Bikov, Dusan and Bouyukliev, Ilija and Stojanova, Aleksandra (2016) Benefit of Using Shared Memory in Implementation of Parallel FWT Algorithm with CUDA C on GPUs. In: ITRO 2016, 10 June 2016, Zrenjanin, Serbia.
Preview |
Text
Zbornik ITRO 2016-split-merge.pdf Download (2MB) | Preview |
Abstract
GPUs have different memory hierarchy than
CPU and with their proper use, we can achieve effective
implementation and improve the performance. In this paper
we discuss how to use shared memory on GPUs and how
does it affect the implementation and performance. For a
more detail clarification of benefit of using shared memory, we take into account parallel algorithm for calculation of the Walsh spectrum on graphics processor unit (GPU) and its parallel implementation in CUDA C. Using shared memory is a good optimization strategy, which gives faster time of execution of the parallel program.
Item Type: | Conference or Workshop Item (Paper) |
---|---|
Subjects: | Natural sciences > Computer and information sciences |
Divisions: | Faculty of Computer Science |
Depositing User: | Александра Стојанова Илиевска |
Date Deposited: | 14 Jul 2016 11:28 |
Last Modified: | 14 Jul 2016 11:32 |
URI: | https://eprints.ugd.edu.mk/id/eprint/16061 |
Actions (login required)
![]() |
View Item |