| OBJECTIVE | |
Increasing the understanding and use of hybrid parallel computer
architectures through design of algorithms, implementation of applications,
and outreach and training for interesed new users.
|
|
|
| |
| EXPERIENCE | |
|
| Advanced Micro Devices (AMD) |
Austin, TX |
|
7/2008 -
Present | |
Senior Member of Technical Staff
Duties:
- Using CPU performance counters to understand application bottlenecks in the microarchitecture
- Pre-sales analysis and support
- Competitive analysis of Intel systems
- Contribution of a new application to SPEC MPI benchmarks
|
|
NUMA and CPU Affinity
| |
- Adjustable NUMA and CPU affinity aware Pthreads implementation of STREAMS benchmark
- Implemented benchmark for nested heterogenous parallelism using OpenMP
- Wrote parallel counting sort benchmark for testing atomic updates and nested
parallel regions
|
|
Oil & Gas Applications
| |
- 4-5x speedup on Reverse Time Migration benchmarks
- Pre-sales application and competitive analysis and customer support
|
|
Bio Informatics
| |
- Parallelization of Bio Informatics applications using OpenMP
- Benchmarking and analysis of BioBench
|
| |
| |
|
|
4/2006 -
5/2008 | |
Applications Manager
Duties:
- Analysis and implementation of user codes for FPGA accelerated computers
- Market analysis for business, government, and academic markets
- Authoring and editing of manuals, documentation, code examples, and running training workshops
|
|
Optimization & Implementation
| |
- Helped develop Mitrion BLAST which achieved a 15-60x speedup over the software only version
- Implemented convolution and other stencil algorithms
- Implemented parallel reduction algorithms
|
|
Customer and Partner Support
| |
- Naval Research Laboratory -- Analysis and parallelization of user applications for FPGA based machines. Managed sales and support of user account.
- Raytheon -- Assist in the implementation of remote sensing and image processing algorithms, organized and ran training workshops.
- Reservoir Labs -- Support compiler re-targeting project to enable the R-Stream compiler to target the Mitrion Virtual Processor.
- Training Workshops: University of Arizona, George Washington University, McGill University, Reconfigurable Systems Summer Institute, French Ministry of Defense, Raytheon
|
| |
| |
|
|
| 1998 - Present | |
CEO, Owner
Parallel computing consulting and software
- Subtractor -- A restructuring/optimizing data extraction tool and calculator
for weather and climate models and other gridded datasets.
User queries
are restructured and optimized for unindexed sequential record files
generated by terabyte scale climate simulations.
Optimized queries are executed using dynamic nested parallel regions
implemented using Pthreads.
- Hum -- Interactive gigapixel hyperspectral motion picture program.
Display latency is masked
using multi-threaded load-decompress-decode-display pipelines.
Speculative frame decoding is used to keep pipelines full,
and images are stored in a modified LRU cache.
- IOTA -- Input/Output Test Apparatus. A filesystem performance and correctness
test which implements a parallel-serial-parallel nested loop over files,
sequential I/O operations, and number of stripes per operation, respectively.
- HMV -- X-windows 2D slice viewer and processor for
model output viewing and relative debugging.
|
| |
| |
|
|
| Tera Computer Company/Cray Inc. |
Seattle, WA |
|
1/1999 -
10/2005 | |
Senior Applications Analyst
Duties:
- Parallelization of user applications, system tools and runtime
- Design and implement new algorithms and tools for Tera/Cray and customers
- Work with customers to identify, prioritize, and communicate problems
Achievements and Projects:
|
|
Parallelization and Optimization
| |
- Parallelization of user codes ranging in size from 5 line benchmark loops
to 6 million line FORTRAN dusty decks. Applications include
climate, weather and ocean stand alone and coupled models,
image processing,
gas, fluid, solid, and electromagnetic dynamics,
scientific visualization, linear algebra,
and signal processing.
- Scalable, efficient graph problem solutions --
Worked on design and implementation of
linearly scalable parallel routines to solve for the shortest path and
connected components of a graph by exploiting both task and data parallelism.
- Parallel processing of record oriented files --
Re-factored programs
to afford parallel processing of multi-gigabyte formatted data files
by rewriting I/O as several parallel phases.
- Speedups in data consistency checks --
Achieved 1500-fold speedup in filesystem consistency tests
by moving dependencies in a linked list data structure
out of a loop and parallelizing updates to a large bit vector.
|
|
Site Applications Analyst
| |
- Naval Research Laboratory -- Parallelization of user applications
and communication conduit to Cray management.
- Applied Technical Systems -- Consulted on implementation of parallel non-relation database.
- San Diego Supercomputer Center -- Parallelization of user applications.
|
| |
| |
|
|
| Science Systems and Applications, Inc. |
NASA -- Goddard Space Flight Center |
|
1/1997 -
12/1998 | |
Software Engineer
- Design and implement interactive visualization tools for GOES and TRMM satellites.
- Support implementation of rain assimilation scheme into MM5
|
| |
| |
|
|
| Argonne National Laboratory |
Math and Computer Science Division Argonne, IL |
|
1/1993 -
12/1993
5/1994 -
8/1994
7/1995 -
12/1996 | |
Intern, Research Assistant
- WRF, MM5, PCCM2 -- Parallelization and maintenance of NCAR/PSU weather and climate models
- Remote I/O -- Design and implementation of parallel remote I/O library which later became part of MPI-IO
- Medical Imaging -- Implement parallel digital mammography program for Kurt Rossman Lab at the University of Chicago
- PCN, FortranM, Nexus -- Support runtime system for Program Composition Notation and Fortran based parallel programming languages
|
| |
| |
|
| EDUCATION | |
|
| University of Pittsburgh |
Pittsburgh, PA |
|
| 1995 | |
M.S., Computer Information Science
Research Adviser: Dr. Robert Korfhage
Using Genetic Algorithms for Weighted Key Based Document Retrieval
|
| |
| |
|
|
| Bradley University |
Peoria, IL |
|
| 1992 | |
B.A., Computer Information Systems
|
| |
| |
|
|
|
| |
| COMPUTER | |
|
Programming languages and tools
| |
Fortran, C/C++, MPI, OpenMP, Pthreads, M4, Make,
CVS, LaTeX, csh, awk, sed, perl, HTML, SQL
|
|
Operating systems
| |
Linux, AIX, UNICOS, Solaris/SunOS, Irix, MTX (Tera MTA), NEC
Super/UX, Mac OS-X
|
|
Hardware
| |
- Vector -- Cray X1, T90, SV1, NEC SX
- Shared Memory -- Tera MTA; SGI Origin & Altix; x86, Sparc and IBM Power SMPs and multicore
- Cluster -- Cray XD1, T3D; Linux clusters; IBM Power series
|
|
|
|
| |
| PUBLICATIONS | |
- "EnVis/Hum: High-resolution Ocean Model Visualization and Display", Michael
A. Chupa, Jace A. Mogill, Derek W. Irby, Robert J. Moorhead, Jay F. Shriver,
and Peter M. Flynn. Association of American Geographers 100th Annual Meeting, 2004.
- "Remote I/O: Fast Access to Distant Storage", Ian Foster, David Kohr,
Rakesh Krishnaiyer, Jace Mogill. Proceedings of the Fifth Annual
Workshop on I/O in Parallel and Distributed Systems (IOPADS'97).
- "Multiple Load Images in PCN", Jace A Mogill. DOE/Argonne National
Laboratory Technical Memorandum, 1993. Also distributed with PCN
compiler and runtime releases.
| |
|
|
| |