Hi,

I develop on XBurst JZ4740, it's a mips32.
I already get all my environment on it (linux, directfb...)
As I'm not familiar with python/scons... can you please tell me how to cross compile and provide my own compiler (mipsel-linux-g++) ?

I'm looking at something similar to --host=mipsel-linux-g++

Thanks for your support.
Best regards,
Nouch64

I think I found it with the variable : env['CXX'] env['LINK'] on SConstruct
I will let you know later if I need more support.

Thanks.
Best regards,
Nouch64

Hi Nouch64,

you're right. These variables can be overridden by setting CXX and/or LD in your environment.

Best regards,
Matthias

Hi,

Based on that, I get now disko and morphine.tv (whithout plugins for now) working on my board.
Almost all the tutorials, demos and benchmakrs are working also.
For information, with the db_blend2layer I turn at 19fps, which one is not so powerful...
These tests are based on directfb 1.3.0, with RGB32 480x272

One of my concerning is that I get different behavior depending if I use outputtype dfb or x11, extendedaccel true or false, allocmethod dfb or malloc.
The differences are either I get the text or not, either the screen refresh or not (if not, I have to move the cursor on the surface and the surface refresh just under). Let's said it's a bit strange.
I don't have X11 on my platform, but to change outputtype from dfb to x11, I get or not some animations.

Beside this, thanks for your work. It's really on the good way.
If I get time, I will try to help you for the devs/debugs.

For reference, this is the board I use : http://www.ingenic.cn/eng/productServ/kfyd/Hardware/pffaqQuestionContent.aspx?Category=2&Question=3

Best regards,
Nouch64

Dear Nouch64,

DirectFB is missing some features in X11. Thats why we implemented an
own X11 backend. The X11 backend of DirectFB is merely intended to be
used for development purposes and lacks the possibility to use the XV
extension (better video quality, vsync, hardware-stretch) is missing a
full screen feature and the usage of expose events.

Our X11 backend ist quite new. If you use disko from git you will
probably get a new x11 feature every other day. To utilize X11 we also
did some groundwork to replace directfb.

We wrote a set of blitting routines (stretch, blend, colorize, nofx) for
various pixelformats. There are still many combinations missing. If
there is a combination missing (when you dont see the font), you
should see some messages on stdout of your program. The easiest way is
to announce such a missing combination here, that we can quickly
implement it. These blitting routines are triggered by the
"extendedaccel" parameter. It is called exetendedaccel because our
blitting routines are in some cases considerably faster than the
Directfb software fall backs.

To utilize the own blitting routines it was also necessary to create
an own surface format to be used within disko when our blitting
routines are in effect. This is triggered by the "allocmethod"
parameter.

So there are some combinations that will create strange results, but
thats ok, they are wrong We are working on some checking routines
that will state if there is a flaw in the configuration. Here a short
summary (parameters are from git, 1.5.0 is not aware of all) :

---
backend:
- x11 (our X11 backend)

outputtype:
- xshm (ximage backend like DFB, but also fullscreen)
- xvshm (xv backend for better video quality and vsync)

this requires: extendedaccel=true, allocmethod=malloc

---
backend:
- dfb (directfb backend)

outputtype (could be overridden by directfb
- viafb (special configuration for viafb)
- matroxfb (special configuration for viafb)
- vesafb (no special configuration at all
- x11 (directfb X11 backend)

this requires: extendedaccel=false, allocmethod=dfb
---

We alse have a first implementation of an own framebuffer
implementation in the pipe, but thats not really finished by now.

All other combinations of the backend, outputtype exetendedaccel and
allocmethod will currently lead to an undefined behavior. This will be
fixed in the next major release.

Hope this makes a bit more sense to you.

have fun
Stefan

Hi,

Thanks a lot for this description.
I think you should put this comment somewhere on your documentation, it's very useful for now.

Have a great future.
Nouch64