Now that we use the locking library to make sure running opcodes cannot
step on each other toes we can have a bigger thread size, and
potentially process many opcodes in a parallel manner.

Reviewed-by: iustinp