AFAIK, JavaCL / OpenCL4Java is the only OpenCL binding that is available on all platforms right now (including MacOS X, FreeBSD, Linux, Windows, Solaris, all in Intel 32, 64 bits and ppc variants, thanks to its use of JNA). It has demos that actually run fine from Java Web Start at least on Mac and … Read more
You discover these values experimentally for your algorithm. Use a profiler to get hard numbers. I like to use CL_DEVICE_MAX_COMPUTE_UNITS as the number of work groups, because I often rely on synchronizing work items. I usually run kernels with little branching, so the take the same time to execute in each compute unit. Some multiple … Read more
To compile and run OpenCL code under Linux, you’ll need four things: 1) An NVIDIA Driver which supports OpenCL. The drivers packaged with Ubuntu are somewhat old, but they should still work just fine. Unless you have explicit need for current drivers, you should stick with the ones packaged with Ubuntu. To be clear, these … Read more
when you are linking, the order of your libraries and source files makes a difference. for example for your case, g++ -I/usr/local/cuda/include -L/usr/lib/nvidia-current -lOpenCL opencl.cpp functions defined in the OpenCL library might not be loaded, since there nothing before them asking for a look-up. however if you use, g++ opencl.cpp -I/usr/local/cuda/include -L/usr/lib/nvidia-current -lOpenCL then any … Read more
As you have stated, barriers may only synchronize threads in the same workgroup. There is no way to synchronize different workgroups in a kernel. Now to answer your question, the specification was not clear to me either, but it seems to me that section 6.11.9 contains the answer: CLK_LOCAL_MEM_FENCE – The barrier function will either … Read more
For nvidia (and amd for that matter) gpus the local memory is divided into memorybanks. Each bank can only address one dataset at a time, so if a halfwarp tries to load/store data from/to the same bank the access has to be serialized (this is a bank conflict). For gt200 gpus there are 16 banks … Read more
The OpenCL Runtime is already included in the Nvidia graphics drivers. You only need the OpenCL C++ header files, the OpenCL.lib file and on Linux also the libOpenCL.so file. These come with the CUDA toolkit, but there is no need to install it only to get the 9 necessary files. Here are the OpenCL C++ … Read more