parallel_primes_rs

A command-line program that finds all primes between two numbers passed to it in the command line.

Overview

This is a parallelized Rust program where you can pass it two integers and it will print a vector of all of the primes between the two integers (inclusive) to standard out. It is parallelized and you can change the number of threads with the -N option

For example

$ ./primes_cli 10 19

prints

[11, 13, 17, 19]

The algorithm to determine if a number is prime is naïve brute force, with some care to avoid overflow errors. It uses a work-stealing threadpool provided by rayon and the default is to set the number of threads equal to the number of physical cores available, but you can change this by calling e.g.

$ ./primes_cli -N 1 10 19

More info can be found by typing

$ ./primes_cli -h

Building parallel_primes_rs

This program uses a typical rust project structure containing library ./src/lib.rs and a binary file ./src/bin.rs. To build the program run cargo build --release and you will create an executable at ./target/release/primes_cli

You can also run the automated tests by running cargo test --release

Strong Scaling Test

Because this problem is embarassingly parallel & compute bound, we should expect the program to scale nearly ideally (~num of threads) as long as each thread can have its own core. There is a little bit of a subtlety because if you actually print the output to the terminal, you may become I/O bound. So to test scaling, I commented out the line in bin.rs that prints to stdout, and timed the code e.g., $ time ./primes_cli -N 1 2000000 5000000 on my macbook with a 2.5 GHz Quad-Core Intel Core i7, changing number of threads from 1 to 4. The results are plotted below

As you can see, the problem scales nearly ideally.

Dependencies

• clap for arg parsing.
• rayon for parallelization.
• anyhow for nicer error handling.