Mathist is a math website project created by a mathematician who loves primes. The name is inspired by a great German mathematician known as the Mathologer; I am just a mathist.
The platform is built with nextjs and leverages powerful computation libraries like BigInt, mathjs, and s-bpsw to explore the fascinating world of prime numbers. Deployed at math.ideniox.com, Mathist allows users to experiment with advanced arithmetic, prime testing, and more.
This project began 25 years ago with my first implementation of the Sieve of Eratosthenes on a machine with just 1GB of RAM. Since then, Mathist has evolved into a robust platform for modern mathematics exploration.
At https://math.ideniox.com/files/test.html you can find more informacion about the calculations done in the software.
0 - Improve visualization and data export, example of view:
Path entered: 12121221212101010 Steps computed: 17 Time taken: 264 μs
Generated Fibonacci-like square: [235701, 72811, 381323, 308512]
Pythagorean triple generated: <89878212423, 44926134464, 100481095865>
Pythagorean tree (depth 3): <3, 4, 5> <-- Root <15, 8, 17> <21, 20, 29> <5, 12, 13> <35, 12, 37> <65, 72, 97> <33, 56, 65> <77, 36, 85> <119, 120, 169> <39, 80, 89> <45, 28, 53> <55, 48, 73> <7, 24, 25>
Click here to visualize in GeoGebra: Tree Visualization
1 - An experimental feature involves mersenne primes generation and test. For that I created a separated repository to test and benchmark different languages for the arithmetic computation, https://github.com/JorgeMartinezPizarro/lucas-lehmer-server. Currently I am working on a GPU based implementation. Inspired on the GIMPS project. A lot have been done since the first time I looked at this problem years ago. My current record with CPU goes up to a mersenne prime with about 300000 digits, yet very small. Let's see what can a modern GPU do. Furthermore, LLT is being replaced by PRP in 2021. Besides it, trial divisions and Fermat little theorem may be used before to speed it up.
2 - Avoid main thread usage on the backend. In the first versions I just wanted to try out nextjs. Backend in js may be an issue if it uses the main thread. Nginx load balancer is a workaround but I want to go for multithread single js docker. Let see if I get it working.
3 - Parallelization of Segmented Sieve algorithm, I would like to speed up the process of counting primes up to a number. The current record is about 10**23.
To get it running locally, you need to install the latest node and npm.
To run a develop version:
npm run dev
To build and run a production version:
npm run build
npm run start
Open localhost:3000 with your browser to see the result.
Running in production will fail the sieve to DOWNLOAD, since public/files generated on the fly are not accessible, so we need a webserver serving the public/files directory.
The easiest way to deploy the website is using docker.
There is a version of mathist dockerized under hub.docker.com.
To run it you need docker installed on your system and run:
docker run -d jorgemartinezpizarro/mathist:latest
To generate your own docker image, run:
docker build .
I use an apache2 file server to serve files inside the docker volumes, an example docker-compose.yml:
services:
files:
image: httpd:latest
volumes:
- /VOLUMES_PATH:/usr/local/apache2/htdocs/files
restart: always
ports:
- 2900:80
mather:
image: jorgemartinezpizarro/mathist:latest
restart: always
ports:
- 3000:3000
volumes:
- /VOLUMES_PATH:/app/public/files
and the /etc/nginx/sites-available/default nginx config file:
server {
server_name YOUR_DOMAIN;
location /files {
proxy_pass http://localhost:2900;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
}
location / {
proxy_pass http://localhost:3000;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
}
listen 80;
}
For this to work you need to set up a valid value for YOUR_DOMAIN and VOLUME_PATH. If you plan to host the site on your own, I recommend to use a load balancer with several instances running, since javascript works in single thread.
To start the containers use docker compose up -d.