Giacomo Boldini
PhD Student
Ca' Foscari University of Venice
SSV Research Group
Research Fellow
University of Parma, Parma, Italy
2022-01 - 2023-07
Researched and implemented parallel algorithms on NVIDIA GPUs for high-performance computing tasks.
Research theme: Implementation of parallel algorithms on GPU (transl.)
- Researched and implemented a new algorithm for NVIDIA GPUs (CUDA/C++) for computing eigenvalues and eigenvectors of small symmetric, positive semidefinite matrices.
- Studied and utilized Tensor Core architecture of NVIDIA GPUs for reimplementation of proprietary numerical computation algorithms.
- Implemented two ad-hoc solutions with CUDA/C++ to improve performance, achieving a speedup of x2N in execution time for matrix diagonalization.