I’m developing an application using CUDA 2.0 beta on Debian GNU/Linux with a 8400M GT card.
Lately I’ve got a problem with the following code:
texture< uint2, 1 > table; uint debugInfo; uint h = hash(...); uint2 texel0 = tex1Dfetch( table, 15829 ); uint2 texel1 = tex1Dfetch( table, h ); debugInfo = h; debugInfo = texel0.x; debugInfo = texel0.y; debugInfo = texel1.x; debugInfo = texel1.y;
Then, I memcopy the debugInfo array to host memory and print it. I obtain the following result:
15829 0 2162688 0 0
That is: h is 15829, but the two texture fetches return different results. Why?