[GE users] Parallel 'make'

David Olbersen dolbersen at nextwave.com
Wed Jun 4 19:35:02 BST 2008


Rayson,

OK, so download gmake 3.78, gmake 3.81 (or whatever), diff those to
create patch files.
Then, get the qmake source and apply the patches there, which should
upgrade the gmake parts of qmake to 3.81 (or whatever). Does that sound
about right?

In the mean time, I've run in to some other funky issues that I'd
appreciate help from anybody on:

1) On a 32-bit machine I submit my job, `qmake -cwd -V -pe make 1-10 -q
short -l arch=lx24-x86 --`
2) This gets sent out to our 32-bit nodes, and run, all is good.
3) On a 64-bit machine, I submit my job the same way, but get an error:
    ***Unknown option -inherit
    Usage: qmake [mode] [options] [files] 
4) If I change my command to, `qmake -cwd -V -pe make 1-10 -q short -l
arch=lx24-amd64 --` then I don't get the error for #3, but since I don't
have any 64-bit exec hosts my job can never be scheduled.

This seems to be a bit of a puzzle, does anybody have any hints?

-- 
David Olbersen
 

-----Original Message-----
From: Rayson Ho [mailto:rayrayson at gmail.com] 
Sent: Wednesday, June 04, 2008 9:45 AM
To: users at gridengine.sunsource.net
Subject: Re: [GE users] Parallel 'make'

David,

The only thing that's available now is the changes documented in issue
2544, I don't have patches yet. You will need to follow the discussions
in that issue page and roll your own qmake :-(

With help from the project owners, may be I can post on the project
homepage a tarball with all the changes.

Rayson


On 6/4/08, David Olbersen <dolbersen at nextwave.com> wrote:
> Rayson,
>
> Thanks for the pointers but I was only able to find an incomplete 
> email thread and no patches.
> Am I blind or has there been some nasty bit rot?
>
> --
> David Olbersen
>
>
> -----Original Message-----
> From: Rayson Ho [mailto:rayrayson at gmail.com]
> Sent: Monday, June 02, 2008 3:27 PM
> To: users at gridengine.sunsource.net
> Subject: Re: [GE users] Parallel 'make'
>
> If it works with gmake 3.81, then may be it's because the makefiles 
> are using features that are not available in 3.78. In that case, you 
> may consider compiling a newer qmake using the 3.8.1 source:
>
> http://gridengine.sunsource.net/issues/show_bug.cgi?id=2544
>
> Rayson
>
>
>
> On 6/2/08, David Olbersen <dolbersen at nextwave.com> wrote:
> > Reuti,
> >
> > Yes, it seems to work with the trivial case.
> > Using the simple Makefile below the two calls to gcc happened 
> > sequentially, on different hosts.
> >
> > Modifying that Makefile to make test.3, test.4, ..., test.6 worked 
> > as well, with it running sequentially.
> >
> > Modifying that Makefile further, to make test.1, test.2, ..., test.6

> > their own targets showed the biggest improvement, as all 6 
> > compilations happened at once
> >
> > So now that I can see that this works, I suppose I get to dig 
> > through my users Makefile to figure out what it's trying to do? Or 
> > does anybody have some other suggestions?
> >
> > --
> > David Olbersen
> >
> >
> > -----Original Message-----
> > From: Reuti [mailto:reuti at staff.uni-marburg.de]
> > Sent: Monday, June 02, 2008 3:01 PM
> > To: users at gridengine.sunsource.net
> > Subject: Re: [GE users] Parallel 'make'
> >
> > Am 02.06.2008 um 23:26 schrieb David Olbersen:
> >
> > > Sorry, I misunderstood:
> > >
> > >   [dolbersen at jotan ~]$ qrsh -q ncsim hostname
> > >   sge-alfrodull-065
> > >   [dolbersen at jotan ~]$
> >
> > No prob. Difference is, that in the first case (w/o command) a /usr/

> > sbin/in.rlogind is used, while with a command /usr/sge/utilbin/lx24-

> > x86/rshd (on the nodes).
> >
> > Is your makefile conforming to the rules laid out in `man qmake`? 
> > Can you try with a simple makefile whether it's working in
principle?
> >
> > all:
> >        gcc -o test.1 test.c
> >        gcc -o test.2 test.c
> >
> > -- Reuti
> >
> > >
> > > --
> > > David Olbersen
> > >
> > >
> > > -----Original Message-----
> > > From: Reuti [mailto:reuti at staff.uni-marburg.de]
> > > Sent: Monday, June 02, 2008 2:17 PM
> > > To: users at gridengine.sunsource.net
> > > Subject: Re: [GE users] Parallel 'make'
> > >
> > > Am 02.06.2008 um 22:53 schrieb David Olbersen:
> > >
> > >> Rayson,
> > >>
> > >> Yes, interactive jobs work well.
> > >> qrsh works as well:
> > >>
> > >>   [dolbersen at jotan ~]$ qrsh -q rfic
> > >>   Last login: Mon Jun  2 10:13:00 2008 from jotan.eng.atg.nw.net
> > >>   [dolbersen at sge-alfrodull-069 ~]$
> > >
> > > qrsh with command is also working? - Reuti
> > >
> > >>
> > >> --
> > >> David Olbersen
> > >>
> > >>
> > >> -----Original Message-----
> > >> From: Rayson Ho [mailto:rayrayson at gmail.com]
> > >> Sent: Monday, June 02, 2008 1:52 PM
> > >> To: users at gridengine.sunsource.net
> > >> Subject: Re: [GE users] Parallel 'make'
> > >>
> > >> David,
> > >>
> > >> Does interactive job work?? Can you run "qrsh hostname"??
> > >>
> > >> Rayson
> > >>
> > >>
> > >>
> > >> On 6/2/08, David Olbersen <dolbersen at nextwave.com> wrote:
> > >>> Reuti,
> > >>>
> > >>> No firewalls, no other parallel integration at all.
> > >>> Most of our jobs are numerous runs of single-threaded 
> > >>> applications, with different input for each run.
> > >>> Now one of my users has asked if they can run "make" on the 
> > >>> cluster to
> > >>
> > >>> get some speed benefit.
> > >>> This is to be the first parallel application.
> > >>>
> > >>> --
> > >>> David Olbersen
> > >>>
> > >>>
> > >>> -----Original Message-----
> > >>> From: Reuti [mailto:reuti at staff.uni-marburg.de]
> > >>> Sent: Monday, June 02, 2008 1:30 PM
> > >>> To: users at gridengine.sunsource.net
> > >>> Subject: Re: [GE users] Parallel 'make'
> > >>>
> > >>> Hi,
> > >>>
> > >>> Am 02.06.2008 um 19:34 schrieb David Olbersen:
> > >>>
> > >>>> I must be missing something, but I thought it would be pretty 
> > >>>> straightforward to run `make` via `qmake` to get some parallel 
> > >>>> advantage. It looks like this was a bad assumption. All the 
> > >>>> documentation I could find mentions using LAM, MPI, or some 
> > >>>> other
>
> > >>>> extra application: is that really required to get "make" 
> > >>>> running on
> >
> > >>>> a few machines in my cluster?
> > >>>>
> > >>>> Here's what I've tried so far:
> > >>>>
> > >>>> % qmake -cwd -V -pe test1 1-8 -q rfic -l arch=lx24-x86 -- 
> > >>>> PLATFORM=nw2000e
> > >>>> qmake: *** Warning: File 
> > >>>> `builds/nw2000e-debug-opt-rvct-3.0b617-
> > >>>> threadx42/depends.mak' has modification time in the future
> > >>>> (2008-06-02 10:20:40 > 2008-06-02 10:20:30) 
> > >>>> sge-alfrodull-069.eng.atg.nw.net
> > >>>> error: executing task of job 435644 failed:
> > >>>> qmake: *** remote_status: No child processes.  Stop.
> > >>>> qmake: *** Waiting for unfinished jobs....
> > >>>> qmake: *** remote_status: No child processes.  Stop.
> > >>>>
> > >>>> ... which exits very quickly and doesn't really do any builds.
> > >>>>
> > >>>> % qmake -cwd -V -q rfic -l arch=lx24-x86 -- PLATFORM=nw2000e
> > >>>> qmake: *** Warning: File 
> > >>>> `builds/nw2000e-debug-opt-rvct-3.0b617-
> > >>>> threadx42/depends.mak' has modification time in the future
> > >>>> (2008-06-02 10:16:40 > 2008-06-02 10:16:30) dynamic mode Build
> > >>>> info:
> > >>>> nw2000e-debug-opt-rvct-3.0b617-threadx42 Mon Jun 2 10:16:30 PDT
> > >>>> 2008
> > >>>> dolbersen at sge-alfrodull-071 dynamic mode dynamic mode dynamic 
> > >>>> mode dynamic mode dynamic mode dynamic mode dynamic mode 
> > >>>> dynamic mode dynamic mode dynamic mode dynamic mode dynamic 
> > >>>> mode
> > >>>> error: error waiting on socket for client to connect: 
> > >>>> Interrupted
>
> > >>>> system call
> > >>>
> > >>> other parallel apps are running using a Tight Integration? Any
> > >> firewall?
> > >>> - Reuti
> > >>>
> > >>>>
> > >>>> ... which does each job one at a time, as though I'd just run 
> > >>>> "make"
> > >>
> > >>>> -- no advantage at all.
> > >>>>
> > >>>> Can anybody help? Pointing me to documentation should be more 
> > >>>> than enough -- I've not been able to find what I'm looking for.
> > >>>>
> > >>>> David Olbersen
> > >>>>
> > >>>>
> > >>>
> > >>>
> > >>> ----------------------------------------------------------------
> > >>> --
> > >>> --
> > >>> -
> > >>> To unsubscribe, e-mail: 
> > >>> users-unsubscribe at gridengine.sunsource.net
> > >>> For additional commands, e-mail:
> > >>> users-help at gridengine.sunsource.net
> > >>>
> > >>>
> > >>>
> > >>> ----------------------------------------------------------------
> > >>> --
> > >>> --
> > >>> -
> > >>> To unsubscribe, e-mail: 
> > >>> users-unsubscribe at gridengine.sunsource.net
> > >>> For additional commands, e-mail:
> > >>> users-help at gridengine.sunsource.net
> > >>>
> > >>>
> > >>
> > >> -----------------------------------------------------------------
> > >> --
> > >> -- To unsubscribe, e-mail:
> > >> users-unsubscribe at gridengine.sunsource.net
> > >> For additional commands, e-mail:
> > >> users-help at gridengine.sunsource.net
> > >>
> > >>
> > >>
> > >> -----------------------------------------------------------------
> > >> --
> > >> -- To unsubscribe, e-mail:
> > >> users-unsubscribe at gridengine.sunsource.net
> > >> For additional commands, e-mail:
> > >> users-help at gridengine.sunsource.net
> > >>
> > >
> > >
> > > ------------------------------------------------------------------
> > > --
> > > - To unsubscribe, e-mail: 
> > > users-unsubscribe at gridengine.sunsource.net
> > > For additional commands, e-mail: 
> > > users-help at gridengine.sunsource.net
> > >
> > >
> > >
> > > ------------------------------------------------------------------
> > > --
> > > - To unsubscribe, e-mail: 
> > > users-unsubscribe at gridengine.sunsource.net
> > > For additional commands, e-mail: 
> > > users-help at gridengine.sunsource.net
> > >
> >
> >
> > --------------------------------------------------------------------
> > - To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> > For additional commands, e-mail: users-help at gridengine.sunsource.net
> >
> >
> >
> > --------------------------------------------------------------------
> > - To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> > For additional commands, e-mail: users-help at gridengine.sunsource.net
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list