A quick study of Cantor, Goedel and Turing as well as a consultation of high school physics tells us that when we read/write from hardware, we are in essence instantiating or else transmuting the "paramagnetic states" for each magnetic spin. All this talk about ASIC computing is good and all, but we hardly see people in the community talk about simple RISC architectures. People hardly ever use ALL 32 bits, even on an average machine, and it's due to the mathematical physics principles of these magnetic states.
Since the Halting problem hasn't been solved yet and a computational system CAN simulate physical hardware in these paramagnetic states, you only need ONE daemon which can interface with OpenCL, where OpenCL provides the interface between the CPU/GPU. All the other load balancing and cryptographic functions are taken care of by gcc.
Edit: Compare solving an object called "block" with downloading pieces of a torrent file. You don't necessarily need 100 peers connected over the protocol to download an entire file, sometimes you can get the job done with just 10 peers in the same amount of time.
A quick study of Calvin, Hobbes and Watterson tell me that a transmogrification apparatus could perform the required adjustments and transformations relatively quickly/
