How to get involved with open source Radeon driver development (Newbie Friendly)

[marek]

4) How is work flow of development and testing drivers happening? (Dual boot, virtual machines, etc)

For testing OpenGL, we use piglit:

http://cgit.freedesktop.org/piglit/

[marek]

I'm sure there is a lot of people out there whose inner hackers wants to someday get involved with open source Radeon driver or Mesa development, but lack of introductory and scattered material makes it to look like a black box to them. So I'm asking developers involved with Mesa and open source drivers development to take a part in this discussion and share their experience.

We are going to assume that target audience (newbie) does already have a solid knowledge of C programming language.



1) First things first. What is usual set of documents, tool chain and software required to develop, test and debug drivers by developers?

That's up to you. The source code is your documentation. It's also very useful to have good knowledge of OpenGL, shader languages, etc. If you don't know OpenGL, I can't help you.

The hardware documentation is here, but good knowledge of OpenGL and/or Direct3D is assumed if you want to understand the 3D docs:

http://www.x.org/docs/AMD/

2) Where to access source code, fork and start developing?

http://cgit.freedesktop.org/mesa/mesa/ - OpenGL driver
http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/ - X driver
http://cgit.freedesktop.org/mesa/drm/ - some shared code between the two above
http://cgit.freedesktop.org/~agd5f/linux/ - Alex's kernel tree with latest stuff for Radeon
http://llvm.org/docs/GettingStarted.html - The shader compiler backend for Southern Islands lives in LLVM, so you might need that too.

3) What compiler switches and libraries should be used during development?

That's up to you.

4) How is work flow of development and testing drivers happening? (Dual boot, virtual machines, etc)

I either install the components with "make install" or make symlinks in /usr/lib which point to .so files in my home, so that I don't have to install anything. Drivers are just libraries, so I can compile them and use them immediatelly.

5) Is it mandatory to have access to multiple generations of hardware to get started?

No, you only need the card you want to work with.

6) How is Mesa and open source drivers source code structured? (Related directories and files)

mesa/src/mesa/main - the OpenGL front-end
mesa/src/mesa/state_tracker - the OpenGL middle-end
mesa/src/gallium/drivers/radeonsi - the hardware driver (API-independent back-end)

And there is also a lot of code in various directories of Mesa that the 3 parts above use.

7) How do drivers work internally? (Important functions and their calling order)

The source code is your documentation.

8) Where are the usual hot spots and how to profile the code?

You must know the hardware to be able to see where bottlenecks might occur. For CPU profiling, you can use e.g. sysprof.

9) What is usual coding conventions and styles being used?

It depends on the project. See the source code. Various parts of Mesa have different styles.

10) What kind of tasks should newbies start with? Is there any list?

Feel free to ask on the appropriate mailing list. [email protected] can be a good start.

11) Who and where to ask when in trouble?

Same as above.

[89c51]

I'm sure there is a lot of people out there whose inner hackers wants to someday get involved with open source Radeon driver or Mesa development, but lack of introductory and scattered material makes it to look like a black box to them. So I'm asking developers involved with Mesa and open source drivers development to take a part in this discussion and share their experience.

We are going to assume that target audience (newbie) does already have a solid knowledge of C programming language.



1) First things first. What is usual set of documents, tool chain and software required to develop, test and debug drivers by developers?

2) Where to access source code, fork and start developing?

3) What compiler switches and libraries should be used during development?

4) How is work flow of development and testing drivers happening? (Dual boot, virtual machines, etc)

5) Is it mandatory to have access to multiple generations of hardware to get started?

6) How is Mesa and open source drivers source code structured? (Related directories and files)

7) How do drivers work internally? (Important functions and their calling order)

8) Where are the usual hot spots and how to profile the code?

9) What is usual coding conventions and styles being used?

10) What kind of tasks should newbies start with? Is there any list?

11) Who and where to ask when in trouble?


Feel free to add your own if anything necessary is missed.

Go to the mailing list and ask for any introductory projects. All the rest will be answered once you start working on the code (from git).

Have fun.

[darkbasic]

We are going to assume that target audience (newbie) does already have a solid knowledge of C programming language.

Knowledge of C alone is from little to no help at all

[siavashserver]

Write something...