OpenCL Kernel in Program
The OpenCL Kernel patch allows for general purpose parallel programming across CPUs and GPUs. With this patch you can create and modify OpenCL kernels in realtime using a subset of C with parallel extensions. The OpenCL Kernel patch will automatically create input and output ports based on arguments provided to the OpenCL kernel. For information on writing OpenCL kernels please refer to the OpenCL Specification available at http://www.khronos.org/opencl.
By default, the OpenCL Kernel patch will automatically determine the Local Work Size, Global Work Size and Output Dimensions for the kernel based on the size of input data. These parameters, along with the Compute Device, Image Pixel Format and Color Space, can be set manually by enabling the “Show Advanced Kernel Settings” within the patch settings. Manually setting the work sizes or output dimensions is required when input arrays differ from one another in size or differ from the size of output arrays. The Advanced Settings also allow for Performance Sampling using the “Sample Now” button to get an average execution time for a given kernel.
|
Input Attributes
|
Output Attributes
|