Hi, i’m trying to extend the marching cubes example of the sdk. I want that the dimension of the volume file is not dipendent of the global memory of my device. What i decide to do is to load not all the volume directly to the device memory, but only two slices so i should not have problem on data dependences.
The output array of the program is the merge of the single results of all the pairs of slices.
In theory, i can use only 2 slices because the marching cubes algorithms use two point on the slice i and two point in the slice i+1 to create the cube. The problem is that the result is different from the result that i obtain if i use the original marching cubes program.
I attach the 2 off file, result of my version of the program and the original one.
Can anyone explain me why this happen? How can i fix it?