[haiku-development] Re: gsoc: Gallium 3D infratructure + Ext4 lkl-linux cleanup + integration
- From: marius adrian popa <mapopa@xxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sat, 9 Apr 2011 13:31:33 +0300
On Thu, Apr 7, 2011 at 10:59 PM, Urias McCullough <umccullough@xxxxxxxxx> wrote:
> On Thu, Apr 7, 2011 at 8:15 AM, marius adrian popa <mapopa@xxxxxxxxx> wrote:
>> I will try to cleanup tomorrow and put the timetable, i will be
>> offline after that
>
> AFAICT, this is the first time anyone here has heard from you here -
> so I would like to suggest you perhaps take a look at our student
> guide:
>
> http://www.haiku-os.org/community/gsoc/2011/students
>
> Thanks!
>
> - Urias
>
Could you please allow modifications to gsoc proposal ? so i can add
the timetable and better description
Sorry for the rush i had to decide myself what is doable in that time frame
I will review my two proposal today and fill the gaps and mistakes
I wanted to introduce myself before : so here is the short intro just
in a few words
I was invisible and only as end user mostly
http://mapopa.blogspot.com/search?q=haiku
I do compile my haiku directly from source and as many might know I
have proposed to solve the gallium3d bounty
http://haikuware.com/bounties/gallium
For the moment i want to reuse the code from the linux kernel
in my case radeon/nouveau driver with glue to make it work
It might be a heavy challenge to finish it until the deadline
so this is why i proposed to only bring it to a compilable state parts
of the whole infrastructure
The glue parts can be reused in further upgrades (new linux kernel
stable version with new radeon,nouveau,intel drivers)
Later after all works it can be possible to get to writing clean haiku
3d drivers
but that is a heavy burden and few takers to take
as someone on the kernel mailing list said : the most complex part in
your computer is the video card and i agree on that
I only scratched the surface of voodoo card yesterday (implemented as
software acceleration
http://vogons.zetafleet.com/viewtopic.php?t=25606 )
and is quite hard to understand the hardware parts (3d graphics are not easy)
So yes i want to contribute more in my free time this is why i thought
so much of what can be done in this timeframe
Yes i can do the python/django too but i better contribute with
something that i didn't do before
kernel development
and some on my todo list is better driver for qemu (cirrus)
and who knows maybe a 3d one (after there is a 3d one in qemu)
Understanding kms
http://www.botchco.com/agd5f/?p=50
http://www.botchco.com/agd5f/?p=52
http://revolf.free.fr/FOSDEM/2010/FOSDEM2010_ALTOS_007_Gallium3D.pdf
Another history of the Xserver evolution
http://lwn.net/Articles/434816/
http://linuxconfau.blip.tv/file/4693305/
There is one image with the full Linux graphics stack ~ middle of the movie
Also helpful
http://www.kernel.org/doc/htmldocs/drm/
Same works in hurd
http://wiki.x.org/wiki/Hurd_Porting
Other ideas is to reuse the code from lkl (passing the heavyweight to
linux side)
There are some troubles with interrupts and dmi …
Other related posts:
