[mythtv] [mythtv-commits] mythtv commit: r24386 - in branches/mythtv-hdaudio by jyavenard

[Jean-Yves Avenard]

Hi

On 5 May 2010 09:48, Mark Spieth <mark at digivation.com.au> wrote:
> yes.
> also AVSync in NVP can be improved heaps so estimation of avdelay is
> accurate to 1ms all the time.
>
> also is there any intention to allow support of ARM based system without
> floating point coprocs?

Support for those is there, and there should be no impact on
performance on those systems either (or more accurately, it won't be
any worse)

I explained the matter in the commit:
http://svn.mythtv.org/trac/changeset/24386

When no audio processing is performed, that is: no timestretch, no
upmixing, no resampling etc..
the audio samples stays as fixed accuracy number and pass as-is to the
audio card.

I had this discussion with Danielk a few weeks ago on the matter, and
considerations were made to not impact negatively some hardware
architecture.

In the worse case scenario, it's just as fast (or as slow) as the
previous code would have done.

It is however, unreasonable to expect some platforms to be able to do
upmixing, resampling etc though...

On X86 (both 32 and 64 bits) platform however, the speed gain is up to
an order of magnitude