Numba is compatible with Python 3.6 or later, and Numpy versions 1.15 or later.
Our supported platforms are:
- Linux x86 (32-bit and 64-bit)
- Linux ppcle64 (POWER8)
- Windows 7 and later (32-bit and 64-bit)
- OS X 10.9 and later (64-bit)
- NVIDIA GPUs of compute capability 2.0 and later
- AMD ROC dGPUs (linux only and not for AMD Carrizo or Kaveri APU)
- ARMv7 (32-bit little-endian, such as Raspberry Pi 2 and 3)
- ARMv8 (64-bit little-endian, such as the NVIDIA Jetson)
Automatic parallelization with @jit is only available on 64-bit platforms.
Installing using conda on x86/x86_64/POWER Platforms¶
The easiest way to install Numba and get updates is by using
a cross-platform package manager and software distribution maintained
by Anaconda, Inc. You can either use Anaconda to get the full stack in one download,
or Miniconda which will install
the minimum packages required for a conda environment.
Once you have conda installed, just type:
$ conda install numba
$ conda update numba
Note that Numba, like Anaconda, only supports PPC in 64-bit little-endian mode.
To enable CUDA GPU support for Numba, install the latest graphics drivers from
NVIDIA for your platform.
(Note that the open source Nouveau drivers shipped by default with many Linux
distributions do not support CUDA.) Then install the
$ conda install cudatoolkit
You do not need to install the CUDA SDK from NVIDIA.
Installing using pip on x86/x86_64 Platforms¶
Binary wheels for Windows, Mac, and Linux are also available from PyPI. You can install Numba using
$ pip install numba
This will download all of the needed dependencies as well. You do not need to have LLVM installed to use Numba (in fact, Numba will ignore all LLVM versions installed on the system) as the required components are bundled into the llvmlite wheel.
To use CUDA with Numba installed by pip, you need to install the CUDA SDK from NVIDIA. Please refer to Setting CUDA Installation Path for details. Numba can also detect CUDA libraries installed system-wide on Linux.
Enabling AMD ROCm GPU Support¶
The ROCm Platform allows GPU computing with AMD GPUs on Linux. To enable ROCm support in Numba, conda is required, so begin with an Anaconda or Miniconda installation with Numba 0.40 or later installed. Then:
Follow the ROCm installation instructions.
roctoolsconda package from the
$ conda install -c numba roctools
See the roc-examples repository for sample notebooks.
Installing on Linux ARMv7 Platforms¶
Berryconda is a
conda-based Python distribution for the Raspberry Pi. We are now uploading
packages to the
numba channel on Anaconda Cloud for 32-bit little-endian,
ARMv7-based boards, which currently includes the Raspberry Pi 2 and 3,
but not the Pi 1 or Zero. These can be installed using conda from the
$ conda install -c numba numba
Berryconda and Numba may work on other Linux-based ARMv7 systems, but this has not been tested.
Installing on Linux ARMv8 (AArch64) Platforms¶
We build and test conda packages on the NVIDIA Jetson TX2, but they are likely to work for other AArch64 platforms. (Note that while the Raspberry Pi CPU is 64-bit, Raspbian runs it in 32-bit mode, so look at Installing on Linux ARMv7 Platforms instead.)
Conda-forge support for AArch64 is still quite experimental and packages are limited, but it does work enough for Numba to build and pass tests. To set up the environment:
Install conda4aarch64. This will create a minimal conda environment.
conda-forgechannels to your conda configuration:
$ conda config --add channels c4aarch64 $ conda config --add channels conda-forge
Then you can install Numba from the
$ conda install -c numba numba
On CUDA-enabled systems, like the Jetson, the CUDA toolkit should be automatically detected in the environment.
Installing from source¶
Installing Numba from source is fairly straightforward (similar to other Python packages), but installing llvmlite can be quite challenging due to the need for a special LLVM build. If you are building from source for the purposes of Numba development, see Build environment for details on how to create a Numba development environment with conda.
$ git clone git://github.com/numba/numba.git
Source archives of the latest release can also be found on
PyPI. In addition to
llvmlite, you will also need:
- A C compiler compatible with your Python installation. If you are using
Anaconda, you can use the following conda packages:
ARM: no conda packages, use the system compiler
- Mac OSX:
clangxx_osx-64or the system compiler at
- Windows: a version of Visual Studio appropriate for the Python version in use
Then you can build and install Numba from the top level of the source tree:
$ python setup.py install
Build time environment variables and configuration of optional components¶
Below are environment variables that are applicable to altering how Numba would otherwise build by default along with information on configuration options.
NUMBA_DISABLE_OPENMP (default: not set)¶
To disable compilation of the OpenMP threading backend set this environment variable to a non-empty string when building. If not set (default):
- For Linux and Windows it is necessary to provide OpenMP C headers and runtime libraries compatible with the compiler tool chain mentioned above, and for these to be accessible to the compiler via standard flags.
- For OSX the conda packages
intel-openmpprovide suitable C headers and libraries. If the compilation requirements are not met the OpenMP threading backend will not be compiled
NUMBA_DISABLE_TBB (default: not set)¶
To disable the compilation of the TBB threading backend set this environment variable to a non-empty string when building. If not set (default) the TBB C headers and libraries must be available at compile time. If building with
conda buildthis requirement can be met by installing the
tbb-develpackage. If not building with
conda buildthe requirement can be met via a system installation of TBB or through the use of the
TBBROOTenvironment variable to provide the location of the TBB installation. For more information about setting
TBBROOTsee the Intel documentation.
Numba has numerous required and optional dependencies which additionally may vary with target operating system and hardware. The following lists them all (as of July 2020).
Required build time:
- Compiler toolchain mentioned above
Required run time:
Optional build time:
See Build time environment variables and configuration of optional components for more details about additional options for the configuration and specification of these optional components.
llvm-openmp(OSX) - provides headers for compiling OpenMP support into Numba’s threading backend
intel-openmp(OSX) - provides OpenMP library support for Numba’s threading backend.
tbb-devel- provides TBB headers/libraries for compiling TBB support into Numba’s threading backend
Optional runtime are:
scipy- provides cython bindings used in Numba’s
tbb- provides the TBB runtime libraries used by Numba’s TBB threading backend
jinja2- for “pretty” type annotation output (HTML) via the
cffi- permits use of CFFI bindings in Numba compiled functions
intel-openmp- (OSX) provides OpenMP library support for Numba’s OpenMP threading backend
ipython- if in use, caching will use IPython’s cache directories/caching still works
pyyaml- permits the use of a
.numba_config.yamlfile for storing per project configuration options
colorama- makes error message highlighting work
icc_rt- (numba channel) allows Numba to use Intel SVML for extra performance
pygments- for “pretty” type annotation
gdbas an executable on the
$PATH- if you would like to use the gdb support
- Compiler toolchain mentioned above, if you would like to use
pyccfor Ahead-of-Time (AOT) compilation
r2pipe- required for assembly CFG inspection.
radare2as an executable on the
$PATH- required for assembly CFG inspection. See here for information on obtaining and installing.
graphviz- for some CFG inspection functionality.
pickle5- provides Python 3.8 pickling features for faster pickling in Python 3.6 and 3.7.
To build the documentation:
makeas an executable on the
Checking your installation¶
You should be able to import Numba from the Python prompt:
$ python Python 3.8.1 (default, Jan 8 2020, 16:15:59) [Clang 4.0.1 (tags/RELEASE_401/final)] :: Anaconda, Inc. on darwin Type "help", "copyright", "credits" or "license" for more information. >>> import numba >>> numba.__version__ '0.48.0'
You can also try executing the
numba --sysinfo (or
numba -s for short)
command to report information about your system capabilities. See Command line interface for
$ numba -s System info: -------------------------------------------------------------------------------- __Time Stamp__ 2018-08-28 15:46:24.631054 __Hardware Information__ Machine : x86_64 CPU Name : haswell CPU Features : aes avx avx2 bmi bmi2 cmov cx16 f16c fma fsgsbase lzcnt mmx movbe pclmul popcnt rdrnd sse sse2 sse3 sse4.1 sse4.2 ssse3 xsave xsaveopt __OS Information__ Platform : Darwin-17.6.0-x86_64-i386-64bit Release : 17.6.0 System Name : Darwin Version : Darwin Kernel Version 17.6.0: Tue May 8 15:22:16 PDT 2018; root:xnu-4570.61.1~1/RELEASE_X86_64 OS specific info : 10.13.5 x86_64 __Python Information__ Python Compiler : GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final) Python Implementation : CPython Python Version : 2.7.15 Python Locale : en_US UTF-8 __LLVM information__ LLVM version : 6.0.0 __CUDA Information__ Found 1 CUDA devices id 0 GeForce GT 750M [SUPPORTED] compute capability: 3.0 pci device id: 0 pci bus id: 1
(output truncated due to length)