Litcius/Paper detail

Enabling Fine-Grained Spatial Multitasking on Systolic-Array NPUs Using Dataflow Mirroring

Jinwoo Choi, Yeonan Ha, Jounghoo Lee, Sangsu Lee, Jinho Lee, Hanhwi Jang, Youngsok Kim

2023IEEE Transactions on Computers10 citationsDOI

Abstract

Neural Processing Units (NPUs) frequently suffer from low hardware utilization as the efficiency of their systolic arrays heavily depends on the characteristics of a deep neural network (DNN). Spatial multitasking is a promising solution to overcome the low NPU hardware utilization; however, the state-of-the-art spatial-multitasking NPU architecture achieves sub-optimal performance due to its coarse-grained systolic-array distribution and incurs significant implementation costs. In this paper, we propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dataflow-mirroring NPU (DM-NPU)</i> , a novel spatial-multitasking NPU architecture supporting fine-grained systolic-array distribution. The key idea of DM-NPU is to reverse the dataflows of co-located DNNs in horizontal and/or vertical directions. DM-NPU can place allocation boundaries between any adjacent processing elements of a systolic array, both horizontally and vertically. We then propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DM-Perf</i> , an accurate systolic-array NPU performance model, to maximize the spatial-multitasking performance of DM-NPU. Utilizing the existing performance models achieves sub-optimal performance as they cannot accurately capture the resource contention caused by spatial multitasking. DM-Perf, on the other hand, exploits the per-layer performance profiles of a DNN to accurately capture the resource contention. Our evaluation using MLPerf DNNs shows that DM-NPU and DM-Perf can greatly improve the performance by up to 35.1% over the state-of-the-art NPU architecture and performance model.

Topics & Concepts

Human multitaskingDataflowComputer scienceParallel computingSystolic arrayArtificial intelligenceEmbedded systemCognitive psychologyVery-large-scale integrationPsychologyAdvanced Neural Network ApplicationsAdvanced Memory and Neural ComputingFerroelectric and Negative Capacitance Devices
Enabling Fine-Grained Spatial Multitasking on Systolic-Array NPUs Using Dataflow Mirroring | Litcius