init_paralution() failure for GPU without double precision

Front page Forums General init_paralution() failure for GPU without double precision

This topic contains 0 replies, has 1 voice, and was last updated by  amzs 1 year, 8 months ago.

Viewing 1 post (of 1 total)
  • Author
    Posts
  • #1517

    amzs
    Member

    Hello,

    Thank you for putting in all the effort in making PARALUTION available to anyone. It is greatly appreciated.

    When compiling/running for the first time, all examples displayed the following:

    This version of PARALUTION is released under GPL.
    By downloading this package you fully agree with the GPL license.
    Number of OpenCL devices in the system: 3
    CL_BUILD_PROGRAM_FAILURE
    OCL BuildLog:

    PARALUTION ver B1.1.0
    PARALUTION platform is initialized
    Accelerator backend: OpenCL
    No OpenMP support
    OpenCL is not initialized

    It took me a while to figure out why by looking at the source code. In the init_paralution() -> paralution_init_ocl() the first GPU is selected. It so happens that on my MacBook Pro that is an Intel Iris Pro without double precision floating point support, thus failing the build of that kernel and leading to OpenCL not being initialized.

    By adding set_ocl_paralution(0,2) prior to invoking init_paralution(), where 2 represents the AMD GPU on my system, the problems is resolved.

    Would it be possible to add a paragraph to the User’s Manual to clarify how to select the appropriate platform/device. It would have saved me considerable time. Also, the manual lists select_device_paralution(int dev), while it should be set_device_paralution(int dev).

    I hope that my comment on this forum will help others encountering the same situation.

    Thank you and keep up the great work,
    amzs

Viewing 1 post (of 1 total)

You must be logged in to reply to this topic.