I think the #1 thing that drives me crazy about GC debates is that people don’t realize just how important bump allocation is for perf.
-
-
Which language does that? :) I know the JVM does bump allocation.
- End of conversation
New conversation -
-
-
Usually that's counter-balanced by the need to copy in compacting phase. Some allocators (including Go) use fixed size buckets for many allocs, which translates in pointer bumps for the right size bucket.
-
Only tenured objects have to be copied. Generational hypothesis says majority of objects don’t survive long enough.
End of conversation
New conversation -
-
-
would be interested in an arena vs gc speed comparison. I'm guessing someone along the way has done one
-
Our ECOOP 2017 paper on tree data structures has some of this. Arenas are pretty good.
End of conversation
New conversation -
-
-
(So I'm guessing the answer here is likely to be "not even close"? https://twitter.com/glaebhoerl/status/931072831806492673 …)
-
It’s an increment, test against limit, and maybe branch to the slow path, and usually multiple allocations can be coalesced because it is inline
End of conversation
New conversation -
-
-
I’m pretty sure many GC allocators bump allocate on the fast path. JSC does FWIW
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.