A PRAM-NUMA model of computation for addressing low-TLP workloads | IEEE Conference Publication | IEEE Xplore