Date: Tue, 26 Jan 1999 22:13:54 -0700 From: Nate Williams <nate@mt.sri.com> To: SHUDO Kazuyuki <shudoh@muraoka.info.waseda.ac.jp> Cc: Tom <tom@sdf.com>, java@FreeBSD.ORG Subject: Re: TowerJ for FreeBSD Message-ID: <199901270513.WAA18478@mt.sri.com> In-Reply-To: <199901270132.KAA26867@cafe.muraoka.info.waseda.ac.jp> References: <Pine.BSF.4.05.9901260851280.7314-100000@misery.sdf.com> <199901270132.KAA26867@cafe.muraoka.info.waseda.ac.jp>
next in thread | previous in thread | raw e-mail | index | archive | help
> > > > Yep, this is where FreeBSD (and less so Linux) got the crap kicked out > > > > of it. Linux is a bit better than ours for a couple of reaasons. > > > > > > 2) Our 'Green Threads' implementation uses locking that is not as > > > > effecient as the Linux version. In particular, Linux kernel's have > > > > poll built in, and we emulate it (badly, but effectively) with some > > > > hacks to use select. > > As above, Nate said that Linux kernels have poll built > in so why Linux 'Green Threads' implementation is a bit > better than FreeBSD one. This is incorrect. because, > > > > JDK 1.1.7 for Linux/glibc use poll() and not > > > select(). But glibc 2.0.7 which is used on most linux > > > box replaces calls to poll with calls to select since > > > Linux kernel prior to 2.1.23 doesn't have poll(). > > JDK on most Linux boxes don't make use of poll(). A > system call poll() isn't used by JDK with glibc 2.0.X > even if the Linux kernel has it. Poll() is being > emulated also on Linux boxes. I suspect the boxes that use select aren't the ones being used in the benchmark tests. According to what I've heard, they got much faster performance when they did some poll tests (vs. select). And, as I originally stated, that is just *one* reason it's faster. Nate To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199901270513.WAA18478>