Hi Matthias

Thanks for your reply.
Attached you will find debug.txt and logfile.txt from db_fillrectangle.

I hope this will help you optimizing disko!

Thanks for your help in advance.

I also found out that db_blend2layer doesn't work anymore when disko is built with messages=y. I will also attach debug_db_blend2layer.txt.

Kind Regards
Nahsah

Here the logfile.txt of db_fillrectangle.

And finally debug_db_blend2layer.txt

After looking deeper in the file, I think I found a critical point:

DISKO: Missing following combination in method extendedAccelBlit()
source type: surface
source memory: managed by disko
source pixelformat: ARGB
source premultiplied: yes
destination type: surface
destination memory: managed by disko
destination pixelformat: RGB16
destination premultiplied: yes
destination color: r=0, g=0, b=0, a=192
blitting flags (000003): BLEND_ALPHACHANNEL BLEND_COLORALPHA
*****

Is this the problem?

Regards
Nahsah

Hi nahsah,

that's not the reason for the bad performance.
It simply means, we don't support blending with alpha
channel to target pixelformat RGB16 by now.

We'll take a closer look at the blitting/blending
routines to RGB16 in the next days. Maybe there is some
obvious problem.

Sorry, for not being more helpful by now, but we'll
let you know, if we find something.

Best regards,
Matthias

Hi nahsah,

We contacted Andahammer and he is convinced that our blitting routines are just too slow.

We made some tests on a blackfin platform which is almost comparable to your board in respect to the performance. Our benchmark results were much better than yours (Factor 10-20 better). I am not convinced that our blitting routines responsible for the slow performance.

I think we are stuck here having no board to test this ourselves.

kind regards
Stefan