Dynamic Parallelism in CUDA allows for the support of kernels to create and synchronize new nested kernels. Additionally, for our use case it also allows us to spend more time on the device to process information quickly without constant cudaMemcpy() or cudaMalloc() calls.