> I did some testing, at least for me, I get the most improvements when the > number > of cores or processors equals the -j number. You can make it higher, even > double it, withoout hurting things, but 95% of the improvements come from > matching the number of processes to the number of available CPUs (and that' by > my own testing, not theory).
I've found that on my Q6600 (quad core) system, the optimal is 8 though the improvements after 5 were minimal. It depends if the jobs are I/O bound or not and the scheduler. Although even with ffmpeg, the optimal number of threads with this quad core system is 8. On my previous system with a dual-core chip with the same hardware, the "magic number" was 4 (again, 2x the number of cores). This was with the ULE scheduler, I'm not sure if the same holds true for the 4BSD scheduler or not. And as you said, it's important to use make without -j if the build fails before reporting bugs, since there are no guarantees that world will build properly with multiple jobs. Josh _______________________________________________ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
