Charm++ parallel programming system

[Hrushit Parikh <parikhhrushit AT hotmail.com>]

Hi,
Thank you, upgrading to charm-6.6.1 worked. I am able to link the UTS code with projections and generate analysis data.

However, when I tried to run UTS with -tracemode projections and my modified cldb.workstealing.c "seed balancer" I run into a segmentation fault. I have been trying to do a stack backtrace analysis and debug it since couple of days. Here is what I have:

The CopyMsg function in machine-common-core.c throws an error (address out of bounds) while doing a memcpy:

************************************************
/* Utiltiy functions */
static char *CopyMsg(char *msg, int len) {
    char *copy = (char *)CmiAlloc(len);
#if CMK_ERROR_CHECKING
    if (!copy) {
        CmiAbort("Error: out of memory in machine layer\n");
    }
#endif
    memcpy(copy, msg, len);
    return copy;
}

The values of copy, msg and len are:

copy: 0x2aaac0ce0020""
len: 50331728
msg: 0xe25940""

*************************************************

Seeing the length it is most likely that we are trying to ask for too big a chunk of memory. I tracked back the  char *msg in the above code to envelope *env in function CkCreateGroup(...) in file ck.C

What parameters is env storing that it has a length of 50331728 ? What data from the converse run time system, especially the seed balancer does it take which could be causing this error, because my changes are primarily in the workstealing seed balancer. 

Thank you in advance.

PS: My seed balancer works fine if I do not link projections to the UTS code.

> From: mille121 AT illinois.edu
> Date: Sun, 15 Feb 2015 19:52:36 -0600
> Subject: Re: [charm] Unable to run UTS with projections
> To: parikhhrushit AT hotmail.com
> CC: charm AT cs.uiuc.edu
>
> Hi Hrushit,
>
> The first recommendation I would give is to use the current stable
> release of Charm++, 6.6.1:
> http://charm.cs.illinois.edu/distrib/charm-6.6.1.tar.gz
>
> Numerous bug-fixes and improvements have been made since version 6.5.1
> came out, and it's often unproductive to try to diagnose an error
> that's already been fixed.
>
> Thanks,
>
> Phil Miller
>
> On Sun, Feb 15, 2015 at 7:18 PM, Hrushit Parikh
> <parikhhrushit AT hotmail.com> wrote:
> > Hi,
> > I am trying to get projections analysis on the UTS benchmark (Tree T1L) in
> > charm++ using the "-tracemode projections" option. I use the workstealing
> > seed balancer.
> >
> > I am using charm-6.5.1 version on the UT Austin Stampede cluster. The
> > stampede cluster uses ibrun to launch the executable. I have compiled charm
> > using ./build charm++ mpi-linux-x86_64 mpicxx -j16.
> >
> > With 256 Processors, the T1L run takes only 0.3 seconds when I compile charm
> > with the "--with-production" option (to disable traces). However, with
> > traces enabled the execution time exceeds 15 minutes and I have to kill the
> > program.
> >
> > I tried running NQueens with the same workstealing balancer on 256
> > Processors with traces enabled. For queens upto 19 (grainsize=3) it runs
> > fine and projections data is generated, but with 20 queens it takes forever.
> >
> > Any ideas why UTS takes forever with traces enabled or am I missing
> > something?
> >
> > Thank you,
> > Hrushit
> >
> > _______________________________________________
> > charm mailing list
> > charm AT cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/charm
> >