| Computer Scientist / Section Chief Advanced Computing Section Aviation Branch Global Systems Division Earth System Research Laboratory 325 Broadway Boulder, Colorado 80305 ph (303) 497-6278 |
![]() |
GPU
Development
I have been leading the research and
development activities in GPU computing. I wrote a compiler to
translate Fortran into CUDA, a high level language used by NVIDIA
GPUs. Details of these activities are described in the accelerator web pages.
Modeling Portal Development
I manage the design and development
of two portals that support modeling activities at NOAA and the
DTC. WRF Domain Wizard is used to configure model domains for
WRF-NMM and WRF-ARW. The NOAA Modeling Portal (formerly WRF
Portal) has been used to support model testing of WRF, and the FIM
model. FIM is currently being run at the Texas Advanced Computing
Center (TACC) to improve hurricane predication.
NOAA Data Systems Integration
Grid Computing and Web Services
In the last decade, grid computing has emerged as a viable technology for high performance computing. I have been exploring the appropriateness of using grid technologies to integrate NOAA's data systems, and HPC centers. Initial exploratory research involved developing a prototype NOAA grid which highlighted the challenges of deploying such a grid at NOAA due to security issues and the need to develop trust relationships between the NOAA HPC computing centers. Further research was done to explore the benefits to NOAA of building and utilizing grid technologies at NOAA. I then wrote a white paper on the topic and circulated it to the NOAA CIO, NOAA HPCC office, and many managers and staff across NOAA.Compiler Development / Language Translation
I developed a compiler, called the Parallel Pre-Processor (PPP) that parallelizes Fortran codes so they can be run on distributed-memory high performance computers. Directives, in the form of Fortran comments, are inserted into the serial code and used to pass parallelization information to the compiler. The compiler then analyzes the code to determine how the serial code should be translated and then produces a MPI-based parallel equivalent version. This generated code relies on the run-time libraries in FSL's Scalable Modeling System (SMS) to perform data decomposition, asynchronous parallel I/O, and communications operations. This compiler has been used to parallelize and debug many weather and climate codes over the last ten years. PPP is currently being used to parallelize a new model being developed at ESRL called the Fluid Flow Following Icosohedral Model (FIM).
Publications (out-of-date)
Before my interest in high performance computing, I was heavily involved in database applications. Specifically, I developed two databases that are widely used by researchers around the world. Both data sets are available on CDrom, and current (RAOB) data can be accessed via a web interface that I also developed. More information can be found from the following links:
Education
Early in my professional career, while working as a aerospace machinist, I helped build instruments that were deployed on satellites used by NASA for earth science studies. Little did I know that my career path would bring me almost full circle such that now I work with the weather and climate codes that use many of these observations.
Prepared by Mark Govett, Mark.W.Govett@noaa.gov
Date of last update:
12-Feb-2007