abstract = "Sub-machine-code GP (SMCGP) is a technique to speed up
genetic programming (GP) and to extend its scope based
on the idea of exploiting the internal parallelism of
sequential CPUs. In previous work [20] we have shown
examples of applications of this technique to the
evolution of parallel programs and to the parallel
evaluation of 32 or 64 fitness cases per program
execution in Boolean classification problems. After
recalling the basic features of SMCGP, in this paper we
first apply this technique to the problem of evolving
parallel binary multipliers. Then we describe how SMCGP
can be extended to process multiple fitness cases per
program execution in continuous symbolic regression
problems where inputs and outputs are real-valued
numbers, reporting experimental results on a quartic
polynomial approximation task.",