Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Proposal of Automatic Offloading Method in Mixed Offloading Destination Environment

Yoji Yamato

TL;DR

The paper tackles the challenge of harnessing heterogeneous hardware by proposing environment-adaptive software that automatically offloads computations to mixed destinations (GPU, FPGA, and many-core CPU). It introduces device-specific offloading strategies and a mixed-environment framework powered by evolutionary pattern search to select loop statements and function blocks that maximize performance, while accounting for data-transfer costs and device characteristics. Through evaluations on three representative workloads, the approach demonstrates substantial speedups (e.g., up to ~1120x for GPU loop offloads and ~21x for FPGA function blocks) and provides a practical verification order and constraints to balance performance and cost. This work advances automatic, deployment-time optimization across diverse hardware, reducing manual tuning and enabling broader adoption of heterogeneous computing in edge, IoT, and cloud contexts.

Abstract

When using heterogeneous hardware, barriers of technical skills such as OpenMP, CUDA and OpenCL are high. Based on that, I have proposed environment-adaptive software that enables automatic conversion, configuration. However, including existing technologies, there has been no research to properly and automatically offload the mixed offloading destination environment such as GPU, FPGA and many core CPU. In this paper, as a new element of environment-adaptive software, I study a method for offloading applications properly and automatically in the environment where the offloading destination is mixed with GPU, FPGA and many core CPU. Y. Yamato, "Proposal of Automatic Offloading Method in Mixed Offloading Destination Environment," 2020 Eighth International Symposium on Computing and Networking Workshops (CANDARW 2020), pp.460-464, DOI: 10.1109/CANDARW51189.2020.00094, Nov. 2020. "(c) 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works."

Proposal of Automatic Offloading Method in Mixed Offloading Destination Environment

TL;DR

The paper tackles the challenge of harnessing heterogeneous hardware by proposing environment-adaptive software that automatically offloads computations to mixed destinations (GPU, FPGA, and many-core CPU). It introduces device-specific offloading strategies and a mixed-environment framework powered by evolutionary pattern search to select loop statements and function blocks that maximize performance, while accounting for data-transfer costs and device characteristics. Through evaluations on three representative workloads, the approach demonstrates substantial speedups (e.g., up to ~1120x for GPU loop offloads and ~21x for FPGA function blocks) and provides a practical verification order and constraints to balance performance and cost. This work advances automatic, deployment-time optimization across diverse hardware, reducing manual tuning and enabling broader adoption of heterogeneous computing in edge, IoT, and cloud contexts.

Abstract

When using heterogeneous hardware, barriers of technical skills such as OpenMP, CUDA and OpenCL are high. Based on that, I have proposed environment-adaptive software that enables automatic conversion, configuration. However, including existing technologies, there has been no research to properly and automatically offload the mixed offloading destination environment such as GPU, FPGA and many core CPU. In this paper, as a new element of environment-adaptive software, I study a method for offloading applications properly and automatically in the environment where the offloading destination is mixed with GPU, FPGA and many core CPU. Y. Yamato, "Proposal of Automatic Offloading Method in Mixed Offloading Destination Environment," 2020 Eighth International Symposium on Computing and Networking Workshops (CANDARW 2020), pp.460-464, DOI: 10.1109/CANDARW51189.2020.00094, Nov. 2020. "(c) 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works."

Paper Structure

This paper contains 16 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Proposal of automatic offloading to mixed offloading destination environment with various devices
  3. Basic ideas for various devices offloading
  4. Automatic offloading to each offloading device
  5. Automatic many core CPU offloading of loop statements
  6. Automatic GPU offloading of loop statements
  7. Automatic FPGA offloading of loop statements
  8. Automatic offloading of function blocks
  9. Automatic offloading for mixed offloading destination environment
  10. Evaluation
  11. Evaluation method
  12. Evaluated applications
  13. Experiment conditions
  14. Experimental environment
  15. Performance results
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Automatic many core CPU offloading method of loop statements
  • Figure 2: Experimental environment
  • Figure 3: Results of offloading to mixed offloading destination environment