GC, the simple solution

[Sean Kelly]

Lars Ivar Igesund wrote:
> Frank Benoit wrote:
> 
>>> * The cost of removing the last pointer to an object is unbounded
>> delete can be unbound? I didn't know that.
> 
> The reason is that a delete might trigger an unknown additional amount of
> decrefs leading to more deletes.

And some allocators colaesce adjacent free blocks aggressively (ie. 
during delete).  Not to mention the fact that the allocator data is 
probably protected by a mutex, etc.  In general, if the allocator 
documentation makes no guarantees about progress then you may at least 
be stuck on a mutex while other (potentially low-priority) threads are 
making an (unbounded) allocation.


Sean