So, i'd like to know it i'm to much optimistic, or if work on thread with pd will be obsolet when pd 0.42 will be out.
i didn't check the sources, but from the description in miller's icmc paper, pd~ introduces a pipelined approach, i.e. adding one block of latency when audio data is passed from one thread to another. from my understanding, the pd~ approach is static, so a patch would need to be written explicitly for the number of threads ...
pd~ has different limitations than max5's poly~, but is still far away from a general solution ...
tim