Stella has a relatively narrow focus: Python programs with a medium size core that needs to be executed very efficiently. Simulations are a prime example: In particular stochastic simulations need to be sampled many times to get relevant results which are statistically relevant, but even regular simulations often are extremely slow when run in the regular Python interpreter.
The key features of Stella are:
- compiled with LLVM for execution times similar to C programs
- statically typed; ensured correctness for the computationally expensive parts
- no interaction with the Python run-time to avoid any unintended performance penalties
- completely valid Python code for verification and debugging purposes
- easy integration with C libraries with the use of Cython
- excellent functional test coverage
- compatible with numpy arrays
Note: Stella is currently in an alpha stage: it is a research project, albeit a well tested one. But there is no guarantee that the program you want to run will be completely supported. Please do give it a try, and report back!
Stella requires numpy and llvmlite (which in turn requires LLVM to be installed). I recommend setting up a virtualenv.
git clone https://github.com/squisher/stella.git/
cd stella
# grab branch of llvmlite known to work with stella
git clone https://github.com/squisher/llvmlite.git/
cd llvmlite
git checkout stella
# At least Debian does not create a llvm-config symlink, so explicitly tell
# llvmlite which binary to use. Debian jessie has `llvm-3.5', while wheezy
# only ships with llvm-3.0, which is rather old. You may want to upgrade to
# jessie or install LLVM by hand.
export LLVM_CONFIG=llvm-config-3.5
pip install .
cd ..
# install stella in editable mode
pip install -e .If you intend to run the unit tests, skip the last line and instead run:
cd mtpy
# make sure cython is installed
python setup.py install
cd ..
pip install -e '.[test]'You can execute any Python function in Stella, given that it only uses supported features. Stella will automatically consider the call graph, and ensure that until the entry method returns, everything is executed in Stella. Otherwise an exception will be raised.
def some_math(x, y):
return sqrt(x**2 + y**2 - 42)
import stella
result = stella.wrap(some_math)(13, 7)Or a slightly more complex example:
# Just an example, for illustration only
def compute(x):
return (x+2)/5
def stuff(a, b):
for i in range(len(a)):
b[i] = compute(a[i])
import numpy, stella
a = numpy.array(range(9))
b = numpy.zeros(a.shape, dtype=a.dtype)
stuff(a, b)
print (b)Stella expects all Python function names to be identical to their C version, and the module must implement a function getCSignatures() which returns a dictionary mapping the function name to a ctypes-style signature.
The mtpy library is small C library and demonstrates how to support Stella. It is an implementation of the Mersenne Twist random number generator by Geoff Kuenning. It is licensed under the LGPL. See the included README for more details.
Note that mtpy is used by some of the tests, and serves as an example, but it is NOT part of Stella.
Testing is done with the absolutely great py.test library. If you also want to run the benchmarks, then pystache is additionally required. Some tests and benchmarks use the aforementioned mtpy library.
import stella
stella.run_tests()Please file an issue at github or send me an email at dmohr@cs.unm.edu
Thanks for trying Stella!
Copyright 2013-2015 David Mohr
Unless otherwise noted files are under the following license:
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.