[comp.sys.atari.st.tech] Ultradisks?

hyc@math.lsa.umich.edu (Howard Chu) (10/16/90)

Howdy... Anyone else out there with an Alpha Systems cartridge-port RAMdisk?

I finally figured out why I was getting trashed directories on mine... I had
set up two 2 megabyte RAMdisks, and would lose the root directory sometimes.
Turns out that the program passes a pre-defined BIOS Parameter Block to DOS
when it asks, and always says there are only 5 sectors per FAT. Well, 5
sectors per FAT is only enough to map 1706 clusters, or 1.7Meg of space.
So since the root directory begins right after the FAT, it got used up next...
Lemme tell ya, ya get some bizarre looking stuff when TOS tries to interpret
a root directory sector as FAT data.

I don't understand why TOS would be allowed to grab the next sector on down,
since the BPB says explicitly that only X number of sectors are available,
but so it went. After disassembling the Alpha Systems RAMdisk code, I found
their hardcoded BPBs and changed 'em from 5 to 7, so now my RAMdisks are
happy again.

Of course, now that I understand why this was a problem, (which is why I
split things into 2 disks in the first place, because back when I first
bought it, it only had 2 meg of RAM. I tried to make a single disk with that,
and hey, it didn't work either... How silly of me to think that they would
ship software set to work properly for the configuration of hardware they
sold you... Sheesh.) I can merge it back into a single 4 meg partition, but
now I'm more interested in just using it for cache. I get the effect anyway,
for some tasks, since I have all of MWC sitting on one of the RAMdisks. But
it'd be nice not to have to worry about loading specific programs onto the
RAMdisk, and just let the cache program take care of commonly invoked stuff.

So... Where's a good place to start? Is CACHExxx (in the Rainbow TOS utilities)
really a disk cacheing program? I had set it up to use 1000 buffers at one
point, but I still saw disk activity lights trying to access a program for the
second or 3rd time. This doesn't seem right to me. (For floppy access, I can
understand the need to look for Media Change, but for a fixed hard drive?)

--
  -- Howard Chu @ University of Michigan
  one million data bits stored on a chip, one million bits per chip
	if one of those data bits happens to flip,
		one million data bits stored on the chip...