# Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication

arXiv: https://arxiv.org/abs/2106.10499 · HTML: https://mlsanity.com/paper/2106.10499/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 43 references that are in the map.
- First posted on arXiv 19 Jun 2021.

## Builds on (most cited first, 5 of 5)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [In-Datacenter Performance Analysis of a Tensor Processing Unit](https://mlsanity.com/paper/1704.04760/) (2017)
- [Interstellar: Using Halide's Scheduling Language to Analyze DNN Accelerators](https://mlsanity.com/paper/1809.04070/) (2018)
- [Fast Stencil-Code Computation on a Wafer-Scale Processor](https://mlsanity.com/paper/2010.03660/) (2020)
- [Understanding the Impact of On-chip Communication on DNN Accelerator Performance](https://mlsanity.com/paper/1912.01664/) (2019)

## Built on by (most cited first, 2 of 2)

- [Union: A Unified HW-SW Co-Design Ecosystem in MLIR for Evaluating Tensor Operations on Spatial Accelerators](https://mlsanity.com/paper/2109.07419/) (2021)
- [Accelerating Edge AI with Morpher: An Integrated Design, Compilation and Simulation Framework for CGRAs](https://mlsanity.com/paper/2309.06127/) (2023)

## Abstract (arXiv)

There is a growing interest in custom spatial accelerators for machine learning applications. These accelerators employ a spatial array of processing elements (PEs) interacting via custom buffer hierarchies and networks-on-chip. The efficiency of these accelerators comes from employing optimized dataflow (i.e., spatial/temporal partitioning of data across the PEs and fine-grained scheduling) strategies to optimize data reuse. The focus of this work is to evaluate these accelerator architectures using a tiled general matrix-matrix multiplication (GEMM) kernel. To do so, we develop a framework that finds optimized mappings (dataflow and tile sizes) for a tiled GEMM for a given spatial accelerator and workload combination, leveraging an analytical cost model for runtime and energy. Our evaluations over five spatial accelerators demonstrate that the tiled GEMM mappings systematically generated by our framework achieve high performance on various GEMM workloads and accelerators.