Re: problems running UPC programs

From: Eric Frederich (eric.frederich_at_gmail_dot_com)
Date: Tue Nov 22 2005 - 15:39:11 PST

  • Next message: Dan Bonachea: "Re: problems running UPC programs"
    Actually, I went and copied the same file over to the other machine so now
    that shouldn't be the issue. I am still getting what appears to be the same
    error, a problem with the requested resource, and something about an invalid
    argument.
    
    I ran md5sum on the binaries and they are exactly the same.
    
    Please help me. I don't know what else could be the problem. Why would I be
    able to launch 2 processes on one machine, be able to launch 2 processes on
    another machine, but not be able to launch 1 process on each?
    
    ~Eric
    
    
    eric@penguin27 build $ ./upcrun -n 2 helloWorld
    UPCR: UPC thread 0 of 2 on penguin27 (process 0 of 2, pid=17505)
    UPCR: UPC thread 1 of 2 on penguin27 (process 1 of 2, pid=17514)
    Hello World from thread 2 of 2 ! !
    Hello World from thread 1 of 2 ! !
    eric@penguin27 build $ ./upcrun -n 2 helloWorld
    UPCR: UPC thread 0 of 2 on myth (process 0 of 2, pid=8799)
    UPCR: UPC thread 1 of 2 on myth (process 1 of 2, pid=8801)
    Hello World from thread 1 of 2 ! !
    Hello World from thread 2 of 2 ! !
    eric@penguin27 build $ ./upcrun -n 2 helloWorld
    AMUDP sendPacket returning an error code: AM_ERR_RESOURCE (Problem with
    requested resource)
    from function sendPacket
    at /home/eric/UPC/berkeley_upc-2.2.1/gasnet/other/amudp/amudp_reqrep.cpp:93
    reason: Invalid argument
    AMUDP AMUDP_RequestGeneric returning an error code: AM_ERR_RESOURCE (Problem
    with requested resource)
    at /home/eric/UPC/berkeley_upc-2.2.1
    /gasnet/other/amudp/amudp_reqrep.cpp:1200
    
    GASNet gasnetc_AMRequestShortM encountered an AM Error: AM_ERR_RESOURCE(3)
    at /home/eric/UPC/berkeley_upc-2.2.1/gasnet/udp-conduit/gasnet_core.c:564
    GASNet gasnetc_AMRequestShortM returning an error code: GASNET_ERR_RESOURCE
    (Problem with requested resource)
    at /home/eric/UPC/berkeley_upc-2.2.1/gasnet/udp-conduit/gasnet_core.c:568
    *** FATAL ERROR:
    GASNet encountered an error: GASNET_ERR_RESOURCE(3)
    while calling: gasnet_AMRequestShort4(peer,
    gasneti_handleridx(gasnete_ambarrier_notify_reqh), phase, 0, id, flags)
    at gasnete_barrier_notify() at /home/eric/UPC/berkeley_upc-2.2.1
    /gasnet/extended-ref/gasnet_extended_refbarrier.c:197
    *** Caught a fatal signal: SIGABRT(6) on node 1/2
    eric@penguin27 build $ md5sum /home/eric/UPC/build/helloWorld
    da6cd645cb5562569d38c82173da2ae3 /home/eric/UPC/build/helloWorld
    eric@penguin27 build $ ssh myth md5sum /home/eric/UPC/build/helloWorld
    da6cd645cb5562569d38c82173da2ae3 /home/eric/UPC/build/helloWorld
    eric@penguin27 build $
    
    
    On 11/22/05, Eric Frederich <eric.frederich_at_gmail_dot_com> wrote:
    >
    > Wow, funny how we need a udp expert to figure out something not related at
    > all.
    >
    > One of those little subtle things. Later on this evening when I am at home
    > I will be able to test it out with the correct executable.
    > Hopefully I will have some good news to report.
    >
    > Thanks a bunch,
    > ~Eric
    >
    > On 11/22/05, Dan Bonachea < bonachea_at_cs_dot_berkeley_dot_edu> wrote:
    > >
    > > Hi Eric - I'm the udp-conduit expert..
    > >
    > > I'm not sure why you're seeing that particular error message, although
    > > based
    > > on your message below I suspect you have inconsistent copies of the
    > > executable
    > > on the two nodes - the penguin27 output is "Hello World from thread 1 of
    > > 2"
    > > but the myth output is "Hello World" - which probably means the programs
    > > are
    > > not the same.
    > >
    > > Berkeley UPC requires all nodes to be running the *exact* same binary
    > > executable - if you lack a shared file system then exact copies are fine
    > >
    > > (although error-prone), but it's not OK to recompile one copy and not
    > > the
    > > others. Also, udp-conduit requires all copies of the executable to
    > > reside at
    > > the same absolute pathname on all clients - so make sure the copies are
    > > all
    > > mounted or mirrored to the same absolute path. Also, if the nodes may
    > > differ
    > > in things like shared libraries, you should probably link statically
    > > (upcc
    > > -Wl,-static) just to be safe.
    > >
    > > Give it another try once you're certain the same binary is present and
    > > working
    > > on all nodes. If it still fails, try appending "-v" to the upcrun line
    > > to see
    > > more details about the startup procedure and send us the complete
    > > output.
    > > Please also send the output of "uname -a" and "cat /proc/cpuinfo" on
    > > each
    > > node.
    > >
    > > Hope this helps...
    > > Dan
    > >
    > > At 02:35 PM 11/21/2005, Eric Frederich wrote:
    > > > > It is intersting to note that when the upchostsfile looks like
    > > > >
    > > > > 192.168.1.207 <http://192.168.1.207> <http://192.168.1.207>
    > > > > 192.168.1.207 <http://192.168.1.207> < http://192.168.1.207>
    > > > > 192.168.1.208 <http://192.168.1.208> <http://192.168.1.208>
    > > > >
    > > > > and I run it with -n 2 it works fine and I see the following
    > > > >
    > > > > UPCR: UPC thread 0 of 2 on penguin27 (process 0 of 2, pid=12356)
    > > > > UPCR: UPC thread 1 of 2 on penguin27 (process 1 of 2, pid=12357)
    > > > > Hello World from thread 1 of 2
    > > > > Hello World from thread 0 of 2
    > > > >
    > > > > Also when I have the file say
    > > > >
    > > > > 192.168.1.208 <http://192.168.1.208> <http://192.168.1.208>
    > > > > 192.168.1.208 <http://192.168.1.208> < http://192.168.1.208>
    > > > > 192.168.1.207 <http://192.168.1.207> <http://192.168.1.207>
    > > > >
    > > > > it works fine too and I see the following
    > > > >
    > > > > UPCR: UPC thread 0 of 2 on myth (process 0 of 2, pid=10447)
    > > > > UPCR: UPC thread 1 of 2 on myth (process 1 of 2, pid=10446)
    > > > > Hello World
    > > > > Hello World
    > >
    > >
    >
    >
    > --
    > ------------------------
    > Eric L. Frederich
    >
    
    
    
    --
    ------------------------
    Eric L. Frederich
    

  • Next message: Dan Bonachea: "Re: problems running UPC programs"