view article

Figure 4
Amdahl's law (established in 1967) appeared to be a fundamental limit on the maximum speedup or performance gain achievable by parallelization of an algorithm to multiple CPUs. The fraction of parallelizable code drives this law. The law was formulated under the assumption of a fixed problem size. This assumption means that the ratio between the serial and parallel code fraction is constant. Under this assumption, the efficiency of parallelization is constantly decreasing with the growing number of CPUs. Until recently, this was still accepted in the range between two and eight CPUs. However, when increasing the problem size while not linearly increasing the serial fraction of the algorithm, Amdahl's law has to be re-evaluated and a much higher scalable speedup or performance gain can be achieved.

Journal logoBIOLOGICAL
CRYSTALLOGRAPHY
ISSN: 1399-0047
Follow Acta Cryst. D
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds