radeon RV280 stretchtblit in YUY2 mode

On Mittwoch 28 Juni 2006 17:59, Michel Dänzer wrote:
> On Wed, 2006-06-28 at 17:51 +0200, Claudio Ciccani
wrote:
> > Stefan Lucke wrote:
> > > On Dienstag 27 Juni 2006 18:20, Michel
Dänzer wrote:
> > > 
> > >>On Tue, 2006-06-27 at 12:09 +0200, Claudio
Ciccani wrote:
> > >>
> > >>>Stefan Lucke wrote:
> > >>>
> > >>>>has someone stretchblit mode
working for a radeon RV280 ?
> > >>>
> > >>>I have a Radeon 9200SE and
YUY2->RGB blits work.
> > >>
> > >>RV250s are known to have issues with this,
maybe Stefan actually has one
> > >>of those?
> > > 
> > > 
> > > Is RV250 the same (in progamming) as RV280 ?
> > > Mine is a 9200 (not SE) with 256MB ram if
that makes a difference.
> > > Card details could be found here:
> > > http://mail.directfb.org/pipermail/directfb
-dev/2006-June/001929.html
> > > 
> > > Perhaps I should give directfb stretchblit
mode a chance, when it
> > > is not called from an X_term .
> > 
> > I heard about a non working YUV->RGB in the
RV250,
> > but that only applies to the RV250, not to the
RV280.
> > Of course, if it works on my 9200SE, it must work
on the 9200 as well
> > since the two cards only differ in the memory
bandwidth.
> > I think the source of the problem is somewhere
else.
> 
> Stefan, the above sounds like you're not using the
radeon gfxdriver but
> the SDL or X11 backend?

No, I'm not using SDL and not an X11 backend. Having X11
working on
directfb is on my wishlist  .

Meanwhile I did the test from console (vdr & softdevice)
and the result is the
the same garbage on screen as soon YUY2 is selected in
stretchblit mode.
This mode of softdevice is known to work with matrox cards.

When returning to console, console view is corrupted. Had to
type reboot blind.
All radeon registers have to be restored (even overly
registers) on directfb-exit.

My RV280 works with mesa YUY2 (GL_YCBCR_MESA) textures even
it is rather
cpu consuming compared to X11-Xv (60% vs 30%).

-- 
Stefan Lucke

_______________________________________________
directfb-dev mailing list
directfb-devdirectfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/
directfb-dev

Stefan Lucke wrote:
> Meanwhile I did the test from console (vdr &
softdevice) and the result is the
> the same garbage on screen as soon YUY2 is selected in
stretchblit mode.
> This mode of softdevice is known to work with matrox
cards.
> 
> When returning to console, console view is corrupted.
Had to type reboot blind.
> All radeon registers have to be restored (even overly
registers) on directfb-exit.
> 
> My RV280 works with mesa YUY2 (GL_YCBCR_MESA) textures
even it is rather
> cpu consuming compared to X11-Xv (60% vs 30%).
> 

That 60% of CPU load sounds like a software fallback is
being used.

Please, could you tell me the device id detected by the
radeon driver?
To do that, search a line similar to the following in the
log messages:
(*) DirectFB/Graphics: ATI Radeon 9200 SE (5964) 1.0
(Claudio Ciccani)


-- 
Regards,
      Claudio Ciccani

klanusers.sf.net
http://directfb.org
http://sf.net/pro
jects/php-directfb

_______________________________________________
directfb-dev mailing list
directfb-devdirectfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/
directfb-dev

Quoting Claudio Ciccani :

> Stefan Lucke wrote:
> > Meanwhile I did the test from console (vdr &
softdevice) and the result is the
> > the same garbage on screen as soon YUY2 is
selected in stretchblit mode.
> > This mode of softdevice is known to work with
matrox cards.
> >
> > When returning to console, console view is
corrupted. Had to type reboot blind.
> > All radeon registers have to be restored (even
overly registers) on directfb-exit.
> >
> > My RV280 works with mesa YUY2 (GL_YCBCR_MESA)
textures even it is rather
> > cpu consuming compared to X11-Xv (60% vs 30%).
> >
>
> That 60% of CPU load sounds like a software fallback is
being used.

Or waiting for v-retrace in a loop. Messages I get indicate
direct rendering == YES

>
> Please, could you tell me the device id detected by the
radeon driver?
> To do that, search a line similar to the following in
the log messages:
> (*) DirectFB/Graphics: ATI Radeon 9200 SE (5964) 1.0
(Claudio Ciccani)

Thereis a difference. Quoting from:
http://mail.directfb.org/pipermail/directfb
-dev/2006-June/001929.html

it is a:
(*) DirectFB/Graphics: ATI Radeon 9200 (5961) 1.0 (Claudio
Ciccani)

Stefan Lucke

_______________________________________________
directfb-dev mailing list
directfb-devdirectfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/
directfb-dev

On Donnerstag 29 Juni 2006 08:33, Claudio Ciccani wrote:
> Stefan Lucke wrote:
> > Meanwhile I did the test from console (vdr &
softdevice) and the result is the
> > the same garbage on screen as soon YUY2 is
selected in stretchblit mode.
> > This mode of softdevice is known to work with
matrox cards.
> > 
> > When returning to console, console view is
corrupted. Had to type reboot blind.
> > All radeon registers have to be restored (even
overly registers) on directfb-exit.
> > 
> > My RV280 works with mesa YUY2 (GL_YCBCR_MESA)
textures even it is rather
> > cpu consuming compared to X11-Xv (60% vs 30%).
> > 
> 
> That 60% of CPU load sounds like a software fallback is
being used.
> 
> Please, could you tell me the device id detected by the
radeon driver?
> To do that, search a line similar to the following in
the log messages:
> (*) DirectFB/Graphics: ATI Radeon 9200 SE (5964) 1.0
(Claudio Ciccani)
> 

I've seen your update in df_dok, but strangely neither
fonts nor pngs work.
Only --stretch-blit-colorkeyed (with gif) works. 
--stretch-blit-ycbcr shows only a green rectangle with a few
flickering black
horizontal bars.
I checked my configuration by rerunning autogen.sh:

Building Image Providers:
  GIF                       yes
  JPEG                      yes
  MPEG2 I-Frame             yes
  PNG                       yes

Building Video Providers:
  Video4Linux               yes (v2: no)

Building Font implementations:
  FreeType2                 yes
  Default font              yes

So this should work.

Complete log from df_dok:
(*) DirectFB/Config: Parsing config file
'/etc/directfbrc'.

       ---------------------- DirectFB v0.9.26
---------------------
             (c) 2000-2002  convergence integrated media
GmbH
             (c) 2002-2004  convergence GmbH
       
-----------------------------------------------------------

(*) DirectFB/Core: Single Application Core. (2006-06-29
17:34)
(*) Direct/Memcpy: Using MMXEXT optimized memcpy()
(*) Direct/Thread: Running 'VT Switcher' (CRITICAL,
11042)...
(*) Direct/Thread: Running 'PS/2 Input' (INPUT, 11043)...
 (!!!)  *** UNIMPLEMENTED [fusion_reactor_set_lock] ***
[reactor.c:853]
(*) DirectFB/Input: IMPS/2 Mouse 1.0 (Convergence GmbH)
(*) Direct/Thread: Running 'Keyboard Input' (INPUT,
11044)...
(*) DirectFB/Input: Keyboard 0.9 (convergence integrated
media GmbH)
(*) DirectFB/Genefx: MMX detected and enabled
(*) DirectFB/Graphics: ATI Radeon 9200 (5961) 1.0 (Claudio
Ciccani)
(*) DirectFB/Core/WM: Default 0.2 (Convergence GmbH)
(*) Direct/Interface: Loaded 'PNG' implementation of
'IDirectFBImageProvider'.
(*) Direct/Interface: Loaded 'FT2' implementation of
'IDirectFBFont'.
(*) Direct/Interface: Using 'GIF' implementation of
'IDirectFBImageProvider'.
 (!!!)  *** WARNING [still objects in 'Layer Context
Pool'] *** [object.c:226 in fusion_object_pool_destroy()]

Benchmarking with 256x256 in 24bit mode... (32bit)

Anti-aliased Text                              3.002 secs (*
803.464 KChars/sec)
Anti-aliased Text (blend)                      3.000 secs (*
805.200 KChars/sec)
Fill Rectangle                                 3.004 secs (*
922.827 MPixel/sec)
Fill Rectangle (blend)                         3.025 secs (*
207.982 MPixel/sec)
Fill Rectangles [10]                           3.014 secs (*
934.986 MPixel/sec)
Fill Rectangles [10] (blend)                   3.151 secs (*
207.984 MPixel/sec)
Fill Triangles                                 3.003 secs (*
399.370 MPixel/sec)
Fill Triangles (blend)                         3.016 secs (*
202.083 MPixel/sec)
Draw Rectangle                                 3.000 secs (*
 95.300 KRects/sec)
Draw Rectangle (blend)                         3.000 secs (*
 43.466 KRects/sec)
Draw Lines [10]                                3.001 secs (*
346.551 KLines/sec)
Draw Lines [10] (blend)                        3.003 secs (*
186.480 KLines/sec)
Fill Spans                                     3.001 secs (*
801.456 MPixel/sec)
Fill Spans (blend)                             3.017 secs (*
232.427 MPixel/sec)
Blit                                           3.005 secs (*
505.968 MPixel/sec)
Blit colorkeyed                                3.016 secs (*
323.768 MPixel/sec)
Blit destination colorkeyed                    3.638 secs ( 
  1.801 MPixel/sec)
Blit with format conversion                    3.015 secs (*
330.397 MPixel/sec)
Blit with colorizing                           3.012 secs (*
276.330 MPixel/sec)
Blit from 32bit (blend)                        3.025 secs (*
190.650 MPixel/sec)
Blit from 32bit (blend) with colorizing        3.023 secs (*
190.776 MPixel/sec)
Stretch Blit                                   3.025 secs (*
381.502 MPixel/sec)
Stretch Blit colorkeyed                        7.698 secs ( 
 42.206 MPixel/sec)
Stretch Blit YCbCr                             3.036 secs (*
400.051 MPixel/sec)

-- 
Stefan Lucke

_______________________________________________
directfb-dev mailing list
directfb-devdirectfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/
directfb-dev

On Donnerstag 29 Juni 2006 22:43, Stefan Lucke wrote:
> On Donnerstag 29 Juni 2006 08:33, Claudio Ciccani
wrote:
> > Stefan Lucke wrote:
> > > Meanwhile I did the test from console (vdr
& softdevice) and the result is the
> > > the same garbage on screen as soon YUY2 is
selected in stretchblit mode.
> > > This mode of softdevice is known to work with
matrox cards.
> > > 
> > > When returning to console, console view is
corrupted. Had to type reboot blind.
> > > All radeon registers have to be restored
(even overly registers) on directfb-exit.
> > > 
> > > My RV280 works with mesa YUY2 (GL_YCBCR_MESA)
textures even it is rather
> > > cpu consuming compared to X11-Xv (60% vs
30%).
> > > 
> > 
> > That 60% of CPU load sounds like a software
fallback is being used.
> > 
> > Please, could you tell me the device id detected
by the radeon driver?
> > To do that, search a line similar to the following
in the log messages:
> > (*) DirectFB/Graphics: ATI Radeon 9200 SE (5964)
1.0 (Claudio Ciccani)
> > 
> 
> I've seen your update in df_dok, but strangely neither
fonts nor pngs work.
> Only --stretch-blit-colorkeyed (with gif) works. 

After switch to my G550 I've seen that even this one did
not work correct.
With rv280 there was only one zooming in and not multiple.
With G550 png works and fonts work too, so I can see the
name of actual
running demo.


-- 
Stefan Lucke

_______________________________________________
directfb-dev mailing list
directfb-devdirectfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/
directfb-dev