Re: Wanted: a faster protocol for IEEE devices

From: Michał Pleban <lists_at_michau.name>
Date: Wed, 18 Sep 2013 09:59:34 +0200
Message-ID: <52395D66.2090002@michau.name>
Hello!

Baltissen, GJPAA (Ruud) wrote:

> I just realized that this won't work :( The above idea is not mine, this is the way Star Commander works. But to be able to read the byte and to send it the next moment, all should be done by one single CPU. But, with the exception of the 2031, the IEEE drives have two CPUs: one handling the disk, the other handling the IEEE bus. How does the disk CPU tell the bus CPU that one byte has been read without loosing too much time?
> An idea would be reading one whole track, but then we would need at least 8 KB of RAM just for buffering the data. And as I wrote to Gordon, we only have 4 KB on board.
> An in between solution: telling the bus CPU during the gap between two sectors that a sector has been read. While the disk CPU is filling another part of RAM with data, the bus CPU can send the data of the first sector. 

Did I understand correctly that you are actually building some custom
hardware for reading the disks and transferring their images to the PC?
If so, why not equip the 8250 with 1 MB of RAM to read the whole disk?
;-) Or better 2 MB - so one CPU could read one disk, and the other CPU
could transfer a previous disk to the PC at the same time :-)

A cheaper solution: why not simply experiment with disk sector
interleave in your disk transfer software, and see what value gives the
best results? The interleave value can have a really huge impact on the
transfer speed, and you can change it programmatically (at least with
cbmlink).

Regards,
Michau.

       Message was sent through the cbm-hackers mailing list
Received on 2013-09-18 08:00:49

Archive generated by hypermail 2.2.0.