hm still choppy :( maybe because its using double-precision samples everywhere.. there is a macro SAMPLE_IS_DOUBLE that ostensibly controls this but setting it to 0 breaks the build in a million places. also int64_t is used for a lot of things where size_t would make more sense