Roy Spliet, MSc.

Postgraduate Researcher
University of Cambridge
rspliet <at> eclipso <dot> eu
Download CV (NL)

About me

I recenty submitted my dissertation for the degree of Doctor of Philosophy, titled A SIMD accelerator for hard real-time systems. Currently I am a postgraduate researcher in the Computer Architecture research group of the University of Cambridge, under the supervision of Dr. Robert Mullins.

Previously I finished a research internship at the Max Planck Institute for Software-System, working in the field of Real-Time Systems after graduating from Delft University of Technology in the area of GPGPU.

My area of interest spans the intersection of architecture and systems, ranging from understanding of the hardware details of computational devices like GPUs to understanding and desiging hardware support for operating systems. With a background in parallel systems I am particularly motivated in understanding the different architectures in the GPGPU space.

Publications

July 2018
Cite Code
"The case for limited-preemptive scheduling for GPUs in hard real-time systems", R. Spliet, R.D. Mullins, Operating System Platforms for Embedded Real-Time Systems (OSPERT).

May 2016
Cite
"Conquering the complexity mountain: Full-stack computer architecture teaching with FPGAs", A.T. Markettos, S.W. Moore, B.D. Jones, R. Spliet, V.A. Gavrila, European Workshop on Microelectronics Education (EWME).

December 2014
Cite
"Fast on Average, Predictable in the Worst Case: Exploring Real-Time Futexes in LITMUS^RT", R. Spliet, M. Vanga, B.B. Brandenburg, S. Dziadek, Real-Time Systems Symposium (RTSS).

March 2014
Cite Code
"KMA: A Dynamic Memory Manager for OpenCL", R. Spliet, A. Varbanescu, B.R. Gaster, L.W. Howes, GPGPU7 Workshop.

Presentation

September 2016
Link
"Motivating preemptive GPU scheduling for real-time systems", R. Spliet, X.org Developers Conference (XDC).

Other contributions

Linux kernel Nouveau OSS driver for NVIDIA GPUs: Contributions in understanding power management functionality, implementation of frequency scaling for NVIDIA ION and late Tesla. Various compiler fixes and optimisations.

Allwinner SoC drivers: Contributions to implement NAND boot and improvements to NAND chip support.

LITMUS^RT downstream real-time scheduling extension: Enhancements and fixes for ARM and other platforms.

Ramulator (DRAM simulation model): Fixes for DDR4 timing constraints.

GPGPU-Sim (GPU simulator): Improvements to OpenCL support.

Drupal (OSS Content Management System): Module prototypes for extended functionality.