Full Parallelism in Uniform Nested Loops Using Multi-Dimensional Retiming

Authors:
Nelson Luiz Passos;Edwin Hsing-Mean Sha
Affiliations:
University of Notre Dame, USA;University of Notre Dame, USA
Venue:
ICPP '94 Proceedings of the 1994 International Conference on Parallel Processing - Volume 02
Year:
1994

Citing 0
Cited 8

Fully Parallel Hardware/Software Codesign for Multi-Dimensional DSP Applications

CODES '96 Proceedings of the 4th International Workshop on Hardware/Software Co-Design
Pursuing a Petaflop: Point Designs for 100 TF Computers Using PIM Technologies

FRONTIERS '96 Proceedings of the 6th Symposium on the Frontiers of Massively Parallel Computation
Maximize Parallelism Minimize Overhead for Nested Loops via Loop Striping

Journal of VLSI Signal Processing Systems
Optimizing parallelism for nested loops with iterational and instructional retiming

Journal of Embedded Computing - Selected papers of EUC 2005
Optimal loop parallelization for maximizing iteration-level parallelism

CASES '09 Proceedings of the 2009 international conference on Compilers, architecture, and synthesis for embedded systems
Loop striping: maximize parallelism for nested loops

EUC'06 Proceedings of the 2006 international conference on Embedded and Ubiquitous Computing
Optimizing nested loops with iterational and instructional retiming

EUC'05 Proceedings of the 2005 international conference on Embedded and Ubiquitous Computing
Low power engineering

Embedded Systems Design

Quantified Score

Hi-index	0.00

Visualization

Abstract

Most scientific and DSP applications are recursive or iterative. Uniform nested loops can be modeled as multi-dimensional data flow graphs (DFGs). To achieve full parallelism of the loop body, i.e., all the computational nodes executed in parallel, substantially decreases the overall computation time. It is well known that for one-dimensional DFGs retiming can not always achieve full parallelism. This paper shows an important and counter-intuitive result, which proves that we can always obtain full-parallelism for DFGs with more than one dimension. It also presents two novel multi-dimensional retiming techniques to obtain full parallelism.