mathematics
and
computer
science

LANS

LANS Publications

"Hybrid Static/Dynamic Activity Analysis"

B. Kreaseck, L. Ramos, S. Easterday, M. Strout, P. Hovland

. Also Preprint ANL/MCS-P1312-0106

Preprint Version: [pdf]

Automatic Differentiation is the process of translating one program that computes a function f and generating a different program that computes the derivative of that function, f'. Activity analysis is important for AD. Our results show that a dynamic activity analysis, checking at run-time, incurs an average overhead of 55% when all independent variable are active. When as few as half of the independent variables are active, dynamic activity analysis enables an average speedup of 28%. We investigate static activity analysis combined with dynamic activity analysis as a technique for reducing the overhead of dynamic activity analysis.