GradPIM: A Practical Processing-in-DRAM Architecture for Gradient Descent
Authors:
- 1 February 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 249-262
Abstract
In this paper, we present GradPIM, a processingin-memory architecture which accelerates parameter updates of deep neural networks training. As one of processing-in-memory techniques that could be realized in the near future, we propose an incremental, simple architectural design that does not invade the existing memory protocol. Extending DDR4 SDRAM to utilize bank-group parallelism makes our operation designs in processing-in-memory (PIM) module efficient in terms of hardware cost and performance. Our experimental results show that the proposed architecture can improve the performance of DNN training and greatly reduce memory bandwidth requirement while posing only a minimal amount of overhead to the protocol and DRAM area.Keywords
All Related Versions
Funding Information
- National Research Foundation
This publication has 56 references indexed in Scilit:
- ShiDianNao2015
- PIM-enabled instructions2015
- A scalable processing-in-memory accelerator for parallel graph processing2015
- Memristor Crossbar Architecture for Synchronous Neural NetworksIEEE Transactions on Circuits and Systems I: Regular Papers, 2014
- NDC: Analyzing the impact of 3D-stacked memory+logic devices on MapReduce workloads2014
- SAP HANA databaseACM SIGMOD Record, 2012
- Optimization of Collective Communication Operations in MPICHThe International Journal of High Performance Computing Applications, 2005
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- A case for intelligent RAMIEEE Micro, 1997
- The Accuracy of Floating Point SummationSIAM Journal on Scientific Computing, 1993