Projects per year
Abstract
Transport triggered architectures are used for implement-
ing bio-inspired systems due to their simplicity, modularity and fault-
tolerance. However, producing ecient, optimised machine code for such
architectures is an extremely dicult task, as the computational complex-
ity has moved from the hardware-level to the software-level. This paper
presents the application of Cartesian Genetic Programming to the evo-
lution of machine code for a simple implementation of a transport trig-
gered architecture. The eectiveness of the algorithm is demonstrated by
evolving machine code for a 4-bit multiplier with three dierent levels of
parallelism. The results show that 100% successful solutions were found
by CGP and by further optimising the size of the solutions, it is possible
to nd ecient implementations of the 4-bit multiplier. Further analysis
of the solutions showed that the use of loops within the CGP function set
could be benecial and was demonstrated by repeating the earlier 4-bit
multiplier experiment with the addition of a loop function. The further
results showed that the use of a loop function did not signincatly aect
the performance of the CGP algorithm but a reduction of up to 55%
in the solution size was observed, which could have the potential to be
classed as \human competitive".
ing bio-inspired systems due to their simplicity, modularity and fault-
tolerance. However, producing ecient, optimised machine code for such
architectures is an extremely dicult task, as the computational complex-
ity has moved from the hardware-level to the software-level. This paper
presents the application of Cartesian Genetic Programming to the evo-
lution of machine code for a simple implementation of a transport trig-
gered architecture. The eectiveness of the algorithm is demonstrated by
evolving machine code for a 4-bit multiplier with three dierent levels of
parallelism. The results show that 100% successful solutions were found
by CGP and by further optimising the size of the solutions, it is possible
to nd ecient implementations of the 4-bit multiplier. Further analysis
of the solutions showed that the use of loops within the CGP function set
could be benecial and was demonstrated by repeating the earlier 4-bit
multiplier experiment with the addition of a loop function. The further
results showed that the use of a loop function did not signincatly aect
the performance of the CGP algorithm but a reduction of up to 55%
in the solution size was observed, which could have the potential to be
classed as \human competitive".
| Original language | English |
|---|---|
| Pages (from-to) | 32-50 |
| Number of pages | 19 |
| Journal | Journal of Adaptive, Resilient and Autonomic Systems |
| Volume | 3 |
| Issue number | 4 |
| DOIs | |
| Publication status | Published - 2012 |
Projects
- 1 Finished
-
SABRE: Self-healing Cellular Architectures
Tyrrell, A., Liu, J., Qadir, O., Tempesti, G. & Timmis, J.
1/10/08 → 30/09/11
Project: Research project (funded) › Research