John Gunnels

John Gunnels is a distinguished engineer, working on mathematical libraries at NVIDIA. His current focus is on matrix-based floating-point emulation methods. John has a Ph.D. in Computer Science from The University of Texas at Austin.
Avatar photo

Posts by John Gunnels

Developer Tools & Techniques

Unlocking Tensor Core Performance with Floating Point Emulation in cuBLAS

NVIDIA CUDA-X math libraries provide the fundamental numerical building blocks that enable developers to deploy accelerated applications across multiple... 11 MIN READ