I am trying to implement a Minimum Spanning Forest Algorithm in OpenCL. Though I know how to implement Kruskal's algorithm, it is sequential in nature, so I think Boruvka's algorithm or the Data Parallel Kruskal's algorithm which is specially made for parallel programming should be implemented in OpenCL.
But in this algorithm(as per mentioned in the paper : "Fast and Memory-Efficient Minimum Spanning Tree on the GPU") we need Graph Primitives like "sort", "split graph", "connect components", "find mins", "contract graphs" for the implementation. Does OpenCL have its Data Parallel Primitives library as the CUDA does...." CUDAPP"..? Or is there any source in OpenCL where I can get these functions...?
There are a few libraries AMD has also been developing and being developed by others as open-source projects.
This is a opensource parallel primitives library : http://code.google.com/p/clpp/
Not sure how much development has been done here though.
BOLT from AMD, has also been open-sourced recently, and you can check out there plans further. They might have a few routines you need already implemented.