<I guess it's fortunate there was only one DMA process going on at the time
<else it might have been real sticky figuring out what had been overwritten
<already.. If you were doing a read in order to do a write, using DMA, you
<might actually get tangled up. Fortunately that showed up while the vendo
<was debugging his code, so I didn't have to deal with that.
Multiple DMA streams are doable too though hard to apply usefully.
<That's quite so. Fortunately one wasn't required to load data at the
<granule size, but rather at the sector size, so you could get by with a rea
<of a 1K sector. Of course you had to read it before you could write it, s
<you had to wait for the next revolution of the disk. All this went by so
<fast, and, since I didn't run big databases requiring sorts to and from
<disk, I didn't perceive much delay, as it only takes a few revolutions to
<load up a program. So each drive had six logical drives on it.
I have. Running a pair of drives and using ramdisk and my own smartdisk
system. That was the speed order as well, the smartdisk system was fasest
as it hard its own CPU and DMA channel to processor ram using hidden cycle
stealing plus caching to 4x physical track size. That and a 6MHz z80 and
dust flew.
<This all sounds like it could be fun if, for example, you're running it al
<on classic and unmodified hardware. I'm not sure I'd want to try to earn m
<living that way, though.
it's more fun on current hardware like 33mhz Z180s and it's still in use
in odd pockets here and there. I don't (never did) make a living off it.
Allison
Received on Tue Mar 30 1999 - 17:41:13 BST