Adding Embarrassingly Parallel Algorithms in NetworkX to nx-parallel

[Schefflera-Arboricola]

Embarrassingly parallel algorithms

Embarrassingly parallel algorithms are those that can be easily and efficiently divided into smaller, independent tasks that can be executed concurrently across multiple processing units, such as CPUs, GPUs, or nodes, without requiring significant communication or synchronisation between them. This type of parallelism is often referred to as “embarrassingly parallel” because it’s so straightforward and easy to implement. Refer this to see how to implement embarrassingly parallel algorithms using joblib.

Characteristics of Embarrassingly Parallel Problems(Algorithms):

• Independent tasks: Each task can be executed independently, without relying on the output of other tasks.
• No communication: Tasks do not require frequent or significant communication with each other.
• No data dependencies: Tasks do not rely on shared data or intermediate results from other tasks.
• Easy to divide: The problem can be easily split into smaller, parallelizable tasks.

Adding a new algorithm to nx-parallel

Step 1 : Find a NetworkX algorithm that is embarrassingly parallel

• You can find such an algorithm from the NetworkX's references - https://networkx.org/documentation/latest/reference/index.html
• Make sure there is a _dispatchable decorator(in the networkx's codebase) on the algorithm you have selected to parallelise.

Step 2 : Complete the checklist in Contributor's guide

This checklist in Contributor's guide for adding a new algorithm to the package.

Some references:

It would be better to look at currently existing algorithms than previous PRs that add parallel algorithms(like #60 , #37 , #45 and many more) because we have changed a few things.

Feel free to ask questions if you get stuck anywhere or if anything is not clear or if you want something to be explained in a bit more detail here.

Thank you :)

[RohitP2005]

Hey , i would like to work on this issue

[Schefflera-Arboricola]

Hi @RohitP2005 , let us know if you have any questions :)

Also, writing parallel versions of functions that return a non-generator(i.e. functions that return and not yield) would be easier to begin with.

[RohitP2005]

Thank you @Schefflera-Arboricola , I will reach out to you if i need anything

[RohitP2005]

@Schefflera-Arboricola i have made a pull request traversal
algorithm added