This message is from the T13 list server.
Today I connected with an asic-head, who explained my English so far has been sloppy
enough to falsely deny:
An analogous, and apparently well-known, vulnerability to make byte counts inaccurate
was technically present in:
standard multiword Dma when moving data in,
I think I now remember this fact was mentioned to me by another guru by phone a couple
of days back as I was just clueing into this stuff.
But in practice, popular devices took care to turnaround fast enough to get DMARQ
deasserted fast enough to duck this issue there.
What's different about UDma in/out vs. MWDma in is that UDma senders much more
commonly do clock more data after a request to pause.
(((As I mentioned, personally I care about UDma after successfully ignoring
SwDma/MwDma for all these years, because UDma is the first Dma that offers burst rates
Pio cannot achieve.)))
Pat LaVarre
>>> Pat LaVarre 12/05/01 08:54AM >>>
...
> > UDma indeterminacy ...
> pause ... only matters ... in the middle.
That would be true only if in advance the sender and the receiver always agreed where
the middle ends ...
...
Subscribe/Unsubscribe instructions can be found at www.t13.org.
