Marek Continues Baking More Mesa Optimizations

[phoronix]

Phoronix: Marek Continues Baking More Mesa Optimizations

AMD developer Marek Olšák continues leading their recent charge on looking to boost the performance potential out of Mesa / Gallium3D...

http://www.phoronix.com/scan.php?page=news_item&px=Marek-Minor-Mesa-ST-Opts

[pewspewpew]

Well, even if it did not budge framerates for me, I believe it is more important to go the way of optimizing existing things, than adding something new just for the sake of it. For some people(like me), current state is important factor to choose the next device to buy, and when the time comes I would not want to risk buying something, that will work equally good comparing to competitors on new games, but will be unplayable with loved and owned 5+ year old ones.
So big thanks to marek for not forgetting the good oldies, this gives me hope.

[Azpegath]

I think I need to write a song called "I love Marek, my hero and saviour". ;-)

[jf33]

I'm currently working on my private game engine and one thing that would greatly improve performance is support for GL_ARB_bindless_texture. Unfortunately, it's not yet implemented in mesa. There is a branch by Dave Airlie

https://cgit.freedesktop.org/~airlied/mesa/log/?h=arb_bindless_texture

The last activity is nine months ago. When can we expect the extension to get done? airlied

[dungeon]

Next week will be my next Mesa Git comparison with RX 460/470 coverage.

Hm, from where that bolded one coming from?

[Drago]

3D and especially 3D drivers are tough nuts to crack. Lots of things to know: maths, memory workings, cache workings, parallel computing, ISA pipelining. I would recommend enter some other area with lower entry level. There are plenty of places one can help Linux ecosystem. Polishing desktop environments and making them even more user friendly is honored and more easier, still helping people to embrace Linux. Hell, even nice unified GUI for GPU setting/overclocking would be nice.

[bridgman]

There's a school of thought that goes the other way as well, basically "if you're going to do anything with any 3D API then learning some OpenGL isn't a bad first step, and learning OpenGL lets you cherry pick exactly the math & HW functionality you need to get your bearings".

Trying to program 3D drivers without being comfortable with the 3D API (or at least *a* 3D API is really hard though.

[imirkin]

Trying to program 3D drivers without being comfortable with the 3D API (or at least *a* 3D API is really hard though.

Dunno, I became comfortable with the OpenGL API precisely through programming 3D drivers (and trying to figure out wtf failing piglits were doing, and much later, writing a few of my own).

I believe the key to getting involved in any project, including mesa, is having a concrete goal in mind.

Looking for "easy" tasks won't work too well -- either those tasks are important, in which case they would have been done by now, or they're low-use, in which case you won't be too motivated to do them, no one will be motivated to review, etc.

[mlau]

Replacing for example linked lists with maps will boost certain parts of Mesa. It is easy - it hasn't been done.

What's stopping you?

[Adarion]

RX 460 Hm, from where that bolded one coming from?

I was wondering the same. Since release was scheduled for today (Mo., Aug. 8th) it nearly sounds like he is already in possession of an exemplar. But that would mean someone sent him one already. That would be great. Means: Early Linux-related coverage. Seeing how the free driver works on that one.


getting into development
I don't know. I am "a bit married" to computers for 25+ years, but hardly got around to development. In the early days, when I still had time and interest (I was 9 - 16) years, I lacked money for compilers / environments, books. Well, I did some BASIC on elderly machines (Warshaw union / soviet block processors), later on i486 and also a little Pascal. Tiny amounts of QuakeC, but it was more reading things, trying to understand and modifying existing stuff. Late in school also one scientific project, even with little animated graphics in 320 x 200, but in BASIC. Later I had no time (university) to get myself into serious coding, but at least I'd have a gcc then. Trying to get some C done on AVR-microcontrollers now; to make LEDs blink. Oh, well.
I guess one needs a lot of time, constant practise and the right synaptic connections to be close with maths and logic, good books, a compiler and probably also a good tutor (well, GSoC is one option, but I guess it is targeted at already "grown" students of coding). I haven't given up hopes totally, but I also found different ways to aid the free software movement: Donations, helping out at booths, organizing Linux events, or e.g. doing catering for these events, advertising the use of freedom software, advising people, helping people with installtions or system care.
And besides, it needs coders on this world, but it also need bakers, plumbers, rocket scientists (okay, those are probably coders, too), medical staff, chemists, car engineers, farmers and so on. So everybody can find a field to contribute something to the advancement of the world.