2014 Poster Sessions : Energy Efficient High Performance Computing

Student Name : Milad Mohammadi
Advisor : William Dally
Research Areas: Computer Systems
The goal of this project is to design an energy efficient high performance computing architecture. Processor on-chip data movement and out of order execution bookkeeping overheads are the key energy consumption factors in modern supercomputers. We aim to minimize these energy overheads via introducing clever compiler-generated latency hiding techniques that enable higher performance while maintaining low energy overhead. Specifically, we introduce the notion of Phraseblock Execution where code blocks can run out of order while instructions within them run in program order.

Milad Mohammadi is a PhD candidate in Computer Architecture at Stanford University and is being advised by Professor Bill Dally. He completed his MS degree in Computer Software at Stanford and his BS degree in Microelectronics and Nano-Systems at the University of British Columbia in Vancouver, Canada. He is the recipient of Stanford Electrical Engineering Fellowship and Natural Sciences and Engineering Research Council of Canada (NSERC) MS and PhD fellowships. Milad can be contacted at milad@cva.stanford.edu.