Example of Master-slave design pattern implementation for matrix multiplication

Introduction

This repository contains an example of master-slave design pattern, used for matrix multiplication.

There is a Matrix class (immutable from outside) and two matrix calculators in the project.
SimpleMatrixCalculator goes row by row and column by column and calculates the result.
ParallelMatrixCalculator uses master-slave design pattern and firstly divides the work for slaves: each slave calculates multiplication of 1 row with the second matrix (e.g. one row of the product) and starts them in parallel (using Tasks) Then, it waits until all the slaves finish, creates a product from the slave results and returns it.

Both matrix calculators inherit from abstract class MatrixCalculator which contains common methods for all of the calculators.

There is also a simple benchamrk in Master_slave_design_pattern.Benchmark project. You can set matrix dimension, number of repetitions there and compare the time of both calculators.

Requirements

• .NET 5.0

Run

Simple example

dotnet run --project Master_slave_design_pattern

Benchmark

dotnet run --project Master_slave_design_pattern.Benchmark

Benchmark result

• average computation times measured with randomly generated matrices of the given size with 10 repetitions

My PC

Matrix size	simple algorithm time (ms)	parallel algorithm time (ms)
10 x 10	4	10
50 x 50	8	16
100 x 100	37	30
200 x 200	356	157
300 x 300	702	314
500 x 500	2540	1256

MFF lab (u-pl0)

Matrix size	simple algorithm time (ms)	parallel algorithm time (ms)
10 x 10	< 1	< 1
50 x 50	1	< 1
100 x 100	11	4
200 x 200	89	28
300 x 300	266	89
500 x 500	1215	428