Implementing reverse execution

Chris Jones jones.chris.g at gmail.com
Wed Feb 18 11:44:47 PST 2015


​Sounds great! :)​

On Mon, Feb 16, 2015 at 7:02 PM, Robert O'Callahan <robert at ocallahan.org>
wrote:

> 1a) if we hit too many breaks between checkpoint A and checkpoint B,
> disable breaking, run forward to about halfway between A and B (say C),
> reenable breaking and keep going from C to B. If no breaks are found
> between C and B, refocus on A-to-C and retry. If breaks are found between C
> and B, refocus on C-to-B and retry.
>

​Also note that the A-c and c-B searches can (recursively) parallelize.​
(And potentially cleverer parallel algorithms are possible.)  We'll have to
let experience see how much this needs to be optimized, but framing this
problem as N-way parallel search is probably useful.

Cheers,
Chris
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/rr-dev/attachments/20150218/b66bcb35/attachment.html>


More information about the rr-dev mailing list