Re: [Gnash-dev] Re: Bitwise stream reading performance

[strk]

On Sat, Aug 25, 2007 at 12:35:36PM +0200, strk wrote:
> On Sat, Aug 25, 2007 at 12:05:54PM +0200, Udo Giacomozzi wrote:
> > For an introduction to this discussion, see
> > http://www.gnashdev.org/wiki/index.php/NinjaProfile
> ...
> > of a local file). Maybe it would be better to implement a 
> > internal buffered memory stream (sort of a read-ahead buffer for the
> > "stream" class.
> 
> Yes, I think we need this.

I committed a slight optmized read_uint and an additional read_bit()
for furhter optimization and profiling options.

Now, for the cache implementation:

gnash::stream exposes a get_underlying_stream, which is currently used
by some tag readers (embedded bitmaps usually). If we can remove any
use of the underlying file, cache can be enabled at the gnash::stream level,
otherwise we're stuck in doing it in the tu_file directly.

Now, doing it in tu_file might be good, in case some tu_file implementation
have caches themselves, to avoid duplication, and could also mean we might tweak
cache size based on input channel.

Finally, we might want to implement bitcache as a separate class, which could
be used by an higher level (say, shape reading).

To recap, here are the levels:

        - tu_file
        - stream
        - stream users

Where to cache ?

--strk;