ahmad-alkadri/curious-julia-numpy-matrix-solver

Repo created as exercise and out of curiosity: benchmarking Python's Numpy against default Julia matrix solver

NumPy and Julia Benchmarking

This repository contains scripts to benchmark the performance of matrix equation solving (Ax = B) using NumPy (Python) and Julia.

Overview

The benchmarking process is divided into three main parts:

1. Matrix Generation: Random matrices and vectors are generated using NumPy.
2. NumPy Benchmarking: The numpy.linalg.solve function is used to solve the equation and its performance is recorded.
3. Julia Benchmarking: The \ operator in Julia is used to solve the equation and its performance is recorded.

All the generated matrices, vectors, and results of the benchmarking are stored as .txt and .csv files respectively.

Dependencies

• Python 3
• NumPy
• Julia
• Python memory_profiler library
• Julia packages: BenchmarkTools, DelimitedFiles, CSV, DataFrames

Directory Structure

• matrix_generation.py: Python script to generate matrices and vectors.
• numpy_side.py: Python script for benchmarking with NumPy.
• julia_side.jl: Julia script for benchmarking with Julia.
• Makefile: Automates the setup, matrix generation, benchmarking, and cleanup processes.

How to Run the Benchmark Tests

Ensure you have Python 3, NumPy, and Julia installed on your machine.

Step-by-step Instructions:

1. Clone the Repository:

git clone https://github.com/ahmad-alkadri/curious-julia-numpy-matrix-solver.git julia_numpy_matrix
cd julia_numpy_matrix

2. Run the Benchmark Tests:

Use the provided Makefile to run the benchmark tests:

make all

Explanation of major Makefile targets:

• genmatrix: Generates matrices and vectors used for testing.
• numpy: Runs the NumPy benchmark.
• julia: Runs the Julia benchmark.
• combine_results: Combines the results of NumPy and Julia benchmarks into a single .csv file.
• all: Runs all of the above targets in order.

3. View Results:

Once the tests have run, the results will be available in the results.csv file.

4. Clean Up (Optional)

To clean up generated files and reset the environment, run:

make reset

This will remove generated matrices, vectors, results, and virtual environments.