Ana Balevic – LIACS


AnaBalevicGTC2013Ana Balevic is passionate about parallel programming with strong interest in CUDA and High Performance Computing. As an active CUDA developer since the early CUDA days, she worked on acceleration of electromagnetic simulations on GPUs, designed and implemented novel parallel algorithms for entropy coding, such as variable length encoding (Huffman), multimedia codecs, and also lead the team that developed CUJ2K, which resulted in the world’s first publicly available JPEG2000 CUDA encoder. In addition, she organized a workshop on Multicore Programming and Threading Games for Performance with Intel, gave CUDA lectures & tutorials, and supervised numerous smaller CUDA projects.

Since 2009, Ana has been working on automatic parallelization techniques in the polyhedral model in close cooperation with Compaan Design b.V. and ACE Compilers, within MEDEA+ European project on Tera-Scale Multi-Core Processor Architectures (TSAR). During the project, Ana designed and implemented a Java-based tool for automatic transformation and mapping of Kahn Process Networks onto GPU accelerators, called KPN2GPU, as an extension to the Compaan compiler. She is currently writing up her thesis, which deals with automatic mapping of streaming multimedia applications onto heterogeneous platforms with data parallel accelerators.

Ana is highly enthusiastic about accelerating applied research problems on GPUs and helping to achieve groundbreaking results faster by parallelizing computationally demanding algorithms. In her free time, Ana contributes to GPU acceleration and code-optimizations in the SpikeFun project, which is a biological neural network simulator of the mammalian thalamo-cortical system.


E-mail: ana.balevic (at)
CUDA Codecs:
SpikeFun Project:

Talk: Mapping Streaming Multimedia Applications onto Heterogeneous Platforms with Accelerators