[Openal-devel] Faster buffer model, and more random stuff

Chris Robinson chris.kcat at gmail.com
Mon Mar 3 13:28:22 PST 2008


On Monday 03 March 2008 03:35:21 am Sherief N. Farouk wrote:
> >>: For the love of whatever gods people worship, I need to make this point
> >>: clear:
>
> The proposed MapBuffer NEED NOT RETURN A POINTER TO THE ACTUAL BUFFER IN
> INTERNAL FORMAT.

And as I've been saying, this is not how map buffer should behave. It 
shouldn't be a guess if you're getting the memory pointer to the buffer or 
not. Mapping a buffer should give you a pointer to the data it holds.. that's 
the point of it, so you can read/write the memory directly.

Obviously this won't necessarilly work on AL buffers, which is why a 
new "sample buffer object" would be used. So you then have three 
possibilities:

Normal users taking the easy (possibly accelerated) path: alBufferData
Users who have/want strict memory requirements: alBufferDataStatic
Users who want to use possible quick buffering and acceleration:
alMapBufferObject/alUnmapBufferObject/alBindBufferObject/alBufferData

We could add your locking-like mechanism, but I personally think it'd be more 
confusing (both stand-alone and when put side-to-side with mapping).


More information about the Openal-devel mailing list