Daniel Arndt
> Now for the other side of the coin: are we worried that the lock on every kernel launch is going to present unacceptable added overhead? https://github.com/kokkos/kokkos/pull/6151#issuecomment-1564392815 shows that locking...
> Does SYCL need locks added as well? I think `SYCL` is fine using atomics since we are never setting the status instead of querying it from the runtime.
> [#6151 (comment)](https://github.com/kokkos/kokkos/pull/6151#issuecomment-1564392815) shows that locking and unlocking a mutex without contention guarding `Serial` execution of an empty kernel causes for me a 4% slowdown. I would expect that to...
> @masterleinad @tpadioleau I think the error in [#6954 (comment)](https://github.com/kokkos/kokkos/pull/6954#issuecomment-2102675742) may be a legitimate bug unrelated to deduction guides. I was sure that it was unrelated to this pull request....
> In the meantime the OpenACC issue is solved I could have a look at how to slightly change the test. I have in mind to remove the inner parallel_reduce....
> is this PR close to being ready, or is it still, in fact, a draft? It's ready but not the highest priority. An open question is what the default...
> OK, thanks! Would you like to discuss enable / disable relocatable device code on Wed.? No, we can do that here.
> Don't you need to update the Makefile too? I could do that if requested.
Retest this please.
Retest this please.