Website issues / discussion

Install OpenCilk

users guideinstall

Requirements

OpenCilk 1.1 runs on Intel x86 64-bit processors (Haswell and newer), AMD x86 64-bit processors (Excavator and newer), and Apple M1 and other 64-bit ARM processors. It has been tested on the following operating systems:

  • Ubuntu (20.04), including via WSL2 (Windows Subsystem for Linux 2)
  • FreeBSD (13)
  • Fedora (32)
  • macOS (10.15, 11.2)

Download

Use the links below to download the appropriate OpenCilk precompiled shell archive (.sh), tarball (.tar.gz), or Docker image.

You can also build OpenCilk from source, which is the recommended approach for Ubuntu 18.04 and other operating systems not listed above.

Linux

Note for non-macOS Linux systems: You need to install GCC (or equivalent) to provide necessary system include files and system libraries.

macOS

Note for macOS users: Unless you are using the OpenCilk Docker image, you must also install XCode or the XCode Command Line Tools, which provide standard system libraries and header files needed by the OpenCilk compiler.

Docker

Install

Shell archive

Execute the shell script to extract OpenCilk 1.1 into the current directory.

$ sh OpenCilk-1.1-LLVM-12.0.0-Ubuntu-20.04-x86_64.sh

You will be prompted to accept the OpenCilk license and whether or not to append a version-specific subdirectory to the installation prefix (by default the current directory).

Optionally, you can select a different installation directory by passing the option --prefix=/path/to/install/dir to the script. To see all options, pass the --help argument to the script.

The OpenCilk C (or C++) compiler can be invoked via bin/clang (or bin/clang++) from within the installation directory.

Example

The following shows the process of installing OpenCilk into a version-specific subdirectory within /opt/opencilk/.

$ sh OpenCilk-1.1-LLVM-12.0.0-Ubuntu-20.04-x86_64.sh --prefix=/opt/opencilk
OpenCilk Installer Version: 12.0.0, Copyright (c) OpenCilk
This is a self-extracting archive.
The archive will be extracted to: /opt/opencilk

If you want to stop extracting, please press <ctrl-C>.

[...license text omitted...]

Do you accept the license? [yn]:
y

By default the OpenCilk will be installed in:
  "/opt/opencilk/OpenCilk-12.0.0-Linux"
Do you want to include the subdirectory OpenCilk-12.0.0-Linux?
Saying no will install in: "/opt/opencilk" [Yn]:
y

Using target directory: /opt/opencilk/OpenCilk-12.0.0-Linux
Extracting, please wait...

Unpacking finished successfully

The OpenCilk C compiler can now be run as /opt/opencilk/OpenCilk-12.0.0-Linux/bin/clang. Or you can add /opt/opencilk/OpenCilk-12.0.0-Linux/bin to your PATH environment variable and invoke the OpenCilk C/C++ compiler simply as clang or clang++.

Tarball

As an alternative to using the self-extracting shell archive above, you can extract OpenCilk 1.1 from the .tar.gz tarball. For example,

$ tar xvzf OpenCilk-1.1-LLVM-12.0.0-Ubuntu-20.04-x86_64.tar.gz

will extract the OpenCilk installation into a subdirectory OpenCilk-12.0.0-Linux/ within the current working directory.

Extracting the tarball as above is equivalent to running the corresponding shell script with options --skip-license --include-subdir.

Docker image

OpenCilk 1.1 is also available as a Docker image based on Ubuntu 20.04. The OpenCilk C and C++ compilers are available as clang and clang++ in the image. To use the OpenCilk 1.1 Docker image, download the docker-opencilk-v1.1.tar.gz file, load the image, and run a container. For example:

$ docker load -i docker-opencilk-v1.1.tar.gz
$ docker run -it opencilk:v1.1 /bin/bash

Test

The OpenCilk 1.1 installation includes LLVM 12 with the following OpenCilk components:

  • OpenCilk compiler (with the clang and clang++ front-ends)
  • OpenCilk runtime library
  • Cilksan race detector
  • Cilkscale scalability analyzer and visualization script

See Getting started for steps to verify that your installation is working.

Additional resources

External OpenCilk-powered libraries

The following third-party libraries are known to work with OpenCilk out of the box for parallel execution.

  • SG-t-SNE-Π: Low-dimensional embedding of sparse stochastic graphs.
  • FGLT: Fast graphlet transform.
  • RecFMM: Adaptive fast multipole method.

Miscellaneous developer tools