Accurately modeling GPGPU frequency scaling with the CRISP performance model
R. Nath; D. Tullsen University of California, San Diego, La Jolla, CA, United States
Abstract
CRItical Stalled Path (CRISP) is the first runtime analytical model of general-purpose computation on graphics processing unit (GPGPU) performance that models the effect of changing frequency. Prior models not targeted at a GPGPU fail to account for important characteristics of GPGPU execution, including the high degree of overlap between memory access and computation and the frequency of store-related stalls. CRISP provides significantly improved accuracy, being within 4% on average when scaling frequency by up to 7×. Using CRISP to drive a runtime energy ...
Get Advances in GPU Research and Practice now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
