[Vorbis-dev] Flash Vorbis player

[Arek Korbik]

Hi Tor,

On Fri, Oct 3, 2008 at 12:53 PM, Tor-Einar Jarnbjo
<tor-einar at jarnbjo.name> wrote:
> Arek Korbik schrieb:
>>
>> http://people.xiph.org/~arek/bzr/fogg.dev
>>
>> The code is based on Cortado's JOrbis, hence the license is LGPL.
>
> Hi Arek,
>
> that looks great, at least the performance is much better than with my AS3
> approach for Flash 9. The sources are however not available on the link you

That is a bzr branch, which looks like an empty folder if you just go
there with a browser. With bzr installed do something like:

    bzr branch http://people.xiph.org/~arek/bzr/fogg.dev fogg

And that will checkout the sources under fogg directory.

> mentioned. If you have any interest in releasing your player under a less
> restricted license (Mozilla PL), I would be happy to share my AS3 Vorbis

I could definitely use other people's opinion on the licensing of that
code. I would like to make it as non-restrictive as possible,
obviously, but the code is pretty clearly based on cortado's jorbis,
which is lgpl-ed itself. Unless I go and do my own port from C, which
I actually started at some point - with Tremor, hoping to get a better
performance with non-floatingpoint implementation - I don't think
there is much I can do about the current fvorbis license. Though,
asking jorbis's authors to re/double-license their code would be one
option.

> decoder with you. Perhaps it's enough to replace the arrays I've been using
> with the new vector to achieve a similar performance boost with my code,

I think it could be useful for you to try. In my benchmarks Vector
access is ~4 times faster on reading and ~3 times faster on writing
then Array. And I think flash developers indicate that JIT compiler
might be able to do more with Vectors in less trivial code.

> although the Flash 10 player seem to generally execute AS slightly faster. I
> don't have much time at the moment to play around with it myself, but could
> of course give you a helping hand if you need.

That would be great. Have you thought or tried porting the Tremor to
java? Flash floating point operations seem to be 2-4 times slower than
integer ones. But my quick port of Tremor's mdct part showed almost no
speed gain. That could be due to the fact that tremor relies on
pointer pointing inside arrays, while in haxe I had to use more base
offsetting index operations to achieve similar effect, which I think
canceled possible speed improvements. But maybe that could be written
in a better way.

Arek