On Fri, 07 Sep 2001 22:21:42 +0100, Philip Blundell <philb@gnu.org> wrote:
>
> >It's interesting that in all cases 2 or 3 bytes are added, then 32
> >bytes later one is skipped.
>
> Yeah. Well, if I'm reading them right, you are always getting two bytes
> repeated and then, 32 bytes later, two bytes dropped. So:
>
> ***** 4226: 0a should be 32
> ******
> 30,0a,31,32,33,34,35,36,37,38,39,30,0a,31,0a,
> 0a <===
> 31,32,33,34,35,36,37,38,39,30,0a,31,32,33,34
> ******
>
> here you've got `0a, 31' repeated, and
>
> ***** 4259: 31 should be 30
> ******
> 37,38,39,30,0a,31,32,33,34,35,36,37,38,39,31
> 31 <===
> 32,33,34,35,36,37,38,39,30,0a,31,32,33,34,35
> ******
>
> here you've lost `30, 0a'.
>
> Remind me whether you see the same behaviour if you disable DMA but leave
> hardware assist enabled? (I.e. if you use parport_pc_fifo_write_block_pio.)
>
> p.
>
That's something I hadn't thought of trying. I've done so (by setting
dma=none in /etc/modules.conf) and it works with no errors. The
interesting thing is that it's *faster* without DMA than with DMA --
on one printer I get 980 kBytes/sec without DMA vs 490 kBytes/sec with
DMA (in ECP mode); on the other printer I get 540 kBytes/sec without
DMA vs. 340 kBytes/sec with DMA. In compatibility mode it is also
faster without DMA but not as dramatically so.
I think at least for now I think I'm just going to turn DMA off. I'm
still open to trying other experiments, though, if anybody wants to
suggest something.
-- Dave Strauss
-- To unsubscribe, send mail to: linux-parport-request@torque.net --
-- with the single word "unsubscribe" in the body of the message. --
This archive was generated by hypermail 2b29 : Mon Sep 10 2001 - 14:13:24 EDT