use int2 to get double texture

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi, I got a problem during the rumtime, the error is:

Cuda error in file ‘texture.h’ in line 20 : invalid texture reference.

I used int2 to get double texture, it work in CUDA2.3, but error in CUDA3.1

the code in my texture.h is:

texture<int2, 1>  tex_x_double;

void bind_vec(double *x)

{  

<b>line 20</b>:   CUDA_SAFE_CALL(cudaBindTexture(NULL, tex_x_double, x));  

}

__inline__ __device__ double fetch_vec(const int &i, double *x)

{

	int2 v = tex1Dfetch(tex_x_double, i);

	return __hiloint2double(v.y, v.x);//avoid banks conflict

}
2

Don’t know, works for me. Is this really the file texture.h? If yes, maybe consider placing executable code in .cu files, not header files.

3

Don’t know, works for me. Is this really the file texture.h? If yes, maybe consider placing executable code in .cu files, not header files.