Program does not run after processor change

[Freeza 2]

Hello,
I develop my own 3D engine, which is beginning to be consistent even if it is far from finished. So far I develop and carry out my tests with a PC with an NVidia card (GeForce FX 5600), I had no problem. Having switched from PC now with ATI GPU (HD 5750) I realized that the engine compiled with my old PC to load properly, load the meshes properly, but displays absolutely nothing. However, there is no error (an appeal glGetError after each directive OpenGL to check). Even better: after rebuilding my engine on my new PC with the same Visual Studio before and without touching either the code or adjacent to bookshops, it works properly on my new pc but not my pc tif (which NVidia GPU). Going up another notch: I wanted to test my engine to a developer with a machine with an ATI Radeon 7000 and any version that I pass it (compiled on ATI or NVidia compiled on) meet the problem of having no display without error. Does anyone have an idea to help me in my research on this bug? Because here we are a bit (totally) confused.

[Gillian Anderson]

I advise you to precisely target maps where you want your engine running or not, without being too ambitious. Buy the card you used minimum for each vendor, ATI and NVIDIA see Intel if you're brave. Today I'll tell you to target the cards above Geforce 8 / ATI HD 2400, but you might want to do better. Support the ATI R600 is below the super painful. The drivers are buggy and will never be repaired. NVIDIA is generally too lax, so badly ATI implements the specification. Intel I prefer not to talk. Then take the latest drivers but still does not change too often either because something breaks often to each driver. Wait until the others take this risk for you. Inspire yourself Ogre3D code that's been there too. For example for the creation of texture it is best not to use the texture Npot (yes, even if the card says it supports them, in fact, not really), and create each mipmap level one by one. Try to understand the separation between state and state Texture Sampler, if you use the multi-texturing. Used the glGet in all directions and, doubting everything, and eventually you will find compatibility. You might want to check error codes everywhere, but if the driver plant is useless .

[Messenger]

It seems to point to a problem on your side, it is not because a buggy code works "by accident" on a machine that will work anywhere, anytime. In short, trying to start from simple tutorials, which you have the assurance that they walk everywhere.

[NetworkeR]

The best is to set the card type minimum to run the program and buy to make his tests above, I would say an ATI preferences that are much less flexible and laid on you a pinch. You're more likely if your program runs under it will work in ATI nVidia than the reverse. After solving your problem so that your engine is already well advanced I wish you good luck . Try to do things really minimalist quickly detect errors incompatibilities.

[Freeza 2]

I come to you, always with the same problem, but with advanced targeting. In fact I managed (I think) to reproduce on my machine, but it is quite disconcerting. So I compiled with Visual Studio. In Debug, by running with Visual Studio or directly by double-clicking on the exe, so good. In Release, by running with Visual Studio, I have no problem, cons double-clicking on the exe the problem appears. The throwing with Visual Studio, what does that make? It initializes the variables? It sets up a debug environment.

[GlassFish]

Since the problem occurs randomly depending on the machine and never debugging, I'd say it's actually a problem initializing the variable, or a writing end of the table. A check with "Application Verifier". For the initialization of variable, now we must look. The only way I see it, is to log the important parts of the program and look what that does not run in release or attach the debug on the program during the crash.