On 9/19/07, mabshoff <[EMAIL PROTECTED]> wrote: > > John's code doesn't leak memory, Sage doesn't outright leak any memory > (well, more than usual, the is still some "possibly lost" and "still > reachable". After about an hour on sage.math we are approaching about > 1GB of memory consumption. I am not sure how much longer the > computation will take, but my take on the situation:
It might be interesting to find out where Sage runs out of memory during the computation. More generally, there are probably numerous situations where Sage does *not* fail gracefully enough when it runs out of memory. Do you have any thoughts about good ways of automatically testing for such things, besides say running various things (e.g., make big matrices, do linear algebra, etc.) in a virtual machine without very little RAM? > a) the job might not fit in the 32bit memory space (depending on the > kernel userspace gets between 2 and 3 GB address space, meccah has > probably 2), but it crashes too early on meccah, so (c) might be a > better explanation. Yep. > b) the amount of memory allocated per allocation might cause failures > if the memory is very fragmented. On the meccah that might have > happened, but I haven't looked at the code to see if that really is > the problem. > c) meccah itself has only about 1 GB free memory (physical+swap), but > since swap+physical is 5 GB totoal userspace might have run out of > memory. There are two very long running gp processes on meccah itself, > so it might be a good idea to take the computation to one of the less > busy nodes, push the gp jobs back into ram and let them finish much > quicker that way. > > Overall we ought to make sure that all malloc's and friends check > their return values to throw an appropriate error message about the > out of memory condition. > > John, maybe you should ask William about an account on sage.math :) John Voight *does* have an account on sage.math. John, you should use that instead -- it's a 64-bit machine with 64GB physical RAM and 36GB of swap space. It has 16 processor cores, and e.g., at this moment only 3 are being used for serious work -- one being using by Michael for the computation discussed above. William --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---
