Vyacheslav Klochkov

The corresponding LIT test verifying the fix: https://github.com/intel/llvm-test-suite/pull/1130

> Even though I see how this can be useful, it might cause confusion if a function called printf is behaving differently from the normal printf. Yes, that is a...

> people might be confused why they can print vec[i] but can't pass vec[1] to a SFINAE function (e.g. template> void f(T);). Yes, they can be confused, but that kind...

> I think we should understand why for the GPU BE requires the intrinsics for the element access and why the solution we use for `sycl::vec` isn't OK. I don't...

The fix and test for the problem have been rejected.

If you add additional fix for this comment https://github.com/intel/llvm/pull/6413#discussion_r916425618 then the fix is good, IMO. I would feel safer though if one more reviewer could take a look too, perhaps...

> Given how easy it is to get into a merge conflict and how many approvals are required (and hard to get quickly enough), I don't think it will ever...

@aelovikov-intel - can you please review this patch as the only active developer of reduction these days?

> Thoughts Personally, I would proceed with the current patch that has the division on host side, but that opinion is not confirmed by performance benchmarking 'div on host +...

Please add a test for the new env variable specifying the preferred group size.