Quantum Program Synthesis (Qupsy)

Qupsy is a tool for quantum program synthesis: given a set of input/output quantum state test cases, it automatically generates a quantum program to satisfy those specifications.

---

🚀 Getting Started

Requirements

Install the required packages with pip:

pip install numpy cirq

Running the Tool

Use the following command to synthesize a quantum program:

python qpsynth.py [specification_path] [search_mode: baseline]
# Example: python qpsynth.py benchmarks/ghz.json baseline

• specification_path: Path to the .json specification file (see below).
• search_mode: Choice of algorithm:
  • baseline: simple baseline search

---

📄 Specification File Format

Your specification MUST be a JSON file with the following structure:

• gates (optional):
  List of gate names to use in synthesis (if omitted, all available gates are used).
  Available gates:H, X, Y, Z, Ry, CX, CY, CZ, CRy
  You can define more gates in QPSynthesis/synthesizer/language.py.
• examples:
  • qubit: Number of qubits
  • input (optional): Comma-separated real/complex amplitudes for initial state $|\text{in}\rangle$. Defaults to $|0\dots0\rangle$.
  • output: Comma-separated real/complex amplitudes for the expected output $|\text{out}\rangle$.

Example Specification

{
  "gates": ["H", "CX"],
  "examples": {
    "1": {
      "qubit": "3",
      "output": "0.70710677,0,0,0,0,0,0.70710677,0"
    },
    "2": {
      "...": "..."
    }
  }
}

See more examples in benchmarks/specification/.

---

🗂️ Overview of Key Modules

Modules are under QPSynthesis/synthesizer/ path.

• search.py
  Implements search algorithms:

  • search_base: Baseline algorithm
  

• language.py
  Defines the quantum small language to synthesize a quantum program.

  

• worklist.py
  Priority queue (worklist) data structure for candidate program states during search.

• transition.py
  Defines the transition relation—how current candidate programs evolve into next candidates.

• prune.py
  Implements basic pruning strategies to speed up synthesis and reduce search space.