Closed loyd closed 3 months ago
@wvwwvwwv, what does "Formally verified EBR" in README mean? Is it about https://github.com/wvwwvwwv/scalable-concurrent-containers/blob/main/src/tests/model.rs?
I also know that the loom
tests in the crate is not ideal. At my previous workplace, I used a generalised atomic interface to avoid any use of compile-time flags, so I may try this approach here as well.
Please don't pay much attention to Formally verified EBR
part :-) It just means that I developed ebr
along with those loom
tests; doesn't mean that I used Coq
or the likes to formally prove the algorithm. Note that the use of every memory barrier/fence was justified internally in my previous company, though, I cannot disclose any part of it.
Let me try to implement some loom
-based tests as a sponsored work backed by @loyd.
I close this issue in favor of https://github.com/wvwwvwwv/scalable-delayed-dealloc/issues/1 due to https://github.com/wvwwvwwv/scalable-concurrent-containers/issues/131
I understand this requires a vast amount of work. However, at least an issue should exist.
Except for using loom's primitives for testing, I think it should use them under
--cfg scc_loom
(in the same way ascrossbeam_loom
exists) to support testing in implementation of structures based on scc-ebr implementation.