Invited Lecture: Optimizing the Memory System Performance

Primary tabs

Date: Thursday, December 8, 2022
Time: 11:00 a.m. - 12:00 p.m.
Location: Klaus 2447
Speaker:  Dr. Nishil Talati
Speaker’s Title: Assistant Research Scientist
Speaker’s Affiliation: Computer Science and Engineering at the University of Michigan
​​​​​​​Seminar Title: Optimizing the Memory System Performance

Today’s explosive data growth has ushered a new generation of applications that transform massive, unstructured, heterogeneous data into actionable knowledge. Data is increasing exponentially in volume, velocity, variety, and complexity. On the other hand, the latency of memory units that store and access this data has remained almost constant throughout the years. Therefore, today’s memory system cannot keep up with the growing demands of data-intensive applications. As a result, commodity hardware  platforms are spending most of the time stalled on retrieving data from memory. In this talk, I will present my research effort in optimizing the memory system performance of a variety of data-intensive application. First, I will present Prodigy [HPCA 2021 Best Paper] that uses a hardware-software co-designed solution to improve the memory system performance of data-indirect irregular workloads. Prodigy proposes a compact, yet efficient representation of program semantics that communicates key workload information from software to hardware. Using compiler analysis and hardware prefetching, Prodigy improves the end-to-end performance of irregular workloads by more than 2.5x on CPUs. Second, I will present Mint [MICRO 2022]that builds an asynchronous programming model and a hardware accelerator architecture for speeding up an emerging graph workload of mining temporal motifs. Mint presents a task-centric programming model to unlock high degrees of parallelism, and a programmable accelerator architecture that improves application performance by 10-2500x.

Biographical Sketch of the Speaker: 
Dr. Nishil Talati will be joining the CSE department of University of Michigan as an Assistant Research Scientist (a research faculty position) starting January 2023. He recently completed his PhD from University of Michigan, where he worked with Ronald Dreslinski and Trevor Mudge. Nishil’s research interests span computer architecture and compiler design for improved memory system performance. During PhD, Nishil’s work mostly focused on hardware-software co-design to optimize a variety of graph applications. His first PhD work, Prodigy, was recognized as the Best Paper at HPCA 2021.



  • Workflow Status:Published
  • Created By:dwatson71
  • Created:12/01/2022
  • Modified By:dwatson71
  • Modified:12/01/2022