Scaling simulation of the fusing-restricted reconfigurable mesh

Using Bus Linearization to Scale the Reconfigurable Mesh

“…This approach has a log N simulation overhead, which improves on the best previous deterministic simulation overhead [8] of log P log N P . Furthermore, the simulating LR-Mesh uses only exclusive writes, whereas all prior scaling simulations required concurrent writes (see Table 1).…”

“…The best previous results on scaling the unrestricted R-Mesh were due to the authors [8] and Matias and Schuster [15] (see Table 1). We developed a scaling simulation for the unrestricted R-Mesh with a simulation overhead of log P log N P .…”

“…The authors [7,8] have proved that the fusing restricted reconfigurable mesh (FR-Mesh), another restriction of the R-Mesh, has a strong scaling simulation with a simulation overhead of log P (see Table 1). We also use bus linearization to derive an FR-Mesh scaling simulation that runs on a weaker simulating model (see Table 1); the previous simulation requires a CRCW FR-Mesh, while the new simulation needs only a CREW LR-Mesh.…”

“…The authors [8] previously developed a scaling simulation of an N × N FR-Mesh on a P × P FR-Mesh. The simulating machine processes one P × P sized window of the big FR-Mesh at a time.…”

“…Since the above FR-Mesh simulation [8] requires the simulating LR-Mesh to find this equivalent configuration a constant number of times per window, the simulation overhead of this new simulation is still O(log P) (see Table 1). …”