llgo

llgo - A Go compiler based on LLVM

LLGo is a Go compiler based on LLVM in order to better integrate Go with the C ecosystem including Python. It's a subproject of the Go+ project.

C standard libary support

package main

import "github.com/goplus/llgo/c"

func main() {
	c.Printf(c.Str("Hello world\n"))
}

This is a simple example of calling the C printf function to print Hello world. Here, c.Str is not a function for converting a Go string to a C string, but a built-in instruction supported by llgo for generating a C string constant.

The _demo directory contains some C standard libary related demos (it start with _ to prevent the go command from compiling it):

• hello: call C printf to print Hello world
• concat: call C fprintf with stderr
• qsort: call C function with a callback (eg. qsort)

To run these demos (If you haven't installed llgo yet, please refer to How to install):

cd <demo-directory>  # eg. cd _demo/hello
llgo run .

See github.com/goplus/llgo/c for more detials.

Python support

You can import a Python library in LLGo!

And you can import any Python library into llgo through a program called llpyg (see Development tools). The following libraries have been included in llgo:

• builtins
• sys
• os
• math
• json
• inspect
• statistics
• numpy
• pandas
• pytorch
• matplotlib

Note: For third-party libraries (such as pandas and pytorch), you still need to install the library files.

Here is an example using the Python math library:

package main

import (
	"github.com/goplus/llgo/c"
	"github.com/goplus/llgo/py"
	"github.com/goplus/llgo/py/math"
)

func main() {
	x := math.Sqrt(py.Float(2))
	c.Printf(c.Str("sqrt(2) = %f\n"), x.Float64())
}

Here, We call py.Float(2) to create a Python number 2, and pass it to Python’s math.sqrt to get x. Then use x.Float64() to convert x to Go's float64 type, and print the value through the C printf function.

Let's look at a slightly more complex example. For example, we use numpy to calculate:

package main

import (
	"github.com/goplus/llgo/c"
	"github.com/goplus/llgo/py"
	"github.com/goplus/llgo/py/numpy"
)

func main() {
	a := py.List(
		py.List(1.0, 2.0, 3.0),
		py.List(4.0, 5.0, 6.0),
		py.List(7.0, 8.0, 9.0),
	)
	b := py.List(
		py.List(9.0, 8.0, 7.0),
		py.List(6.0, 5.0, 4.0),
		py.List(3.0, 2.0, 1.0),
	)
	x := numpy.Add(a, b)
	c.Printf(c.Str("a+b = %s\n"), x.Str().CStr())
}

Here we define two 3x3 matrices a and b, add them to get x, and then print the result.

The _pydemo directory contains some python related demos:

• callpy: call Python standard library function math.sqrt
• pi: print python constants math.pi
• statistics: define a python list and call statistics.mean to get the mean
• matrix: a basic numpy demo

To run these demos, you need to set the LLGO_LIB_PYTHON environment variable first.

If Python is in the search path for clang linking, then LLGO_LIB_PYTHON only needs to be set to the name of the Python library. For example:

export LLGO_LIB_PYTHON=python3.12

You can also specify the path to tell llgo where the Python library is located:

export LLGO_LIB_PYTHON=/foo/bar/python3.12

For example, /opt/homebrew/Frameworks/Python.framework/Versions/3.12/lib/libpython3.12.dylib is a typical python library location under macOS. So we should set it like this:

export LLGO_LIB_PYTHON=/opt/homebrew/Frameworks/Python.framework/Versions/3.12/lib/python3.12

Note that the file name must be written in a platform-independent format, using python3.12 instead of libpython3.12.dylib.

Then you can run the demos:

cd <demo-directory>  # eg. cd _pydemo/callpy
llgo run .

See github.com/goplus/llgo/py for more detials.

Other frequently used libraries

LLGo can easily import any libraries from the C ecosystem. Currently, this import process is still manual, but in the future, it will be automated similar to Python library imports.

The currently supported libraries include:

• llama2.c
• cjson
• sqlite

Here are some examples related to them:

• llama2-c: inference Llama 2 (It's the first llgo AI example)
• mkjson: create a json object and print it
• sqlitedemo: a basic sqlite demo

Go syntax support

Most of the Go syntax is already supported. Except for the following, which needs to be improved:

• map (Very limited support)

Here are some examples related to Go syntax:

• concat: define a variadic function
• genints: various forms of closure usage (including C function, recv.method and anonymous function)
• errors: demo to implement error interface
• defer: defer demo
• goroutine: goroutine demo

Defer

LLGo defer does not support usage in loops. This is not a bug but a feature, because we think that using defer in a loop is a very unrecommended practice.

Garbage Collection (GC)

By default, LLGo implements gc based on bdwgc (also known as libgc).

However, you can disable gc by specifying the nogc tag. For example:

llgo run -tags nogc .

Go packages support

Here are the Go packages that can be imported correctly:

• unsafe
• unicode
• unicode/utf8
• unicode/utf16
• math
• math/bits
• math/cmplx
• sort
• strconv
• sync/atomic
• sync (partially)
• syscall (partially)
• errors (partially)
• io (partially)
• io/fs (partially)
• os (partially)
• fmt (partially)
• reflect (partially)
• time (partially)

Dependencies

• Go 1.20+ (build only)
• LLVM 17
• LLD 17
• Clang 17
• pkg-config 0.29+
• bdwgc/libgc 8.0+
• cJSON 1.7+ (optional, for github.com/goplus/llgo/c/cjson)
• SQLite 3 (optional, for github.com/goplus/llgo/c/sqlite)
• Python 3.11+ (optional, for github.com/goplus/llgo/py)

How to install

Follow these steps to generate the llgo command (its usage is the same as the go command):

on macOS

brew update # execute if needed
brew install llvm@17 pkg-config libgc
brew install cjson sqlite python@3.12 # optional
export PATH=$(brew --prefix llvm@17)/bin:$PATH # you may want to add this to your shell RC file, e.g. ~/.zshrc
export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
git clone https://github.com/goplus/llgo.git
cd llgo
export LLGOROOT="/path/to/llgo" # Replace this with the root directory of the llgo project
go install -v ./...

on Linux (Debian/Ubuntu)

echo "deb http://apt.llvm.org/$(lsb_release -cs)/ llvm-toolchain-$(lsb_release -cs)-17 main" | sudo tee /etc/apt/sources.list.d/llvm.list
wget -O - https://apt.llvm.org/llvm-snapshot.gpg.key | sudo apt-key add -
sudo apt-get update # execute if needed
sudo apt-get install -y llvm-17-dev clang-17 lld-17 pkg-config libgc-dev
sudo apt-get install -y libcjson-dev libsqlite3-dev python3.12-dev # optional
export PATH=/usr/lib/llvm-17/bin:$PATH # you may want to add this to your shell RC file, e.g. ~/.bashrc
export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
git clone https://github.com/goplus/llgo.git
cd llgo
export LLGOROOT="/path/to/llgo" # Replace this with the root directory of the llgo project
go install -v ./...

on Windows

TODO

Development tools

• pydump: It's the first program compiled by llgo (NOT go) in a production environment. It outputs symbol information (functions, variables, and constants) from a Python library in JSON format, preparing for the generation of corresponding packages in llgo.
• pysigfetch: It generates symbol information by extracting information from Python's documentation site. This tool is not part of the llgo project, but we depend on it.
• llpyg: It is used to automatically convert Python libraries into Go packages that llgo can import. It depends on pydump and pysigfetch to accomplish the task.
• llgen: It is used to compile Go packages into LLVM IR files (*.ll).
• ssadump: It is a Go SSA builder and interpreter.

How do I generate these tools?

export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
go install -v ./...  # compile all tools except pydump
cd chore/_xtool
llgo install ./...   # compile pydump
go install github.com/goplus/hdq/chore/pysigfetch@v0.8.1  # compile pysigfetch

Key modules

Below are the key modules for understanding the implementation principles of llgo:

• llgo/ssa: It generates LLVM IR files (LLVM SSA) using the semantics (interfaces) of Go SSA. Although LLVM SSA and Go SSA are both IR languages, they work at completely different levels. LLVM SSA is closer to machine code, which abstracts different instruction sets. While Go SSA is closer to a high-level language. We can think of it as the instruction set of the Go computer. llgo/ssa is not just limited to the llgo compiler. If we view it as the high-level expressive power of LLVM, you'll find it very useful. Prior to llgo/ssa, you had to operate LLVM using machine code semantics. But now, with the advanced SSA form (in the semantics of Go SSA), you can conveniently utilize LLVM.
• llgo/cl: It is the core of the llgo compiler. It converts a Go package into LLVM IR files. It depends on llgo/ssa.
• llgo/internal/build: It strings together the entire compilation process of llgo. It depends on llgo/ssa and llgo/cl.