FORGE Language — VS Code Extension
forgelangdev.forge-lang — Official VS Code extension for the FORGE programming language.
FORGE: Zero-GC auto-parallel compiler. 7,797× faster than Python. 4.57× @parallel speedup.
Write sequential code. The compiler parallelises everything. No threading primitives. No GC overhead.

Features (Phase 7 Complete)
Syntax Highlighting
Full syntax highlighting for .forge files:
- Keywords —
fn, let, module, struct, pub, import, loop, match, if, else, for, while, return
- Annotations —
@parallel ⚡, @gpu 🖥️, @stream 🌊, @zero_copy, @compute
- Built-in types —
i8–i128, u8–u128, f32, f64, bool, str, void
- Constants —
true, false, null
- Strings, numbers, comments
- Function declarations and calls
- Operators and arrows (
->, =>, ..=)
- Generic type parameters (
<T>, []T)
Code Snippets
| Prefix |
What it creates |
module |
Module definition with exports |
fn |
Function definition |
pfn |
@parallel auto-parallelised function |
gfn |
@gpu + @parallel GPU-offload function |
sfn |
@stream + @zero_copy stream function |
struct |
Struct definition |
let |
Variable declaration |
for |
For loop |
while |
While loop |
match |
Match expression with arms |
main |
Full module with main function entry point |
ecs |
Entity-component system scaffold |
import |
Module import statement |
LSP Integration
- Type-check on save — Runs
forge check automatically
- Diagnostics — Inline error and warning squiggles
- Hover info — Type information on hover
- Inlay hints — Shows
@parallel and @gpu annotation hints
- Go to definition — Navigate to symbol definitions
Language Configuration
- Bracket matching
{}, [], ()
- Auto-closing pairs
- Line comments
//
- Block comments
/* */
- Word pattern for double-click selection
- Indentation rules for
fn, module, struct, match
Example
module GameEngine {
// Auto-parallelised across all CPU cores
@parallel
fn update_entities(entities: []Entity) -> []Entity {
return entities.map(e => Entity{
x: e.x + e.velocity * 0.016,
y: e.y + e.velocity * 0.016,
})
}
// GPU-offloaded compute — one annotation
@gpu
fn apply_physics(entities: []Entity) -> []Entity {
return entities.map(e => simulate_physics(e))
}
// Zero-copy 50GB world stream — two annotations
@stream
@zero_copy
fn load_world(path: str) -> void {
let stream = Stream.open(path)
for chunk in stream {
process_chunk(chunk)
}
}
fn main() -> void {
let entities: []Entity = load_entities()
loop {
let updated = update_entities(entities)
let physics = apply_physics(updated)
render(physics) // Vulkan / Metal / DX12
}
}
}
The FORGE Ecosystem (11 Live Sites)
CLI Commands (13 Total)
| Command |
Description |
forge build |
Compile to native ELF binary |
forge run |
Build and execute |
forge check |
Type-check without emit |
forge emit-ir |
Dump LLVM IR |
forge bench |
Run benchmark suite |
forge profile |
Profile execution |
forge fmt |
Format source code |
forge repl |
Interactive REPL |
forge share |
Share code snippets |
forge game |
Game engine scaffold |
forge llm |
LLM inference tools |
forge pkg |
Package management |
forge lsp |
LSP server |
Phase Roadmap
| Phase |
Status |
Description |
| Phase 0 |
✅ Complete |
Benchmarks: 9.47 GB/s, 19× Python |
| Phase 1 |
✅ Complete |
Runtime: memory regions, fibers, streams |
| Phase 2 |
✅ Complete |
Compiler: lexer, parser, typechecker, LLVM codegen |
| Phase 3 |
✅ Complete |
Game engine, ForgeHub, LSP server |
| Phase 4 |
✅ Complete |
WebGPU, incremental cache, debug info |
| Phase 5 |
✅ Complete |
Native binary, web IDE, package registry, VS Code |
| Phase 6 |
✅ Complete |
@parallel → real speedup, forge pkg publish |
| Phase 7 |
✅ Complete |
docs.forgelang.dev, games, studio, cloud |
| Metric |
FORGE |
Comparison |
| Stream throughput |
9.47 GB/s |
7,797× faster than Python |
| 100M records (8 cores) |
416 ms |
28× faster than Python |
| @parallel speedup |
4.57× |
8 cores, auto-distributed |
| HTTP server memory |
212 KB |
212× leaner than Node.js |
| Compiler size |
11,922 lines Zig |
Auditable, lean codebase |
Links
FORGE v0.1.3 — Syntax highlighting, snippets, LSP. Zero-GC auto-parallel compiler. 7,797× faster than Python.
| |