Finally I got disko working on the MINI2440 board (400 MHz Samsung S3C2440A ARM9 processor).
I was able to run the benchmark appliactions db_fillrectangle and db_blend2layer. db_fade2layer doesn't seem to work properly on my board.

Here are the results (all examples run with fbdev and 800x480 pixel):

db_fillrectangle:
- Fill window with alpha=0xa0 -> 4mpps
- Fill window with alpha=0xff -> 20mpps
- Clear window area with r=0 g=0 b=ff a=0 -> 20mpps
- Clear window area with r=0 g=0 b=0 a=0 -> 20mpps

db_blend2layer:
- Area with alpha channel 0xa0 and picture -> 4fps
- Area with alpha channel 0xff and picture -> 5fps
- Area with alpha channel 0x00 and picture -> 5fps
- Area with alpha channel 0xa0 no picture -> 3fps
- Area with alpha channel 0xff no picture -> 5fps
- Area with alpha channel 0x00 no picture -> 5fps

How do you interprete these results ?
Is it right that this benchmarks simulate a OSD with a running background application (e.g. tv)?
For my understanding a graphical application should run with at least 20fps (my experience with games).
So I'm a bit confused about the results.
Do you have any tips to get better results?
I'm eager to see how a "real" application would perform.

Regards
Nahsah

Hi Nahsah,

its imperative that you try the benchmarks with disko compiled without debug or messages. In case of debug you will get code with -O0 and the extensive messages from disko also slow the execution down considerably.

If you didn't already have, I encourage you to do so. Please tell us your results.

If you already have compiled it without messages and debug, we have to go deeper into your setup. The platform isn't very powerful, but the results should not be that bad

kind regards
Stefan

Hi Stefan

Unfortunately I already compiled without debug and messages.
What else would you like to know about my setup?

I don't want to use disko for playback of movies or tv. The main purpose is more some kind of an information display with mail notification, missed phone calls, "remote control" of my dreambox, ...
So I think this should reduce the performance requirements for the hardware.

Regards
Nahsah

Hi nahsah,

please give us some time to analyse/evaluate the board. We are convinced, that you'll receive better results but we have to discuss the next steps.

It's unaccustomed for us to port Disko to a board over a forum . The classic way is, to send a board to us, that we can evaluate the hardware.

regards

Guido

Hi Guido

Thanks for your reply and spending so much time with my problems.
Unfortunately I'm not able to send you my hardware or sponsor you a MINI2440 as this is a "hobby" project in my spare time.
The disko framework looks promising for me because it's not too big and fulfills my requirements almost perfect. I was evaluating Qt Embedded but for me it's too overloaded.

So I don't expect a special MINI2440 port but I would be happy if you could give me some hints.

If you're really interested in porting disko to the MINI2440 (I hope so ) there may be another solution:
You could contact Andahammer (www.andahammer.com). They spent some hardware to a guy (buserror) who ported the actual kernel to the MINI2440. Maybe they have a spare MINI2440.

Again, thanks for spending so much time for me!

Kind regards
Nahsah

Hi nasah,

can you please compile disko with messages=y and send us the output from
stdout, when running your test application? It lists the blitting routines
that are used on your board.

Then we'll have a look on how to optimize them.

Best regards,
Matthias