Ji Kim
Ph.D. Student, ECE
Office: 364 Upson Hall, Ithaca, NY 14853 USA
Email: ji(at)csl(dot)cornell(dot)edu
My research focuses on fine-grain heterogeneous architectures with support for seamless adaptive execution of amorphous data parallel tasks across general-purpose and specialized accelerator cores. While we are reaching the limits of what we can accomplish by simply adding more cores onto a chip, computer architects are turning towards new abstractions to take advantage of the growing transistor density. Although current approaches of coarse-grain heterogeneity are promising, there are significant challenges with unifying the parallelism abstraction, support for programming APIs and compilers, as well as efficient adaptivity. I am interested in exploring potential unifying architectures for exposing, scheduling, and executing fine-grain parallel tasks to improve performance and energy efficiency on applications with varying levels of amorphous data parallelism. Part of this vision is innovating designs of data-parallel accelerators using mechanisms to eliminate redundancy and improve load balancing.
Research Interests
Fine-Grain Heterogeneous Architectures, Energy-Efficient Data-Parallel Accelerators, Amorphous Data Parallelism, Vertically Integrated Design Methodology
Education
- Doctor of Philosophy (Ph.D.), Electrical & Computer Engineering (Computer Architecture), 2010-Present
- Masters of Science in Engineering (MSE), Electrical Engineering, 2006-2010
- Bachelor of Science in Engineering (BSE), Electrical Engineering, 2006-2010
Publications
- Ji Kim and Christopher Batten. Accelerating Irregular Algorithms on GPGUs Using Fine-Grain Hardware Worklists. 47th IEEE/ACM Int'l Symposium on Microarchitecture (MICRO-47), December 2014.
- Ji Kim, Chris Torng, Derek Lockhart, Shreesha Srinath, and Christopher Batten. Microarchitectural Mechanisms to Exploit Value Structure in SIMT Architectures. 40th ACM/IEEE Int'l Symposium on Computer Architecture (ISCA-40), June 2013.
Presentations
- Ji Kim. Exposing, Scheduling, and Executing Fine-Grain Parallel Tasks. Thesis Proposal, Cornell University, October 2014.
- Ji Kim. Morph Algorithms on GPGPUs. Paper Presentation for Cornell Systems Lunch, Cornell University, September 2013.
- Ji Kim. Complexity-Effective Heterogeneous Specialization for Explicit-Loop Architectures. Poster for ECE Visit Day, Cornell University, April 2012.
Previous Research Experience
- Implementation of Computation Group, Professor Andre DeHon, 2009-2010,
University of Pennsylvania
- Research on lifetime performance modeling of self-correcting FPGAs based on incremental timing-analysis of routing faults.
Professional Experience
- Emcore Broadband, Summer Intern for R&D Division, Summer 2008, Alhambra, CA, USA
- Applied Optoelectronics, Summer Intern for R&D Divison, Summer 2007, Houston, TX, USA
Teaching Experience
- ECE 4750 - Computer Architecture, Fall 2010, Fall 2011
Coursework
- ECE 5715: Computer Hardware Design Practicum
- ECE 5720: Parallel Computer Architecture
- ECE 5750: Advanced Computer Architecture
- ECE 5950: Complex Digital ASIC Design
Honors
- 2012 NDSEG Fellowship
- 2012 ECE Director's TA Award
- Awarded to best TA of academic year.
- 2010 Cornell University Fellowship
- Selective one-semester graduate fellowship.
- 2010 Honorable Harold Berger Award
- For technical innovation and entrepreneurial possibility of senior design project, Voice over IP with Enhanced Resiliency (VIPER).
- 2010 SEAS Senior Design Competition Honorable Mention
- Hugo Otto Wolf Memorial Prize
- Awarded to one senior student in each engineering major.
- IEEE Student Award
- For great contribution to the IEEE student branch.
- President of IEEE student branch at the University of Pennsylvania
- President of Eta Kappa Nu Lambda Chapter
- Tau Beta Pi
- Dean's List 2006-07, 07-08, 08-09, 09-10
