RFR: 8178836: Improve PtrQueue index abstraction

[Kim Barrett]

> On May 8, 2017, at 9:43 AM, Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> 
> Hi Kim,
> 
>   thanks for keeping on improving the ptrqueue stuff, much appreciated
> :)
> 
> […]
> Looks good in case you were still looking for somebody to look at.

Thanks.

> Some minor issues in case you need something for the next change:

I pushed 8178836 changes this morning, before your review.  I’ve added
these items to my todo list.  Not sure whether they’ll be in the very next
changeset in this area (I’ve got a large pending stack), but I’ll get to them
eventually.  (Some might have already been addressed later in that stack.)

> - I would prefer to keep the use of pointer_delta() at
> satbMarkQueue.cpp:141. While unusual and probably the VM won't get to
> this point when done, buffers can be of uintx size (see e.g.
> G1SATBBufferSize). I.e. the result of this calculation (ptrdiff_t) may
> overflow.
> Alternatively limit G1SATBBufferSize etc. to something realistic.
> 
> Pre-existing issues:
> 
> - maybe document for PtrQueueSet::_buffer_size that it is in elements.
> 
> - ptrQueue.hpp:258: remove the "All these variables are are protected
> by the TLOQ_CBL_mon. XXX ???" sentence. Apart from having two
> predicates, this sentence refers to a non-existing monitor.
> 
> - ptrQueue.hpp:273: mentions non-existing TLOQ_FL_lock.
> 
> - PtrQueueSet::buf_free_list_sz may need expansion of the "sz".
> 
> Thanks,
>   Thomas