On Sat, 2 Aug 1997, Kim G. S. yhus wrote:

> > On Fri, 1 Aug 1997, S.J. Borrill wrote:
> > 
> > > Has anyone got any experience of CD writers under RISC BSD (or even RISC
> > > OS at a push)? Are there any particular ones recommended or any caveats in
> > > their use?
> > 
> > You cannot do it under RiscBSD running on a RiscPC due to the speed of the
> > drivers unless you have a writer with a build in HD.
> 
> Lets quantify this.
> Say one use a phillips CDwriter, with 1MB cache, and 300KB writing speed,
> and have an old SCSI I card, writing 1MB/s.
> 
> To fail, RiscBSD would have to stop sending data to the CDwriter for
> at least 3 seconds. This is a large latency. I have worked with old
> 16MHz 68020 Unix machines that had latencies at least an order of
> magnitude less than this.
> 
> The writing itself, would require the moving of 300KB of data to the
> SCSI card. This is very little of the DEBI-bus bandwith of a RiscPC.
> 
> The SCSI card would only have to work at 1/3 of its capacity,
> which gives very large room for overhead.
> 
> So, why is RiscBSD so bad at handling this?
RiscBSD in principle has no problem (I have a system that could write
CDROM's easily) the problem is with SCSI CDROM writers under RiscBSD on a
RiscPC. Yes it should be able to do however none of the SCSI drivers have
the speed. With your argument above you are assuming that RiscBSD gets 1MB
per sec out of the SCSI card. At the moment it doesn't.
This would indicate that the SCSI driver need work and I would not
disagree. Note: This does not just apply to RiscBSD. I can think of a
number of SCSI cards that will not support a CDwriter under RISCOS.
For example the DMA SCSI cards are quoted as being nice and fast however
what happens if you turn DMA off ? You can end up with similar performance
to the RiscBSD SCSI drivers that are not using DMA.
You could say well if you have DMA then you will always be using it. That
is not true I know people who have problems with corrupts when using DMA
transfers on SCSI cards etc. etc.

Cheers,
				Mark