Hi,

I hope you don't mind that I Cc'ed the list. I've accidentally
forgotten to do so in my reply but I think we should continue this
thread on the mailing-list...

Frank Mehnert <[EMAIL PROTECTED]> writes:

> On Thursday 01 August 2002 11:34, you wrote:
> > Frank Mehnert <[EMAIL PROTECTED]> writes:
> > > I'm interested in the current state of the savagefb kernel driver patch.
> > > Does anyone works on it? README.savagefb states that acceleration does
> > > not work with the 0.4.0 version. Does that mean that the machine hangs
> > > or that nothing appears at the screen?
> >
> > the accelerated functions simple don't seem to have any effect which
> > means that the screen is not cleared and not scrolled.  Fortunately
> > this is harmless and acceleration can be turned off using the command
> > 'fbset -accel false' which makes the 0.4.0 version of the driver
> > behave like the 0.3.2 version.
> 
> Did you test it only with Savage4 graphics cards? I'm have a Thinpad T20
> here with a Savage/IX-MX. There is another problem with the chipset of that
> Thinkpad which causes me some trouble using XFree86 some months ago. The
> Savage/IX-MX freezes the whole system under some strange conditions. After
> many hours of debugging and talking with Tim Roberts (XFree86 Savage driver
> developer) we developed a solution for that problem.

I'm using the very same IBM model with the Savage/IX-MX card.

> There are functions WaitForFifo and WaitForIdle which wait until the chip
> is idle in some way, polling a memory mapped status register. The check of
> the register leads to freeze the whole machine on very few Thinkpads under
> some spare circumstances. The freeze does not happen if the status register
> is shadowed into the RAM or even into the video RAM. For that reason the
> ShadowStatus option was introduced into the late XFree86 savage drivers.

yes, I've seen DirectFB applications freeze my hardware. I haven't
tracked it down in detail but I had the impression that it happens
if the command buffer fills up. It can be easily triggered by
running df_dok with a longer benchmark duration (--duration 10000).

For this very reason, I'm almost always using the no-hardware option
to turn off acceleration on my notebook but of course I'd like to make
use of the hardware capabilities. We should probably try to get a
similar fix into the savage frame buffer and directfb drivers then.

> That is the reason that I'm still not tested the Savage fb driver -- I'm
> afraid to loose data under Linux. 

If you use EXT3 or another journaling filesystems and sync a lot, your
data should be pretty safe. Note that you can put the sync option into
your directfbrc.

> Unfortunately, the directfb driver is not able to program the "new" Savage
> stream engine as needed by the Savage/IX-MX. Because I not own the Savage
> documentation, I had to use the slow YUY2 video mode for yuv2rgb conversion
> from the XFree86 driver.

yes, it would be cool to have support for the new streams engine.

> So my intention is to get the documentation for the Savage Chipset from
> somewhere (wrote to the EMail address at s3graphics.com -- no answer, no
> no answer from Denis Kropp who seems to have the docs, too) and to improve
> my driver.

oh, I thought that Dok gave you the docs we have. Didn't he?

> I also would like to help developing the directfb Savage driver, too. In 
> some weeks I have more time to setup a "crash" partition where I could test
> the directfb Savage driver. I'm confident that I can help to get the directfb
> driver get work -- but not in the next days because I have to put further
> work into a paper which is accepted at a conference.

any help is greatly appreciated.


Salut, Sven


-- 
Info:  To unsubscribe send a mail to [EMAIL PROTECTED] with 
"unsubscribe directfb-dev" as subject.

Reply via email to