Support and General Use > Audio Playback, Database and Playlists

Discussion: Decoder Plugin for non-realtime codecs and benchmarking!

(1/3) > >>

Davide-NYC:
To all DEVS: I wish to point out an untapped resource. The users!
Currently the forums contains a lot of "audio skips/hangs/stops" posts, as in: "My audio is skipping, why?".

Many of these posts can be individually addressed by simply telling the OP to turn off peakmeters or crossfade or whatever to reduce CPU load. But it does not address the core issue.

The core issue is codec optimization, right?

What is needed is a decoder plugin with a benchmarking mode,
      (see flyspray request here)
and a standardized sample set of files for testing,
      (see Twiki page here)
so that we could turn a group of confused/disgruntled users into a group of useful testers!

What we need is a decoder plugin so we can solicit easy testing from the users.

An encoder plugin would be beneficial too and seemingly easier since we already have two encoder plugins: wav2wv and mp3.encoder (which I have requested be renamed here)

Discuss!

{edit} I changed the title from "Request" to "Discussion" before I got yelled at by Llorean.  ;D

Llorean:
I actually won't yell as long as there's a feature request in the tracker, clearly linked from within the topic. :-P

linuxstb:
I'm not sure how useful this would be - we know Rockbox needs optimising on the ipods, and it's not just codecs - it's also LCD drawing, disk I/O, GUI code...  We don't really need testers, we need coders who can spend time improving things.

Davide-NYC:
I can give one concrete example of usefulness even if the testing is deemed unnecessary.

A user could, with this imagined plugin, decode a wma to wav and then listen to it in real-time.
Clearly this task is better served by a PC, but on the go, away from a PC in a pinch it would be great.

And here's another (less-concrete) example.
At what point is a codec deemed "fast enough" for a specific target? This would be easier to determine if a larger number (anything more than 1 currently) of users posted data on the matter in the wiki.

Of course I concede that this all might be futile, but it is an honest attempt, by someone that speaks no ASM and little C to find a way to contribute in concrete terms.

If this is deemed totally useless I will drop it, but I'd like the matter discussed at some length first just for confirmation.

The plugin is envisioned as an eventual 'transcoder interface' allowing users to convert from any readable (even non-realtime) format to any supported encoded format.  Those input formats are the currently supported ones plus WMA.  The output formats are MP3, Wavpack and WAV only.  I personally don't see the need to expand the number of output formats.

preglow:

--- Quote from: linuxstb on October 24, 2006, 05:04:26 PM ---I'm not sure how useful this would be - we know Rockbox needs optimising on the ipods, and it's not just codecs - it's also LCD drawing, disk I/O, GUI code...  We don't really need testers, we need coders who can spend time improving things.

--- End quote ---
Not all attempts at optimising, as I've quite often found out, work out as you'd think. For this reason I very much miss the ability to measure codec performance accurately (as we could with wav2xxx plugs). The only current alternative is looking at the boost factor, which I'm sure I don't need to explain is quite inaccurate and cumbersome ;)

Navigation

[0] Message Index

[#] Next page