On 1 April 2011 03:31, Justin C. Walker <jus...@mac.com> wrote: > Hi, all, > > I've just upgraded my home system, to a Mac Pro (Dual 6-core Xeon, 24GB > memory), and figured (hoped) I'd see 24 processes hammering away during the > build & test of a Sage release. > > Turns out not to happen. For the record, these new cores from Intel sport 2 > "threads" each, which act almost like separate cores, capable on this system > of running 24 processes at a time. However, both for the build itself, and > the test, there are generally at most 12 runnable processes at any time. > > Anyone have any idea why the parallelism is limited this way? > > FWIW, I started with "-j24", and when I saw only half the threads active, > talked with Apple about it. Their suggestion (based on their experience with > kernel builds) is to request double the number of available threads. I reran > the build/test with "-j48", and saw no improvement: the number of runnable > threads remained at ~12. > > Thoughts? > > Justin
If you look in the Makefile, there's a comment that the number should be limited to 8, so I'm surprised you see 12. The reason this was introduced was the assumption that any machine with 8+ cores is a multi-user machine, not owned by one individual, so he/she should not be able to accidentally cause havoc by having too many threads. So they need to specifically edit the Makefile. On t2.math for example, there are 128 threads. If one was to compile Sage with 128or 256, it would run out of memory. Hence Sage has a bit of a safety factor built in. Dave -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
