Inspect-GPU: A Software to Evaluate Performance Characteristics of CUDA Kernels Using Microbenchmarks and Regression Models

Gargi Alavani, Santonu Sarkar

2023

Abstract

While GPUs are popular for High-Performance Computing(HPC) applications, the available literature is inadequate for understanding the architectural characteristics and quantifying performance parameters of NVIDIA GPUs. This paper proposes “Inspect-GPU”, a software that uses a set of novel, architecture-agnostic microbenchmarks, and a set of architecture-specific regression models to quantify instruction latency, peakwarp and throughput of a CUDA kernel for a particular NVIDIA GPU architecture. Though memory access is critical for GPU performance, memory instruction execution details, such as its runtime throughput, are not revealed. We have developed a memory throughput model providing unpublished crucial insights. Inspect-GPU builds this throughput model for a particular GPU architecture. Inspect-GPU has been tested on multiple GPU architectures: Kepler, Maxwell, Pascal, and Volta. We have demonstrated the efficacy of our approach by comparing it with two popular performance analysis models. Using the results from Inspect-GPU, developers can analyze their CUDA applications, apply optimization, and model GPU architecture and its performance.

Paper Citation

in Harvard Style

Alavani G. and Sarkar S. (2023). Inspect-GPU: A Software to Evaluate Performance Characteristics of CUDA Kernels Using Microbenchmarks and Regression Models. In Proceedings of the 18th International Conference on Software Technologies - Volume 1: ICSOFT; ISBN 978-989-758-665-1, SciTePress, pages 59-70. DOI: 10.5220/0012079200003538

in Bibtex Style

@conference{icsoft23,
author={Gargi Alavani and Santonu Sarkar},
title={Inspect-GPU: A Software to Evaluate Performance Characteristics of CUDA Kernels Using Microbenchmarks and Regression Models},
booktitle={Proceedings of the 18th International Conference on Software Technologies - Volume 1: ICSOFT},
year={2023},
pages={59-70},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012079200003538},
isbn={978-989-758-665-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 18th International Conference on Software Technologies - Volume 1: ICSOFT
TI - Inspect-GPU: A Software to Evaluate Performance Characteristics of CUDA Kernels Using Microbenchmarks and Regression Models
SN - 978-989-758-665-1
AU - Alavani G.
AU - Sarkar S.
PY - 2023
SP - 59
EP - 70
DO - 10.5220/0012079200003538
PB - SciTePress