Paper title: AN EVOLUTIONARY COMPUTATIONAL SYSTEM ARCHITECTURE BASED ON A SOFTWARE TRANSACTIONAL MEMORY
Author(s): BRANISLAV KORDIC, MARKO POPOVIC, MIROSLAV POPOVIC, MOSHE GOLDSTEIN, MOSHE AMITAY, DAVID DAYAN, ERICK FREDJ,
Abstract: In the last decade software transactional memory has become a prominent programming paradigm, which aims to improve the
execution performance of concurrent programs. However, most research in the field is done in programming languages such as
C, C++ and JAVA. In this paper, we present a PSTM-based architecture of DEEPSAM, a computational chemistry program
written in Python language. The PSTM-based architecture aims to improve execution performance of the original computational
chemistry system architecture based on evolutionary programming and to provide transactional-memory-based means for its
future optimizations. The metric analysis includes system execution time and problem size scalability. The experimental results
show that the new PSTM-based version gained better execution time results relative to the original version. Likewise, the results
did not reveal any architectural bottleneck of the new architecture.
Keywords: Software transactional memory, Python, Diffusion equation evolutionary programming simulated annealing method
(DEEPSAM), Python software transactional memory (PSTM), Evolutionary programming (EP) Year: 2021 | Tome: 66 | Issue: 1 | Pp.: 47-52
Full text : PDF (2813 KB) |