The philosophy (and technicalities) behind GPU acceleration
#1
Question 
This is a reply to moderator shut down of a GPU acceleration thread...
http://forum.xbmc.org/showthread.php?tid=41530&page=3

<Begin>
You shut down a good discussion on why GPU acceleration is important. Your reasoning is that the original post (GPU assisted ffmpeg decoder) was no longer being discussed.

But it was...

Perhaps you should consider that good conversations evolve. While there was a touch of philosophy in the posts, there was also technical discussion related to GPU acceleration via different methods. Other than 2 users ranting about 'no gpu acceleration', all other pro-GPU acceleration posts discussed DXVA, Open CL and other render methods while touching on environmental benefits of energy savings.

Perhaps I'm being oversensitive but there seems to be this fanatical opposition to GPU acceleration by some members. Is it because they can't code it or simply have this puritanical insistence on all code being cross platform?

My two cents say that neither reason for blocking GPU acceleration is adequate given the benefits that could be made available to users by enabling GPU acceleration. Insisting on CPU upgrades for better software decode is bass ackwards...that's the Microsoft/Intel/CE industry way.

The Linux/OpenSource/XBMC/Nirvana way should be all about harmony between software and hardware...and saving little kiddies in China from E-trash caused by too-early functional obsolescence, coral reefs from global warming, and your hdd from high case temps caused by inefficient software decoding.
Reply
#2
Please don't thread jack a feature request with this crap. If you want this somewhere else, post as such and we'll move it.
Reply
#3
DXVA is not the only way...

Firstly, XBMC is not the only 'player' out there that would see the benefit of cross-platform Open CL routines for video decode.

However, considering just how puritanical the XBMC developers are about cross platform code (until something like VDPAU pops up for Linux...ahem) then it would stand to reason that Open CL gpu acceleration code would be adopted by XBMC. And if the code would be adopted by a major open source project, then someone will write it for all the glory (and wealth and women) that such a contribution would bring.

Are those safe assumptions?

To summarize, Open CL seems to be the best way of pushing GPU acceleration forward:
1. It's cross-platform and has wide support
2. It's technically feasible
3. Did I mention it's cross-platform and doesn't involve Microsoft?!?

If anyone says otherwise, let them step forward (better if they have a few code snippets)

I could have avoided all of these useless postings by simply writing the code myself. Alas, to quote an old country song:

'I shoulda been a programmer...
shoulda learned to code while high...
Wearing my dockers, riding my ten-speed to DEFCON five,
Stealing young girl's hearts
Just like Linus and Roy,
Singing those open source songs...
Oh I shoulda been a programmer"
Reply
#4
Deleted
Reply

Logout Mark Read Team Forum Stats Members Help
The philosophy (and technicalities) behind GPU acceleration2