Date: Mon, 24 Dec 2001 23:47:20 -0700 (MST) From: "M. Warner Losh" <imp@village.org> To: peter.jeremy@alcatel.com.au Cc: phk@critter.freebsd.dk, dg@root.com, bde@zeta.org.au, dillon@apollo.backplane.com, silby@silby.com, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/dev/sio sio.c Message-ID: <20011224.234720.15269884.imp@village.org> In-Reply-To: <20011224162247.B85044@gsmx07.alcatel.com.au> References: <20011222193231.D24034@nexus.root.com> <61654.1009097922@critter.freebsd.dk> <20011224162247.B85044@gsmx07.alcatel.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
: >My guess would be that only four people in the project who have : >ever measured our interrupt latency: Bruce, Louie, Warner and me. : : That's definitely an under-estimate. Andrew Gallatin posted his : latency measurements on a Alpha h/w. I think I've seen one or two : other people indicate that they'd made measurements on -hackers. I've : also done some limited latency testing on both i386 and Alpha's. I'm only able to measure the interrupt latency of ISA devices, and even then I'm not sure my results are good yet (which is why I haven't published them). I hope to present the findings at usenix. Bruce has a cool way to measure these things in software. : >If our interrupt latency is so bad that we cannot run a serial port : >as well on a PIII/Athlon 1GHz cpu as we could on a 486/66 a few : >years back, then the solution is not to make the serial driver more : >defensive, the solution is to fix the interrupt latency. : : Unfortunately, I can't see this happening. : 1) The migration of the PC-card interface to Cardbus means that PC- : card devices no longer have the option of a fast interrupt. This : will bite everyone who uses a laptop without a real builtin modem. Must bite slow laptop users. :-). My 900MHz Dell is using a 56k link as I type this w/o a single sio overflow. Also, the interrupt latency tolerence for PCI based devices is a few microseconds longer than for ISA devices because reading the status registers takes less time on the PCI bus. However, I have had complaints from a few people that have see silo overflows on pccard (both with fast interrupts, and slow ones now basically forced by 4.4). I don't know if the latencies in normal interrupts will all be solved by the HIGH -> MEDH change (since it takes us from 2 character times (32us) to 8 character times (128us). I've seen interrupt latency occasionally be as high as 20-100ms (depending on the machine we ran it on) for the parallel port (since I'm measuring pps from an atomic clock with it, we can tell these huge outliers very easily). Warner To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011224.234720.15269884.imp>