[m-rev.] For review: Protecting backward live regions for disjunctions.

[Quan Phan]

On Mon, Oct 15, 2007 at 04:04:56PM +1000, Zoltan Somogyi wrote:
> On 15-Oct-2007, Quan Phan <quan.phan at cs.kuleuven.be> wrote:
> > The reason for this is because generally we do not locally know what regions
> > will be needed when the program backtracks to the disjunction. Details of
> > this modification will be in a separate document.
> 
> It would be hard to evaluate this diff without those details.
> 
I updated the advanced_support document in papers/rbmm.
Generally, it can be summarised that instead of adding a field
disj_protected into the region data structure and setting it to the disj
frame when a region needs to be protected by the disj frame, we save the
current sequence number in the frame. This then is used to protect all
the regions with sequence numbers less than the saved values. Other
things follow from it.
 
> Go ahead and commit your diff. I will past my review afterwards.
I commited the change.

Quan.
> 
> --------------------------------------------------------------------------
> mercury-reviews mailing list
> Post messages to:       mercury-reviews at csse.unimelb.edu.au
> Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
> Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
> --------------------------------------------------------------------------
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------