The Quantification and Analysis Method of Algorithm Essential Parallelism

Abstract

The invention discloses a method for quantifying and analyzing the intrinsic parallelism of an algorithm (intrinsic?parallelism?of?algorithm), which is suitable for configuring a processing unit to quantify and analyze the intrinsic parallelism of an algorithm, which includes the following steps: configuring the processing The unit is used to obtain several operation groups of the algorithm, the processing unit is configured to obtain a corresponding Laplacian matrix according to the operation group, and the processing unit is configured to calculate the eigenvalue and eigenvector group of the Laplacian matrix and configuring the processing unit to obtain an essential parallelism data of the algorithm according to the eigenvalues ​​and eigenvectors.

Description

technical field [0001] The present invention relates to a method for quantifying and analyzing the degree of parallelism of an algorithm (intrinsic parallelism of algorithm), in particular to a method for quantifying and analyzing the degree of intrinsic parallelism of the algorithm. Background technique [0002] G.M.Amdahl proposed an algorithm parallelization method based on the ratio of the continuous part of the algorithm ("Validity of single-processor approach to achieving large-scale computing capability," Proc.ofAFIPSConference, pages483-485, 1967), the shortcoming of Amdahl's method is that the parallelism of an algorithm is estimated by the above method is dependent on the target platform on which the method is executed, and does not need to be attached to the algorithm itself, so the degree of parallelism obtained with Amdahl's method is not essential to the algorithm and is offset by the target platform. [0003] A. Prihozhy et al. proposed a method based on the r...

AI Technical Summary

Problems solved by technology

[0002] G.M.Amdahl proposed an algorithm parallelization method based on the ratio of the continuous part of the algorithm ("Validity of single-processor approach to achieving large-scale computing capability," Proc.ofAFIPSConference, pages483-485, 1967), the shortcoming of Amdahl's method is that the parallelism of an algorithm is estimated by the above method is based on the target platform on which the method is executed, and does not need to be attached to the algorithm itself, so the parallelism obtained with Amdahl's method is not essential to the algorithm and is offset by the target platform

[0003] A. Prihozhy et al. proposed a method based on the ratio between the complexity of the algorithm and the length of the critical path to evaluate the parallelization possibility of the algorithm ("Evaluation of the parallelization potential for efficient multimedia implementations: dynamic evaluation of algorithm critical path," IEEETrans.onCircuitsandSystemsforVideoTechnology, pages593-608, Vol.15, No.5 , May2005), its complexity is the total number of operations in the algorithm, and the critical path length is the maximum number of operations that need to be processed consecutively due to operational data dependencies, although this method can be characterized in that there is a mean value of parallelism embedded in algorithm, but not enough to specifically feature volatile multi-grain parallelism embedded in the algorithm

Images