I'm getting an intermittent segfault in the scalable_allocator in the call chain sequence:
internalFree() isLargeObject() getBackRef()
I think I saw this in unaltered scalable_allocator code, but I've been fiddling with the values for blockSize and memReqSize, so I can't gurantee it.
Looking at struct BackRefMaster, there are some comments that seem to indicate that it may be limited to dealing with 256 GB of memory. I'm using several terabytes. Should I be increasing the size of BackRefMaster::bytes?
I know that TBB is about multithreading, but can we please keep the allocator
experiment discussion together in one thread, if only to avoid alarming
titles that suggest problems in TBB itself? I suggest going back to "Customizing the scalable allocator" for this.