Ayaz's Homepage

Ayaz Ali received his PhD in Computer Science from the University of Houston in 2008. He earned his MS from the same university in 2006. He finished his BS in Computer Systems Engineering from GIK Institute in 2000 and worked for TPS as Systems Consultant till 2003. Since 2004, he has been affiliated with the Texas Learning & Computation Center as a Research Assistant. He has also held internship positions with Hewlett Packard and Intel Corporation during the summer of 2006 and 2007, respectively. Broadly, his research interests lie in the area of high performance and parallel computing. One of the main goals of his current research is to develop new techniques for portable, adaptive high performance libraries and code generators for multi-core/many-core and hybrid architectures. (cv.pdf, cv.txt)

Research

With the growing gap between the on-board processing speed and the memory speed, achieving high efficiency in scientific codes has become a big challenge for the high performance computing community. The challenge to achieve high efficiency is now becoming more complex through the emergence of multi-core architectures and architectures using accelerators or special purpose processors for certain tasks, such as FPGAs, GPUs or the Cell Broadband Engine. To address the challenge in performance critical functions; domain specific tools and compilers have been developed. Over the last several years an evolution has taken place towards carrying out final optimizations at runtime.
Through my current and future research, I will attempt to address the issues related to the development of new techniques for adaptive and portable libraries for scientific applications in the context of current technological shift in scalable computing towards multi-core/many-core architectures. This shift presents a particularly difficult challenge to the HPC community, which has been relying on scientific codes that were written more than a decade ago. While there is a need to port existing systems to the new generation of architectures, better programming techniques, paradigms and languages need to be researched that can withstand any future hardware transitions to ensure longer lifespan of performance portability. (full research statement)

UHFFT

Publications and Presentations

Useful Links

Course Work and Projects

Fall 2004
Advanced Parallel Processing. [UHFFT-smp: OpenMP and PThreads Implementation]
Advanced Numerical Analysis. [ODE & PDE solvers: FORTRAN code]
Bioinformatics. [GABA: Genetic Algorithm Based Sequence Alignment Implementation]

Spring 2006
Advanced Compilation for Vector & Parallel Processors @ Rice University.

References

Dr. Lennart Johnsson
Director Texas Learning & Computation Center
University of Houston

Dragan Mirkovic
Asst. Professor of Radiation Physics
M. D. Anderson Cancer Center
The University of Texas

Peter Tang
Sr. Principal Engineer
Software and Solutions Group
Intel Corporation

Joe Hoch
Manager, Thin Client Software Development
Personal Systems Group
Hewlett Packard Company, Houston