Steffen Larsen

Since the ABI-breaking window is opening, I took the liberty of removing the unused ctors and have marked this as ABI-breaking.

Since the ABI break will need separate approval, I have decided to roll it back here and do it as a follow-up.

> You don't seem to go through `ProgramManager::addImage/removeImages` for SYCLBINs -- do you still get support for eliminated arguments, and are the PM's datastructures cleaned when a SYCLBIN kernel bundle...

@jchlanda - It looks like this made Plugin/sycl-ls.cpp fail on the HIP machine in post-commit. Could you please have a look?

> If there is already a decided upon way to deal with unittests going forward that I can add at this point I would be willing to do so. Alternatively...

Optimally we would consider any device USM operation valid with a mapped virtual memory pointer. To my knowledge, both L0 and CUDA allow that.

> > > @intel/llvm-gatekeepers ping > > > > > > Please fix: > >  > > I haven't seen this before. Where is it from and how can...

This looks reasonable to me, but I trust @KseniyaTikhomirova and @tovinkere more than myself when it comes to XPTI.

@mdtoguchi - Driver/sycl.c on Windows has been failing in post-commit as of yesterday for unrelated changes (https://github.com/intel/llvm/pull/15297) and this PR is the first case I've seen it fail in pre-commit....

Failure in Driver/sycl.c for Windows is known and will be disabled in https://github.com/intel/llvm/pull/15416.