[gmpi] Re: Reqs draft

  • From: Steve Harris <S.W.Harris@xxxxxxxxxxxxxxx>
  • To: GMPI list <gmpi@xxxxxxxxxxxxx>
  • Date: Sat, 8 Nov 2003 13:59:13 +0000

On Fri, Nov 07, 2003 at 12:40:48 -0800, Tim Hockin wrote:
> * How are plugins discovered on the system?
>   - registration required in a central DB?
>   - a directory that must be scanned?

I dont like the idea of a db that might get out of sync, but it might make
parsing plugin metadata faster - is this a serious concern?

Filesystem based discovery seems to perform acceptably, and is easier to
implement.

> * What info/services must the host provide at instantiation vs. runtime?

sample rate for one.

> * What must the plugin provide the host at instantiation?

port descriptions, inplace-safeness (if we go that route)

> * How do wrapper plugins (e.g. VST in GMPI) fit into this?

They should be transparent to the host, surely.
It probably requires that one physical plugin (.dll, .so, whatever) can
contain morte than one logical plugin.

> DSP state
> -------
> * When must/may the host reset a plugin?

I'd like to see reset on sample rate changes.

> non-FPU systems
> ------------
> * Do we REALLY want to support non-FPU systems?
>   - means we need to have float and int sample types
>   - means 'real' parameters can not be float
>   - means that anywhere we want to use a float or double needs to change

I'm slightly concerned that we dont have a good enough grasp on the
requirements and systems (I know I dont).

Defining a GMPI-based plugin system for non-FPU systems could be left as
an exercise for future working-groups - if the demand is sufficient. The
important point is that the nomenclature and APIs are similar, for
familiariaty. I dont believe we will never have (practical) source-level
portability between int and float based systems.

> * If we do support it -- profiles -- what falls into a profile?
>   - samples (int16 or int24 or 'real')
>   - definition of 'real' and operations for reals

To retain any kind of parrallel with a float based dsesktop system the
reals would have to have headroom, eg. 16.16 fixedpoint.

I think if real is well enough defined in the profile then the operations
do not need to be defined.

- Steve

----------------------------------------------------------------------
Generalized Music Plugin Interface (GMPI) public discussion list
Participation in this list is contingent upon your abiding by the
following rules:  Please stay on topic.  You are responsible for your own
words.  Please respect your fellow subscribers.  Please do not
redistribute anyone else's words without their permission.

Archive: //www.freelists.org/archives/gmpi
Email gmpi-request@xxxxxxxxxxxxx w/ subject "unsubscribe" to unsubscribe

Other related posts: