From: Derek.Glidden@f42.n377.z1.fidonet.org (Derek Glidden) Subject: Serial Performance under Linux Date: 14 Jul 1992 09:27:36 GMT
Just to stick my own $0.02 into the hat:
Just the other night, while using Linux 0.96c (NO PATCHES) and Xv1.1,
while IN an xterm, and while doing two other xterms at the same time (As
well as having xclock and xeyes active) I d/l at 14.4K/v.32bis under
xcomm with not a single error or retry. I'd say I was abusing the
system
resources pretty heavily, even on a 486/33 with 8MB RAM.
Of course, I do have a '550, but I believe I'd seen other references to
this not being a wholly entirely big deal with the "problems" with the
serial stuff.
I've noticed some things about the behaviour of Linux and hard drive
interrupts also... Some hard drives (Specifically Kalok, in my recent
experience) utilize "non-standard" IDE style interfaces to achieve
greater performances. I couldn't tell ya exactly how they're
non-standard, other than most DOS-based (gasp) utility type programs
refuse to operate at a "really" low-level, and most diagnostics and
benchmarks are unable to read seek-time performance because the
controller ignores pure seek commands.
What this is getting to is that, with these same hard drives, running
Linux will come up with scads and scads of "Unexpected HD Interrupt"
messages.
I'm wondering if this may not have something to do with some of the
current talk on interrupt disabling and whatnot. If you're having this
sort of problem, maybe you could post what model/make of HD you've got,
and whether it's IDE, ESDI, etc... Perhaps something could be
determined
that the ones that're having lost interrupts are all using some of the
same IDE style drives that don't entirely conform to the normal
interface
specifications.
Of course, this may have already been all worked out and figured on and
everything else and I'm just blowing wind...
* Origin: The Slut Club -- 813-689-1150 (1:377/42.0)