On Tuesday, August 30, 2016 at 1:23:47 PM UTC-7, Volker Braun wrote: > > > ulimit -v is useless anyways >
In my experience it is quite useful on multi-user machines to limit the effects of runaway computer algebra processes. Many packages (maple, magma at least) will segfault earlier with a ulimit -v set than without. With a properly set limit, this will often prevent the machine from thrashing the swap (which tends to affect all users). My impression was that sage benefitted from similar protection. > > >> and in practice requires to allow vm-overcommit. > > > In the default mode (overcommit_memory==0) calls to mmap with > MAP_NORESERVE are not checked, so you don't have to do anything special. > Really, restricting virtual memory on a 64-bit system is only for > specialist applications. > Isn't that exactly what makes the above use of ulimit -v useful? Presumably, once pages DO get allocated, the operating system will check that the total virtual address space that is actually allocated (to swap or memory) fits within the "ulimit -v" bound. In that case we don't particularly need to care about the huge mappings that sage seems to request either. In any case, I hope we can keep sage's memory management configured in such a way that one can set standard operating system limits in order to ensure that overeager memory-consuming processes will crash before they start affecting other processes too much. -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at https://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/d/optout.
