P0652 Concurrent associative data structure with unsynchronized view

[jensmaurer]

P0652R2 Concurrent associative data structure with unsynchronized view (Sergey Murylev, Anton Malakhov, Antony Polukhin)

[jensmaurer]

SG1 in Kona

[ogiroux]

SG1 results:
"We believe it is important to permit implementations that do not protect
values from concurrent visits. (note: P0652 does protect values from
concurrent visits)"
SF F N A SA
6 3 1 3 0

"We would like to see a policy-based design that allows customization of 
concurrent map (e.g., whether deletion is permitted)."
SF  F  N  A  SA
3   5  4  0  2

SA: authors are not willing to do this because it would over-complicate their
design.  and they believe that LEWG would be against such a design.


"We believe it is important that the concurrent map permits concurrent reads
without contention."
Unanimous consent.
(note: we also agree that we would like to see evidence in the form
a reference implementation.)

[wg21bot]

P0652R3 Concurrent associative data structure with unsynchronized view (Sergey Murylev, Anton Malakhov, Antony Polukhin)

[ogiroux]

SG1 in Cologne
We would like a class template for concurrent map:

• Does not have removal support
• Does not guarantee synchronized visitation
• Does not guarantee multi-element access
  SF F N A SA
  3 6 3 0 3 - Consensus
  SA: wouldn't trust users to use correctly
  SA: usage scope reduced to much
  SA: ditto, not user friendly, less useful

We would like a class template for concurrent map:

• Has removal support
• Does not specify synchronous destruction of elements
• Does not guarantee synchronized visitation (except against destruction)
• Does not guarantee multi-element access
  SF F N A SA
  7 5 2 0 1 - Consensus
  SA: wouldn't trust users to use correctly

We want both of these class templates, under different names TBD
SF F N A SA
5 6 2 1 2 - Consensus
SA: wouldn't trust users to use correctly
SA: not usable and not worth putting in the stdlib