On 26 September 2011 19:50, Christian Robottom Reis kiko@linaro.org wrote:
On Mon, Sep 26, 2011 at 03:21:06PM +0300, Ilias Biris wrote:
- libpng optimization
I've got interest in this one -- can anyone articulate potential wins?
http://lists.linaro.org/pipermail/linaro-multimedia/2011-September/000074.ht...
- Better video rendering integration in UI, like X11 proposal for wayland, extended dri2
Needs describing.
Whatever is ends up being, something needs to be done about video rendering. What we need is a solution providing 1) zero-copy, 2) hardware overlays, and 3) window system integration. Current interfaces allow any two of those but not all three. This ties in with ongoing UMM work.
- Audio DTS decoding - could be tricky, involves legal aspects which need to be carefully looked at, already done in libav? - Compressed data sound support (as in http://www.linuxplumbersconf.org/2011/ocw/proposals/633)
Are these not the same thing? Or is the first "Optimized Software DTS decoding"?
Yes, I believe the first was referring to software decoding. A DTS decoder exists in libav already, and it is mostly NEON optimised. The remaining parts are only relevant in some specific situations.
The second item would be more correctly described as "hardware- accelerated audio decoding". I'd love to see some more detail than the link above provides.
- Realvideo on ARM (popular in China) - needs optimization for 720p playback, VGA is ok
Needs describing.
Libav has decoders for all Realvideo formats, but they are not NEON optimised. Patches with some optimisations have been submitted and are awaiting review.
- armv6 optimizations for vp8 in libav
A first patch for this already exists, but it needs some work to become acceptable for mainline libav. This should take a few days, a week at most.
- 10-bit h264 optimization needed in libav
Current ARM cores are still too slow to handle 10-bit H264, so I consider this low priority for now. Once Cortex-A15, or even upcoming A9 chips, hit the market, this is something we will need.
The work consists of adapting existing NEON code to operate on wider pixels. For this fairly large piece of work, I estimate a few weeks, maybe a month. Not all of it has to be done at once, however. It would be possible to convert only the most important functions first and defer the less-used ones until later.
- audio library optimization
I believe this refers to libraries like libsamplerate used by ALSA and pulseaudio.