How to use OpenCL GPU Kernel debugger for Windows?

[Colin_Reinhardt]

Hi, 

I want to get the GPU kernel debugger set up and running, but not finding much info on how to do this and what is/isn't supported.

Can you point to any articles, KBs or anything describing:

1. what version of Intel OpenCL SDK is required for GPU kernel debugging on Windows (Skylake chipset)?

2. description of how to setup proper system configuration for GPU kernel debugging, and supported configurations (stand-alone, 2 system,...?)

3. any additional info on using the GPU kernel debugging feature to debug OpenCL kernels - tutorial, tech note, video??

Thanks, Colin

[SailingDreams]

I know I'm replying to an old post, but I thought I'd add some info that I've found w.r.t. this topic

This is a good starting point

https://www.intel.com/content/www/us/en/developer/tools/opencl-sdk/overview.html

The following page shows what features are available in the 2020 release. Unfortunately they have removed "Step-by-step debugging for GPU kernels" in the 2020 Gold Update 1. IMHO, this is more of a downgrade than an improvement update so I'd call it a "brass" update not a "gold" update 

https://www.intel.com/content/www/us/en/develop/documentation/openclsdk-devguide/top/introducing-intel-sdk-for-opencl-applications.html

And another tool was removed: "Image and memory view". This seems like an integral part of debugging so not sure why they are going backwards in GPU debuggability.

The SDK download version is selected on this page after you enter your information

https://software.seek.intel.com/intel-opencl

As per videos, there are a ton of Intel OpenCL for FPGAs. Go to youtube and search "Opencl on intel fpgas". I'm not sure why there isn't much information about "Opencl on Intel GPUs". It's either super simple or they are not focusing on that. I get the sense they are focusing more on DPC++, but what if you are in a company where they have a huge OpenCL code base?