Muhammad Usman Ashraf
Papers
1
Total Citations
11
H-Index
1
About
Dr. Muhammad Usman Ashraf is a leading researcher in high-performance computing (HPC) and sparse matrix computations, with a particular focus on optimizing the sparse matrix–vector product (SpMV)—one of the seven numerical dwarfs critical to scientific and analytical applications. His most cited work, "AAQAL: A Machine Learning-Based Tool for Performance Optimization of Parallel SpMV Computations Using Block CSR" (2022, 11 citations), introduces an innovative machine learning-driven approach to automatically select the best block size for the Block Compressed Sparse Row (BCSR) format, significantly accelerating SpMV on modern parallel architectures. This contribution addresses a fundamental bottleneck in solving large sparse linear systems, directly impacting fields from computational physics to data analytics. Dr. Ashraf’s research bridges the gap between traditional HPC optimization and modern AI techniques, demonstrating how machine learning can automate performance tuning for irregular computations. His work has been recognized for its practical impact, providing tools that enable researchers and engineers to achieve near-optimal performance without manual tuning. By advancing the efficiency of core numerical kernels, Dr. Ashraf continues to shape the future of scalable scientific computing.
Research Focus
Key Achievements
Top Papers
- 1