[Read Paper] Eyeriss: A Spatial Architecture for Energy-Efficient Dataflow for Convolutional Neural Networks

Published: December 29, 2019 by SingularityKChen (Last updated: March 07, 2020 )

• Categories:
• ReadPaper 30

• Tags:
• DLA 19
• DNN 12
• Eyeriss 4

• Eyeriss: A Spatial Architecture for Energy-Efficient Dataflow for Convolutional Neural Networks
  • Framework for Energy Efficiency Analysis
    • Storage Area
    • Input Data Access Energy Cost
    • Partial Sum Accumulation Energy Cost

Eyeriss: A Spatial Architecture for Energy-Efficient Dataflow for Convolutional Neural Networks

Compared to the Eyeriss v2, this article provides a more detailed explanation of Row Stationary, a baseline storage area for a given number of PEs and the energy cost estimation for RS reuse pattern.

---

This article proposed RS dataflow which can adapt to different CNN shape configurations and reduces all types of data movement through maximally utilizing the processing engine (PE) local storage, direct inter-PE communication and spatial parallelism.

Also, an analysis framework that compares energy cost under the same hardware area and processing parallelism constraints.

Framework for Energy Efficiency Analysis

Storage Area

The baseline storage area for a given number of PEs is calculated as

#PE×Area(512B RF)+Area((#PE×512B) global buffer).

Input Data Access Energy Cost

Input data access energy cost estimation:

a×EC(DRAM)+ab×EC(global buffer)+abc×EC(array)+abcd×EC(RF)

Partial Sum Accumulation Energy Cost

Partial Sum accumulation energy cost estimation:

(2a−1)×EC(DRAM)+2a(b−1)×EC(global buffer)+ab(c−1)×EC(array)+2abc(d−1)×EC(RF)

where the EC(*) is shown as follows: