Subject: Re: NetBSD with 4mb ram (or on MVII?)
To: None <NA4G@aol.com>
From: Lord Isildur <mrfusion@uranium.vaxpower.org>
List: port-vax
Date: 05/29/2002 11:34:40
this slowness of compiles is not netbsd's fault, but because of something
out of our control: gcc. gcc gets slower and more bloated with every
release. Yes, gcc's optimizer might generate code thats a bit faster than
pcc (weve had that thread last year), but overall, i find that gcc and
the gnu C library are a poor performer. Using the gnu c library, things
are much slower despite gcc's better optimizer on VAX. Try doing a million
printf()s from the old berkeley c library and from the gnu library,
compiled with their respective favored compilers, pcc for one and gcc for the
other. the gcc-compiled code using the gnu printf will be way slower than
the pcc compiled code using the bsd printf, for example. Given that gcc is
orders of magnitude slower, (it took an hour to compile kermit! yikes!) i
think we really need to do something about this. The last time this
subject came up, i recall some people disagreeing, but most people
agreeing, and of those who agreed most found pcc to be a decent
alternative, with the main detractors from that idea citing the less
optimal optimizer, and one or two people saying 'but pcc is only K&R',
and then some others saying that that is a good thing (i'd be one of
those.. :) At any rate, if we could beef up the optimizer with pcc, i
think at least for code that pcc can compile (i.e. good clean C with no
weird gnu-isms in it) (like the kernel, for example!) we would have a
major win. I dont know if pcc is in the BSD only portion of BSD or if its
covered by the USL license.
As for something fast on your uv2, yeah, tahoe is the best thing i think.
isildur
On Tue, 28 May 2002 NA4G@aol.com wrote:
> If someone has something fast on an MVII, kindly
> tell me where I can get my hands on it. 1.5 is
> slower than 1.4.3 in my machines, by some 20% or
> so. I am waiting to try the new 1.5.3, 1.4.4,
> and anything related to 1.6. The elf snap just
> would not go anywhere after the initial boot.
> I have two 4 drive MVII's waiting for something
> zippy. So far, in my machines, 1.4.3 seems to
> be about as good as it gets, or I am doing
> something very wrong with these 1 horsepower
> end tables......
>
> Gosh... 14 hours to compile kernel on a 4000/60.
> That makes my 9 hour MVII compiles zippy! But,
> why does Tahoe compile in less than 1 hour on
> the same MVII? (:+\\.....
>
> Anyone get the elf snap or something remotely
> approaching current up on an MVII, yet? Tell me
> how. Mine locks up in the boot process. I
> get the boot netbsd msg, and then it totally
> locks up the machine, requiring a hard reset.
>
> Yeah, I know, MVII's are not exactly production
> type boxes, but, when I get them for a buck or
> two each, they are just plain fun to play with!
>
> Bob
>