Determining Registers Per Work-Item and Shared Memory Per Work-group

How can i determine Registers per work-item and Shared Memory Per work-group? I know that is possible with CUDA, what about OpenCL?

Use NVIDIA’s Compute Profiler, it works for both CUDA C and OpenCL applications. Parallel Nsight might also be worth a look.

Danke schön :biggrin: -Thank you very much