The loop fusion for data localization
Автор: Steinberg Boris Yakovlevich, Steinberg Oleg Borisovich, Vasilenko Alexandr Alexandrovich
Журнал: Программные системы: теория и приложения @programmnye-sistemy
Рубрика: Математические основы программирования
Статья в выпуске: 3 (46) т.11, 2020 года.
Бесплатный доступ
The loop merging can be used to improve data localization. Merging loops that share variables can speed up execution by reducing the number of cache misses. This transformation has been known for a long time, but compilers perform it only for the simplest cases.Our improved algorithms use pre-transformations to correctly merge loops with different numbers of iterations and informational dependencies.
Loop fusion, optimizing compiler, program transformations, data locality, memory access optimization
Короткий адрес: https://sciup.org/143172950
IDR: 143172950 | DOI: 10.25209/2079-3316-2020-11-3-17-31
