On Solving the Decycling Problem in a Torus Network

Modern supercomputers are massively parallel systems: they embody thousands of computing nodes and sometimes several millions. The torus topology has proven very popular for the interconnect of these high-performance systems. Notably, this network topology is employed by the supercomputer ranked number one in the world as of November 2020, the supercomputer Fugaku. Given the high number of compute nodes in such systems, efficient parallel processing is critical to maximise the computing performance. It is well known that cycles harm the parallel processing capacity of systems: for instance, deadlocks and starvations are two notorious issues of parallel computing that are directly linked to the presence of cycles. Hence, network decycling is an important issue, and it has been extensively discussed in the literature. We describe in this paper a decycling algorithm for the 3-dimensional k -ary torus topology and compare it with established results, both theoretically and experimentally. (This paper is a revised version of Antoine Bossard (2020)).

Hierarchical Hexagon: A New Fault-Tolerant Interconnection Network for Parallel Systems

A new interconnection network topology called Hierarchical Hexagon HH(n) is proposed for massively parallel systems. The new network uses a hexagon as the primary building block and grows hierarchically. Our proposed network is shown to be superior to the star based and the hypercube networks, with respect to node degree, diameter, network cost, and fault tolerance. We thoroughly analyze different topological parameters of the proposed topology including fault tolerance routing and embedding Hamiltonian cycle.

Enhanced Ant Colony-Inspired Parallel Algorithm to Improve Cryptographic PRNGs

We present and motivate a parallel algorithm to compute promising candidate states for modifying the state space of a pseudo-random number generator in order to increase its cycle length. This is important for generators in low-power devices where increase of state space to achieve longer cycles is not an alternative. The runtime of the parallel algorithm is improved by an analogy to ant colony behavior: if two paths meet, the resulting path is followed at accelerated speed just as ants tend to reinforce paths that have been used by other ants. We evaluate our algorithm with simulations and demonstrate high parallel efficiency that makes the algorithm well-suited even for massively parallel systems like GPUs. Furthermore, the accelerated path variant of the algorithm achieves a runtime improvement of up to 4% over the straightforward implementation.1