Scheduling of uniform multidimensional systems under resource constraints

Abstract

Multidimensional (MD) systems are widely used to model scientific applications such as image processing, geophysical signal processing, and fluid dynamics. Such systems, usually, contain repetitive groups of operations represented by nested loops. The optimization of such loops, considering processing resource constraints, is required in order to improve their computational time. Most of the existing static scheduling mechanisms, used in the high-level synthesis of very large scale integration (VLSI) architectures, do not consider the parallelism inherent to the multidimensional characteristics of the problem. This paper explores the basic properties of MD loop pipelining and presents two novel techniques, multidimensional rotation scheduling and push-up scheduling, able to achieve the shortest possible schedule length. These new techniques transform a multidimensional data flow graph representing the problem, while assigning the loop operations to a schedule table. The multidimensional rotation scheduling is an iterative "heuristic" method, depending upon an user input, while the push-up scheduling algorithm is able to compute the new schedule in polynomial time. The optimal resulting schedule length and the efficiency of the algorithms are demonstrated by a series of practical experiments.

A Survey of Loop Parallelization: Models, Approaches, and Recent Developments

“…Multi-dimensional data-flow graph (MDFG) was introduced by Passos et al [49]. It is an extended PDG, which can be represented by a tuple of four elements,…”

Optimal partitioning and balanced scheduling with the maximal overlap of data footprints

“…For the ALU schedule, the multi-dimensional rotation scheduling algorithm [7] is used to generate a static schedule for one iteration. Then the entire ALU schedule can be formed by simply replicating this schedule for each iteration in the partition.…”

“…The traditional software-directed prefetching generally has little consideration on how to increase the program's parallelism. On the contrary of the prefetching techniques, multi-dimensional loop pipelining techniques [7] can be used to explore more parallelism, even to the extent of achieving full-parallelism for uniformly nested loops. These techniques, however, do not consider the memory latency in £ This work is partially supported by CAM Graduate Fellowship their algorithms.…”

“…A loop body can be represented by a multi-dimensional data flow graph (MDFG) [7]. Each node in the MDFG is a computation.…”

PROSIDIS: a special purpose processor for protein similarity discovery

“…The theoretical foundations of the PHG derive from the works on systolic arrays conducted by many researcher in the last years [13][14][15][16]. Such works are mainly based on the projection of a regular computation domain onto a time-processor space in order to exploit the algorithm parallelism.…”