🧠 Toy Virtual Machine

A fully functional virtual machine and interpreter for a dynamically typed language, implemented in Java.

---

🚀 Overview

This project implements a stack-based virtual machine capable of parsing, interpreting, and optimizing execution of a toy high-level language (Simple/ToyLang).
It follows key design principles from Truffle/GraalVM and modern language runtime architectures.

✨ Core Features

• Bytecode-based execution engine
  Custom compiler emits compact bytecode instructions interpreted by the VM loop.

• Hidden Classes / Shapes Efficient object representation inspired by JavaScript engines like V8 and Truffle.

• Inline Caching (IC) Monomorphic and polymorphic inline caches to optimize dynamic property access.

• Array Strategies Adaptive array storage depending on content types (int[], double[], Object[]).

• Lazy Strings / Ropes Memory-efficient string concatenation and slicing without immediate flattening.

• (Optional) JIT Compilation Minimal experimental JIT tier to compile hot functions or loops.

---

🏗️ Architecture

Source Code (.sl)
     ↓
 Parser & Compiler
     ↓
 Bytecode (Toy BCI)
     ↓
 Toy Virtual Machine
     ↓
 [Optimizations: ICs, Shapes, Strategies]
     ↓
 Execution / Profiling

Key components:

• ToyBciLoop – the main interpreter loop
• VObject, Shape – object model and property storage
• ToyCompiler – frontend bytecode generator
• ToyBuiltins – built-in library functions

---

🧪 Example

function main() {
    o = new();
    o.x = 42;
    println(o.x);
}

Execution:

$ toy examples/object.sl
42

---

📈 Benchmarks

Includes micro-benchmarks comparing interpreted vs inline-cached execution:

Monomorphic IC benchmark: 10 runs, 1,000,000 iterations/run
Without IC: 2300 ms
With IC:    1850 ms
Speedup:    ~1.25×

---

🔍 Technologies & Skills Demonstrated

• Java 15+
• Virtual machine internals
• Bytecode design and interpretation
• Dynamic object models (hidden classes)
• Inline caching and specialization
• Memory-efficient data structures
• Performance benchmarking and tracing

---

📚 References

• Truffle & GraalVM Papers
• V8 “Shapes” and Inline Cache design
• TU/e Virtual Machines course materials

---

👤 Author

Adrià Martín
📧 [adria.martinma@autonoma.cat]
🔗 github.com/adriamartin

Pau Jaso 📧 [pau.jaso@autonoma.cat]
🔗 github.com/pau-jaso

---

“A VM is not just an interpreter — it’s a dynamic runtime capable of learning how your program behaves.”