[MPlayer-dev-eng] NUT pts/startcode optimization proposal

[Ville Saari]

On Sun, May 02, 2004 at 02:57:18PM -0400, D Richard Felker III wrote:
> >From the nut spec:
> 
> coded_timestamp
>         if coded_timestamp < (1<<msb_timestamp_shift) then its a
>         lsb timestamp, otherwise its a full timestamp + (1<<msb_timestamp_shift)
>         lsb timestamps are converted to full timesamps by:
>         mask = (1<<msb_timestamp_shift)-1;
>         delta= last_timestamp - mask/2
>         timestamp= ((timestamp_lsb-delta)&mask) + delta
>         a full timestamp must be used if there is no reference timestamp
>         available after the last frame_startcode with the current stream_id

Is there some reason for the definition of the lsb timestamps to be so
complicated? It gives the exact same range of lsb-encodable timestamps
as would the much more obvious definition:

   bias = -((1<<msb_timestamp_shift-1)-1)
   timestamp = last_timestamp + bias + timestamp_lsb

In this case "delta" would be more appropriate name than "lsb".

A more useful range could be accomplished by adding some
educated guess of the delta to the bias. For example the median of
initial_timestamp_predictors. Or would the average of minimum and
maximum initial_timestamp_predictor be better?

-- 
 Ville