https://github.com/pmocz/nbody-performance-comparison

Directly compare performance of vectorized and for-loop implementations of the N-body problem in Matlab, Julia, and Python
https://github.com/pmocz/nbody-performance-comparison

Last synced: 3 months ago
JSON representation

Directly compare performance of vectorized and for-loop implementations of the N-body problem in Matlab, Julia, and Python

• Host: GitHub
• URL: https://github.com/pmocz/nbody-performance-comparison
• Owner: pmocz
• License: gpl-3.0
• Created: 2020-09-13T17:25:04.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2020-09-13T17:33:38.000Z (about 4 years ago)
• Last Synced: 2024-07-04T02:15:07.020Z (4 months ago)
• Language: Python
• Size: 85.9 KB
• Stars: 6
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# nbody-performance-comparison

Directly compare performance of vectorized and for-loop implementations of the N-body problem in Matlab, Julia, and Python

### Philip Mocz (2020) Princeton Univeristy, [@PMocz](https://twitter.com/PMocz)

Run the various versions:

```

matlab nbody_vec.m

julia nbody_vec.jl

python nbody_vec.py

matlab nbody_for.m

julia nbody_for.jl

python nbody_for.py

python nbody_for_numba.py

```

![Simulation](./comparison.png)

## Some Takeaways

* Python for loops are horribly SLOW!! Use **numba** to improve performance. Decorate bottleneck functions with:

```python

from numba import jit

@jit(nopython=True)

def myFunction( myInput ):

```

* Vectorized codes are FAST in all languages, and easier to read. The downside is that they require significantly more memory.