Hello,
Can anyone explain the relationship between -b -t and -p and GPU work groups? I tried doing some research to try and optimize this but couldnt figure out the relationship and I dont have the coding know how to code a simple benchmarking utility that would test the ranges.

My understanding is that GPUs have compute units and workgroups are distribute across these units...and each workgroup consists of blocks and I think threads.

I assumed that if I learned a little about workgroup sizing as well as knowing the compute units for my gpu (I have an RX 580 so it has 36 CU with 2304 SMs) I could better determine a good configuration.

Thanks