Manual LLVM build

Build dependencies

Depending on which OS you use the command to install the dependencies needed to build LLVM will differ. Skip over to the command for your Operating System (OS) in this section (OS names are bolded):

Linux Debian and Ubuntu users may install all required tools this way:

sudo apt-get install build-essential git cmake ninja-build

Linux Fedora users may install all required tools with:

sudo dnf groupinstall "Development Tools"
sudo dnf install cmake ninja-build

MacOS users may install these tools using Homebrew:

brew install cmake ninja

Windows users may install build dependencies using Chocolatey. Install Chocolatey first, then run the following in a command prompt or PowerShell with administrative privileges:

choco install --confirm git golang mingw make cmake ninja python

Windows users should also use Git Bash (installed above) to run all the build commands like make. The TinyGo build system expects a Unix-like environment that is not normally provided by Windows but is included already in Git Bash.

Choco doesn’t seem to add CMake automatically to the $PATH variable. You can do this manually if needed, in Git bash:

export PATH="$PATH:/c/Program Files/CMake/bin"

Building LLVM

The following instructions are common to all operating systems.

Background: LLVM is a library and collection of tools for building compilers. It is the base of many newer compilers, like Clang, Swift, and the Rust compiler so they can all build on the same high-quality base. Similarly, TinyGo uses LLVM for all its low-level optimization and machine code generation so that it can produce code that’s roughly as small and efficient as all these other modern compilers.

We build LLVM from inside the TinyGo repository we cloned in the previous step. The following command when run inside the TinyGo repo will take care of first downloading the LLVM source code to later build it in the next step. It places the source code in llvm-project/ and the build output in llvm-build/. It only needs to be done once until the next LLVM release (every half year).

Note that the build step may take some time- feel free to grab a drink meanwhile. Warnings emitted through the compilation in this part are normal as of LLVM 17.

make llvm-source llvm-build

When building on Windows, add CCACHE=OFF.

make llvm-source llvm-build CCACHE=OFF

Building TinyGo

Once this is finished, you can build TinyGo against this manually built LLVM:

make

This results in a tinygo binary in the build directory:

$ ./build/tinygo version
tinygo version 0.32.0-dev-d4189fec linux/amd64 (using go version go1.22 and LLVM version 18.1.2)

You have successfully built TinyGo from source. Congratulations! What’s left now is to complete the additional requirements

Adding tinygo to your path

Linux users may choose to run TinyGo from any directory you may want to move the built binary to a location on your path or add the ./build directory to your path. The following shell command moves the TinyGo binary to /user/bin so that any user can run TinyGo. You may need root privileges to complete this step.

mv ./build/tinygo /usr/bin/

If you ran the above command, uninstalling TinyGo is as easy as running rm /usr/bin/tinygo.

If you’d prefer a user-based install you can move TinyGo to your $HOME/go/bin directory, where the Go compiler installs binaries by default. The directory should be added to your PATH if it hasn’t already been added for this to work.