There is an interesting capability of ANSYS Fluent CFD code, to run user defined functions on a GPU.
I am quoting here the only available information on how to do it: (taken from ANSYS literature)
“UDFs can be compiled in ANSYS FLUENT with OpenCL support for execution on General Purpose Graphics Processing Units (GPGPUs) on lnamd64 and win64 platforms. To enable OpenCL support, you should use the following text user interface (TUI) command: /define/user-defined/enable-udf-on-gpu before compiling the UDF. The UDF library is linked with libOpenCL.so on lnamd64 and OpenCL.dll on win64. For execution on GPU systems, the appropriate environment variable (LD_LIBRARY_PATH on lnamd64 or %path% on win64) should be correctly set so that libOpenCL.so/OpenCL.dll is loaded along with the UDF library.”
Unfortunately, I do not understand the requirement of the last sentence ("For execution on GPU systems…)
What does it mean to set %path% to OpenCL.dll? (I am working with Winx64)
Any help will be highly appreciated.