怀沙'志

There is a known issue/BUG with MFC, were MFC makes a call to
_CrtDumpMemoryLeaks() in the destructor of the _AFX_DEBUG_STATE, followed by _CrtSetDbgFlag() which sets it to ~_CRTDBG_LEAK_CHECK_DF (therefor disabling memory leak test at *true* program exit) This destructor is called at exit (i.e. atexit()), but before statics residing in dlls and others are destroyed, resulting in many false memory leaks are reported

As to fix any real memory leaks, you have the source ... also you can do a
google to see how others have gotten around this issue to get at any real
leaks.

The MFC memory leak will not go away as Microsoft have no reason to fix it( it been there for many years) as MFC is a deprecated API as far as they are concerned


Another solution to the MFC and wrong memory leak reports. You need to make your MFC dll load first during application start up and unload last during application exit. I managed to do this by making the following change.

In your MFC application.

. Goto project settings. In there, make the following changes for theDebug build.
. General->Use of MFC->Use Standard Windows Libraries.
. Add _AFXDLL to C/C++->Preprocessor->Preprocessor Definitions.
. Add mfc80??.lib (in my case it is mfc80ud.lib) as a first dependency or at least before osg libs to Linker->Input->Additional
Dependencies.

You can find what the other mfc libs are in your visual studio install directory\VC\atlmfc\lib. You don't need to do this release mode.


Other solutions possible :

If you add a MFC dependency to osg dlls that should also create the desired effect. Change the Use of MFC setting to General->Use of
MFC->Use MFC In a Shared DLL in your debug builds of osg libraries.

This should add MFC dll dependency to your osg dll in debug so forcing MFC dll to unload after the osg dlls but as I said I did not actually
try this.

Rebuild the MFC dlls with the following change. At least remove the "_CrtDumpMemoryLeaks()" call in
_AFX_DEBUG_STATE's destructor. It is in VS_INSTALL\VC\atlmfc\src\mfc\dumpinit.cpp. The makefile is in VS_INSTALL\VC\atlmfc\src\atlmfc.mak. I did not try this, either.

　　貌似很久没来更新这个BLOG了，其实一直没有离开，只是本科的时候有时候.NET方面的东西，现在一心一意搞图形学了。博客圆的计算机图形学版我一直都很关注，这里也是国内为数不多的可以讨论图形学的地方，只是一直没有写BLOG，太懒了，呵呵。
　　之前很长一段时间一直在搞虚拟展示的项目，做项目着实是一件很烦心的事情，用到的技术倒不复杂，只是要关注的东西很多，功能、稳定性，以至于有一段时间一直在改Bugs，有时候跟踪了很久发现是OSG的Bug。呵呵。不过暂时告一段落了。可以说做项目是一件积累经验的事情，旁枝末节的东西很多，涉及的东西又很杂，一个研究性的DEMO就要简单的多，如果项目里要用到一个效果，拿一个DEMO过来就想把它马上用上去，有时候不是一件容易的事情，一来DEMO考虑的实验环境相对比较简单、单纯，而不像项目里那样是一个完全动态、高度可定制的环境，二来还要考虑会不会和其它技术相互冲突。中间几个月和中国美术学院新媒体系的一伙人参加了2007年的上海电子艺术节，做了一些互动的艺术表演。这个东西很有意思。



　　
过完年开始在做一些GPU全局光照方面的东西，本科的时候做过基于GPU加速的光线跟踪，当时是基于Stanford University的Timothy J. Purcell的相关论文，采用Uniform Grid预处理方法做了一个实现，做得比较简陋，速度上稍有问题，后来也就扔在那里没去改进了。后来发现ShaderX4上Martin Christen有一篇Implementing Ray Tracing on the GPU，采用了类似的方法。现在再去看就觉得老了，对动态场景的预处理也蛮耗时。倒是GPU Gem3里介绍的Baking Normal Maps on the GPU方法，采用Unform Grid这种加速结构还比较适用。

想来居然有一年没有更新这边的BLOG了，也是自己惰性所致，不过还是每天都有到博客园来逛。
最近一直在做一个家具虚拟展示的项目，投入了比较大的精力。
这两天做了一下镜面反射的实时绘制