Jason McBrien wrote:
It's genetic/cellular automata style computation.
It cycles through various
logic configurations until it finds the optimal setup. The bad news is that
it means that the "compile" cycles could take a while. The good news is
once
an optimal algorithm is found, it's saved for later. This scheme should
theoretically work well for compute-intensive operations, especially those
requiring lots of variables, such as DSP and fluid dynamics.
They have had that years ago -- And star treks's computers even had blinking lights.
But oddly you made me wonder about a FPGA system can handle the problem
since most problems off hand don't seem to be integer math, rather floating
point of some sort. I just don't trust the words computer science uses
not the algorithms.