Rogier> We could try to issue the next "read" AS FAST AS POSSIBLE,
Rogier> discarding the data just read. This probably requires some
Rogier> serious hacking.
As of sane-0.62, this is already what's happening for Linux. That's
the reason I introduced the asynchronous scsi interface to SANE. I'm
sure it would be possible to shave off a few cycles here and there,
but control-flow wise, the right thing is happening. If somebody gets
serious about this, I'd recommend to first find out how much slack one
has with issueing the next "read data" command. E.g., modify the
backend to not transfer any data at all to the frontend and see
whether that helps. Once you get to the point where no backtracking
occurs, you can start inserting delays until backtracking starts
occuring again. At that point, you should have a pretty good idea of
what should be and can be done.
--david
-- Source code, list archive, and docs: http://www.azstarnet.com/~axplinux/sane/ To unsubscribe: mail -s unsubscribe sane-devel-request@listserv.azstarnet.com