From: Michael Sevakis <jethead71_at_comcast.net>
Date: Sat, 4 Feb 2012 15:36:11 -0500

----- Original Message -----
From: "Boris Gjenero" <boris.gjenero_at_gmail.com>
To: "Rockbox development" <rockbox-dev_at_cool.haxx.se>
Sent: Saturday, February 04, 2012 1:28 PM
Subject: Re: PP502x cache bug and ATA DMA test build


> On 03/02/2012 8:43 PM, Michael Sevakis wrote:
> > For H10 20GB, dircache init and database init seem to pass alright where
> > as before with DMA it would fault frequently. It's running UDMA 2.
>
> Thanks for testing. It's nice to see that DMA is working better now.
>

I should point out that I never had issues on PP5024 devices with any sort
of test. They don't use DMA for storage I/O but do for audio, but it is a
different DMA controller for the ATA than the general-purpose "SIDMAC". If
some particular models don't need the extra operations, then it would be
better not to have the overhead. I'll give the newer cache tester a spin on
5024 with an unmodded build.

As far as UDMA2, it was just a verification of DMA mode being used. The
beast uses UDMA4 but it doesn't seem to buy any actual speed over UDMA3 in
testing (just pointing out that faster modes don't always mean much anyway
on a real disk).

Mike
Received on 2012-02-04