profiling OpenCl applications

[RamyaSri01]

Hi,

Can anyone help me to execute OpenCL benchmarks on intel integrated UHD 620 graphics and profile them to get timeseries data of low level counters like memory bandwidth, IPC, cache statistics? 

currently, I am thinking to use system analyzer tool to profile while executing OpenCL benchmarks on gpu in the backend. Can you please confirm whether it works or not ? 

It will be helpful if you provide any related documentation also.

Thanks and regards

RamyaSri

[Pamela_H_Intel]

Moving this post to the Graphics Forum.

[Ben_A_Intel]

Hi, have a look at the OpenCL Intercept Layer.  This is a tool we use internally to collect statistics like these:

https://github.com/intel/opencl-intercept-layer

Specifically, have a look at:

https://github.com/intel/opencl-intercept-layer/blob/main/docs/mdapi.md

If you have follow-on questions or have trouble getting this to work definitely let us know - thanks!