I believe I found a bug in this new nvcc/toolkit. Adding a texture inside a namespace cause errors in cudaMalloc (error 10500), and probably in many other functions. In 2.3 toolkit this works fine.
To reproduce the bug, just add in any .cu file:
[codebox]namespace MyNamespace{
texture<float,1,cudaReadModeElementType> bugTest;
}[/codebox]
Its a runtime bug.
I`ve tested it only on a GTS 360M, sm_12, driver 197.13, visual studio 2008, Windows 7 x64, but running CUDA x86.
I can confirm that the bug also exists in CUDA 3.1b and 3.1 release. Is there any further progress on this bug? Also, since we spent several days tracking down the reason why a single texture declaration ruins the entire program and finally found that it’s compiler bug, I start wondering if there’s any public bug tracker system for CUDA so that we can check it first before wasting time.