A low-level compiled alternative to C, C++, and more!

Why? • Goals • Performance • Static Analysis & Ownership • Status • Getting Started • Join Us

Luma is a modern systems programming language focused on explicit control, fast compilation, and static verification.

It is designed for developers who want C-level performance and transparency, but with stronger compile-time tooling to catch common memory errors early — without a borrow checker, lifetimes, or runtime overhead.

Luma uses manual memory management with static analysis. Memory allocation and deallocation are always explicit, while the compiler performs ownership-aware checks during type checking to detect common classes of bugs before code generation.

Luma is intentionally memory-unsafe by design, but provides static checks for specific error patterns such as use-after-free, double-free, and forgotten deallocations.

Modern systems programming often involves a trade-off between performance, safety, and developer experience.
Luma aims to bridge that gap by providing:

• Manual memory control with compile-time static analysis
  — The type checker validates use-after-free, double-free, and unfreed allocations before codegen.
• Direct hardware access and predictable performance
• Readable, minimal syntax that doesn't hide control flow or introduce lifetimes
• Zero runtime overhead — all verification is done statically
• Fast, transparent tooling that stays close to the metal

Unlike Rust, Luma doesn't use lifetimes or a borrow checker. Instead, developers can annotate functions with lightweight ownership hints like #returns_ownership and #takes_ownership so the analyzer can reason about ownership transfers — for example, when returning an allocated pointer.

The result: C-level control with targeted compile-time checks, and no runtime or hidden semantics.

• Minimal & Explicit Syntax – No hidden control flow or implicit behavior
• Lightning-Fast Compilation – Sub-100ms builds for rapid iteration
• Zero-Cost Abstractions – No runtime overhead for safety or ergonomics
• Tiny Binaries – Comparable to C in size and efficiency
• Manual Memory Control – You decide when to alloc() and free()
• Static Verification – The type checker validates common memory errors (use-after-free, double-free, leaks) before codegen
• Optional Ownership Annotations – Use #returns_ownership and #takes_ownership to make ownership transfer explicit

Luma performs ownership-aware static analysis at the end of type checking to detect common memory management errors — without introducing runtime overhead.

• Use-after-free
• Double free
• Memory allocated but never freed
• Ownership transfer across function boundaries (when annotated)

Luma does not claim full memory safety. The following are currently not prevented:

• Out-of-bounds memory access
• Uninitialized memory reads
• Pointer arithmetic misuse
• Aliasing violations
• Invalid pointer dereference

These are deliberate design trade-offs to preserve simplicity, performance, and low-level control.

Current Phase: Early Development

Luma is currently in active development. Core language features are being implemented and the compiler architecture is being established.

What Works:

• Complete lexer and parser
• Full type system with structs, enums, functions
• Static memory analysis with ownership tracking
• LLVM backend for native code generation
• Standard library (math, memory, strings, terminal effects)
• Real-world applications (3D graphics, memory management)

Check out the todo to see what is being worked on or that is done.

You'll need the following tools installed:

• Make - Build automation
• GCC - GNU Compiler Collection
• LLVM - Compiler infrastructure (Version 20.0+ required)
• Valgrind (optional) - Memory debugging

Important: Luma requires LLVM 20.0 or higher due to critical bug fixes in the constant generation system.

Known Issues:

• LLVM 19.1.x: Contains a regression that causes crashes during code generation (illegal hardware instruction errors)
• LLVM 18.x and older: Not tested, may have compatibility issues

If you encounter crashes during the "LLVM IR" compilation stage (typically at 60% progress), this is likely due to an incompatible LLVM version.

llvm-config --version

Arch Linux:

sudo pacman -S llvm
# For development headers:
sudo pacman -S llvm-libs

Fedora/RHEL:

sudo dnf update llvm llvm-devel llvm-libs
# Or install specific version:
sudo dnf install llvm20-devel llvm20-libs

Ubuntu/Debian:

sudo apt update
sudo apt install llvm-20-dev

If that does not work take a look at this as well https://blog.wellosoft.net/update-llvm-from-18-to-21-in-ubuntu-24

macOS (Homebrew):

brew install llvm

"illegal hardware instruction" during compilation:

• This indicates an LLVM version incompatibility
• Upgrade to LLVM 20.0+ to resolve this issue
• See LLVM Version Requirements above

Missing LLVM development headers:

# Install development packages
sudo dnf install llvm-devel        # Fedora/RHEL
sudo apt install llvm-dev          # Ubuntu/Debian

Install the required tools using Scoop:

# Install Scoop package manager first if you haven't: https://scoop.sh/
scoop install python ninja cmake mingw

1. Clone the LLVM repository:

git clone https://github.com/llvm/llvm-project.git
cd llvm-project

1. Configure the build:

cmake -S llvm -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_PROJECTS="clang;lld" -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ -DCMAKE_ASM_COMPILER=gcc

1. Build LLVM (adjust -j8 based on your CPU cores):

ninja -C build -j8

• Build time: 30 minutes to several hours depending on hardware
• RAM usage: Can use 8+ GB during compilation
• If you encounter memory issues, reduce parallelism: ninja -C build -j4 or ninja -C build -j1

The compiled binaries will be located in build/bin/

To use clang, lld, and other LLVM tools from anywhere, add the build directory to your PATH:

set PATH=%PATH%;C:\path\to\your\llvm-project\build\bin

1. Open System Properties → Advanced → Environment Variables
2. Edit the PATH variable for your user or system
3. Add the full path to your build\bin directory (e.g., C:\Users\yourname\Desktop\llvm-project\build\bin)

[Environment]::SetEnvironmentVariable("PATH", $env:PATH + ";C:\path\to\your\llvm-project\build\bin", "User")

After adding to PATH, open a new command prompt and test:

clang --version
lld --version
llvm-config --version

@module "main"

pub const main = fn () int {
    output("Hello, World!\n");
    return 0;
}

Compile and run:

$ luma hello.lx
[========================================] 100% - Completed (15ms)
Build succeeded! Written to 'output' (15ms)

$ ./output
Hello, World!

See tests/3d_spinning_cube.lx for a complete 3D graphics application that:

• Renders rotating 3D cubes
• Uses sine/cosine lookup tables for performance
• Uses defer and ownership annotations for correct memory cleanup
• Compiles in 51ms to a 24KB binary

Interested in contributing to Luma? We'd love to have you!

• Check out our GitHub repository
• Join our Discord community
• Look at the doxygen-generated docs for architecture details
• If you would like to contribute, please read our contribution guidelines.

Built with ❤️ by the Luma community