Hi Karl I'm not really having a problem running code as you are. From Andreas's answer to my original post, I am under the belief that my buffer objects are being resent each time I run a kernel with parameter arguments. Hence the reason why I would like to experiment splitting the setting of arguments and the enqueueing of a kernel. I'm not entirely sure if this correct.
Would it be possible if you could try running the sample code I sent earlier on your system and see if you receive the same error? Perhaps it's a Windows specific bug. Thanks Blair On Tue, Oct 6, 2015 at 11:55 PM, Karl Czajkowski <[email protected]> wrote: > Blair: > > I am using version 2014.1 on Fedora Linux with Intel > opencl-1.2-4.5.0.8 on CPUs and > xorg-x11-drv-nvidia-cuda-355.11-1.fc21.x86_64 on Kepler GPUs. > > I don't use the enqueue_nd_range_kernel() but run my own custom OpenCL > kernels in sequence on the OpenCL device by initializing input arrays > with pyopencl.array.to_device() and retrieving output arrays with > array_dev.map_to_host(). > > For the kernels running on those device arrays, I invoke them like: > > ctx = pyopencl.create_some_context() > clq = pyopencl.CommandQueue(ctx) > > ... > > program = cl.Program(ctx, opencl_sourcecode).build() > > program.foo( > cl_queue, (workshape,), None, > dst_dev.data, src_dev.data, > cl_array.vec.make_int3(*dst_dev.shape), > numpy.uint32(other_param) > ) > > where the arguments starting with "dst_dev.data" are the actual kernel > parameters accepted by my OpenCL kernel called "foo". > > I hope that helps. > > > Karl > >
_______________________________________________ PyOpenCL mailing list [email protected] http://lists.tiker.net/listinfo/pyopencl
