From: Marcelo E. M. <mar...@bi...> - 2001-08-28 13:57:10
|
>> "Mike A. Harris" <mh...@re...> writes: > Is this something that is not implemented due to technical > complexity, or is it something that just hasn't happened yet due > to more interesting projects, improving Mesa, etc..? That. > What all would it take in order to get this going? The DRI design document gives a suggestion regarding a possible implementation method. > It would be at least a semi-useful feature to play with, and I'm sure > many out there would probably get great usage out of it for some > interesting projects. Indirect hardware accelerated rendering is interesting in the case when the client and the server are *not* the same machine, and in that case there's a fair ammount of stuff that will be slower besides GLX rendering (e.g., you are sending all sorts of events across the network; you can't use MIT-SHM anymore, and then you find yourself not only pushing pixmaps and images over a socket but also across the network[0]). It made lots of sense ten years ago, back when a 10x increase in the hardware cost actually meant 10x as many features and power. Today it's probably reduced to the case of "we can't afford to port it (or the possibility doesn't even exist) but we still need to run it". > What would need to be done to get this going, perhaps as a 'copious > spare time' project? Something simple, or something fairly difficult > involving deep OpenGL/DRI knowledge? Actually it's one of the simple open issues. You need to understand how the X protocol works, the differences between the direct and indirect cases and how that affects GLX, the requirements GLX imposes on the client and the server in the indirect case. And you need to understand some details about how the DRI is actually implemented (which, incidentally, I've found FBDRI makes a lot easier). > What approximate timeframe (guestimate) do you think it would > take to implement this to play with? I'm bad at that kind of guessing. I'll just sit and wait for other opinions. Marcelo [0] Which is a charge I proudly confess being guilty of, in a different context. |