On Fri, 24 Jul 2015, Jason Power wrote:
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/2989/#review6824
-----------------------------------------------------------
I'm not sure how I feel about this patch (and the other ones posted
today) in general, but if we are going to move away from reference
counting, then this patch needs to include very specific information on
which objects are responsible for deleting any heap variables. As it
stands, it would be very hard for someone to modify this code without
introducing spurious memory bugs.
The question of responsibility arises when multiple objects may hold
pointer to a heap allocated object. As per my understanding, this is not
the case for any of the protocols and on-chip network implementations.
So the current owner can choose to delete the message as soon as it
realizes that the message is no longer required.
--
Nilay
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
