Dmitriy Sobolev

@akukanov, could you check the PR as well? It modifies the API, and I suppose it needs an additional approval from the architect.

Honestly, the amount of changes is terrifying, and because of that I need to take a closer look at the PR, even though these are mechanical changes. @SergeyKopienko, I have...

There are two major reasons of using the outdated barrier API: 1. Performance. You can find the details here: https://github.com/intel/llvm/issues/12531. 2. Non-compatible semantics, as mentioned in the code comments, which...

I tried to understand the impact of the replacement of SYCL 1.2.1 barrier with SYCL 2020 one. Below is a table with speed-ups (times) after switching to SYCL 2020 barrier...

Leaf size was enlarged as a part of https://github.com/oneapi-src/oneDPL/pull/1735. The performance benefit was significant for GPUs. However, the question regarding adjustment of a leaf size to avoid extra copy still...

I see these factors as reasons for removal AOT commands from our CMake: - We should introduce many CMake options to support possible AOT targets. - The options are not...