Compute Error Exit Code 1073741819


Advanced search

Message boards : Number crunching : Compute Error Exit Code 1073741819

AuthorMessage
Profile ritterm
Avatar
Send message
Joined: Nov 23 11
Posts: 5
Credit: 173,293
RAC: 0
Message 213 - Posted 2 Feb 2013 5:17:22 UTC

    Last modified: 2 Feb 2013 5:18:22 UTC

    I'm getting compute errors on this host (AMD FX8150, Win7-64, 16GB RAM). Recently, 8 of 12 results are showing an error similar to this.

    The stderr output includes:

    - Unhandled Exception Record -
    Reason: Access Violation (0xc0000005) at address 0x00411022 read attempt to address 0x1E6BFB7F


    Is this a memory error? This host isn't having problems on any of the several other projects that it's running, and three other hosts aren't having any problems on SAT.
    ____________

    Dagorath
    Avatar
    Send message
    Joined: Oct 15 11
    Posts: 11
    Credit: 63,227
    RAC: 0
    Message 214 - Posted 7 Feb 2013 0:54:08 UTC - in response to Message 213.

      It's a memory related error but not an error that indicates faulty or failing RAM. Every program that runs on your computer has an address space assigned to it. No program should ever need to read or write memory addresses outside of its assigned space. If a program attempts to read/write outside its space then the OS will usually shut it down. That's what happened in your case... the SAT app attempted to read a memory location (address) that belonged to another program so the OS killed it.

      I used to get lots of those with the SAT app, so many I quit running SAT for a while. Recently I tried it again and am getting far fewer such errors, almost zero.

      Why do such errors happen? Lots of reasons and none of them are easy to track down. Typically the program computes the addresses it reads/writes and sometimes it computes an address incorrectly and tries to read from memory that is outside it's assigned space. With read type errors it's often that the program thinks it owns say 200 bytes beginning at address X but it actually owns only 150 bytes so when it tries to read the 151st byte the OS says, "Nope!" and you get an error.

      Profile Oleg Zaikin [SAT@home]
      Forum moderator
      Project administrator
      Project developer
      Project scientist
      Send message
      Joined: Sep 15 11
      Posts: 133
      Credit: 4,826,453
      RAC: 0
      Message 216 - Posted 7 Feb 2013 15:11:56 UTC - in response to Message 213.

        I'm getting compute errors on this host (AMD FX8150, Win7-64, 16GB RAM). Recently, 8 of 12 results are showing an error similar to this.

        The stderr output includes:

        - Unhandled Exception Record -
        Reason: Access Violation (0xc0000005) at address 0x00411022 read attempt to address 0x1E6BFB7F


        Is this a memory error? This host isn't having problems on any of the several other projects that it's running, and three other hosts aren't having any problems on SAT.


        I see, we will try to fix it.

        Profile Oleg Zaikin [SAT@home]
        Forum moderator
        Project administrator
        Project developer
        Project scientist
        Send message
        Joined: Sep 15 11
        Posts: 133
        Credit: 4,826,453
        RAC: 0
        Message 217 - Posted 7 Feb 2013 15:12:36 UTC - in response to Message 214.

          It's a memory related error but not an error that indicates faulty or failing RAM. Every program that runs on your computer has an address space assigned to it. No program should ever need to read or write memory addresses outside of its assigned space. If a program attempts to read/write outside its space then the OS will usually shut it down. That's what happened in your case... the SAT app attempted to read a memory location (address) that belonged to another program so the OS killed it.

          I used to get lots of those with the SAT app, so many I quit running SAT for a while. Recently I tried it again and am getting far fewer such errors, almost zero.

          Why do such errors happen? Lots of reasons and none of them are easy to track down. Typically the program computes the addresses it reads/writes and sometimes it computes an address incorrectly and tries to read from memory that is outside it's assigned space. With read type errors it's often that the program thinks it owns say 200 bytes beginning at address X but it actually owns only 150 bytes so when it tries to read the 151st byte the OS says, "Nope!" and you get an error.


          Thanks for the advices!

          Profile ritterm
          Avatar
          Send message
          Joined: Nov 23 11
          Posts: 5
          Credit: 173,293
          RAC: 0
          Message 218 - Posted 7 Feb 2013 15:28:05 UTC - in response to Message 214.

            Dagorath wrote:

            It's a memory related error but not an error that indicates faulty or failing RAM...

            Thanks for the tutorial...Excellent info and explanation. :-)

            Oleg wrote:
            I see, we will try to fix it.

            Thanks for checking this and doing what you can.

            Regards,

            MarkR
            ____________

            Profile ritterm
            Avatar
            Send message
            Joined: Nov 23 11
            Posts: 5
            Credit: 173,293
            RAC: 0
            Message 222 - Posted 8 Feb 2013 14:56:47 UTC

              In case it helps... 4 of 5 recent tasks on this one host ended in compute errors. The other is pending validation. I'm still not having any issues on my other hosts.
              ____________

              Dagorath
              Avatar
              Send message
              Joined: Oct 15 11
              Posts: 11
              Credit: 63,227
              RAC: 0
              Message 223 - Posted 9 Feb 2013 10:03:22 UTC - in response to Message 222.

                Thanks for the kind words in your previous post. When crunchers post their problems respectfully and by respectfully I mean without all that "I've lost X credits because of this" and "I'm gonna quit" and "you're gonna lose all your crunchers" nonsense I try hard to help.

                If this is happening on just one host then it suggests a memory problem. On the other hand, the fact that it happens only with this project suggests a problem with this project's app. When I am faced with indicators that point in different directions it's very difficult to deduce the cause via logic. I find the best way to start diagnosing is via process of elimination. I hate shutting everything down to run memtest for 24 hours but sometimes it's necessary. Check temperatures and fan speeds too, check disks via SMART reports (it could be virtual memory) or whatever other disk diagnostics you can find.

                Yesterday I said it's a problem with the app addressing memory it does not own but I think sometimes hardware issues can make it look like that is what's happening and possibly trick the OS into thinking it's happening. I don't know enough to explain from first principle how that can happen but experience suggests strongly to me that it does.

                Why does this alleged memory problem affect only the SAT application? Well there are 2 plausible theories for that. The first is that it is affecting other apps but you just don't realize it because it does something innocuous such as make a pixel on the screen blue instead of red or it fails to detect a keystroke and you just pass that off as you failing to hit the key hard enough. The other explanation is that the SAT app tickles things just the right way whereas other apps do not. That's not likely but not likely isn't the same as impossible. So again, process of elimination helps narrow it down somewhat

                Finally, I want to add a little to what I said yesterday about applications trying to read/write memory it does not own. What I've said so far on that topic makes it look like its the programmer's fault and technically it is but one needs to appreciate how and why programmer's make such errors. First know that when writing a program in C, C++ or any language that has the ability to access memory directly, one can play it the safe way and simply let the compiler and OS handle storage addresses. Unfortunately that approach has overhead which means CPU cycles must be spent on assigning storage and keeping all the details straight. That slows the program down somewhat. You can make the program run a little faster (sometimes a lot faster) by managing memory "manually". The problem with that approach is that it can be very difficult to do and sometimes unexpected or unpredictable conditions arise that break your otherwise perfect memory management scheme. It can be an absolute nightmare to determine what that unexpected or unpredicted condition is and sometimes it's even more fun finding a way to deal with it.

                Profile Oleg Zaikin [SAT@home]
                Forum moderator
                Project administrator
                Project developer
                Project scientist
                Send message
                Joined: Sep 15 11
                Posts: 133
                Credit: 4,826,453
                RAC: 0
                Message 231 - Posted 26 Feb 2013 7:56:02 UTC - in response to Message 223.

                  I am working on new version of client application. Memory using will be changed, may be it will help.

                  Profile ritterm
                  Avatar
                  Send message
                  Joined: Nov 23 11
                  Posts: 5
                  Credit: 173,293
                  RAC: 0
                  Message 248 - Posted 21 May 2013 19:05:43 UTC

                    After some time off, I'm back here as I dedicate some time to the math projects. Unfortunately, I'm still experiencing these errors on my AMD 8150/Win7-64 machine... :-(

                    Not a big deal, as I'm happy to take it out of the mix and crunch with the hosts that aren't having problems. I just wanted to pass this on as an FYI in case you are interested in looking at it again.

                    Cheers,

                    MarkR
                    ____________

                    Profile Oleg Zaikin [SAT@home]
                    Forum moderator
                    Project administrator
                    Project developer
                    Project scientist
                    Send message
                    Joined: Sep 15 11
                    Posts: 133
                    Credit: 4,826,453
                    RAC: 0
                    Message 251 - Posted 24 Jun 2013 16:01:41 UTC - in response to Message 248.

                      New application released. Please try it on your OS.

                      Post to thread

                      Message boards : Number crunching : Compute Error Exit Code 1073741819


                      Home | My Account | Message Boards


                      Copyright © 2019 Institute for System Dynamics and Control Theory of SB RAS and Institute for Information Transmission Problems of RAS