Previous | Next --- Slide 34 of 69
Back to Lecture Thumbnails

To explain the conflict graph, a node in the graph represents an edge in the mesh. An edge in the graph signifies that the CUDA threads assigned to those two edges in the mesh will be trying to modify the same flux field (for the same vertex). So for example, the CUDA threads assigned to edges 1 and 5 in the mesh both write to the flux field for vertex C, and so there is an edge in the conflict graph between the nodes labeled 1 and 5.


This is basically like a graph coloring problem.