Subject: Re: kernel compile times
From: "Ethan Dicks" <ethan.dicks at gmail.com>
Date: Wed, 24 Oct 2007 15:19:15 +1300
To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at
classiccmp.org>
On 10/24/07, Sridhar Ayengar <ploopster at gmail.com> wrote:
> My
compile on the VAXstation 2000 was at least a few days. Maybe a week.
...it was swapping like crazy during the compile, and
swapping to the same disk where the sources were stored. That machine
is S*L*O*W. It would have probably gone a whole lot faster if it had
enough RAM.
Oh, yeah. The CPU is nominally as fast as a MicroVAX II (and over 50%
faster than an 11/750), but if you were swapping, that'd be a *huge*
difference in compile times. I was doing this on an 8MB 11/750 with
nothing else going on (or on a 5MB 11/730 in a similar state), which
is, BTW, fully loaded, thus the difference.
Actually since there is no Qbus the basic VS2000 CPU is generally faster.
However.. It lacks the external hardware support for fast IO for mass
storage and any bulk transfers. It also has a few implmentation items
that actually slow it for IO.
I haven't had the pleasure to use a fully-loaded VS2000 - mine are
around 6MB - enough to boot and run, but not a lot of empty RAM
sitting around.
I have three one each 6, 8mb and 12mb. I use the 6mb as hot spare
and for formatting disks as the differnce from 6 to 8mb is fairly
noticeable even with V5.44.
I think a VS2000 disk vastly underperforms compared to
an RA81 on a
UDA50, so the swapping makes it much, much worse.
No question. MFM disks have less than half the serial data rate and
the RA/UDA has a lot more smarts to make it happen (Buffers, LRU cache,
silos). The 9224 HDC is fairly dumb it's basicically a 765 +similar
HDC with DMA chip built in. On the VS2000 the DMA only goes to a
16k segment (hardware implementation limitation) so to get data
elsewhere the CPU or another DMA chip moves data again from the
buffer area to wherever. So in the end The VS2000 disk system
is slower than most PCs running the WD1003 with the same RD54.
A small insight as to why a MVII with the same chip is generally
faster.
Allison