JavaSockets is still in development, and won't be officially supported until the upcoming release. Perhaps you can get the default communication library to stop giving you a 104 error, but I suspect that putting 128 places on a 32-core machine is causing too much latency (your threads are busy performing other work, and are not interacting with the network frequently enough), leading to the error. Adding explicit calls to Runtime.probe() in your program may help, but it may not. The correct way to run X10 on a 32 core machine is with 32 total X10 worker threads, so I'm not sure there's much more help we can provide if you are having problems when going well beyond that.
- Ben From: SUYASH GUPTA <dynamic.suy...@gmail.com> To: Benjamin Herta/Watson/IBM@IBMUS, Cc: x10-users@lists.sourceforge.net Date: 08/02/2013 09:07 AM Subject: Re: [X10-users] Regarding ERRNO=104 I am running version x10-2.3.1 and still getting the issue On Fri, Aug 2, 2013 at 6:23 PM, Benjamin Herta <bhe...@us.ibm.com> wrote: What version of X10 are you running? JavaSockets is available only in the most recent release (v2.3). - Ben Inactive hide details for SUYASH GUPTA ---08/02/2013 01:41:30 AM---I tried running again and found actually that this issue wasSUYASH GUPTA ---08/02/2013 01:41:30 AM---I tried running again and found actually that this issue was coming: "Unable to load JavaSockets. Fo From: SUYASH GUPTA <dynamic.suy...@gmail.com> To: Benjamin Herta/Watson/IBM@IBMUS, Cc: x10-users@lists.sourceforge.net Date: 08/02/2013 01:41 AM Subject: Re: [X10-users] Regarding ERRNO=104 I tried running again and found actually that this issue was coming: "Unable to load JavaSockets. Forcing single place execution" Is there any solution for this issue. Thanx On Thu, Aug 1, 2013 at 12:37 PM, SUYASH GUPTA <dynamic.suy...@gmail.com> wrote: Hello again Well I tried using the option, but something strange is occurring.. When I am running on the cluster --- Cluster configuration : Each node of the cluster has two Intel E5-2670 2.6GHz processors, each processor has eight cores and each core can make use of (up to) two hardware threads --- instead of one output I am getting several outputs. I somehow believe that instead of a single copy of the code, multiple copies of the program are getting executed. Such a scenario was never encountered by me earlier. Thanx On Wed, Jul 31, 2013 at 11:38 PM, Benjamin Herta <bhe...@us.ibm.com > wrote: You are now using a different library for communications. How much this affects the timing of your program depends on how your program uses the network, and whether it's on the critical path. This is a new communication library, so if you end up re-running benchmarks and find it makes a difference in performance, I would be interested to know if those differences are positive or negative. - Ben Inactive hide details for SUYASH GUPTA ---07/31/2013 01:52:35 PM---Hello again Well this method works... There is no more ERRNoSUYASH GUPTA ---07/31/2013 01:52:35 PM---Hello again Well this method works... There is no more ERRNo=104 problem. From: SUYASH GUPTA <dynamic.suy...@gmail.com> To: Benjamin Herta/Watson/IBM@IBMUS, Cc: x10-users@lists.sourceforge.net Date: 07/31/2013 01:52 PM Subject: Re: [X10-users] Regarding ERRNO=104 Hello again Well this method works... There is no more ERRNo=104 problem. But I wish to know how much affect to the timing of the program shall occur now. I had previously collected few results without using this option and my further results using this option could differ upto what extent. Thanx On Wed, Jul 31, 2013 at 11:03 PM, Benjamin Herta < bhe...@us.ibm.com> wrote: You won't see an OutOfMemoryError if the memory problem is reached in the native portion of the code, not the java portion. Try adding the argument "-DX10RT_IMPL=JavaSockets" after the x10 command, and let me know if it makes any difference. - Ben Inactive hide details for SUYASH GUPTA ---07/31/2013 01:16:51 PM---Hello again Well my code works well with 32 and 64 instancesSUYASH GUPTA ---07/31/2013 01:16:51 PM---Hello again Well my code works well with 32 and 64 instances and I am running on a From: SUYASH GUPTA <dynamic.suy...@gmail.com> To: Benjamin Herta/Watson/IBM@IBMUS, x10-users@lists.sourceforge.net, Date: 07/31/2013 01:16 PM Subject: Re: [X10-users] Regarding ERRNO=104 Hello again Well my code works well with 32 and 64 instances and I am running on a cluster where the nodes have upto 64Gb memory for utilization. Also I am not receiving an Out of memory error (which I generally received when some of my programs used to run out of memory). This code of mine does not require large memory. Also I am working with Managed X10. if some of my programs used to run short of memory On Wed, Jul 31, 2013 at 7:38 PM, Benjamin Herta < bhe...@us.ibm.com> wrote: The ERRNO 104 suggests that one or more of your processes is exiting unexpectedly. The exit might not be due to the number of threads or network links, but instead due to something else, such as running out of memory. Have you checked that you have enough memory to run 128 instances of your program on your machine? Does it run ok with 32, or 64 instances, and if it does, how much memory is free? Are you running native or managed X10? - Ben Inactive hide details for SUYASH GUPTA ---07/31/2013 12:47:52 AM---Hello All I am trying to execute an X10 program by setting XSUYASH GUPTA ---07/31/2013 12:47:52 AM---Hello All I am trying to execute an X10 program by setting X10_NPLACES=128 while the From: SUYASH GUPTA <dynamic.suy...@gmail.com> To: x10-users@lists.sourceforge.net, Date: 07/31/2013 12:47 AM Subject: [X10-users] Regarding ERRNO=104 Hello All I am trying to execute an X10 program by setting X10_NPLACES=128 while the number of hardware threads I have are 32. I know best performance can be achieved only when number of hardware threads and X10_NPLACES are set equal but just for some analysis I need such an arrangement. My program is getting crashed by ERRNO=104 problem. Is there some way to avoid it or some patch file to use. Thanx -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras ------------------------------------------------------------------------------ Get your SQL database under version control now! Version control is standard for application code, but databases havent caught up. So what steps can you take to put your SQL databases under version control? Why should you start doing it? Read more to find out. http://pubads.g.doubleclick.net/gampad/clk?id=49501711&iu=/4140/ostg.clktrk _______________________________________________ X10-users mailing list X10-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/x10-users -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras -- Regards Suyash Gupta MS Scholar Pace Lab, Department of Computer Science & Engineering IIT Madras
<<inline: graycol.gif>>
------------------------------------------------------------------------------ Get your SQL database under version control now! Version control is standard for application code, but databases havent caught up. So what steps can you take to put your SQL databases under version control? Why should you start doing it? Read more to find out. http://pubads.g.doubleclick.net/gampad/clk?id=49501711&iu=/4140/ostg.clktrk
_______________________________________________ X10-users mailing list X10-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/x10-users