Gerwin Klein

@axel-h this PR should either be closed or merged — happy to merge it if it’s useful for you, otherwise I’d close it.

Interesting, I have not seen that one yet. Do you have a pointer to a run?

Oh right, of course, they would be in successful runs. That explains why I haven't seen them. Yes, that would definitely be good to address.

👍 That'd actually be nice for verification as well.

Well, it looks like the number of times this function is called is not 4 on the sabre platform..

It's 8, for which the bit operations are also still safe. There is something sub-optimal going on, though: it's 8, because it locks 2 TLB entries and does that each...

> Only suggestion I have is to remove it altogether and see if it makes any difference in the benchmark results. It'd be fine to do this in a separate...

I think we can merge this as an improvement and look at the dependencies again separately. @axel-h could you file an issue for that so we don't forget about it?

@kent-mcleod could you please add the sign-off to your commit?

It's a bit early to review the PR (should have left it on draft). We should sort out the API in the RFC discussion first.