[Soot-list] DeadAssignmentEliminator and assignments x=null

[Andrey_Poliektov at amat.com]

I think you are right.

No matter what you do, please take into account that sometimes people 
assign ref to another object instead of null.

Object o = ....

A: ..... something requiring o ...

o = "";

B: ... nothing uses o...

Would you be worried that the program could use more memory during 
execution of B after your optimization that it might have been without ? 
:)








"Eric Bodden" <eric.bodden at mail.mcgill.ca> 
Sent by: eric.bodden at googlemail.com
09/11/2007 02:18 PM

To
Andrey Poliektov/APPLIED MATERIALS at AMAT
cc
"Patrick Lam" <plam at sable.mcgill.ca>, "Soot list" 
<soot-list at sable.mcgill.ca>, soot-list-bounces at sable.mcgill.ca
Subject
Re: [Soot-list] DeadAssignmentEliminator and assignments x=null






 


> As far I remember java spec, the result of your "test" is undetermined.

Yes, I remember that - thanks for pointing that out again, though. The
question is just whether we should change the probability of having
that value actually GC'ed or not. I guess it does not matter too much,
actually.

Eric

-- 
Eric Bodden
Sable Research Group
McGill University, Montréal, Canada

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.CS.McGill.CA/pipermail/soot-list/attachments/20070911/a5546941/attachment.htm