Date: Sun, 22 Mar 1998 15:56:35 +0800 From: Peter Wemm <peter@netplex.com.au> To: "John S. Dyson" <toor@dyson.iquest.net> Cc: sos@FreeBSD.ORG, smp@FreeBSD.ORG Subject: Re: is it just me, or... Message-ID: <199803220756.PAA10655@spinner.netplex.com.au> In-Reply-To: Your message of "Sat, 21 Mar 1998 19:42:02 EST." <199803220042.TAA00321@dyson.iquest.net>
next in thread | previous in thread | raw e-mail | index | archive | help
"John S. Dyson" wrote:
> > In reply to Peter Wemm who wrote:
> > > It is just me or has the smp kernel taken a nose dive over the last month
> > > or so?
> >
> > I havn't had any problems until the latest breakage John provided :)
> >
> > I run SMP on a TYAN s1662d with 128M and two Maxtor EIDE disks, no SCSI,
> > so thats at least one difference.
> > The system behaves very nicely, even under very heavy load, interactive
> > response (here under X) is still exelent, no long delays or anything...
> >
> I do most of my testing, verification and day-to-day usage with SMP systems.
> It is only an after-thought that I check out UP systems. If there is anythin
g
> that you can point to where performance has decreased, let me know.
Sorry if this sounded like an accusation, it was meant as a fact finding
excercise to see if just I had a problem or whether everybody else was too
and had gone to UP or something.
Anyway, I got a couple of replies saying that they were having no trouble
at all. One even fired up a rc564/rc5des key searcher (since that was one
of the easiest ways I could provoke problems), but nothing unexpected
happened.
So, it must be something I'm doing. Is *anybody* running parallel rc564 or
rc5des under SMP? (the rc5 cores are (apparently) not reentrant, so you
need two or more copies running at once to get max throughput, even in the
multithread version)
The only other things I could think of:
- the machine having trouble is a pure-ELF system, even though the rc5
key search problems are from a.out executables.
- I had VM86 enabled (being compiled out now)
- I had USER_LDT enabled (going too)
- I had a sound driver enabled (going too).
- I have AHC_TAGENABLE, AHC_SCBPAGING_ENABLE and AHC_ALLOW_MEMIO (works
fine under UP).
I'm going to take out the pci vga card. it's currently being allocated the
same IRQ and APIC pin as the onboard ahc0. It shouldn't be a problem
(the screen and keyboard are not connected anyway), but would eliminate
the possibility of shared interrupt problems.
I guess the other thing to do is to recompile the system for a.out again.
I suspect that there are problems with ELF executable startup..
Cheers,
-Peter
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-smp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199803220756.PAA10655>