fix for intel's compiler
Locked
fix for intel's compiler
 
|
Singleton pattern doesn't work with Intel's C++ compiler 11.1 on
Windows. I tested it both in 32 and 64 bit modes. When optimization is enabled (I used O3), each translation unit ends up creating its own instance of Singleton. For example, this code fragment doesn't do what it is supposed to do. // file1.cpp Settings::instance().evaluationDate() = date1; // file2.cpp Settings::instance().evaluationDate() = date2; // file1.cpp still sees date1 // fie3.cpp Date date3 = Settings::instance().evaluationDate(); // not equal to date2 Overwriting evaluation date in file2.cpp doesn't actually overwrite it. The above code will create 3 instances of Settings class and the program will eventually crash. Somehow static non-const local variables and templates don't mix on ICC. I modified the original implementation to uses const static variable and everything worked well. The new implementation is attached. Regards, Kakhkhor Abdijalilov. //----------------------------------------- template <class T> class Singleton : private boost::noncopyable { public: static T& instance() { #if defined(QL_ENABLE_SESSIONS) Integer id = sessionId(); #else Integer id = 0; #endif boost::shared_ptr<T>& p = (*instances_)[id]; if (!p) p.reset(new T); return *p; } protected: Singleton() {} private: typedef std::map<Integer, boost::shared_ptr<T> > map_type; static const boost::scoped_ptr<map_type> instances_; }; template <typename T> const boost::scoped_ptr<Singleton<T>::map_type> Singleton<T>::instances_(new Singleton<T>::map_type); //----------------------------------------- ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Sat, 2010-08-14 at 20:14 -0500, Kakhkhor Abdijalilov wrote:
> Singleton pattern doesn't work with Intel's C++ compiler 11.1 on > Windows. I tested it both in 32 and 64 bit modes. > > When optimization is enabled (I used O3), each translation unit ends > up creating its own instance of Singleton. [...] Somehow static > non-const local variables and templates don't mix on ICC. I modified > the original implementation to uses const static variable and > everything worked well. The new implementation is attached. Ok, thanks. Luigi -- Ogden's Law: The sooner you fall behind, the more time you have to catch up. ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
I uploaded C++ code into the path system.
------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Tue, 2010-08-17 at 18:17 -0500, Kakhkhor Abdijalilov wrote:
> I uploaded C++ code into the path system. Ok, thanks. Luigi -- The nice thing about standards is that there are so many of them to choose from. -- Andrew S. Tanenbaum ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Kakhkhor Abdijalilov
On Tue, 2010-08-17 at 18:17 -0500, Kakhkhor Abdijalilov wrote:
> I uploaded C++ code into the path system. Kakhkhor, sorry, but may you upload the whole modified file instead of only the modified functions? With the whole file, I can easily use the diff tool to merge your patch (and be sure that it doesn't conflict with other changes, if any is made between now and when I do the merge); with only part of the file, it becomes more difficult. Thanks, Luigi -- Academic: a term of opprobrium applied to those that do their job well by those who cannot. -- Sir Ernest Gowers ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
> may you upload the whole modified file
It is done. Regards, K.A. ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
In reply to this post by Kakhkhor Abdijalilov
On Sat, 2010-08-14 at 20:14 -0500, Kakhkhor Abdijalilov wrote:
> Somehow static non-const local variables and templates don't mix on > ICC. I modified the original implementation to uses const static > variable and everything worked well. The new implementation is > attached. Kakhkhor, do you also need the scoped_ptr below? Would a simple static const map_type work? Luigi > //----------------------------------------- > template <class T> > class Singleton : private boost::noncopyable { > ... > private: > typedef std::map<Integer, boost::shared_ptr<T> > map_type; > static const boost::scoped_ptr<map_type> instances_; > }; > > template <typename T> > const boost::scoped_ptr<Singleton<T>::map_type> > Singleton<T>::instances_(new Singleton<T>::map_type); -- The First Rule of Optimization: Don't do it. The Second Rule of Optimization (For experts only): Don't do it yet. -- Michael Jackson ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
I think it is broken just like the design using scoped_ptr or the
original design. Consider what happens when Singleton's instance is used in two implementation files. // file1.cpp namespace { SingletonObject* const ptr = &SingletonObject::instance(); } // file2.cpp namespace { SingletonObject* const ptr = &SingletonObject::instance(); } Who get's to build the singleton? The only way to avoid this problem is to make the instance itself a static local and forget about session ID. Each session would have to reset the singleton, if needed. A half-solution is to make sure that no instance is crated before main() is launched. That means no static objects using singleton. Which way we go? ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Wed, 2010-09-01 at 05:24 -0500, Kakhkhor Abdijalilov wrote:
> I think it is broken just like the design using scoped_ptr or the > original design. Ok. One way or the other, Singletons are all broken. But trying to avoid breaking interfaces for the time being, and in "normal" usage (i.e., no Singletons initialized before main() starts): was it needed to put the map into a scoped_ptr, as in: private: static const boost::scoped_ptr<map_type> instances_; or would the following work, too? private: static map_type instances_; Luigi -- Though this be madness, yet there is method in't. -- Hamlet, Act II, scene II ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
private:
static map_type instances_; It works. No need to for scoped_ptr. I tested it with Intel compiler. ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
I could run several examples but the test suite crashes instantly with
memory access violation error. Can someone try the test suite with the singleton attached below? //////////////////////////////////////////////////////////////////// #ifndef quantlib_singleton_hpp #define quantlib_singleton_hpp #include <ql/types.hpp> #include <boost/noncopyable.hpp> #include <boost/shared_ptr.hpp> #include <map> namespace QuantLib { #if defined(QL_ENABLE_SESSIONS) Integer sessionId(); #endif template <class T> class Singleton : private boost::noncopyable { public: static T& instance() { #if defined(QL_ENABLE_SESSIONS) Integer id = sessionId(); #else Integer id = 0; #endif boost::shared_ptr<T>& p = instances_[id]; if (!p) p.reset(new T); return *p; } protected: Singleton() {} private: typedef std::map<Integer, boost::shared_ptr<T> > map_type; static map_type instances_; }; template <typename T> typename Singleton<T>::map_type Singleton<T>::instances_; } #endif ------------------------------------------------------------------------------ This SF.net Dev2Dev email is sponsored by: Show off your parallel programming skills. Enter the Intel(R) Threading Challenge 2010. http://p.sf.net/sfu/intel-thread-sfd _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Thu, 2010-09-02 at 02:49 -0500, Kakhkhor Abdijalilov wrote:
> I could run several examples but the test suite crashes instantly with > memory access violation error. Can someone try the test suite with the > singleton attached below? Tried it on an Ubuntu system with gcc 4.4.3. Worked fine... Luigi -- The shortest way to do many things is to do only one thing at once. -- Samuel Smiles ------------------------------------------------------------------------------ Automate Storage Tiering Simply Optimize IT performance and efficiency through flexible, powerful, automated storage tiering capabilities. View this brief to learn how you can reduce costs and improve performance. http://p.sf.net/sfu/dell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
Thanks a lot. That is encouraging.
What if the instance() method is synchronized with mutex, could it solve static singleton problem? ------------------------------------------------------------------------------ Automate Storage Tiering Simply Optimize IT performance and efficiency through flexible, powerful, automated storage tiering capabilities. View this brief to learn how you can reduce costs and improve performance. http://p.sf.net/sfu/dell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Fri, 2010-09-10 at 10:25 -0500, Kakhkhor Abdijalilov wrote:
> Thanks a lot. That is encouraging. > > What if the instance() method is synchronized with mutex, could it > solve static singleton problem? I'm not sure it's a synchronization problem. Are you? (Also, we'd have the brand new problem of choosing a threading library.) Luigi -- Testing can never demonstrate the absence of errors in software, only their presence. -- W.E. Dijkstra ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
It worked with all examples, but the test suite crashes with access
violation error. I suspect it as to do something with compiler settings. Bjam uses different compiler settings for icc and it must be run with preset environment variables. Some of the settings could be different from Visual Studio project settings. I will recompiled boost and QuantLib and try it again. ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
Did anyone run the test suite with boost 1.44? It is failing under
Visual Studio 2008 as well as under ICC. In the past v1.01 test suite run fine under Visual Studio 2008, but now it is crashing with memory access violation error message. Maybe something is wrong with boost? ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
|
|
On Thu, 2010-09-16 at 05:01 -0500, Kakhkhor Abdijalilov wrote:
> Did anyone run the test suite with boost 1.44? It is failing under > Visual Studio 2008 as well as under ICC. In the past v1.01 test suite > run fine under Visual Studio 2008, but now it is crashing with memory > access violation error message. Maybe something is wrong with boost? It's running with gcc on Linux. I might try a Windows box, but it's going to take me some time. If anyone could report success with VC2008 and Bppst 1.44, that would be of help. Luigi -- All generalizations are dangerous, even this one. -- Alexandre Dumas ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
On Thu, Sep 16, 2010 at 1:50 PM, Luigi Ballabio
<[hidden email]> wrote: > If anyone could report success with VC2008 > and Bppst 1.44, that would be of help. no problem with the current trunk on my Windows XP box using VC2008 Express and boost 1.44 release configuration ciao -- Nando ------------------------------------------------------------------------------ Start uncovering the many advantages of virtual appliances and start using them to simplify application deployment and accelerate your shift to cloud computing. http://p.sf.net/sfu/novell-sfdev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
In reply to this post by Kakhkhor Abdijalilov
Unfortunately, I couldn't get it to pass the test suit. The suite
fails even with the stock QL on Visual Studio (it crashes instantly). I suspect there is some conflict between compiler setting used to compiler boost and QL. Reinstalling and recompiling didn't help. I would have to reformat HD or get a new box at some point in the future. Maybe someone could test it for us on Visual Studio? The file is singleton.hpp and available from sourceforge. The problem is with the test suite only. Everything else works fine on ICC and Visual Studio. If the new singleton passes all tests on Visual Studio, we could keep it. Complete ICC support would require more than just fixing singleton. Let's postpone it until v1.2. I am planing to work on it after transition to VS2010. Regards, Kakhkhor Abdijalilov. ------------------------------------------------------------------------------ Nokia and AT&T present the 2010 Calling All Innovators-North America contest Create new apps & games for the Nokia N8 for consumers in U.S. and Canada $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store http://p.sf.net/sfu/nokia-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
Re: fix for intel's compiler
|
|
Sounds good to me. Also, please feel free to rename interface
functions if you think it fits better QuantLib's naming convention. Regards, Kakhkhor Abdijalilov. ------------------------------------------------------------------------------ Nokia and AT&T present the 2010 Calling All Innovators-North America contest Create new apps & games for the Nokia N8 for consumers in U.S. and Canada $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store http://p.sf.net/sfu/nokia-dev2dev _______________________________________________ QuantLib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
«
Return to quantlib-dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |