We are looking for an efficient convolution reverb algorithm (music).  It must be opimized for Windows, OSX (Intel) and OSX (ppc).  The algorithm should take advantage of SSE (or altivec) instructions.  Language: C/C++.  This algorithm cannot use any open-source code, as it will be used in a commercial product. 



The code should be in the form of a simple library, with  init(), setup() and  process() functions.  The input will be stereo floats.



It should be as CPU-efficient as other convolution reverb plugins on the market, such as Voxengo Pristene Space, TL Labs TL Space, and Waves IR-1



We don't need multiprocessor support, single processor code is fine.