[XviD-devel] dev-api-4: plugins api

[suxen_drol]

On Sat, 15 Mar 2003 16:08:40 +0100 (CET) Christoph Lampert <chl at math.uni-bonn.de> wrote:

> On Sat, 15 Mar 2003, Edouard Gomez wrote:
> 
> > suxen_drol (suxen_drol at hotmail.com) wrote:
> > > plugins api progress is coming along smoothly, in addition to general
> > > dev-api-4 bug fixing and house cleaning. i've included two plugin samples
> > > in the /xvidcore/src/plugins directory:
> > > * xvid_plugins_psnr: which prints sse to stdout.
> > 
> > Would it be better to return 3 floats so every application can format it
> > the way it wants and more important, don't rely on console output (i
> > think of win32 users) ?

plugins are not really intended for console io. this was just an example.
after chatting with ed, i am conviced libxvidcore should continue to
calculate psnr internally, instead of a seperate plugin.

currently, xvid returns "long sse_[y,u,v]". why didnt we use floats for
this? i presume to keep the api stricly integer-only.
is there any reason why we cant use floats, instead of raw sse values?

> Sorry, I didn't check plugin API, but shouldn't two parameters of the
> call to plugin be void* pointers for input and output, which is then
> read from (e.g. a filename) and/or filled with output data, e.g. PSNR
> values? That way, we can have arbitrarily much input and output (or none).

plugins can be passed an pointer at encoder creation time. the pointer
could point to a plugin-specific configuration structure.

On Sat, 15 Mar 2003 15:36:25 +0100 Edouard Gomez <ed.gomez at free.fr> wrote:
> Fine :> this removes a bit more of code from xvid_encraw/xvid_decraw

again its only an example, using hardcoded filenames "orig-%05i.pgm"
and "mpeg-%05i.pgm".

-- pete; life is like a box of ammo