Path: utzoo!utgpu!water!watmath!clyde!att!mtunx!rutgers!njin!princeton!udel!rochester!cornell!batcomputer!itsgw!steinmetz!uunet!mcvax!uva!maarten
From: maarten@uva.UUCP (Maarten Carels)
Newsgroups: comp.sys.mac
Subject: Re: SCSI Accelerator
Message-ID: <329@uva.UUCP>
Date: 7 Jun 88 07:28:16 GMT
References: <8586@dhw68k.cts.com>
Reply-To: maarten@uva.UUCP (Maarten Carels)
Organization: Faculteit Wiskunde & Informatica, Universiteit van Amsterdam
Lines: 33


I just looked at SCSI accelerator (just posted in comp.binaries.mac) and
thought you folks would like to share my experiences. 

What it does:
Basicly, SCSI Accelerator patches itself into the scsi manager to do
some loop unfolding. This causes some real speedup, especially if you
reformat the disk with another (lower) interleave. I was able to lower
the DistTimer II results of one disk from about 140 to 66 by installing it
and reformatting the disk with an interleave of 2 instead of 3.
So, the speedup is remarkable. On another disk, no speedup was achieved
(but I did not reformat it, backup and restore of some 50 Megs just to try...).

What it does not do:
Although some code is present to patch the write part of the scsi mgr, that
is not installed. The write results of my victim disk increased from 160 to
220. This is to be expected, the interleave is just plain wrong then.

CAUTION:
By the way it works, SCSI accelerator will not work on some disks. Especially
Rodime drives, to be found in SE's (well, some of them) and HD20SC's.

What you can expect of me:
An INIT like this, expected to work on all (well, maybe not all, but on most)
disks, equally active during read and write.
Stay tuned!

--maarten
In real life:	Maarten Carels
		Computer Science Department
		University of Amsterdam
email:		maarten@uva
		{philabs, decvax, seismo}!mcvax!uva!maarten