Pyxis: An Open-Source Performance Dataset of Sparse Accelerators

TL;DR

PYXIS is an open-source dataset that captures real execution performance data of various sparse accelerators, aiding research and design in specialized hardware for sparse data applications.

Contribution

It introduces PYXIS, a comprehensive, growing dataset of accelerator designs and performance metrics for sparse data, filling a gap in analytic modeling capabilities.

Findings

Contains 73.8K instances of accelerator performance data.

Provides a valuable resource for benchmarking and designing sparse accelerators.

Supports diverse research in architecture, algorithms, and performance analysis.

Abstract

Specialized accelerators provide gains of performance and efficiency in specific domains of applications. Sparse data structures or/and representations exist in a wide range of applications. However, it is challenging to design accelerators for sparse applications because no architecture or performance-level analytic models are able to fully capture the spectrum of the sparse data. Accelerator researchers rely on real execution to get precise feedback for their designs. In this work, we present PYXIS, a performance dataset for specialized accelerators on sparse data. PYXIS collects accelerator designs and real execution performance statistics. Currently, there are 73.8 K instances in PYXIS. PYXIS is open-source, and we are constantly growing PYXIS with new accelerator designs and performance statistics. PYXIS can benefit researchers in the fields of accelerator, architecture, performance, algorithm, and many related topics.