Abstract— The problem of in-operation fault detection in mesh-connected multicore and many-core VLSI multiprocessors is considered. A novel approach to the multiprocessor test based on the combination of self-test and mutual inter-unit test performed in a multiplexed mode is presented; it allows for an increase in the successful fault detection probability. Formal rules are defined for forming sets of testing and tested neighbors for each processor node of the mesh that is invariant to the location of the node within the mesh and its dimension. In contrast to the mutual inter-unit test mechanism, the same testing neighbor is alternately used to test the two processors in a multiplexed mode; and the test result is formed by applying the majority operator to the individual faulty/healthy tags calculated by all testing neighbors and the current unit itself in the course of its self-test. The formulae for determining the number of testing neighbors for each node depending on the dimension of the mesh are given. The successful fault detection probability is evaluated in the case when the proposed approach is used. The successful fault detection probability versus the individual test unit reliability dependencies is investigated. For all practically significant cases, the proposed approach is shown to provide an increased successful fault detection probability compared to the mutual inter-unit test and self-checking.