Skip to content

vlad-perevezentsev/sdc

 
 

Intel® Scalable Dataframe Compiler

Travis CI Azure Pipelines

Numba* Extension For Pandas* Operations Compilation

Intel® Scalable Dataframe Compiler (Intel® SDC) is an extension of Numba* that enables compilation of Pandas* operations. It automatically vectorizes and parallelizes the code by leveraging modern hardware instructions and by utilizing all available cores.

Intel® SDC documentation can be found here.

Note

For maximum performance and stability, please use numba from intel/label/beta channel.

Installing Binary Packages (conda and wheel)

Intel® SDC is available on the Anaconda Cloud intel/label/beta channel. Distribution includes Intel® SDC for Python 3.6 and Python 3.7 for Windows and Linux platforms.

Intel® SDC conda package can be installed using the steps below:

> conda create -n sdc-env python=<3.7 or 3.6> pyarrow=0.17.0 pandas=1.0.5 -c anaconda -c conda-forge
> conda activate sdc-env
> conda install sdc -c intel/label/beta -c intel -c defaults -c conda-forge --override-channels

Intel® SDC wheel package can be installed using the steps below:

> conda create -n sdc-env python=<3.7 or 3.6> pip pyarrow=0.17.0 pandas=1.0.5 -c anaconda -c conda-forge
> conda activate sdc-env
> pip install --index-url https://pypi.anaconda.org/intel/label/beta/simple --extra-index-url https://pypi.anaconda.org/intel/simple --extra-index-url https://pypi.org/simple sdc

Building Intel® SDC from Source on Linux

We use Anaconda distribution of Python for setting up Intel® SDC build environment.

If you do not have conda, we recommend using Miniconda3:

wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
chmod +x miniconda.sh
./miniconda.sh -b
export PATH=$HOME/miniconda3/bin:$PATH

Note

For maximum performance and stability, please use numba from intel/label/beta channel.

It is possible to build Intel® SDC via conda-build or setuptools. Follow one of the cases below to install Intel® SDC and its dependencies on Linux.

Building on Linux with conda-build

PYVER=<3.6 or 3.7>
NUMPYVER=<1.16 or 1.17>
conda create -n conda-build-env python=$PYVER conda-build
source activate conda-build-env
git clone https://github.com/IntelPython/sdc.git
cd sdc
conda build --python $PYVER --numpy $NUMPYVER --output-folder=<output_folder> -c intel/label/beta -c defaults -c intel -c conda-forge --override-channels conda-recipe

Building on Linux with setuptools

export PYVER=<3.6 or 3.7>
export NUMPYVER=<1.16 or 1.17>
conda create -n sdc-env -q -y -c intel/label/beta -c defaults -c intel -c conda-forge python=$PYVER numpy=$NUMPYVER tbb-devel tbb4py numba=0.49 pandas=1.0.5 pyarrow=0.17.0 gcc_linux-64 gxx_linux-64
source activate sdc-env
git clone https://github.com/IntelPython/sdc.git
cd sdc
python setup.py install

In case of issues, reinstalling in a new conda environment is recommended.

Building Intel® SDC from Source on Windows

Building Intel® SDC on Windows requires Build Tools for Visual Studio 2019 (with component MSVC v140 - VS 2015 C++ build tools (v14.00)):

It is possible to build Intel® SDC via conda-build or setuptools. Follow one of the cases below to install Intel® SDC and its dependencies on Windows.

Building on Windows with conda-build

set PYVER=<3.6 or 3.7>
set NUMPYVER=<1.16 or 1.17>
conda create -n conda-build-env -q -y python=%PYVER% conda-build conda-verify vc vs2015_runtime vs2015_win-64
conda activate conda-build-env
git clone https://github.com/IntelPython/sdc.git
cd sdc
conda build --python %PYVER% --numpy %NUMPYVER% --output-folder=<output_folder> -c intel/label/beta -c defaults -c intel -c conda-forge --override-channels conda-recipe

Building on Windows with setuptools

set PYVER=<3.6 or 3.7>
set NUMPYVER=<1.16 or 1.17>
conda create -n sdc-env -c intel/label/beta -c defaults -c intel -c conda-forge python=%PYVER% numpy=%NUMPYVER% tbb-devel tbb4py numba=0.49 pandas=1.0.5 pyarrow=0.17.0
conda activate sdc-env
set INCLUDE=%INCLUDE%;%CONDA_PREFIX%\Library\include
set LIB=%LIB%;%CONDA_PREFIX%\Library\lib
git clone https://github.com/IntelPython/sdc.git
cd sdc
python setup.py install

Troubleshooting Windows Build

  • If the cl compiler throws the error fatal error LNK1158: cannot run 'rc.exe', add Windows Kits to your PATH (e.g. C:\Program Files (x86)\Windows Kits\8.0\bin\x86).
  • Some errors can be mitigated by set DISTUTILS_USE_SDK=1.
  • For setting up Visual Studio, one might need go to registry at HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\VisualStudio\SxS\VS7, and add a string value named 14.0 whose data is C:\Program Files (x86)\Microsoft Visual Studio 14.0\.
  • Sometimes if the conda version or visual studio version being used are not latest then building Intel® SDC can throw some vague error about a keyword used in a file. So make sure you are using the latest versions.

Building documentation

Building Intel® SDC User's Guide documentation requires pre-installed Intel® SDC package along with compatible Pandas* version as well as Sphinx* 2.2.1 or later.

Intel® SDC documentation includes Intel® SDC examples output which is pasted to functions description in the API Reference.

Use pip to install Sphinx* and extensions:

pip install sphinx sphinxcontrib-programoutput

Currently the build precedure is based on make located at ./sdc/docs/ folder. While it is not generally required we recommended that you clean up the system from previous documentaiton build by running:

make clean

To build HTML documentation you will need to run:

make html

The built documentation will be located in the ./sdc/docs/build/html directory. To preview the documentation open index.html file.

More information about building and adding documentation can be found here.

Running unit tests

python sdc/tests/gen_test_data.py
python -m unittest

References

Intel® SDC follows ideas and initial code base of High-Performance Analytics Toolkit (HPAT). These academic papers describe ideas and methods behind HPAT:

About

Intel® Scalable Dataframe Compiler for Pandas*

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 89.9%
  • C++ 9.4%
  • C 0.5%
  • Batchfile 0.1%
  • Shell 0.1%
  • CMake 0.0%