What is the connection between a worm about a millimeter long and OpenCL?

    image

    Recently on a habr there was an article about the OpenWorm project . The project uses OpenCL to calculate the hydrodynamics of smoothed particles on both CPU and GPU. Among the researchers / developers are our compatriots Andrei Palyanov and Sergei Khayrulin from the Siberian branch of the Russian Academy of Sciences.

    Under the cut, the geography of the OpenWorm community, as well as the popularity rating of development on OpenCL by country and links to examples for the standard version 2.0.

    Judging by the publications, Andrey Palyanov and Sergey Khayrulin participate in the NeuroML project and optimized the PCI-SPH algorithm for GPGPU. And I learned that this project is going to use OpenCL, in March last year from their project colleague Matteo Cantarelli.

    Geography of the openworm community according to the site :

    image

    The situation with the popularity and use of OpenCL in Russia is strange. Judging by the statistics of my blog and links to it from the forum of the khronos.org consortium website and JavaCL documentation , the GPGPU topic is not popular in our country. Perhaps not very representative, but the rating of popularity of development on OpenCL by country (based on my blog and not taking into account the latest habr effect):

    1. United States
    2. Germany
    3. Ukraine
    4. United Kingdom
    5. Poland
    6. Latvia
    7. France
    8. Canada
    9. China
    10. Russia

    It's sad ... While OpenCL is not only a GPU, a CPU, but an FPGA. Why do you think this situation developed?

    Finally, news for those who are interested. The release of the new version of AMD APP SDK 3.0 Beta for OpenCL 2.0. Which also includes the Bolt C ++ template library and an optimized version of OpenCV (Open Computer Vision) for OpenCL .

    Examples in the SDK archive will allow you to deal with features that appeared in version 2.0: SVM Coarse Grain, Pipe , New Workgroup Built-in APIs, Image Read and Write, Program Scope Variable, Generic Address Space, Shared Virtual Memory pointer with offset, SVM Fine Grain Buffer , C ++ 11 Atomics, Device-side Enqueue, Depth Image.

    A detailed description of examples on the AMD blog and SDK.

    Also popular now: