> On Tue, Jun 08, 2010 at 11:21:08AM -0700, Garrick Staples alleged:
> > I'm not going to be able to resolve these conflicts. Work has been
> > going on
> > with CSA code in both the NUMA and trunk. I don't know which code to
> > keep.
> >
> > Which is prefered? ENABLE_CSA or USEJOBCREATE code?
> >
> > On Tue, Jun 08, 2010 at 10:57:23AM -0600, David Beer alleged:
> > > Garrick,
> > >
> > > I don't think NUMA has been synced with the trunk in a while. I
> > > had been doing it, but I've gotten sidetracked. If you don't mind,
> > > I'd really appreciate it.
> > >
> > > David
> > >
> > > ----- Original Message -----
> > > > What about NUMA?
> > > >
> > > > On Tue, Jun 08, 2010 at 09:16:04AM -0600, David Beer alleged:
> > > > > Garrick
> > > > >
> > > > > Thanks for helping keep things synced.
> > > > >
> > > > > David
> > > > >
> > > > > ----- Original Message -----
> > > > > > Last sync from trunk was in April. You are up to date now.
> > > > > >
> > > > > > I have a nifty script that makes this easy, so I'm happy to
> > > > > > keep the branched synced for you.
> > > > > >
> > > > > > On Mon, Jun 07, 2010 at 08:12:01PM -0400, Glen Beane
> > > > > > alleged:
> > > > > > > We've been synching array_changes with trunk. It may have
> > > > > > > been a
> > > > > > > couple months now because development in trunk has slowed
> > > > > > > down.
> > > > > > >
> > > > > > > Go for it if you want. We wanted to merge it into trunk by
> > > > > > > the 15th of this month but there are a few bugs right now.
> > > > > > > Let me
> > > > > > > know if you
> > > > > > > are interested in helping to debug. We want to get these
> > > > > > > changes merged and do a 2.5.0 beta snapshot for testing.
> > > > > > >
> > > > > > > Sorry for the top post, sending from my iPhone
> > > > > > >
> > > > > > > On Jun 7, 2010, at 7:07 PM, Garrick Staples
> > > > > > > <garrick at usc.edu>
> > > > > > > wrote:
> > > > > > >
> > > > > > > > Should array_change and NUMA be following trunk? Would
> > > > > > > > you like a
> > > > > > > > hand tracking
> > > > > > > > changes in trunk?
> > > > > > > >
> > > > > > > > -- Garrick Staples, GNU/Linux HPCC SysAdmin
> > > > > > > > University of Southern California
> > > > > > > >
> > > > > > > > Life is Good!
> > > > > > > > _______________________________________________
> > > > > > > > torquedev mailing list torquedev at supercluster.org
> > > > > > > > http://www.supercluster.org/mailman/listinfo/torquedev
> > > > > > > _______________________________________________ torquedev
> > > > > > > mailing list torquedev at supercluster.org
> > > > > > > http://www.supercluster.org/mailman/listinfo/torquedev
> > > > > >
> > > > > > -- Garrick Staples, GNU/Linux HPCC SysAdmin
> > > > > > University of Southern California
> > > > > >
> > > > > > Life is Good!
> > > > > >
> > > > > > _______________________________________________ torquedev
> > > > > > mailing list torquedev at supercluster.org
> > > > > > http://www.supercluster.org/mailman/listinfo/torquedev
> > > > >
> > > > > -- David Beer | Senior Software Engineer
> > > > > Adaptive Computing
> > > > > _______________________________________________ torquedev
> > > > > mailing list torquedev at supercluster.org
> > > > > http://www.supercluster.org/mailman/listinfo/torquedev
> > > >
> > > > -- Garrick Staples, GNU/Linux HPCC SysAdmin
> > > > University of Southern California
> > > >
> > > > Life is Good!
> > > >
> > > > _______________________________________________ torquedev
> > > > mailing list
> > > > torquedev at supercluster.org
> > > > http://www.supercluster.org/mailman/listinfo/torquedev
> > >
> > > -- David Beer | Senior Software Engineer
> > > Adaptive Computing
> > > _______________________________________________ torquedev mailing
> > > list torquedev at supercluster.org
> > > http://www.supercluster.org/mailman/listinfo/torquedev
> >
> > -- Garrick Staples, GNU/Linux HPCC SysAdmin
> > University of Southern California
> >
> > Life is Good!
> 
> 
> 
> > _______________________________________________ torquedev mailing
> > list torquedev at supercluster.org
> > http://www.supercluster.org/mailman/listinfo/torquedev
> 
> 
> -- Garrick Staples, GNU/Linux HPCC SysAdmin
> University of Southern California
> 
> Life is Good!
> 
> _______________________________________________ torquedev mailing list
> torquedev at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torquedev

-- 
David Beer | Senior Software Engineer
Adaptive Computing


More information about the torquedev mailing list