E-Mail	idooley2@uiuc.edu
Phone(O):	217-333-5827
Phone(H):
S-mail:	4103, Siebel Center for Computer Science, 201 N. Goodwin Avenue, Urbana, IL 61801-2302

Isaac Dooley is a Department of Energy High Performance Computer Science Fellow who received his M.S. Degree from the University of Illinois in Computer science and his B.S. in Mathematics and Computer Science from Birmingham-Southern College. He now is pursuing a Ph.D. at the University of Illinois. He works with his advisor Professor Kale in the Parallel Programming Laboratory.

Here is a printable resume.

Existing parallel programming models and languages focus on the decomposition of data structures and control flow. The parallel portion of the application is often a library (e.g. MPI) used by the application. The library simply performs actions on behalf of the application such as parallel task creation and communication of data through messages or a shared namespace. In these existing parallel systems, the parallel runtime libraries do not influence or change the behavior of the application itself. Runtime systems, however, are uniquely poised to observe characteristics of a parallel program's execution as it runs in order to dynamically change the behavior of the application to increase its performance. Some existing parallel systems such as Charm++ already observe characteristics of a parallel program's execution in order to dynamically balance computation or communication load, but no mechanisms exist for the runtime system to control other behaviors of the application. My dissertation research investigates the creation of mechanisms for applications to expose information about their behavior to the runtime system. The runtime system can tune these application parameters to modify the behavior of the application in response to observed performance characteristics.

I have developed portions of the ParFUM framework and the SIERRA framework. Currently ParFUM is the only known parallel framework which supports asynchronous incremental mesh modification in parallel. Additionally, I have developed tools for predicting the performance of real world applications on new not-yet-existing supercomputer designs. My approach uses partial-execution on existing processors or cycle-accurate simulators to build a model for the execution time for sequential execution blocks. The model is then used to predict the new times for all sequential execution blocks.

Publications:

1. 05-12 Orion Lawlor, Hari Govind, Isaac Dooley, Michael Breitenfeld, and Laxmikant Kale; Performance Degradation in the Presence of Subnormal Floating-Point Values;  in In OSIHPA Workshop, at PACT05, September 2005.
2. 05-14 Orion Lawlor, Sayantan Chakravorty, Terry Wilmarth, Nilesh Choudhury, Isaac Dooley, Gengbin Zheng, and Laxmikant Kale; ParFUM: A Parallel Framework for Unstructured Meshes for Scalable Dynamic Physics Applications;  in Engineering with Computers, Journal, 2006
3. 06-13 Isaac Dooley and Laxmikant Kale; Quantifying the Interference Caused by Subnormal Floating-Point Values;  in In OSIHPA Workshop, at PACT06, September 2006.
4. 06-17 Isaac Dooley; Automated Source-to-Source Translations to Assist Parallel Programmers;  in Master's Thesis, Department of Computer Science, University of Illinois 2006
5. 07-07 Isaac Dooley and Sandhya Mangala and Laxmikant Kale and Philippe Geubelle; Parallel Simulations of Dynamic Fracture Using Extrinsic Cohesive Elements;  in PPL Technical Report July 2007
   This paper is superceded by this journal article.
6. 08-04 Isaac Dooley, Chao Mei, Laxmikant V. Kale; NoiseMiner: An Algorithm for Scalable Automatic Computational Noise and Software Interference Detection;  in In Proceedings of HIPS Workshop at IEEE International Parallel and Distributed Processing Symposium 2008
7. 08-14 Isaac Dooley and Laxmikant V. Kale; Control Points for Adaptive Parallel Performance Tuning;  in PPL Technical Report 2008
8. 08-15 Isaac Dooley and Sandhya Mangala and Laxmikant Kale and Philippe Geubelle; Parallel Simulations of Dynamic Fracture Using Extrinsic Cohesive Elements;  in Springer's Journal of Scientific Computing Volume 39 Number 1, April 2009
   This paper supercedes an earlier PPL technical report.
9. 09-07 Aaron Becker, Isaac Dooley, and Laxmikant Kale; Flexible Hardware Mapping for Finite Element Simulations on Hybrid CPU / GPU Clusters;  in SAAHPC : Symposium on Application Accelerators in HPC, 2009
10. 09-09 Isaac Dooley, Chee Wai Lee, and Laxmikant Kale; Continuous Performance Monitoring for Large-Scale Parallel Applications;  in 16th annual IEEE International Conference on High Performance Computing (HiPC 2009)

• My DOE HPCSF Page
• My Personal Page