MANAGING GRAPHICS PROCESSING UNITS' MEMORY AND ITS ASSOCIATED TRANSFERS IN ORDER TO INCREASE THE SOFTWARE PERFORMANCE

Alexandru Pîrjan ()
Additional contact information
Alexandru Pîrjan: Romanian-American University, Bucharest, Romania

Journal of Information Systems & Operations Management, 2017, vol. 11, issue 1, 106-117

Abstract: This paper is focused on analyzing a key aspect, the management of the Graphics Processing Units' (GPUs) memory, which is of paramount importance when developing a software application that makes use of the Compute Unified Device Architecture (CUDA). The paper tackles important technical aspects that can affect the overall performance of a CUDA application such as: the optimal alignment in memory of the data that is to be processed, obtaining optimal memory access patterns that facilitate the retrieving of instructions; aligning to the L1 cache line according to its size, taking into account the balance achieved between single or double precision and the effect on how much memory is being used; joining more kernel functions into a single one in certain situations, benefiting from the increased speedup offered by putting into use the shared and cache memory, adjusting the code to the available memory bandwidth by taking into account the memory latency and the need to transfer data between the host and the device.

Date: 2017
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
http://www.rebe.rau.ro/RePEc/rau/jisomg/SU17/JISOM-SU17-A09.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:rau:jisomg:v:11:y:2017:i:1:p:106-117

Access Statistics for this article

More articles in Journal of Information Systems & Operations Management from Romanian-American University Contact information at EDIRC.
Bibliographic data for series maintained by Alex Tabusca ().