prime

This project is for calculating and storing prime numbers. It uses the Sieve of Eratosthenes to mark bits in a block of memory. As there are an infinite number of prime numbers, this class can apply the sieve to an ever increasing list of blocks. I'm maintaining a wiki with more information.

For version 0.4, OpenMP support was added to greately increase speed by using multiple cores on the CPU. It is limited to size_t (64 bits) numbers. The plan is to extend this out to "__int128" (128 bits) or use one of the "infinite" bit classes for integers. You will likely run out of disk space before using up size_t space.

Storage of the prime numbers, at least the initial numbers, is inefficient as a list. Storage of a bit for each number (true for prime) is initially more efficient. This list of bits can be cut in 1/2 by only storing odd numbers. It can be further cut by 1/3 by eliminating numbers divisible by 3, 1/5 by eliminating numbers divisible by 5 and so on. Since future cuts beyond 5 give diminishing returns, version 0.3 stops here. I invite anyone with a more efficient method to let me know of your solution so I can further compress the data.

P.S. The most efficient way to store these numbers is to just store the algorithm. However, we want to be able to look up the numbers in roughly constant time.

Some useful links:

The Nth Prime Page

How Many Primes Are There?

Some next steps are:

1. Optimization of the C++. There are a number of ways to make if faster. If you see one, let me know or submit a change.
2. More efficient storage of the numbers (ping me if you have ideas).
3. Tools for metadata, extraction of interesting information.

While C++ is efficient, Python is more fun for trying out clever ideas. Some next "fun" steps are:

1. Python code to read the output files. There is a rough version of this now.

2. Python image creation based off prime numbers. See below for an example of an Ulam Spiral

3. Maybe a "primes as a service" implementation :-)

Motivation: This is just a fun project. I got a scholarship to the University of Florida for some work I did with prime numbers in high school. It took me days to calculate a set of prime numbers that today can be calculated in a few seconds.

Changes since Last Version:

For version 0.4: Mainly updated for OpenMP inlcuding a number of new test cases for validation. The numbers have been validated up to 11,020,102,204,020 which contains 380,088,301,203 primes. Thanks to Andrew Booker for this : https://primes.utm.edu/nthprime/index.php#piofx

For verison 0.3 the following changes were made:

1. Require environment variable BOOST_ROOT to be set so that hard coded paths can be removed.
2. Updated to Visual Studio 2019 and v142 toolset.
3. Updated to Boost 1.72
4. Added a few more python functions for creating and searching though Ulam Sprials. These aren't ready for real use but may be a useful example.

For version 0.2, the following changes were made:

1. Use of Boost bitset so that the size of the bitset does not require recompile of the DLL/so
2. Addition of C++ Catch. Creation of new test case project and a number of test cases.
3. A new project called "go" that just keeps running and running looking for new primes. You can stop it and restarted it and it will remember any blocks of primes it already completed.
4. Python files for reading in the prime files and creating some prime bases images. This is still rough and in development but I inlcude it here in case anyone wants a parcer for the files.