Issue No. 03 - May-June (2013 vol. 33)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MM.2013.25
Melanie Kambadur , Columbia University
Kui Tang , Columbia University
Martha A. Kim , Columbia University
Parallel block vector profiles (PBVs) establish a mapping between a multithreaded application's basic blocks and the degree of parallelism the application exhibits each time a block executes. PBVs offer a new perspective that helps users both reason about parallel programs' hardware and software interactions and identify opportunities for performance improvements. Here, the authors present two PBV applications for architectural design and discuss further opportunities to apply PBVs in other fields. They also demonstrate how the open-source tool Harmony lets programmers collect PBVs with minimal programmer effort and application perturbation.
Computer architecture, Parallel processing, System performance, Computer programs, Performance evaluation, Programming, Hardware, Measurements, parallel block vector, parallel programming, programming techniques, software engineering, performance analysis and design aids, performance and reliability, hardware, performance measures, metrics/measurement
M. Kambadur, M. A. Kim and K. Tang, "Parallel Block Vectors: Collection, Analysis, and Uses," in IEEE Micro, vol. 33, no. , pp. 86-94, 2013.