Subject: Re: silo overflows, was fifo overruns
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Jukka Marin <email@example.com>
Date: 11/13/1997 13:26:21
On Thu, Nov 13, 1997 at 02:40:00AM -0800, Jonathan Stone wrote:
> >I know that all too well. Someone just suggested that the interrupt latency
> >problem affects all ports (which I think is true) and many platforms _do_
> >have parallel ports..
> No, this problem doesn't affect all ports. It doesn't affect 68k
> macs, which use the `MI' [sic] 8530 driver. The motherboard interrupt
> structure on most 68k macs is fairly hostile to good performance. so
> this is just _not_ an MI problem.
Well, i386 has been suffering from serial port problems. I don't know
if the new drivers solves the problems on slower machines (like 386/486).
I guess a fast pentium with 16-byte serial FIFOs forgives relatively
high interrupt latency.
Under 1.2, I am having occasional problems with both the com driver and
the Cyclades serial port driver at 38400 bps (the Cyclades card has 12-
byte FIFOs and I'm seeing these problems on 120 and 166 MHz crashiums..
I mean, pentiums).