FPGA implementation of 2D Convolution using OneAPI and OpenCL - CentraleSupélec Access content directly
Conference Papers Year : 2022

FPGA implementation of 2D Convolution using OneAPI and OpenCL

Daouda Diakite
Nicolas Gac

Abstract

Thanks to High-Level Synthesis (HLS) tools, FPGAs have become an alternative to GPUs for compute-intensive applications. These tools have been developed to provide flexibility in FPGA design at a higher abstraction level than hardware description languages. Major FPGA manufacturers have proposed many HLS tools depending on the target audience. In this paper, we propose a hardware architecture on FPGA for 2D convolution designed through two software-like development tools based on oneAPI and OpenCL languages. This paper also focuses on comparing the oneAPI and OpenCL HLS tools in terms of performance and productivity with the case study of the 2D convolution operator using an Intel Stratix 10 device.
Fichier principal
Vignette du fichier
SOC2_2022_paper_8630.pdf (160.15 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03695100 , version 1 (14-06-2022)

Identifiers

  • HAL Id : hal-03695100 , version 1

Cite

Daouda Diakite, Nicolas Gac. FPGA implementation of 2D Convolution using OneAPI and OpenCL. 16ème Colloque National du GDR SOC2, Jun 2022, Strasbourg, France. ⟨hal-03695100⟩
134 View
133 Download

Share

Gmail Facebook X LinkedIn More