home *** CD-ROM | disk | FTP | other *** search
- Path: valinor.mythical.com!valinor!n5ial!jim
- From: jim@n5ial.mythical.com (Jim Graham)
- Newsgroups: comp.dcom.modems
- Subject: Re: Astounding high CPS rate!
- Message-ID: <1996Jan18.003712.2799@n5ial.mythical.com>
- Date: Thu, 18 Jan 1996 00:37:12 GMT
- References: <4cvsii$h0@horus.infinet.com> <4db4ra$56@hg.oro.net> <DLAG1v.v8@giskard.demon.co.uk>
- Organization: Okaloosa Island dep't of fishing/brewing/ham radio/computers
-
- In article <DLAG1v.v8@giskard.demon.co.uk> dale@giskard.demon.co.uk
- (Dale Shuttleworth) writes:
-
- >If you are having problems with "overruns" then
- >you may be able to solve it by changing the FIFO settings, but this
- >cannot give you better performance than a carefully tuned system using
- >8250 or 16450 UARTS.
-
- Well, maybe. I agree that, unless you're seeing buffer overrun errors
- (i.e., the CPU can't keep up), you aren't going to see improved
- throughput. However, if you're running a multitasking system (e.g., some
- UNIX variant), you *MIGHT* see an overall increase in system performance
- (how much, I'm not sure---that would, I suspect, vary from one system to
- the next) by increasing the number of characters that the 16550 waits for
- before issuing an interrupt. On the other hand, you might also start
- seeing FIFO overruns if you set this too high.... It's all a question of
- balancing the context switching overhead with the need for giving the FIFOs
- breathing room to allow the CPU time to get around to servicing the
- interrupt.
-
- I know, minor point, but it seemed worth mentioning at the time. :-)
-
- Later,
- --jim
-
- --
- 73 DE N5IAL (/4) MiSTie #49997 < Running Linux 1.2.13 >
- jim@n5ial.mythical.com ICBM or Hurricane: 30.23N 86.32W
- || j.graham@ieee.org Packet: --OFFLINE-- (Ft. Walton Beach, FL)
- E-mail me for information about KAMterm (host mode for Kantronics TNCs).
-
-