Exploiting More Parallelism from Applications Having Generalized Reductions on GPU Architectures | IEEE Conference Publication | IEEE Xplore