On Thu, Jun 12, 2003 at 11:37:28 -0700, Tim Hockin wrote: > > There is no evidence that we need more than float on desktop CPUs, modulo > > high bit width dithering. I dont think that thats significant enough to > > warrant the extra complexity in all other plugins. > > OK, and you agree we DO need to support Integer types? If we support 2 > types, we support N types, in my book. Not ant once, I think its better to specify a profile where the gmpi_sample type is int24 (or whatever Chris and co. decide they want in 2 years time) and one where its float. Those would not interoperate (why would you want them to) and could be two completly seperate SDKs, though they might only vary in the typedef gmpi_sample line of gmpi.h. In practice there are other things that paltops can't do (eg. threads historically, maybe not now)> > > Not if the profiles are linked to a CPU architecture. eg. all PPC and x86 > > (or Alitvec and SSE if you want it at that granularity) plugins use the > > float profile, ARM uses int24. This could tie up the binary compatibility > > issue on multi platform OSs to. > > I'm sorry to say this: GAG! We are the WRONG people to make decisions like > this. Um, sorry? Who else is going to make it? I dont think there are any people on the planet more qualified to decide what makes a good and efficient sample representation for realtime audio than audio plugin devleopers. > My main point is that IF we need to support more than one datatype (and by > the palmtop, it sounds as if we do) then we should support N datatypes. But we dont need to support more than one _at the same time_, and there is no need for the spec to say how they interoperate, as the plugins designed for one profile, will not execute on the other machine. That makes a huge difference. > If you don't want to support it in your host, DON'T. We don't need to > define a Float64 type for now. One day, almost certainly, we will. When > that day comes, it will be easy. Most hosts won't support it (yet). Many > NEVER will. Calling a datatype a profile, and forcing me to pick a > closest-fit is not a solution to the problem. But in the case of mobile and desktop use there are exact fits, int24 and float. - 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