Abstract
Existing runtime parallelization techniques impose severe performance penalties when a speculative parallelization is attempted and fails. Some techniques require a sequential restart of the specular tive execution while others only disregard the work after the first point of failure. This paper introduces a new technique that reduces the performance overhead of failure to less than 1% on standard processors through a combination of hoisting the failure path and partitioning work to a Coinspector Thread. © Springer-Verlag Berlin Heidelberg 2007.
Cite
CITATION STYLE
Luján, M., Gustafson, P., Paleczny, M., & Vick, C. A. (2007). Speculative parallelization - Eliminating the overhead of failure. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4782 LNCS, pp. 460–471). Springer Verlag. https://doi.org/10.1007/978-3-540-75444-2_45
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
