Installation

The Numba/llvmlite stack consists of the following major components:

  • Numba is the compiler package, this depends on llvmlite.

  • llvmlite is a lightweight binding package to the LLVM APIs, it depends on LLVM.

  • LLVM is the JIT compiler framework for producing executable code from various inputs.

All components must be compiled in order to be used. And, since each component on the stack depends on the previous one, you need to compile LLVM in order to compile llvmlite in order to compile Numba. The LLVM package is a significant size and may take significant time (magnitude, roughly an hour) and skill to compile depending on the platform.

Pre-built binaries

As mentioned above, building LLVM for llvmlite is challenging. Installing a binary package that has been built and tested is strongly recommend.

Official Conda packages are available in the Anaconda distribution:

conda install llvmlite

Development releases are built from the Git main branch and uploaded to the Numba development channel on Anaconda Cloud:

conda install -c numba/label/dev llvmlite

Binary wheels are also available for installation from PyPI:

pip install llvmlite

Development releases of binary wheels are not made available.

Contrary to what might be expected, the llvmlite packages built by the Numba maintainers do not use any LLVM shared libraries that may be present on the system, and/or in the Conda environment. The parts of LLVM required by llvmlite are statically linked at build time. As a result, installing llvmlite from a binary package from the Numba channel does not also require the end user to install LLVM. (For more details on the reasoning behind this, see: Why Static Linking to LLVM?). Note however also that llvmlite packages compiled by other parties, e.g. conda-forge may split this into and llvmlite and llvm package and link dynamically.

Conda packages:

The Numba maintainers ship to the Numba channel:

  • Numba packages

  • llvmlite packages

  • llvmdev packages (this contains a build of LLVM)

The llvmdev packages are not needed at runtime by llvmlite packages as llvmlite’s dynamic libraries are statically linked (see above) at compile time against LLVM through the dependency on the llvmdev package.

The Anaconda distribution and conda-forge channels ship:

  • Numba packages

  • llvmlite packages

  • LLVM split into runtime libraries (package called llvm) and compile time libraries/headers etc this contains a build of LLVM (package called llvmdev)

At compile time the llvmdev and llvm packages are used to build llvmlite and llvmlite’s dynamic libraries are dynamically linked against the libraries in the llvm meta-package. This means at runtime llvmlite depends on the llvm package which has the LLVM shared libraries in it (it’s actually a package called libllvm that contains the DSOs, but the llvm package is referred to so as to get the run_exports).

Using pip

The Numba maintainers ship binary wheels:

  • Numba wheels (x86* architectures)

  • llvmlite wheels (x86* architectures)

Note that the llvmlite wheels are statically linked against LLVM, as per the conda packages on the Numba channel. This mitigates the need for a LLVM based binary wheel. Note also that this, as of version 0.36, does not include the aarch64 architectures, for example installation on a Raspberry Pi is not supported.

The Numba maintainers ship an sdist for:

  • Numba

  • llvmlite

Note that there is no sdist provided for LLVM. If you try and build llvmlite from sdist you will need to bootstrap the package with your own appropriate LLVM.

How this ends up being a problem.

  1. If you are on an unsupported architecture (i.e. not x86*) or unsupported Python version for binary wheels (e.g. Python alphas) then pip will try and build Numba from sdist which in turn will try and build llvmlite from sdist. This will inevitably fail as the llvmlite source distribution needs an appropriate LLVM installation to build.

  2. If you are using pip < 19.0 then manylinux2010 wheels will not install and you end up in the situation in 1. i.e. something unsupported so building from sdist.

Historically, this issues has manifested itself as the following error message, which included here verbatim for future reference:

FileNotFoundError: [Errno 2] No such file or directory: 'llvm-config'

Things to “fix” it…

  1. If you are using pip < 19.0 and on x86*, then update it if you can, this will let you use the manylinux2010 binary wheels.

  2. If you are on an unsupported architecture, for example Raspberry Pi, please use conda if you have that available.

  3. Otherwise: you will probably need to build from source, this means providing an LLVM. If you have conda available you could use this to bootstrap the installation with a working llvm/llvmdev package. Learn more about compiling from source in the section on Building manually below. and in particular note the use of the LLVM_CONFIG environment variable for specifying where your LLVM install is.

What to be aware of when using a system provided LLVM package.

When using a system provided LLVM package, there are a number of things that could go wrong:

  1. The LLVM package may not work with Numba/llvmlite at all.

  2. If it does work to some degree it is unlikely the carry the correct patches for Numba/llvmlite to work entirely correctly.

  3. Since the Numba/llvmlite maintainers may not know how the package was compiled it may be more difficult to get help when things do go wrong.

Building manually

Building llvmlite requires first building LLVM. Do not use prebuilt LLVM binaries from your OS distribution or the LLVM website! There will likely be a mismatch in version or build options, and LLVM will be missing certain patches that are critical for llvmlite operation.

Prerequisites

Before building, you must have the following:

  • On Windows:

    • Visual Studio 2015 (Update 3) or later, to compile LLVM and llvmlite. The free Express edition is acceptable.

    • CMake installed.

  • On Linux:

    • g++ (>= 4.8) and CMake

    • If building LLVM on Ubuntu, the linker may report an error if the development version of libedit is not installed. If you run into this problem, install libedit-dev.

  • On Mac:

    • Xcode for the compiler tools, and CMake

Compiling LLVM

If you can build llvmlite inside a conda environment, you can install a prebuilt LLVM binary package and skip this step:

conda install -c numba llvmdev

The LLVM build process is fully scripted by conda-build, and the llvmdev recipe is the canonical reference for building LLVM for llvmlite. Please use it if at all possible!

The manual instructions below describe the main steps, but refer to the recipe for details:

  1. Download the LLVM source code. You can download the complete “project” package, or llvm, ldd, and libunwind.

  2. Download or git checkout the llvmlite source code.

  3. Decompress the LLVM tar files and apply the appropriate patches from the llvmlite/conda-recipes/ directory. You can apply each patch using the Linux patch -p1 -i {patch-file} command. Patches are prefixed with the LLVM version they apply cleanly to.

  4. For Linux/macOS:

    1. export PREFIX=desired_install_location CPU_COUNT=N ( N is number of parallel compile tasks)

    2. Run the build.sh script in the llvmdev conda recipe from the LLVM source directory.

  5. For Windows:

    1. set PREFIX=desired_install_location

    2. Run the bld.bat script in the llvmdev conda recipe from the LLVM source directory.

Compiling llvmlite

  1. To build the llvmlite C wrapper, which embeds a statically linked copy of the required subset of LLVM, run the following from the llvmlite source directory:

    python setup.py build
    
  2. If your LLVM is installed in a nonstandard location, set the LLVM_CONFIG environment variable to the location of the corresponding llvm-config or llvm-config.exe executable. This variable must persist into the installation of llvmlite—for example, into a Python environment.

    EXAMPLE: If LLVM is installed in /opt/llvm/ with the llvm-config binary located at /opt/llvm/bin/llvm-config, set LLVM_CONFIG=/opt/llvm/bin/llvm-config.

  3. If you wish to build against an unsupported LLVM version, set the environment variable LLVMLITE_SKIP_LLVM_VERSION_CHECK to non-zero. Note that this is useful for e.g. testing new versions of llvmlite, but support for llvmlite built in this manner is limited/it’s entirely possible that llvmlite will not work as expected. See also: why llvmlite doesn’t always support the latest release(s) of LLVM.

Installing

  1. To validate your build, run the test suite by running:

    python runtests.py
    

    or:

    python -m llvmlite.tests
    
  2. If the validation is successful, install by running:

    python setup.py install
    

Installing from sdist

If you don’t want to do any modifications to llvmlite itself, it’s also possible to use pip to compile and install llvmlite from the latest released sdist package. You’ll still need to point to your llvm-config if it’s not in the PATH:

LLVM_CONFIG=/path/to/llvm-config pip3 install llvmlite

This should work on any platform that runs Python and llvm. It has been observed to work on arm, ppc64le, and also pypy3 on arm.

x86 users will need to pass an extra flag (see issue #522):

LLVM_CONFIG=/path/to/llvm-config CXXFLAGS=-fPIC pip3 install llvmlite

This is known to work with pypy3 on Linux x64.

It’s also possible to force pip to rebuild llvmlite locally with a custom version of llvm :

LLVM_CONFIG=/path/to/custom/llvm-config CXXFLAGS=-fPIC pip3 install --no-binary :all: llvmlite