Date: Mon, 02 Jul 2007 19:35:11 -0500 From: Stephen Montgomery-Smith <stephen@math.missouri.edu> To: freebsd-current@freebsd.org Subject: great job with multithreaded programs Message-ID: <468999BF.9040006@math.missouri.edu>
next in thread | raw e-mail | index | archive | help
I did a recent upgrade of current, and a multithreaded program which used to take 33s suddenly takes 18s. Somebody did something very right!!! I can provide details (but please email me directly as I don't subscribe to current), but basically the program does these massively large 3D Fourier transforms using the math/fftw3 port. I have two dual Xeon's with HT running, and I run fftw3 with 4 threads. I use the ULE scheduler, and have witness and invariants all switched off. I should add, perhaps disappointingly, that this now only brings FreeBSD on par with Linux with respect to this program - Red Hat Linux also ran this at 18s. But hopefully this is part of some rather skillful reworking of multithreading on FreeBSD and we will see great gains in the future. Best regards, Stephen
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?468999BF.9040006>