Ana Lucia Varbanescu – Delft University of Technology


Dr. Ana Lucia Varbanescu got her PhD in 2010 from Delft University of Technology, on “Effective Parallel Programming
of Multi-Core Processors”. She has worked on performance analysis and prediction for various multi- and many-core
architectures, including the somewhat exotic SpaceCAKE (from Philips) and STI Cell/BE (from IBM), and later on multi-core
general purpose architectures (from Intel/AMD) and GPUs (fron NVIDIA/AMD). She has been an intern at IBM TJ Watson, USA (twice),
working on the Cell/B.E., and at NVIDIA, USA, working on highly optimized FFTs. She has received a HPC-Europa Young Researcher
Grant to work at the Barcelona Supercomputing Center, Spain, where she analyzed several programming models for many-cores.
In 2008, she has also been a Finalist of Google’s Anita Borg competition.

In 2009, she received an NWO personal grant (in the IsFast Call for Proposals) for the project “ARTCube – From Performance to
Accuracy in Astronomy Computing: A Case for Multi-Core Processors”, a project that investigates the correlation between the Accuracy
of scientific astronomy applications, available Resources, and expected execution Time. The target applications are astronomy
image processing applications (e.g., galaxy fitting and radio-astronomy image synthesis).

Since 2009, she works both at TUDelft and VU University Amsterdam. In the Parallel and Distributed Systems group at TUDelft,
she works in a team that includes eight PhD and MSc students, with interests ranging from memory allocation for massively parallel
heterogeneous architectures to application characterization, large scale graph processing on multi-layered parallel architectures
and programming models. In the Computer Systems group at VUAmsterdam, she is part of a team that works on parallel application
design and implementation for many-core architectures, focusing on performance and energy efficiency, and on programming models
for many-cores.

Her research interests include (in random order): (1) analyzing programming models for many-/multi-cores, (2) performance analysis and prediction of
parallel applications on many-/multi-cores, (3) the use of modern hardware architectures for (radio-)astronomy, and (4) the use
of massive parallelism for large scale graph processing.

Talk: OpenCL against the world: a battle for performance