fastpy

LLVM JIT compiler as a function decorator


Keywords
fastpy
License
MIT
Install
pip install fastpy==0.1.1

Documentation

fastpy

Documentation Status Updates

Python made fast. Decorate your functions with @fast, we will infered the types you used, compile to machine code, and execute.

Biased test showing how fast fastpy is:

Initial code:

def long_loop(a):
  for i in range(100000):
    for j in range(10000):
      a += 1
return a
print long_loop(0)
$ time python loop.py
1000000000
python test.py  39.24s user 0.01s system 99% cpu 39.420 total
$ time pypy loop.py
1000000000
pypy test.py  0.92s user 0.01s system 99% cpu 0.937 total

Now we modify the code to use fastpy

from fastpy import fast

@fast
def long_loop(a):
  for i in range(100000):
    for j in range(10000):
      a += 1
  return a
print long_loop(0)
$  time python loop.py
1000000000
python test.py  0.11s user 0.00s system 99% cpu 0.117 total

Credits

Based on this tutorial http://dev.stephendiehl.com/numpile/