ffmpeg (mplayerc) h.264/VC1 GPGPU decoding found
#16
ashlar Wrote:he consensus by whom? I, for once, I've never encountered a file that I wasn't able to play with software based decoding, while finding plenty of files that were not decodable by the hardware decoders.

Well, that's why it should be considered as optional into the XBMC options... to play using the GPU hardware or not...

just my two cents...

regards
#17
Sorry, didn't mean to sound like an ass. And maybe I did.

It's just that I feel that hardware acceleration is overblown in importance and that Popcorn hour is used by a friend of mine and the interface sucks big time when compared to XBMC.
#18
ashlar Wrote:Go use a popcorn hour and come back. The user interface is sluggish to say the least. Not even speaking about functionality.

Yeah, but the Video decoder(=core functionality) is done in hardware and that part is working like lightning.. IMO the power of XBMC should be: And a awsome+original interface ánd blazing fast and smooth real HD video.. I'm fully convinced XBMC can be the top mediacenter for years to come...

Quote:The consensus by whom? I, for once, I've never encountered a file that I wasn't able to play with software based decoding, while finding plenty of files that were not decodable by the hardware decoders.

Haha.. Only the smaller vendors like Intel, ATI and Nvidia, but that's been a hidden feature for some appearantly Wink But.. Again, XBMC on Windows can do both: Software decoding for almost every codec you can dream of and hardware for the more demanding parts.

I fully agree with fidoboy.. Maybe a fork like the "plex" project would be in place for the Windows XBMC also? However i pesonally think that XBMC should be made adaptable to the platform it runs on by choosing the right version of ffmpeg on architecture..
#19
us1111 Wrote:Maybe a fork like the "plex" project would be in place for the Windows XBMC also?
Go for it, the source is free for the forking (just make sure you understand the GPL). We've made it clear that our intentions are to remain platform neutral.
#20
us1111 Wrote:Yeah, but the Video decoder(=core functionality) is done in hardware and that part is working like lightning..
Surely not better than my $60 E2160 running at 3GHz.
Quote:Haha.. Only the smaller vendors like Intel, ATI and Nvidia, but that's been a hidden feature for some appearantly Wink
Surprise!!! Hardware vendors try to sell you extra hardware you don't need, never heard of it! Rofl
The hidden feature has been, for more than one year, how to accelerate everything, without caring for L4.1, subtitles rendering with this video renderer and only this video renderer, Reclock being unable to work with hardware acceleration, and so on and so forth.
Have fun with that, I'm sticking to hassle-free software decoding, thank you very much.
#21
Ok, thanks ashlar... you doesn't like GPU acceleration, all right. But may be that this feature could be VERY useful for many users with less powerful CPU chips or having problems with certain types of HD videos.... why not add it to XBMC if it can be done? As us111 has said, XBMC could be the best, by far, media center software in the world (much better than the buggy MediaPortal, the long time waited MeediOS, etc, etc)

kind regards,
#22
Fido, I have nothing against hw acceleration, per se. It's just that, so far, it's been confined to proprietary technologies and it's been full of limitation that software decoding does not have.
If I weigh pros and cons I'm all for software decoding. If there was an easy solution for hw decoding, great. But there isn't. It's not as if this question hasn't been asked ten times or more.
And a fast processor, as I've demonstrated is now cheaper than ever and will get faster and cheaper as time passes. In a year's time, quite likely, entry level dual core will have the performance needed to decode 1080p without too much effort (it's already the case if you overclock). That would leave hw acceleration exactly where?
#23
yoshiofthewire Wrote:The code in question was offered by the developers of ffmpeg for backporting into the main codebase. The code was REJECTED by ffmpeg for not being cross platform.

Hence why it won't make it into XBMC. No

I'd point out that the XBMC developers can and HAVE pulled "fringe" code from ffmpeg into XBMC in the past. They do not stick to the main ffmpeg stuff only and seem to be pretty open minded. The CABAC patch is a good example - I think if you look it was first rejected by ffmpeg guys but incorporated into XBMC - to great effect!

Ashlar does have a point though - some of the hardware decoding stuff for the GPU does seem awful profile dependantAngry Hopefully the OpenCL stuff just announced will be helpful but it's only a set of standards starting it looks like and will be even longer along than the existing ffmpeg patches for Linux acceleration I fear. At least it's multivendor...
Openelec Gotham, MCE remote(s), Intel i3 NUC, DVDs fed from unRAID cataloged by DVD Profiler. HD-DVD encoded with Handbrake to x.264. Yamaha receiver(s)
#24
ashlar Wrote:Fido, I have nothing against hw acceleration, per se. It's just that, so far, it's been confined to proprietary technologies and it's been full of limitation that software decoding does not have.
If I weigh pros and cons I'm all for software decoding. If there was an easy solution for hw decoding, great. But there isn't. It's not as if this question hasn't been asked ten times or more.
And a fast processor, as I've demonstrated is now cheaper than ever and will get faster and cheaper as time passes. In a year's time, quite likely, entry level dual core will have the performance needed to decode 1080p without too much effort (it's already the case if you overclock). That would leave hw acceleration exactly where?

I would leave it probably here:

Quote:The BBC is taking part in trials of a new high definition technology, Super Hi-Vision, which is 16 times the resolution of standard HD.

Super Hi-Vision, or Ultra High Definition (UHD), provides 7680×4320 resolution images and 22.2 channel immersive audio.

But actually I'm gonna withdraw from this conversation.. Don't like the "non-constructive mode" of Ashlar..
#25
us1111 Wrote:But actually I'm gonna withdraw from this conversation.. Don't like the "non-constructive mode" of Ashlar..
Huh Non constructive? I'm telling people to forget about a technology not supported by the program on whose forum we are discussing because there are viable alternatives that could be actually better for a variety of reasons... and I'm non-constructive?

Also, concerning super high resolutions, maybe you overlook the fact that 1080p takes a theoretical 75+" screen for you to be able to fully resolve the detail at 10 feet of distance (which is a relatively average seating distance, I'd say). To fully resolve *that* resolution at 10 feet you would probably need a movie theatre sized screen. I'd say it's a non problem for home use, really.
#26
us1111 Wrote:But actually I'm gonna withdraw from this conversation.. Don't like the "non-constructive mode" of Ashlar..

Don't leave...it's refreshing to see someone else picking the good fight.


My 2 cents to Ashlar...simply because you don't like HW acceleration, it does not mean others do not have very good reasons for liking it.

Want a good reason? The environment.

The user in question has said that GPU acceleration would enable him to keep 3 machines in use. That's 3 machines not ending up in a landfill in China where they leach toxic chemicals into kid's drinking water.

As for myself, I'm currently in the process of creating the lowest energy consumption integrated Home Theater system (for real world release). For audio playback, I've hit on XBMC with an Iphone/N800 remote for music playback without a screen. As for video, I'm not through with my calculations, but I predict that GPU acceleration could save 15 watts idle power and 60 watts during playback.

The math is simple and it does lead to GPU acceleration. Beyond energy savings for mainstream systems, a cheaper processor could mean a cheaper motherboard which could mean a more affordable low energy unit which could mean more people saving power.

I strongly encourage XBMC developers to consider a higher principle than 'cross platform development' when considering GPU acceleration. While software decoding is more versatile, the reality is that focusing on doing a few formats (mkv w/ h.264) better (great GUI + gpu acceleration) would serve more people. Of course, that argument is moot considering that it's not either/or...software decode can certainly live alongside gpu acceleration.

Of course, the big trump card here is Open CL. Considering the state of Plex and Apple's own implementation of Open CL, it looks like XBMC on Linux/Windows will yield the performance and innovation crown to Plex/Apple quite soon. That's a pity as Apple is clearly not concerned enough about low power consumption - and I don't like Apple's closed ecosystem.
#27
That will be a power saver for sure. Listen to music across the airwaves from a media center on a tablet or Iphone. Why not just download the music using a usb cable onto 8 GB+ sd card to carry around with you for days and days and days. Save plenty of power, oil, coal, etc. Smile
#28
speed32219 Wrote:That will be a power saver for sure. Listen to music across the airwaves from a media center on a tablet or Iphone. Why not just download the music using a usb cable onto 8 GB+ sd card to carry around with you for days and days and days. Save plenty of power, oil, coal, etc. Smile

Because the Iphone doesn't play lossless 96/24 multichannel flac through a high efficiency surround sound amp does it?

The point is to offer an energy efficient platform that offers minimal compromise.
#29
I don't understand why people are still bringing this discussion up in new threads. It's already in like 4000 threads. I just don't see any kind of upside to having the gpu decode the video. Why not just invest in a better processor? If developers want the support they will prolly add it. I think they have heard about it by now!

I really don't understand what the energy saving is about? And plz don't say "saving the enviroment"!
#30
dmidi Wrote:Of course, the big trump card here is Open CL. Considering the state of Plex and Apple's own implementation of Open CL, it looks like XBMC on Linux/Windows will yield the performance and innovation crown to Plex/Apple quite soon. That's a pity as Apple is clearly not concerned enough about low power consumption - and I don't like Apple's closed ecosystem.

OpenCL is interesting but someone has to write the routines to adapt OpenCL to video decode algorithms. OpenCL is not aimed at that so you will not see Apple or other major players adapting it for video decode.

Only if Plex uses QuickTime player as an external application for video content decode. The hardware acceleration is currently not available outside of QuickTime Player. Otherwise, it's a level playing field with everyone using ffmpeg. Linux with nvidia hardware will get the first performance boost with VDPAU.

And Apple is concerned enough low power consumption, it's called sleep and it works quite well. The AppleTV is the only product that does not sleep and that's due to the power supply which is just a plain switching power supply.

We will see tomorrow if Apple pops out some new toys Smile

Logout Mark Read Team Forum Stats Members Help
ffmpeg (mplayerc) h.264/VC1 GPGPU decoding found2