Some Obscure Problem

[Jerome Glisse]

On 5/10/05, Bora Sahin <bora.sahin at ttnet.net.tr> wrote:
> Hello,
> 
> Jerome> I think we face the same problem visual problem with r300 & 2d acceleration..
> Jerome> Having a picture (even digital camera could help a lot to see it exactly :))
> 
> Jerome> Something like this  ?
> Jerome> http://dj.planet-d.net/Capture.png
> 
> Jerome> Thus if it's same problem it's simply a endianness issue. If you see
> Jerome> not artifact
> Jerome> in 640*480*8 this is most likely because this is a 8bits modes...
> 
> I tried 8 bit depth(Xorg uses PSEUDOCOLOR) and no success. But my working area is a xterm:
> background - white, foreground - black(Characters). So if my assumptations are correct, depth is
> not problem in this case. The problem is pixel locations. I didnt put e.g. the pixel which has
> (50,50) coordinate to the correct point in the frame buffer. Is this a correct approach to the
> problem?
> 
> Secondly, If my card supports byte switching, I should do that in SMI_MapMem() function before the
> the first [in|out][16/32] call? I am asking this because I didnt do any video driver work before and
> dont know Xorg stuff much.(My first tries) Although there are other function calls before
> SMI_MemMap(), I dont think they do in/out but I want to be consistent with general guidelines before
> doing it...

I don't know much neither on X driver, IIRC in the radeon driver
endianness is setup
during the initialization and in some bit blit functions. Thus i guess
you should setup this kind of stuff in the function that initialize
your card and then
you could have to do some more work, depending on case, in bit blit function.
Thought i may be wrong, advice from a true X hacker would surely be better then
mines which are made from a little understanding of how it works for radeon :)

Jerome Glisse