Comment by Jasper_

Conversions like this are basically as fast as a memcpy, which will always be required in this system. Did either you or dcode try actually profiling the extra time spent?