Might be of interest to people with 'memory rich' systems...

	On a NetBSD/sparc IPX with 40MB of ram (I know, luxury :)
	building a kernel was speeded up by 10% by increasing the
	size of the buffer cache, specifically adding:

		options         BUFPAGES=512
		options         NBUF=512

	to the config file.
	Given the CISC nature of the m68k, you might find some benefits
	from increasing the buffer cache even with 20MB. 
	Sigh.. I _really_ have to dig out my si & get NetBSD onto it :)

		David/absolute

Definition of the day:
'Disinterested': Unbiased, with nothing to gain... _not_ 'uninterested'.
(It would be preferable if more of the world could grasp the above)

On Sun, 1 Feb 1998, Michael R Zucca wrote:

> Date: Sun, 1 Feb 1998 15:40:22 -0500 (EST)
> From: Michael R Zucca <mrz5149@cs.rit.edu>
> To: Mason Loring Bliss <mason@acheron.middleboro.ma.us>
> Cc: port-mac68k@NetBSD.ORG
> Subject: Re: Daystar Accelerator Works!
> 
> > > This is very exciting! I can actually build kernels within the span of a
> > > human lifetime now :) (The 16 megs of RAM also helped :-)
> > 
> > How long's it take? I recently put together a cheap 40mHz 486 system, and
> > it still takes me almost five hours to build a small kernel. (It only has
> > eight megs of RAM, though - that probably DOESN'T help!)
> 
> Time reported that a full compile from scratch (make depend, make) took
> about an hour and 40 minutes. That's with 20 megs of RAM.
> 
> The disk thrashing is what's killing you. My IIvx with just it's 030 could
> build a kernel in 5 hours with 20 megs of RAM (If I remember right). With
> 8 megs is was more of a 12 hour project :-)
> 
> However, while I'm developing kernel stuff I don't usually have to do a
> full recompile. I just need to compile a couple of updated files. That's
> where the real speed benefit comes in.
> 
>