Daisy S. Hollman

Also, if this is changed, `layout_stride::mapping::is_unique()` `layout_stride::mapping::is_always_unique()` need to change.

Well, if we're going to do that, there are about 10 other things in the previous 10 issues that need to be changed also

## Next question: where does this leave us on P0546 There are some issues that still need to be addressed here, but it seems reasonable to leave this as a...

Oh, one more "executors lesson" I forgot: it pays to be very careful and deliberate early on about the difference between properties and their values. For instance, you shouldn't have...

> Hm, how do we define the order in which the hooks get applied? What if they don't commute? I guess they are supposed to be orthogonal; does that really...

> Wouldn't that mean that atomic_access has to let non-atomic operator[] do its thing? I don't see how this would work At least as I understand it, the `atomic_access` implementation...

The implementation of `restrict_access` in terms of hooks is a lot trickier, though. One option is to expose the type of the underlying pointer via a customization point. For instance...

Example of the hook I suggested above for `restrict_access` (seems to work with gcc, but nothing else) [https://godbolt.org/g/tHNBtj](https://godbolt.org/g/tHNBtj)

> Huh, works for recent enough Clang and Intel too, for me at least By "works", I don't mean that it just compiles. If you look at the instruction output,...

> At least as I understand it, the atomic_access implementation we have uses a hash table of sharded locks that is dependent on the address of the mdspan itself (and...