Matthew Parkinson
Matthew Parkinson
@brooksdavis the data structure that makes the requirement is not particularly important in the current design, and could be replaced with a lock without harming perf (this wasn't true in...
@davidchisnall you added these bits are you okay with them going away now?
Thanks for doing this. Are you sure about this. You've put a lot of work into snmalloc-rs, and putting it under the Microsoft org, would give you less personal ownership....
@NeilMonday I wonder if you had any thoughts on this PR?
That is a great observation. I have a few thoughts: 1. I am not happy to put it at the top level. That will cause a branch on the fast...
I thought it would only call it in the case of `malloc` if `_query_new_mode()` was set? But would always call the handler for `new`? We also really need to implement...
#791 means this can now be extended to correctly cover the Windows CRT `_set_new_handler` behaviour. So moving to draft until that is done.
This is addressed by #775.
I think many of the things that have been improved in mimalloc 3 are hard to benchmark. Large long running systems, so a empirical comparison with mimalloc-bench wouldn't say very...
Could we add that to mimalloc-bench? How long does it run for?