Mantas Mikaitis

Department of Mathematics

Alan Turing Building

University of Manchester

Oxford Road, Manchester

M13 9PL, UK


Photo: next to SpiNNaker machine.

I am a Research Associate with the Numerical Linear Algebra Group at the University of Manchester, working with Professor Nicholas J. Higham. I received a B.Sc. (Hons.) degree in Computer Science in 2016 and a PhD degree in Computer Science in 2020, both from the University of Manchester.

During my doctoral degree I was part of two neuromorphic computing projects, called SpiNNaker and SpiNNaker2, that aim to build efficient spiking neural network (SNN) simulators. There I explored numerical aspects and performance of SpiNNaker software and developed hardware accelerators for elementary functions in SpiNNaker2.

My research interests include: computer arithmetic (fixed- and floating-point software/hardware arithmetics), numerical linear algebra, high-performance computing, mathematical software, performance optimization and benchmarking.

Main programming and software skills: C, MATLAB, Python, CUDA, Verilog, ARM assembly, Git, LaTeX, Bash.

I occasionally contribute to the NLA group blog.


Mar 31, 2021 Anymatrix, a MATLAB matrix collection with an interface to organize matrices, was released today in v0.1. Collection developed in collaboration with Prof. Nicholas J. Higham.
Mar 27, 2021 A new paper on efficiently simulating randomized rounding in floating-point arithmetics, co-authored with Max Fasi, accepted for the ARITH 2021 conference. To appear in a special issue of IEEE TETC. Early access here.
Mar 2, 2021 As part of SIAM CSE21, which happened virtually between 1-5 March, 2021, I presented a poster on the arithmetics of the latest GPUs (PDF) and a talk on the implementation of probabilistic rounding (check out the proceedings of a minisymposium on Reduced Precision Arithmetic and Stochastic Rounding here).
Nov 23, 2020 I wrote a new blog post on simulating stochastic rounding in IEEE 754-compliant arithmetics. It is published here.
Nov 12, 2020 A new paper titled “Numerical Behavior of NVIDIA Tensor Cores” that I co-authored with Max Fasi, Nick Higham and Srikara Pranesh was accepted to appear in PeerJ Computer Science. Preprint is available here.