There's no reason it could not, at least the authors think so in principle:
> Compacting garbage collection in managed languages:
> Compacting garbage collection has long been a feature of languages like LISP and Java [12, 14]. Contemporary run- times like the Hotspot JVM [20], the .NET VM [19], and the SpiderMonkey JavaScript VM [7] all implement compaction as part of their garbage collection algorithms. Mesh brings the bene ts of compaction to C/C++; in principle, it could also be used to automatically enable compaction for language implementations that rely on non-compacting collectors.
That is very much a moving garbage collector, marked (live) objects get moved to the start of the heap for compaction. GP is specifically asking about the applicability of Mesh/TFA's non-moving capabilities for GCd systems: moving objects requires the ability to update in-system references somehow, and makes FFI more complex as references either can't be shared/leaked or specific pinning API have to be set up (or some updatable addressing scheme has to be provided but then you don't get straightforward pointers anymore).
No, that clearly updates all the pointers after moving the allocations. Mesh is different as it works on the page level and use the virtual memory to achieve compaction without changing the pointers.
