offset in shared memory of OpenCL

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi

Sorry the title is wrong. It’s about device memory.

On CUDA, with a pointer to a device memory block, p, an offset of p can be used in say cublas function, like p+m where m is an integer.

I wonder how to do this with device memory on opencl type cl_mem, a direct add like the above doesn’t seem to work.

thanks.

2

You should be able to do simple offsets into device memory, although OpenCL does have restrictions on pointer traversals, see our OpenCL jumpstart guide, page 2:
[url=“http://developer.download.nvidia.com/OpenCL/NVIDIA_OpenCL_JumpStart_Guide.pdf”]http://developer.download.nvidia.com/OpenC...Start_Guide.pdf[/url]