Don't know how much of a factor it is in this design, but might also just be expensive to introduce heap compaction to Go now, after going this far with heap objects having stable pointers. Might be a lot of random spots in the code that still assume that.
Not that there aren't benefits to compacting, including the bump allocation, less bloat as programs run a long time, and help with locality. Just, describing where Go's at.
> I'm still confused how they maintain this separation safely
Don't know about this design, but in other places they have some goroutine-local data structures and then do the expensive thread coordination stuff after they run out of resources. Seems plausible that could be the general approach here too.
EDIT: Found the post.
> In the CLR allocation is blazing fast: just increment the heap head pointer. You don't have to take care of problems traditional allocators face: find a big enough fitting free hole. This has a cost: when objects are destroyed, the free memory on the heap becomes fragmented. This needs to be compacted, and this is where the GC pause is needed, as the references for moved objects need to be updated. Basically the cost of finding a free hole in memory at allocation was moved here.
> This was done intentionally, as studies found that most real world applications produce a load where most objects are short lived: this cheap allocation is needed, and a multigenerational GC can take care of them cheaply after destroyed (as for most of the time you only scan and free the youngest generation for unreachable objects, limiting the GC pause time).
First-fit seems like it could be quite slow here.
The implication seems to be that they can at least avoid locking for allocation (and deallocation and promotion) as long as they find a fit within the reserved region for the goroutine, but I'm still confused how they maintain this separation safely.