Genetic Parallel Programming: Design and Implementation

Abstract

This paper presents a novel Genetic Parallel Programming (GPP) paradigm for evolving parallel programs running on a Multi-Arithmetic-Logic-Unit (Multi-ALU) Processor (MAP). The MAP is a Multiple Instruction-streams, Multiple Data-streams (MIMD), general-purpose register machine that can be implemented on modern Very Large-Scale Integrated Circuits (VLSIs) in order to evaluate genetic programs at high speed. For human programmers, writing parallel programs is more difficult than writing sequential programs. However, experimental results show that GPP evolves parallel programs with less computational effort than that of their sequential counterparts. It creates a new approach to evolving a feasible problem solution in parallel program form and then serializes it into a sequential programif required. The effectiveness and efficiency of GPP are investigated using a suite of 14 well-studied benchmark problems. Experimental results show that GPP speeds up evolution substantially.

Savant: Automatic parallelization of a scheduling heuristic with machine learning

“…There is little detail presented however. In [15], [16], [17], the authors use GP to evolve a program in order to achieve parallelism. The programs found are evaluated both in terms of correctness (their purpose) and their degree of parallelism.…”

Exploring the evolution of internal control structure using digital enzymes

“…With respect to the evolution of parallel programs, a variety of approaches exist in the area of genetic programming [1] including Cartesian Genetic Programming [5,8], Parallel Distributed Genetic Programming [6], Genetic Parallel Programming [3], and Concurrent Genetic Programming [7]. However, each of these approaches contain one of the following drawbacks: (1) the maximum size of the genetic program's grid, graph, or tree must be predefined, (2) a crossover operator imposes a high rate of disruption and genetic bloat, (3) instructions are coarse-grained requiring each actuator action to be encoded as an instruction, or (4) their application does not include evolving reactive controllers.…”

High performance genetic programming on GPU

“…A common strategy to tackle this problem is to parallelize or distribute the GP computations, e.g. [24,13,7,4]. Newly introduced graphics processing units (GPUs) provide fast parallel hardware for a fraction of the cost of a traditional parallel system.…”