Linear time complexity GF(256) RaptorQ implementation on GPU

RaptorQ is the most advanced raptor code and has an overhead-failure curve close to the random fountain code over the GF(256) finite field. Theoretically, it is possible to encode and decode with linear time complexity by an inactivation decoding algorithm, which is a hybrid algorithm of belief propagation and Gaussian elimination. However, achieving linear time complexity in a real-world implementation is a challenging task. In this paper, we provide an algorithm and data structure to implement inactivation decoding with near-linear time complexity using graphic processing unit.