Matthew Parkinson

Seems reasonable to me. I do wonder if what the cost of just using locks for that would be.

> Is this behavior (i.e. the need for the range to be committed beforehand) intentional and desired? So that particular configuration was added to support SGX enclaves. They don't have...

This is precisely the kind of thing I expected with my "However, doing something new normally throws up issues, so please let me know how you get on.". The pagemap...

> Now that this fixed-range allocation works, I would need multiple instances of it. > But of course, there's a lot of statics in the types that store the actual...

> Is that expected or can it at least be explained? Not expected. > It looks like empty blocks aren't being merged with what appears to be their buddy counter-part,...

So this is really interesting. I previously had an issue with a service that seemed to be experience fragmentation over time on SGX. I thought it was due to thread...

Also, if you have a chance, would you be able to try #751. That has a simpler thread local state model.

Thanks. I have been looking at resurrecting an old PR I never landed: https://github.com/microsoft/snmalloc/pull/616 If I get this working nicely again, then we might be able to trace your memory...

@Trithek I have update #616 to be usable again. This has a function ``` snmalloc::print_alloc_stats() ``` which should dump to stderr some CSV of various stats that could be used...

@nishiken I agree that looks like a typo. I also agree with @derSteFfi that the test doesn't seem correct. I am fairly convinced for something like that dlmalloc might even...