GPGPUS are an interesting implementation of parallelized computation without having to spell out all the details. Of course, CUDA etc libraries are required to

GPGPUS are an interesting implementation of parallelized computation without having to spell out all the details. Of course, CUDA etc libraries are required to take advantage of the hardware. sufficiently smart compiler is typically said with a smirk in conversation. In recent times I have seen some compilers attempt to "vectorize" array code. But the results vary greatly. Perhaps greater parallelism could be achieved using event-driven / message-driven multi-processing. The original conceptual design of Smalltalk proposed a situation where each object was implemented as a virtual processor. But that all went south with the advent of Java.