Taking the address of a texture why it is not supported? It would be nice to do selectors...

Accelerated Computing CUDA CUDA Programming and Performance
1

I would like to do smth like this and results are not correct (CUDA 2.2):

texture<float2, 1, cudaReadModeElementType> _texA = tex_a;

if(Sel == 1) _texA = tex_b;

else if(Sel == 2)  _texA = tex_c;

for(int i = 0; i < 8; i++) {

	const float2 val = tex1Dfetch(_texA, i);

  ....

Sometimes I receive this:

Error: External calls are not supported (found non-inlined call to _ZN7textureI6float2Li1EL19cudaTextureReadMode0EEC1Ei21cudaTe

xtureFilterMode22cudaTextureAddressMode)

However this is of course supported:

texture<float2, 1, cudaReadModeElementType> _texA = tex_a;

if(Sel == 1) _texA = tex_b;

else if(Sel == 2)  _texA = tex_c;

for(int i = 0; i < 8; i++) {

	const float2 val = Sel == 0 ? tex1Dfetch(_tex_a, i) : (Sel == 1 ? tex1Dfetch(_tex_b, i) : tex1Dfetch(_tex_c, i));

}

I hope that first code could be more efficient :)