Home|AllPDCS'13CC'13HPC-UA'13HPC Day'13PDCS'14HPC-UA'14HPC Day'14
HPC-UA on Facebook

High Performance Computing 2012
Kyiv, October 8-10, 2012

Monday

08:30 – 10:00

Registration

@ Building 1, Conference Hall
10:20 – 11:00

Opening Session

@ Building 1, Conference Hall
Presenter(s):
Ivan SergienkoMykhailo Zgurovsky
11:00 – 12:00

ISC Keynote

@ Building 1, Conference Hall
Chair(s):
Oleksii Koshulko
11:00

20 Years TOP500 - An Insider Report

Hans Meuer
11:45

Looking back to ISC'12 in Hamburg, Germany and facing ISC'13 in Leipzig, Germany

Martin Meuer
12:30 – 18:00

Plenary Session

@ Building 1, Conference Hall
Chair(s):
Oleksii Koshulko
12:30

HPC and the Cloud

Dana Petcu
13:00

Entering the Many-Cores Era: Towards a Uniform Programming Model for Heterogeneous HPC

Sergii Gorlatch
14:30

Quantum computing: typifying processing in living organisms

Sergii Lyshevskyi
15:00

Astrophysical Supercomputing with "Green" GPU Clusters

Peter Berczik
15:30

Instrument Element – Grid Service as a Computational Platform for Metrology Tasks

Pavlo Nejezhmakov
16:30

WLCG – world grid-infrastructure for processing data of Large hadron collider

Eugene Martynov
17:00

A (ir)regularity-aware task scheduler for heterogeneous platforms

Artur Mariano

This paper addresses the design, implementation and validation of an eective scheduling scheme for both regular and irregular applications on heterogeneous platforms. The scheduler uses an empirical per-formance model to dynamically schedule the workload, organized into a given number of chunks, and follows the Heterogeneous Earliest Finish Time (HEFT) scheduling algorithm, which ranks the tasks based on both their computation and communication costs. The evaluation of the proposed approach is based on three case studies { the SAXPY, the FFT and the Barnes-Hut algorithms { two regular and one irregular application. The scheduler was evaluated on a heterogeneous platform with one quad-core CPU-chip accelerated by one or two GPU devices, embedded in the GAMA framework. The evaluation runs measured the eectiveness, the eciency and the scalability of the proposed method. Results show that the proposed model was eective in addressing both regular and irregular applications, on heterogeneous platforms, while achieving ideal (100%) levels of eciency in the irregular Barnes-Hut algorithm

17:30

Parallel computations in insurance business optimization

Bogdan Norkin
18:15 – 20:15

Welcome Party

@ Building 1, Alma Mater Cafe

Tuesday

09:30 – 10:00

Registration

@ Building 1, Conference Hall
10:00 – 15:30

Session 1

@ Building 1, Room 143
Chair(s):
Oleksii Koshulko
10:00

Comparison of Noise-Tolerant Architectures of Logic Gates for Nanoscaled CMOS

Sergii Lyshevskyi

High-performance computing can be accomplished by using nanoscaled CMOS. The nanoscaled microelectronic devices and ICs exhibit noise which leads to faults and errors. These errors may be reduced by using advanced design techniques which are covered in this paper. This paper investigates two novel noisetolerant architectures of logic gates based on Binary Decision Diagrams. One of them realizes the Markov Random Field model, and the other is based on an error-correcting scheme. The simulation experiments for the logic networks, which implement the proposed models, using the 16-nm Berkeley CMOS technology model, are reported and compared.

10:20

Processing of Extremely-Large-Data and HPC

Sergii Lyshevskyi

High performance computing implies the ability to effectively, efficiently, consistently, cohesively and coherently access, represent, manage, store, assess, evaluate, compute and process data, thereby enabling analysis, visualization, data mining, extraction,“knowledge generation”, “learning”, “intelligence”, “cognition”, etc. In various applications, processing is accomplished on very large heterogeneous data sets. Exceptional scalable microelectronic hardware solutions, efficient algorithms, advanced architectures, enabling organizations and other solutions were developed and effectively used. However, the nanoscale microelectronics, microelectromechanical systems, photonics and nanotechnology frequently result in diverse, incomplete, incompatible and heterogeneous extremely-large-scale data. The problems of data quality, data complexity, data consistency and data utilization have not being cohesively addressed and solved. The paper proposes coherent data complexity and data quality analyses. These solutions may enable processing of extremely-large-data, contribute to cyber security and enhance overall computing capabilities, such as robustness, reliability, throughput, etc. The macroscopic and quantum probabilistic information- and entropy-centric analyses are performed. Ability to process highly inconsistent, incomplete and heterogeneous data and information is exhibited by living organisms. Engineered systems can be designed with an attempt to typify natural systems ensuring robust, effective, low-power and cohesive processing. The data complexity and data quantity measures, as well as other relevant estimates, are derived using information theory. These measures could be applied in information assurance, information assessment, cyber security, etc. New fundamental and algorithmic solutions are reported. The results of basic, numerical and empirical studies and evaluation are presented.

10:40

Intel® Cluster Studio XE 2013 – a comprehensive toolset for HPC applications development

Igor Lopatin

This paper describes a set of developer tools called Intel® Cluster Studio XE 2013 aimed to improve application performance and developer productivity for development of applications which effectively run on node and cluster level. The suite includes scalable implementation of MPI runtime library, MPI tracing tools, correctness checkers and performance profiler. Also a new tool included in the suite - Intel® Parallel Advisor XE – is being discussed. Using open-source ray tracing application Tachyon as an example we demonstrate how Cluster Studio XE 2013 suite tools could be applied for various phases of application development to achieve better scalability, correctness and per-thread performance for hybrid parallel implementation of the tracing algorithm. By applying Cluster Studio XE tools we were able to improve performance of the application from 2 FPS to 106 FPS on an 8-node cluster of multi-core machines.

11:00

Desktop Grid for the Deployment and Execution of Highly Loaded Webservice Components of Distributed Information Systems

Mykola Alekseev

The article is devoted to the current issues of creating an environment for the place-ment and execution highly loaded webservice components of distributed information systems, which have a dynamically changing heterogeneous architecture. In the paper we study the example of Institute of Telecommunication Systems of NTUU "KPI" as a small scientific and educational institution possibility of constructing on the existing resource base environment for solving computational problems which can be formulated as a set of small interdependent tasks, and for placement and performance of highly loaded components of distributed information systems, developed under the research and production projects. Obtained practical result confirms that such system of parallel computing on Desktop Grids, involving computing resources of unsteady subscribers by providing them with access to paid information resources (electronic publishing, distance learning courses, internet, etc.) can be very effective for computing tasks, consisting of a few mutually dependent tasks, as well as to accommodate and performance of heavy duty components of distributed information systems.

11:20

Problem-oriented approach to MPI applications development cycle

Sergii Stirenko

Applications, highly-loaded in terms of computation, also tend to take significant time during the whole development cycle, namely while debugging on a real-scale system. Profiling techniques aiming to increase software performance are time-consuming themselves, therefore it is important to propose an approach decreasing the total development time for MPI applications and facilitating where possible the work of programmer. We propose it as a combination of framework for stating tasks in terms of problems rather than parallel computing primitives with an analysis tool to tune MPI applications. Used jointly they allow to take control over abstraction levels in the development process.

12:00

Enabling Clang to statically check MPI type safety

Dmytro Gribenko

This paper introduces a static analysis method to check that the actual type of the data buffer and the type specified by MPI library constants match. The method is based on annotations in MPI library header files. Type checking is performed at compile time and has no run-time overhead. The paper also covers implementation of the method in Clang compiler and MPICH2 implementation of MPI.

12:20

Analysis of Molecular Dynamics (MD_OPENMP) on Intel® Many Integrated Core Architecture

Prasad Wadlakondawar

This paper assesses the behavior of Molecular Dynamics (MD) simulation program by using OpenMP on the x86-based Intel® Xeon PhiTM coprocessor. The coprocessor is used to offload highly parallel, compute intensive tasks from the host CPU and examine the ability of this new hardware architecture to offload the target workloads with minimal developer efforts. The applicability of SEP (Sampling Enabling Platform) and Intel(R) VTuneTM Amplifier XE performance analyzer for different hardware matrices has also being tested.

12:40

Parallel algorithms for solving relaxation problems of filtering consolication

Vsevolod Bogaenko
13:00

Effective use of HPC for solving large plasma physics problems

Evgenyi Berendeyev
14:20

Multi-GPU algorithm for estimation of mutual information for molecular dynamics modelling

Pavlo Grubyi
14:40

Experiment investigation of task planning algorithms for grid systems using simulator

Petr Polezhaev
15:00

Supercomputer with preemptive memory management for continuous processing of large data

Eugeny Bryndin
10:00 – 15:30

Session 2

@ Building 1, Room 149
Chair(s):
Peter BerczikSergii Stirenko
10:00

Optimization of transitions in the binary translation of "Elbrus" architecture

Aleksei Rybakov
10:20

Integration Ukrainian grid infrastructure to EGI

Sergii Svistunov
10:40

High performance heterogeneous computing systems for parallel-hierarchical image processing

Volodymyr Mudryk
11:00

One way to generic ISO/IEC 11404 standard type transformations for use in heterogeneous environment

Andrii Stenyashyn
11:20

Use of modified cascade method for series expansion

Lyudmyla Ilicheva
12:00

Resource planning method for distributed heterogeneous system its implementation study

Sergii Minuhin
12:20

Separable parametric algorithms

Sergei Yermakov
12:40

Processor for error-free calculation of the inverse matrix

Anatoly Sergienko
13:00

Experiment planning in the Data Farming system

Olena Krykovlyuk
14:20

Modeling of matrix multiplication problem on GPU

Kostyantyn Zhereb
14:40

Parallel computing in the implementation of reforms in rings of cut polynomials

Kostyantyn Pogrebnyak

The present paper is devoted to application of the finite element analysis for solving electrostatics problems. The job appears when one reconstructs the rock structure by geomagnetic probing. The 3D axisymmetric problems are computed using the first and second order finite elements. The free library Triangle is used for triangulation, the libraries MKL, CUSP and the technologies OpenMP, MPI are used for parallel computing. The proposed implementation of the method operates faster than COMSOL's one with the same mesh. The maximum solution error is not higher than 1%.

15:00

Ways of parallelization of Pollard discrete logarithm algorithms for common memory systems

Kostyantyn Pogrebnyak
15:40 – 18:40

Excursion

Bus and walking excursion around city center. Explore famous Kyiv churches, Dnieper river, Kreschatyk street and try national Ukrainian food!

Presenter(s):
Andrii GolovynskyiAndrii MalenkoGleb Koshulko

Wednesday

09:30 – 10:00

Registration

@ Building 1, Conference Hall
10:00 – 13:30

Plenary Session

@ Building 1, Conference Hall
Chair(s):
Oleksii Koshulko
10:00

Hybrid algorithm of Nordugrid ARC 2.0 broker

Pavlo Svirin
10:20

Hybrid algorithms for solving sparse triangular systems based on methods

Olexandr Khimich
10:40

Use of grid technology for medical images storing

Andrii Malenko
11:00

Assessing the impact of the overhead of the calculations in BOINC-GRID

Evgenyi Ivashko
11:20

Processing of seismic data based on parallel computing using GPUs.

Oleksandr Opanasenko
11:40

Modeling in FlowVision software package of heat transfer processes in fast reactors with sodium coolant

Vladimir Shmelev
12:30

HPC Innovation for Scientific Research at HP

Philippe Trautmann
13:00

SCIT-4 – the new supercomputer of Institute of Cybernetics NASU

Andrii Golovynskyi
14:20 – 16:40

Section 1

@ Building 1, Conference Hall
Chair(s):
Sergii Stirenko
14:20

Study of protil-tRNA synthetase complex with pre-transfer substratum

Oleksii Rayevskyi
14:40

Forming the choice matrix for task distribution in grid

Olexandr Shcherbyna
15:00

Development of gene regulation networks modeling methods for systematic analysis of liver response to the action of interferon alpha in grid

Alina Frolova
15:20

New methods of high speed data authentification

Anatolii Boyko
15:40

Web portal for modeling and monitoring of man-made situation using grid

Vadym Khramov
16:00

Ways to improve computational efficiency by using the apparatus of best Chebyshev approximation

Angelina Kalenchuk-Parkhanova
16:20

Method for synthesis of functional transformation of the differential maximum entropy

Kyrylo Novotarskyi
14:20 – 16:00

Section 2

@ Building 1, Room 143
Chair(s):
Sergii Stirenko
14:20

Creating a virtual management platform computing environment

Vladimir Soloviev
14:40

Full support for typed properties in OWL EL ontologies based on a parallel processor ELK polynomial

Olexandr Pospishnyi
15:00

Software system "Photophoresis and aerosol dynamics"

Vadim Nikitin
15:20

Modeling of global trends in the Sun

Oleksii Loginov
15:40

Choosing a configurable processor architecture for cloud computing

Anatoly Sergienko
16:40 – 17:00

Closing Session

@ Building 1, Conference Hall
Chair(s):
Sergii Stirenko
Presenter(s):
Olexandr Palagin

Please note! Schedule may change for some reason. Please check it now and then.

Powered by MyConf conference management system