Sharp - Transaction-based Hardware Description Language

Overview

Sharp is a transaction-based hardware description language built on MLIR that enables conflict-free hardware design through explicit conflict matrices and transaction-level modeling. Inspired by Bluespec and Koika, Sharp provides high-level abstractions while ensuring synthesizability to FIRRTL and Verilog.

Key Features:

• Transaction-level modeling with automatic conflict resolution
• Multiple simulation modes (TL, RTL, JIT, concurrent)
• Python frontend for programmatic hardware generation
• Complete FIRRTL/Verilog generation pipeline
• Comprehensive analysis and validation passes

Quick Start

# Build Sharp
curl -fsSL https://pixi.sh/install.sh | bash
git clone https://github.com/uv-xiao/sharp.git
cd sharp
pixi install
pixi run build

# Run tests
pixi run test  # 98/102 tests passing

# Try an example
./build/bin/sharp-opt examples/sharp-tutorial/chapter1/toggle.mlir

Example

txn.module @Counter {
  %reg = txn.instance @count of @Register<i32> : index
  
  txn.rule @increment {
    %val = txn.call @count::@read() : () -> i32
    %one = arith.constant 1 : i32
    %next = arith.addi %val, %one : i32
    txn.call @count::@write(%next) : (i32) -> ()
    txn.return
  }
  
  txn.schedule [@increment]
}

Documentation

• Tutorial - 8-chapter hands-on guide
• Txn Dialect - Language specification
• Execution Model - Three-phase execution semantics
• Python Frontend - PySharp API guide

Project Structure

sharp/
├── lib/           # Core implementation
├── include/       # Headers and TableGen
├── test/          # Comprehensive test suite
├── docs/          # Technical documentation
├── examples/      # Tutorial and examples
└── frontends/     # Python frontend (PySharp)

Development Status

See STATUS.md for detailed feature status. Current highlights:

• ✅ Core dialect and primitives complete
• ✅ All major conversion passes implemented
• ✅ Multiple simulation modes working
• ✅ Python frontend following PyCDE pattern
• 🚧 Runtime Python binding fixes needed
• 📋 IDE support and tooling planned