Date: Sun, 12 Oct 2003 20:57:41 +0200 From: Bernd Walter <ticso@cicely12.cicely.de> To: "M. Warner Losh" <imp@bsdimp.com> Cc: nate@root.org Subject: Re: Interrupt statistics? Message-ID: <20031012185740.GS13791@cicely12.cicely.de> In-Reply-To: <20031012.095503.129593225.imp@bsdimp.com> References: <20031011234314.P23991@root.org> <20031012.095503.129593225.imp@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Oct 12, 2003 at 09:55:03AM -0600, M. Warner Losh wrote: > In message: <20031011234314.P23991@root.org> > Nate Lawson <nate@root.org> writes: > : Does anyone have recent statistics for interrupt latency and arrival > : timings? > > In -stable I know that we service fast interrupts in < 10us on a > 666MHz machine 99.2% of the time. I don't know about non-fast > interrupts, but other results suggest that we'd do 99+% in less than > 100us. This is to the first instruction of the ISR. The ISRs that > I've been running typically run for 2us (since they do 30 instructions > plus 2 ISA I/Os). >From the serial application I did recently I know that writing 8 bytes at 19200bps could pause the output for > 286us and < 429us after 5 bytes in about 50% tries with 16C552 ISA hardware. Given the fact we get the interrupt before the last stop bit goes out the real latency in handling is a multiple of this. However - this was with an old current and I can retry with a recent one once I get over the "no kernel output" problem. If things should be acurate I can modify ISA hardware with a micro- controller to do real measurements on a socket 7 HX board. -- B.Walter BWCT http://www.bwct.de ticso@bwct.de info@bwct.de
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031012185740.GS13791>