I'm worried that this over-fits our hashing strategy to the implementation
details of these particular data structures. If objects added to a Map or
Set
are later added to a WeakMap, they'll end up with exactly the bad behavior
you
describe in the CL description for open-addressed hash tables (though I
suppose
we could use different private symbols for hashing in an ObjectHashTable vs
an
OrderedHashTable).
This hashing strategy may also be over-fit to the microbenchmark, since new
objects are created and inserted once, meaning that the ordering of the hash
codes always matches the ordering of the container. In more varied workloads
(such as adding and removing the same objects from a set), I'd this to be a
wash.
If you disagree with the above, can you help me overcome these concerns?
https://codereview.chromium.org/1157073002/
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
