About Me
I am currently working as a Senior GPU Architect at NVIDIA in Santa Clara, investigating memory subsystem architecture for the next generation of GPUs.
I am an alumnus of The University of Edinburgh, where I completed my PhD in the area of Computer Architecture.
During my PhD, I was advised by Prof. Nigel Topham and co-advised by Dr. Vijay Nagarajan and was affiliated to the Compiler and Architecture Design Group (CArD) within the Institute of Computing Systems Architecture (ICSA) at the School of Informatics.
During my graduate school, I interned at NVIDIA in the GPU Memory Architecture Group in Santa Clara, USA. In the past, I have also worked at Synopsys as an R&D Engineer in Hyderabad, India. During my undergraduate, I was advised by Dr. Shubhajit Roy Chowdhury for my dissertation at IIIT Hyderabad.
In the summer of 2011, I also interned at the University of Cambridge under the supervision of Dr. Neil Collings where I worked on LCOS displays and high-speed display drivers at the Centre for Advanced Photonics and Electronics.
Research Interests
My research interests span across GPU memory systems, shared resource management in throughput-oriented systems and application of machine learning in building computer systems.
Publications
Poise : Balancing Thread-Level Parallelism and Memory System Performance in GPUs using Machine Learning
Saumay Dublish, Vijay Nagarajan and Nigel Topham
IEEE International Symposium on High-Performance Computer Architecture (HPCA)
Washington D.C., USA, February 16-20, 2019
[paper] [slides-pptx] [slides-pdf]
Managing the Memory Hierarchy in GPUs
Saumay Dublish
PhD Thesis
The University of Edinburgh, July 2018
[thesis]
Evaluating and Mitigating Bandwidth Bottlenecks Across the Memory Hierarchy in GPUs
Saumay Dublish, Vijay Nagarajan and Nigel Topham
IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)
Santa Rosa, USA, April 23-25, 2017
[paper] [slides-pptx] [slides-pdf]
Cooperative Caching for GPUs
Saumay Dublish, Vijay Nagarajan and Nigel Topham
ACM Transactions on Architecture and Code Optimization, 13 (4),39, December 2016 (TACO)
[paper] [slides-pptx] [slides-pdf]
Characterizing Memory Bottlenecks in GPGPU Workloads
Saumay Dublish, Vijay Nagarajan and Nigel Topham
IEEE International Symposium on Workload Characterization (IISWC)
Providence, Rhoda Island, USA, September 25-27, 2016
[paper] [poster]
Slack-Aware Shared Bandwidth Management in GPUs
Saumay Dublish
ACM SRC, The 25th International Conference on Parallel Architectures and Compilation Techniques (PACT)
Haifa, Israel, September 11-15, 2016
[paper] [poster]
Awards
ISPASS 2017 Student Travel Grant
2017
IEEEPACT 2016 Student Travel Grant
2016
ACM SIGARCHSchool of Informatics Doctoral Scholarship
2014-2018
The University of EdinburghSaranu International Research Scholarship
2014-2018
The University of EdinburghCambridge-India Partnership grant
2011
University of CambridgeCanSat Satellite Design Launch Competition Award: Rank-1
2010
NASAWorld Embedded Software Contest Award: Rank-4
2010
Ministry of Knowledge Economy, South Korea
Education
PhD
2014-2018
The University of EdinburghBachelor of Technology
2008-2012
International Institute of Information Technology Hyderabad (IIIT-H)
Work Experience
NVIDIA
2020-present
Senior GPU Memory Architect
Synopsys
2019-2020
Senior Microprocessor Architect
NVIDIA
July-October 2017
GPU Memory Architecture Intern
Synopsys
2012-2014
R&D Engineer
University of Cambridge
May-July 2011
Center for Advanced Photonics and Electronics (CAPE)
Research Intern
Teaching
TA, Introduction to Computer Systems
Fall 2016
The University of EdinburghTA, Computer Design
Fall 2014/15/16
The University of EdinburghTA, Computer Architecture
Spring 2016
The University of EdinburghTA, Computer System Organization
Spring 2012
IIIT HyderabadTA, Embedded Systems-1
Fall 2011
IIIT HyderabadTA, Basic Electronic Circuit
Spring 2011
IIIT HyderabadTA, Digital Logic and Processors
Fall 2010
IIIT Hyderabad