معرفی
Dr. Ahmad Abdelfattah is a Research Assistant Professor at the University of Tennessee's Innovative Computing Laboratory (ICL). He holds a PhD in Computer Science from KAUST (2015), an MS in Computer Engineering from Ain Shams University (2010), and a BS in Computer Engineering (2006). His research focuses on high-performance computing, parallel numerical algorithms, and GPU acceleration, particularly in dense and sparse linear algebra libraries. He leads the MAGMA library and has contributed to NVIDIA's cuBLAS and AMD's rocBLAS libraries. His work emphasizes exascale computing, mixed-precision arithmetic, and optimizing batched operations on GPUs.
His research spans algorithm design for GPUs, including batched matrix operations, sparse linear algebra, and high-order finite element methods. He has developed libraries like libCEED and SLATE, targeting exascale systems. His contributions address challenges in memory-bound computations, irregular patterns in matrices, and efficient use of tensor cores.
Key trends in his publications include optimizing GPU kernels for small matrices, exploring mixed-precision techniques for solving linear systems, and enabling portable performance across diverse architectures. His work bridges theoretical numerical methods with practical implementation on modern hardware, aiming to accelerate scientific applications.
Awards and recognitions include acknowledgments from NVIDIA and AMD for contributions to their BLAS libraries. His advising and grants involve advancing exascale computing through collaborations with ECP (Exascale Computing Project), focusing on CEED-enabled applications and performance tuning. He leads initiatives to enhance software tools for high-order discretizations and scalable linear algebra solutions.
Abdelfattah is affiliated with the Innovative Computing Laboratory, contributing to both academic and industrial partnerships. His research impacts fields like computational science, computer vision, and robotics through optimized algorithms and libraries.
