At 04:13 PM 10/13/96 -0500, Dave Huang <khym@bga.com> wrote:

[ ... blah blah ... ]

> >Description:
> People have been telling me to send-pr this, so here it is :)
>
>
> I recently got a Quantum Atlas drive, and was disappointed that using
> dd to read the first 128 megs of the disk, I could only get about
> 2.8MB/s out of it. I've also got a Quantum Capella, which is a slower
> drive, but I was getting 5.8MB/s out of it, so I figured something was
> wrong with the driver... Also, if I used dd to read from both disk
> simultaneously, I only got about 2.2MB/s out of each drive, for
> 4.4MB/s total... much less than the theoretical max of 10MB/s.
>
> Since the FreeBSD ncr driver is supposed to work a lot better, I
> >Description:
> People have been telling me to send-pr this, so here it is :)
> 
> 
> I recently got a Quantum Atlas drive, and was disappointed that using
> dd to read the first 128 megs of the disk, I could only get about
> 2.8MB/s out of it. I've also got a Quantum Capella, which is a slower
> drive, but I was getting 5.8MB/s out of it, so I figured something was
> wrong with the driver... Also, if I used dd to read from both disk
> simultaneously, I only got about 2.2MB/s out of each drive, for
> 4.4MB/s total... much less than the theoretical max of 10MB/s.
> 
> Since the FreeBSD ncr driver is supposed to work a lot better, I
> downloaded a FreeBSD boot floppy and fixit floppy and tried the dd
> test again... got 7.1MB/s out of the Atlas this time (still 5.8MB/s

>From a quick look at the patch included in the PR, I gather that the main
improvement is to replace a maximum number of command queue tags
(SCSI_NCR_MAX_TAGS) with a default number of tags (SCSI_NCR_DFLT_TAGS), so
presumably the number of tags can be increased, and this is what allows the
drive to work at full speed where it couldn't before.

If so, you might attain the same objective simply by increasing the maximum
number of tags (SCSI_NCR_MAX_TAGS) in the existing NetBSD driver to an
appropriate figure.

I can see why a dynamic number might be preferable, but if I'm accurate,
then it really shouldn't be necessary to modify the driver so extensively to
get it working right with your drive.

But then, I'm probably out to lunch, which is not unheard of...


--Jim Howard
  jiho@mail.c-zone.net