Date: Thu, 2 Jul 1998 13:37:22 -0400 From: sbabkin@dcn.att.com To: rminnich@Sarnoff.COM, hackers@FreeBSD.ORG Subject: RE: timeout granularity (was: Re: Console driver...) Message-ID: <C50B6FBA632FD111AF0F0000C0AD71EEFF8B46@dcn71.dcn.att.com>
next in thread | raw e-mail | index | archive | help
> -----Original Message----- > From: Ron G. Minnich [SMTP:rminnich@Sarnoff.COM] > > On Thu, 2 Jul 1998 sbabkin@dcn.att.com wrote: > > The measurements > > have shown that handling 115200 bps transfer caused 11520 > > interrupts per second and ate up about 20% CPU of 20 MHz 386 > > in the interrupt handler. The OS was SCO Unix 3.2.1. > > Interesting. One the 486/25, linux 1.0.xx, the 10k interrupts also > seemed to > eat about 20% of the machine. > That handler worked with plain ring buffers and had no complex processing, at most it called wakeup() to schedule the higher level routine when it reached the low water mark. It was not a generic serial port driver, it used special protocol to use a slow PC as fast terminal. So it worked at 115Kbps or 56Kbps (the XTs were not able to handle over 56Kbps, ATs worked well with 115Kbps) for output and not more that 100 bytes per second (this was part of the protocol) for input from keyboard. This asymmetric design was also the reason why the lost interrupts were not a problem. -SB To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?C50B6FBA632FD111AF0F0000C0AD71EEFF8B46>