Subject: CGAL users discussion list
List archive
- From: Zohar Levi <>
- To:
- Subject: [cgal-discuss] Fw: Memory leak in parameterization
- Date: Sat, 29 Aug 2009 02:49:27 -0700 (PDT)
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Subject:To:MIME-Version:Content-Type; b=kW7CTZTfJQR72CI94yhyMYMGTsym21vD4pqpW5tdhCx7akRH5xQjTVIbrMTFc0CUEtHCdiGRpCeT0kTl7VFYEldOkilHCgStOu7n6TentLsCc9w9iIbCBsAEf0XR0Y4EfMlh6/wXPmlCKdiORTLxEixpbdU6QHrPC9UvlKz7K6g=;
----- Forwarded Message ----
From: Zohar Levi <>
To:
Sent: Saturday, August 29, 2009 8:24:18 AM
Subject: Memory leak in parameterization
Laurent,
- You need to fix the problems with the mailing list. I didn't get your last two replies (read them from archive), and I don't know how to continue the thread. There is also this Chinese guy who keeps sending me mails in Chinese calling my mother in names. I think it's time to open a forum. Being nicer and much more comfortable, I think people could talk with each other more freely.
- About the taucs debug lib making a corrupted exe. Forget the crystal ball, use Tarot cards instead. I'm using vc8 2005.
- About the crash in taucs, it occurred in the 600 iteration in my app. (damn, after running for 5 hours!), and will be harder to reproduce the same configuration. I'm not eager to try the library further before we fix the running time of the solver. I could send you my whole code, but I'm afraid you'll have hard time using it, it's a maya plug-in.
- About taus running very, very slow on small meshes. It's great that you narrowed down the problem to a bug in CGAL, in TAUCS, or in my code. I was starting to think there was a problem with my lucky rabbit foot charm :). I'll try to encapsulate the code snippet I gave you inside a working app. to prove my point. Actually, I tried the default solver and it worked on my example quite well and fast. I can't remember why I ruled it out. Maybe it was when I was dealing with larger neighborhoods and the taucs gave much better results. To my opinion you should simply combine between the two. Small number of vertices use default solver, else use taucs.
- About the memory leak in the parameterization. File "fixed_border_parameterizer_3.h" procedure check_parameterize_preconditions() line 407:
int genus = feature_extractor.get_genus();
It blows up my memory after 700 iterations. I didn't pursue the bug until the end, but I presume it is when you add a tag to the vertex info, used as a visiting flag, and you never free it. Just run the parameterize() procedure in a loop on the same polyhedron and watch how the application memory increases.
From: Zohar Levi <>
To:
Sent: Saturday, August 29, 2009 8:24:18 AM
Subject: Memory leak in parameterization
Laurent,
- You need to fix the problems with the mailing list. I didn't get your last two replies (read them from archive), and I don't know how to continue the thread. There is also this Chinese guy who keeps sending me mails in Chinese calling my mother in names. I think it's time to open a forum. Being nicer and much more comfortable, I think people could talk with each other more freely.
- About the taucs debug lib making a corrupted exe. Forget the crystal ball, use Tarot cards instead. I'm using vc8 2005.
- About the crash in taucs, it occurred in the 600 iteration in my app. (damn, after running for 5 hours!), and will be harder to reproduce the same configuration. I'm not eager to try the library further before we fix the running time of the solver. I could send you my whole code, but I'm afraid you'll have hard time using it, it's a maya plug-in.
- About taus running very, very slow on small meshes. It's great that you narrowed down the problem to a bug in CGAL, in TAUCS, or in my code. I was starting to think there was a problem with my lucky rabbit foot charm :). I'll try to encapsulate the code snippet I gave you inside a working app. to prove my point. Actually, I tried the default solver and it worked on my example quite well and fast. I can't remember why I ruled it out. Maybe it was when I was dealing with larger neighborhoods and the taucs gave much better results. To my opinion you should simply combine between the two. Small number of vertices use default solver, else use taucs.
- About the memory leak in the parameterization. File "fixed_border_parameterizer_3.h" procedure check_parameterize_preconditions() line 407:
int genus = feature_extractor.get_genus();
It blows up my memory after 700 iterations. I didn't pursue the bug until the end, but I presume it is when you add a tag to the vertex info, used as a visiting flag, and you never free it. Just run the parameterize() procedure in a loop on the same polyhedron and watch how the application memory increases.
- [cgal-discuss] Fw: Memory leak in parameterization, Zohar Levi, 08/29/2009
Archive powered by MHonArc 2.6.16.