Compiling Haskell to Massively Parallel Hardware [pdf]
conal.net
conal.net
It seems this goes even further with Haskell, allowing us to compile not only to new CPU architectures, but to even more exotic targets like FPGAs, since the code is closer to a logical description of the problem domain, as opposed to a sequence of state modifications. This will become particularly relevant if end up hitting a wall with regards to CPU performance, and need to move to different hardware in order to gain a speedup.