Re: [Dri-devel] Status Dual Head & acceleration

[Daryll S. <da...@va...>]

On Fri, Jul 21, 2000 at 12:01:04PM -0700, Patrick Ritchie wrote:
> I just got the DRI runnig on my box. K63-450, g400max... it works great!
> ~44fps in Quake3 demo1 (i'll be doing some more comprehensive benchmarks
> soon).

People have to realize how large the scope of work is with respect to
3D. It really is huge, and sometimes things that seem little are really
big projects. We've made a remarkable amount of progress in a very short
time, but there is still huge amounts of work to do.

> I was wondering what I could do to help with the proejct. Dual-head is
> something that really appeals to me so this might be a good place to
> start, unfortunately my coding experience isn't that great in this area.
> Could you point me to some good areas to start learning? What other
> things can I do to help out, like running benchmarks, testing various
> features etc?

For any level of user:
    Run benchmarks and post good detailed bug reports. Every bug should
    be in the bug database so it doesn't get lost, so we know where we
    stand, and so we have all the right data to find the problems.

    Write documentation. Pick anything and write some docs. Build
    guides. Performance tweaking. Interfaces. Anything you understand.

For the developer:
    Work on some of the bugs in the database. Usually these are fairly
    self contained.

    Find a feature that isn't supported and work on it. For example,
    some code path might not be implemented in hardware.

    Do some performance testing and see if you can isolate performance
    bottlenecks. Be careful about correctness.

    Write some OpenGL test cases for Glean. This isn't really DRI, but
    it is really important. Glean is an open source OpenGL test
    suite. We need as many tests as possible to make our versions
    robust.

For the senior developer:
    Look at the DRI 2 document, and see about some of the more difficult
    architecture problems. 

People always ask what they should work on. The answer is always the
same: whatever you think is interesting. If you've got a multihead setup
and you think it would be a cool feature to work on great. If you're
using a CAD program and antiliased lines don't work well on your
hardware, tackle that. You're going to do a better job and enjoy it more
if it is important to you.

Start by digging in the code a little. In the case of multihead it is an
X server issue, so you'll be looking at the 2D X
drivers. (xc/xc/programs/Xserver/hw/xfree86/drivers/...) See if you can
figure out what's going on in there. When you get an idea, work out a
plan for what you think needs to be done. Post it here. If other people
think it is interesting they'll jump in. We'll do what we can to look at
it and comment. When you run into questions post them here. Hopefully
the same thing will happen. We can't say how involved we'll be with
these various projects, because it rather depends on how busy we are,
but the way to grow the community of knowledgable people is to get the
discussions going.

						- |Daryll