chose FPGA

Use the altera C2H compiler it compiles C code into hardware.

Or analize your code to find out how many multiplications and intermidiate memory you use.

I did a complex image processing algorithm in a FPGA 125MHz and it run 20000 faster than a very powerfull 3GHz workstation with lots of cache.

So some classes of algorithms benefit a lot from FPGA implementation.

But it is hard to say how much, unless you really do it.