-
Philippe Gerum authored
We remove the per-heap object lock, which was duplicating the outer locking from a sync object the heap belongs to in most cases, fixing the only caller who did expect the heap object to handle locking internally (i.e. VxWorks's memParLib). Secondly, we get rid of the indirection to invoke the allocator routines via a call table in the pshared case, which brought absolutely zero value, but overhead. At this chance, we also inline most of the calls to TLSF/malloc for private allocation.
7a66719d