svadev AT lists.siebelschool.illinois.edu
Subject: Svadev mailing list
List archive
- From: John Criswell <criswell AT illinois.edu>
- To: svadev AT cs.uiuc.edu
- Subject: Re: [svadev] Linkage Types
- Date: Tue, 19 Oct 2010 09:53:53 -0500
- List-archive: <http://lists.cs.uiuc.edu/pipermail/svadev>
- List-id: <svadev.cs.uiuc.edu>
- Organization: University of Illinois
|
On 10/19/10 9:49 AM, Arushi Aggarwal wrote:
I was asking because, for C++ programs, a lot of the
class functions get linkonce_odr and weak_odr linkages. Which led
to a lot of DSNodes being marked External.
I see. As far as I know, these linkage types make a symbol visible to external code, so their DSNodes should be marked External. But running internalize as preprocessing helps.
I guess that is the correct solution.
Yes, you have to run the internalize pass when doing whole program analysis in order to get any good optimization or analysis results out of any properly written LLVM pass. Even LLVM's libLTO results would be pessimized if it didn't run internalize first. -- John T. -Arushi
On Sun, Oct 17, 2010 at 10:24 AM, John
Criswell <criswell AT uiuc.edu>
wrote:
On 10/17/2010 12:34 AM, Arushi Aggarwal
wrote:
Which other linkage types do you think should be internal?
-- John T.
_______________________________________________ svadev mailing list svadev AT cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/svadev |
- [svadev] Linkage Types, Arushi Aggarwal, 10/17/2010
- Re: [svadev] Linkage Types, John Criswell, 10/17/2010
- Re: [svadev] Linkage Types, Arushi Aggarwal, 10/19/2010
- Re: [svadev] Linkage Types, John Criswell, 10/19/2010
- Re: [svadev] Linkage Types, Arushi Aggarwal, 10/19/2010
- Re: [svadev] Linkage Types, John Criswell, 10/17/2010
Archive powered by MHonArc 2.6.16.