I'm presently disabled.  And I started programming in assembly as well.  I have a degree in Network Systems Administration and I would like to find something along those lines in work, but so far not able to do so.
OpenCL isn't TOO difficult to learn, but I have trouble with the syntax of some commands like prefetch and the like.  I'm thinking about tossing a prefetch or two into the code to see if it'll increase the speed by much.  In particular, just before sharoundC to prepare K and ConstW if it's not already.  And then another prefetch to call the parts of H when it's needed for Vals.  I don't know if this could shave off a few cycles or not, but I plan to find out.