https://hgpu.org/?p=9862
Parallelization Strategies for Local Search Algorithms on Graphics Processing Units