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

[strk]

I reopen this thread as I think it might be time to take care of it...

On Tue, Aug 28, 2007 at 11:22:55AM +0200, Udo Giacomozzi wrote:

> The "stream" class knows the size of a TAG, right? So it could load
> into memory the whole tag (unless direct access is desired).
> 
> I don't expect anything larger than 64k so no big deal. And even if,
> we could still optimize that.

To recap, we're trying to improve parser performance,
and found bitwise reading being one of the problems.
I'll add to the above that also *bytes* reading is a bottlneck
if every byte read needs a call to the underlying I/O functions.
In general, reading from underlying stream in chunks would give 
dramatically improved performance: reading in chunk from stdin reader
improved speed by more then x15.

Where and how to perform the chunked reads is being discussed.

Udo suggests to read tag-long chunks in gnash::stream,
to avoid touching any other code. I'd second this.

Things to take care of would be:

        1) big tags (split reads in chunks?)
        2) interpreters needing chunks of data
           (mem copies or access to underlying cache?)

--strk;