https://github.com/sdiehl/numpile

A tiny 1000 line LLVM-based numeric specializer for scientific Python code.
https://github.com/sdiehl/numpile

compiler ipython-notebook jit llvm-tutorial numba python specializer tutorial

Last synced: 25 days ago
JSON representation

A tiny 1000 line LLVM-based numeric specializer for scientific Python code.

• Host: GitHub
• URL: https://github.com/sdiehl/numpile
• Owner: sdiehl
• License: mit
• Created: 2015-01-22T13:44:56.000Z (almost 10 years ago)
• Default Branch: master
• Last Pushed: 2022-01-02T21:55:16.000Z (almost 3 years ago)
• Last Synced: 2024-09-28T23:23:31.111Z (about 1 month ago)
• Topics: compiler, ipython-notebook, jit, llvm-tutorial, numba, python, specializer, tutorial
• Language: Jupyter Notebook
• Homepage: http://dev.stephendiehl.com/numpile/
• Size: 64.5 KB
• Stars: 403
• Watchers: 16
• Forks: 52
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
Numpile

-------

A tiny 1000 line LLVM-based numeric specializer for scientific Python code.

You really shouldn't use this for anything serious, it's just to demonstrate how

you might build one of these things from scratch. There's a lot of untapped

potential and low hanging fruit around *selective embedded JIT specialization*

for array expression languages in the SciPython space.

Installing

----------

Numpile requires ``numpy`` and ``llvmlite`` (the latter includes needed

LLVM libraries). You can either try to install them using your OS package

manager, or alternatively, using ``pip``:

```bash

$ pip install llvmlite

$ pip install numpy

```

Usage

-----

```python

import numpy as np

from numpile import autojit

@autojit

def dot(a, b):

    c = 0

    n = a.shape[0]

    for i in range(n):

        c += a[i] * b[i]

    return c

a = np.arange(100, 200, dtype='int32')

b = np.arange(300, 400, dtype='int32')

result = dot(a, b) 

print(result)

```

License

-------

Released under the MIT License.

Copyright (c) 2015, Stephen Diehl