Probably you should try to change the value realDim=16384 to something more at your machine level, such as 1024... So that you won't experiment a hung :) My purpose is no to use an already done fft algorithm, but just to try how it is to write one starting from a serial version.. I know that exists 10000kk times better algorithms.. -----Original Message----- From: Jeff Hammond <[email protected]> To: discuss <[email protected]> Sent: Fri, Jun 28, 2013 5:41 pm Subject: Re: [mpich-discuss] mpi assertion error Yes, I can tell your new to programming. Stop reading K&R C as a style guide :-) Your program hung my virtual machine so I cannot help you. I recommend you use FFTW instead of rolling your own FFTs. Best, Jeff On Fri, Jun 28, 2013 at 10:28 AM, Danilo <[email protected]> wrote:
Hi Jeff, the program was tested intensively on the previous cluster. The changes made are in scatter/gather (due to sendbuf and recvbuff that has to be differente in this version it seems..). The other main change is due to hydra, because on the previous cluster there wasn't such a process management system. But I'm quite new to programming, so I don't know...
Thanks for your help.
Regards
-----Original Message----- From: Jeff Hammond <[email protected]> To: discuss <[email protected]> Sent: Fri, Jun 28, 2013 5:12 pm Subject: Re: [mpich-discuss] mpi assertion error
Null buffer assertions are suggestive of incorrect programs. Can you share the source of this program?
As for the inline vs attached files debate, I think that pastebin is a superior option for large output since it is plain-text readable from any internet-enabled device and doesn't lead to huge messages on the list. But for short messages, inlining is definitely good for email reading on phones.
Jeff
On Fri, Jun 28, 2013 at 9:46 AM, Danilo <[email protected]> wrote:
In the last topic I read it was asked more that once to zip the files and I did it.. By the way, this is the first error: Assertion failed in file helper_fns.c at line 361: ((((char *) sendbuf + sendtype_true_lb))) != NULL internal ABORT - process 2
Starting from the second execution I get:
===================================================================================== = BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES = EXIT CODE: 139 = CLEANING UP REMAINING PROCESSES = YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
=====================================================================================
HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): assert (!closed) failed HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): assert (!closed) failed HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event HYD_pmcd_pmip_control_cmd_cb (./pm/pmiserv/pmip_cb.c:928): assert (!closed) failed HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77): callback returned error status main (./pm/pmiserv/pmip.c:226): demux engine error waiting for event HYDT_bscu_wait_for_completion (./tools/bootstrap/utils/bscu_wait.c:70): one of the processes terminated badly; aborting HYDT_bsci_wait_for_completion (./tools/bootstrap/src/bsci_wait.c:23): launcher returned error waiting for completion HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:191): launcher returned error waiting for completion main (./ui/mpich/mpiexec.c:405): process manager error waiting for completion
Regards, Danilo
-----Original Message----- From: Wesley Bland <[email protected]> To: discuss <[email protected]> Sent: Fri, Jun 28, 2013 3:22 pm Subject: Re: [mpich-discuss] mpi assertion error
Can you just copy paste your error into the email? Most of us will probably not be all that excited about opening up strange tarballs attached to an email. Also, we get these emails on our phones and tablets where unzipping source code isn't as much of an option.
Wesley
On Jun 28, 2013, at 8:17 AM, Danilo <[email protected]> wrote:
Good afternoon,
I wrote a little application in C to compute 2D fft. This app was firstly executed on a cluster on which it was installed 2007 mpi version (don't remember the package name) and then adapted for a different cluster with mpi 1.4.1 (had to change the scatter/gather because in the previous version I could use the same buffer for both sendbuff and recvbuff). By the way, when executing with 2 processes it works fine. When trying with 4/8/16/32 and so on it gives firstly an assertion error as shown in the file attached, and starting from the second time you try to run it on more than 2 procs it gives error code 139. The error I'm talking about appears just when you run it with "realDim=16384" (it means that you have 16384 rows and 16384x2 columns since it is designed for real/imaginary numbers). I know the code is working since it was all ok on the previous cluster (even with 4-8-16-32 procs) and I can't find out which is the problem now.. Can you help?
As said attached you can find my application as well as the errors appearing and mpi info..
Regards, Danilo <error+app.tar.gz>_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss
-- Jeff Hammond [email protected] _______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss
-- Jeff Hammond [email protected] _______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss