Seminārs var noderēt gan tiem, kas interesējas par kompilatoru būvi, programmu statisko analīzi, dinamisku koda pielāgošanu, gan tiem, kas interesējas par paralēliem skaitļojumiem un efektīvu lietojumprogrammu izstrādi daudzkodolu procesoriem, GPGGPU iekārtām un datoru klasteriem.   Semināra dalībnieku kodolu veidos Lielbritānijas pētnieku grupa 15-20 cilvēku sastavā, kas pēta un veido kompilatorus paralēlām arhitektūrām.   Zemāk ir Artjoma Šinkarova ielūgums uz pasākumu. Artjoms Šinkarovs 2009.gadā beidza LU maģistratūru. Pašlaik viņš strādā un mācās kā doktorants Hertforširas Universitātē (University of Hertfordshire) Lielbritānijā. Artjoms ir šī pasākuma iniciators.       Dalība ir bezmaksas.       Ielūgumu var pārsūtīt citiem potenciālajiem interesentiem gan Latvijā, gan ārpus tās.       Ar cieņu,   prof. Guntis Arnicāns  

ADVANCE development workshop

 

ADVANCE [1] is an EU framework 7 [2] project which investigates the use of statistical programme properties for dynamic code adaptations.

This workshop constitutes an open meeting where both, project partners as well as interested other parties, can present and discuss their latest developments.

 

Topics of interest include but are not limited to:

   * adaptive mapping of applications to resources

   * dynamically adapting code / just-in-time compilation

   * application-specific language extensions

   * runtime techniques for collecting statistical data

   * novel compilation techniques

   * tool chain infrastructure measures

   * performance analyses

 

The workshop will be rather informal permitting for extensive technical discussions. We expect the majority of talks to be related to the core technologies of the ADVANCE project, i.e., the tool chains for the programming languages SaC [3] and S-Net [4].

 

The tool chains for SaC and S-Net both attempt to tackle the problem of auto-parallelisation.

 

SaC [3] is a functional data-parallel language designed to address computational-heavy problems. SaC follows an idea of implicit parallelisation, meaning that a program is written once and it is up to the compiler to generate an efficient code for a given architecture. Currently SaC can efficiently address multi-core architectures and GPGGPU (specifically CUDA). Computational scientists could treat SaC as an advanced version of Fortran.

 

SNet is a coordinational language which allows to express a program as a network consisting of computational kernels written in an arbitrary language. The SNet program describes how the computational kernels are interconnected. SNet can address clusters using MPI and multi-core machines using threads.

 

However, contributions of a more general nature or in the context of different tool chains are highly welcome as well.

 

If you are interested to give a talk, please send me [5] your name, the title and a brief abstract.

If you are willing to attend, without giving a talk, please send me [5] your name.

 

The workshop is going to be hosted in the University of Latvia on 26.07.2011.

 

 

Thanks,

Artem Shinkarov

 

[1] www.project-advance.eu

[2] cordis.europa.eu/fp7/home_en.html

[3] www.sac-home.org

[4] www.snet-home.org

[5] artyom.shinkaroff@gmail.com

Dalīties