I had a vague idea this was how malloc usually worked. Per-thread arena and pass pointers back to the allocating thread on deallocate. Paper claims freed point

I had a vague idea this was how malloc usually worked. Per-thread arena and pass pointers back to the allocating thread on deallocate. Paper claims freed pointers tend to go in the freelist of the thread that called free, not back to the allocator. Sending pointers back to the original probably has positive effects on cache locality when the block gets handed out again, effectively defragmenting the distribution of freelist stored blocks over time.