Od srede, 16. oktobra, do petka, 18. oktobra 2013, bo na Univerzi v Ljubljani, Fakulteti za računalništvo in informatiko gostoval Nodari Sitchinava, podoktorski raziskovalec na Tehnološkem inštitutu Karlsruhe, ki bo z novim letom prevzel mesto profesorja na Univerzi na Havajih. Nodari se ukvarja predvsem z vzporednimi algoritmi, podatkovnimi strukturami in modeli računanja.
V četrtek ob 14:15 v PR03 na Tržaški 25 bo v sklopu FRI seminarja imel predavanje z naslovom Locality-conscious parallel algorithms.
As we solve problems with bigger input sizes, the time it takes to access data is no longer uniform as the traditional algorithmic models teach us. The modern systems are designed with deep memory hierarchies which include multiple levels of caches, NUMA DRAM design and slow but large disks available via virtual memory. The non-uniformity of memory access is even more apparent in parallel architectures: added computational power of the additional processors/cores increases the gap between the computational throughput and the memory throughput.
In this talk, we will present a theoretical model and several techniques for designing algorithms for multicores under non-uniform memory access. The theoretical analysis of algorithms in this model more closely correlates with the actual runtimes on modern multicores, compared to traditional parallel models, such as PRAM.
We will also show that similar techniques can be used to design and analyze algorithms for general purpose graphics processing units (GPGPUs) and result in better overall GPGPU implementations.
Nodari Sitchinava received Bachelor and Master degrees in Electrical Engineering and Computer Science from MIT and a PhD in Computer Science from the University of California, Irvine. He is currently a postdoctoral researcher at the Institute for Theoretical Informatics at Karlsruhe Institute of Technology. From January he will be starting as an assistant professor at the University of Hawaii.
His research concentrates on developing accurate models of computation for modern parallel architectures and designing algorithms for them. In particular, his PhD dissertation concentrated on combining cache-efficiency with parallelism for multi-core architectures and on the development of a number of fundamental combinatorial, graph and geometric algorithms in the new model.