An Interview With the Developers of FFmpeg
An anonymous reader writes "Following the long-awaited release of FFmpeg 0.5, Phoronix has conducted an interview with three FFmpeg developers (Diego Biurrun, Baptiste Coudurier, and Robert Swain) about this project's recent release. In this interview they talk about moving to a 3/6-month release cycle, the criteria for version 1.0, Blu-Ray support on Linux, OpenCL and GPGPU acceleration, multi-threading FFmpeg, video APIs, their own video codecs, and legal challenges they have run into."
how do you like eating my asshole?
The versions will have to be numbered better if they play to have a release cycle. "0.5" just doesn't work.
"Following the long-awaited release of FFmpeg 0.5"
Define "Long" and who was waiting for something that is still ".5" beta?
What the hell is FFmpeg anyway? Oh, it is just another CODEC library. Yawn!
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
They say they have trouble implementing code due to time and motivational constraints. Seems to me that if Google/YouTube has built a very successful website around the FFmpeg engine then Google ought to pony up some programming help with improving the project. Wouldn't that be "The right thing to do" (TM)?
Anyone know what kind of performance increase VDPAU can give compared with software decoding?
That application I'm working on needs to decode lots of H.264 streams so being able to offload that to a GPU would be a godsend.
for specialized pr0n. Guess not
hall of shame
Pretty good, you can easily decode 1080p H264 video with a cheap nvidia card. It is also consume less power and cooler than using CPU for decoding, so VDPAU = great for HD video on laptop.
I also find it interesting that VDPAU can help decoding H264 video that I can't decode using its counterpart on Windows (DxVA) :)
ffmpeg is a godsend to the FOSS community. Great work to the ffmpeg developers and I know I speak for a lot of people when I say that I'm looking forward to the fruits of their hard work.
X.org has posted some project ideas for Google Summer of Code projects, including "VDPAU state tracker for Gallium. Admit it, it would be pretty cool."
--
make install -not war
In the "Legal Issues" section of the article, I expected to see something about the issue of FFMPEG potentially infringing on existing patents. Instead there's just some stuff about violating the GPL. Seems like a major oversight to me.
What *is* up with the patent issue? Is it possible to use FFMPEG legally in commercial software if you adhere to the GPL and buy licenses for the patents that you're using? Since I already paid for software that includes encoders for some of these patented codecs, does that allow me to legally use FFMPEG? In my lifetime, how many times do I have to pay the patent fees for MPEG2 encoding?
Once again, using GPUs for graphics/video acceleration is not GPGPU. It is just plain hardware acceleration. Stop using this phrase if you don't know what it means.
To anyone involved in maintaining ports of third-party software to a particular OS (or even a distro), working with a release of that software is quite important. Unfortunately, ffmpeg-developers couldn't be bothered with such things. Here is a rather arrogant response I got two years ago from them:
He further explains their stance as:
If, indeed, this idiotic view has been retired for good — great!.
In Soviet Washington the swamp drains you.
The correct title should be: "comments about some (lost?) interview to three ffmpeg developers". I'd like to read the real answers of the developers (and the exact questions), and not that kind of "educational explanation on their supposed responses".
BTW, some months ago I indirectly had to use their software (because of some web portal software related to video conversion.) From my little experience:
1) I found that they have a lot of bugs being accumulated and don't have the needed workforce to analyze nor resolve them
2) Too much effort (and attention priority) is put in the hall-of-fame and license violation "prosecution"
3) As they enforce the (IMO) confusing LGPL, several closed-source developers had passed a very bad time. Some were (again IMO) not trying to steal code, and just misinterpreted the weird licensing terms.
I'm posting anonymously because I'm waiting the resolution of a couple of yet "unassigned" bugs for more than two months and don't want their reprimands for the comments:)
I have some notes on my efforts to compile ffmpeg with all available codecs (using PLF repositories) for Mandriva 2009.0 on my blog, MaximumHoyt.com. Comments appreciated.
"I believe in Karma. That means I can do bad things to people all day long and I assume they deserve it." : Dogbert
-O3 and -funroll-loops are likely to increase code size, and thus increase cache pressure. Particularly for H.264 this in our limited experience is more likely to slow things down that speed them up. Generally it is a bad idea to "speed-optimize" without benchmarking.