I'm trying to use the kernel in the attached file, it is with the tiled convolution example from AMD with few minor changes (
I had to changed the use of pointers to access vector data types. e.g. from :
*(__local float4*)&lclPlane[lcloffset4] = horiz_filter;
The file get compiled ok, but I get CL_INVALID_VALUE -30 for clCreateKernel..
Couldn't figure out why the clCreateKernel fails, and I should note that it works on Intel and Nvidia OpenCL sdks.
I'm using sabre lite board with linaro-ubuntu-desktop
The opencl device:
DEVICE_NAME = Vivante OpenCL Device
DEVICE_VENDOR = Vivante Corporation
DEVICE_VERSION = OpenCL 1.1
DRIVER_VERSION = OpenCL 1.1
Will appreciate your help.
Original Attachment has been moved to: kernel.cl.zip