Hi,
I am having problems debugging my custom CUDA code. I, shortly, simply initialize
all the elements of an array on GPU to 1.0 or 0.0 and copy it back to CPU.
When printing the host array, some elements are neither 1.0 nor 0.0.
I compile everything with arch 2.0., SDK 3.2.16 and I am working on a dual-GPU
(460GTX) box.
Running alignedTypes from SDK tells me all TESTs FAIL. Strange, as when
I installed SDK it was perfectly working.
Adding a printf line to alignedTypes
if( ((char *)&src)[i] != ((char *)&dst)[i] ){
printf( “%c %c\n”, ((char *)&src)[i], ((char *)&dst)[i]);
return 0;
}
produces output as in the attached figure.
How do I fix this?
Can the two things be correlated, so that fixing alignedTypes
will also help my app?
Thanks
