leif wrote: > William Stein wrote: >> On Tue, Aug 30, 2016 at 9:56 AM, Thierry Dumont >> <tdum...@math.univ-lyon1.fr> wrote: >>> I have two computers, and sage installed on both: >>> >>> 1) Ubuntu 12.04 , sage-7.3 on an nfs mount,y >>> >>> 2) Ubuntu 16.04, sage-7.3 on a local system, on a ssd. >>> >>> With the first one, sage starts slowly (as could be expected!), and I >>> have time to look at sage starting with "top". And during the starting >>> phase, before the prompt, sage uses more than 16 GB of virtual memory >>> (VIRT: 16.380g); the same quantity is used when stopping sage (in the >>> console interface). >>> It seems quite large, no ? >>> >>> With the second machine, it's more difficult to see what happens with >>> top, but it does not seems to use more than 7gb. >>> >>> Why such a large amount of memory ? (ok, it's virtual, but it's large, >>> no ?). > > IMHO horrible (as *each* Sage subprocess is claiming that amount of > memory, here usually ~28 to 50+ GB IIRC, as it depends on the physical > memory installed) and dangerous, as this makes the usage of 'ulimit -v' > nearly impossible and in practice requires to allow vm-overcommit. With > the latter, it's presumably quite easy to effectively crash many (if not > most) machines, be it by just some bug in Sage, but also by a malicious > user of course. (I.e., at least the VM a Sage server / installation may > run in.) > > It also makes it more complicated to see how much memory a Sage process > actually *uses*. > > >> I think this is a hack that is used by PARI/GP, due to their >> "interesting" primitive (but fast!) memory design. > > And GAP / libgap. In this recent case I guess the latter is (once > again) to blame.
Nope, GAP / libgap by default "only" contribute about 10% to the (at least!) 48 GB claimed by *every* Sage process here, while indeed the PARI library in Sage takes /one fourth/ of sage.misc.memory_info.MemoryInfo().virtual_memory_limit() which is *total* (not free!) physical RAM + total (not free) swap space, in bytes. That's IMHO a regression, but dates back quite a couple of (beta) releases AFAIK. (The last *commit* regarding that is from January this year, don't know when it got merged.) -leif P.S.: The change regarding the PARI library was made in #19883; according to the ticket's /milestone/ (which doesn't say much) it presumably got merged into Sage 7.1. -- 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.
