Status: Complete
Preprint: Integrated into qco-integration — "End-to-End Fidelity Analysis of Quantum Circuit Optimization" (Submitted to arXiv)
Overview
A production-quality C++ quantum circuit optimizer designed to study cross-layer compilation: how gate-level optimization metrics (circuit depth, gate count) translate to pulse-level fidelity. This project bridges the CUDA Quantum Simulator for validation and QubitPulseOpt for pulse-aware analysis.
Key Research Question: Do circuits with fewer gates actually have higher pulse-level fidelity? Initial evidence suggests gate merging can create arbitrary rotation angles that are harder to implement, potentially reducing fidelity despite "better" gate counts.
Architecture
| Stage | Component | Description |
|---|---|---|
| 1 | Parser | Hand-written recursive descent parser for OpenQASM 3.0 |
| 2 | Circuit IR | DAG-based intermediate representation |
| 3 | Optimization Passes | 4 passes: cancellation, commutation, merging, folding |
| 4 | SABRE Routing | Topology-aware qubit mapping |
| 5 | Output/Export | Optimized circuit for simulation or pulse generation |
Key Features
Parser
- OpenQASM 3.0 Parser: Hand-written recursive descent parser with error recovery
- Supports standard gate set, qubit declarations, and circuit structure
Intermediate Representation
- DAG-based IR: Nodes represent gates, edges represent qubit dependencies
- Enables topological traversal for optimization
- Similar to how LLVM uses SSA for data flow analysis
Optimization Passes
- Identity Cancellation: X·X = I, H·H = I, etc.
- Gate Commutation: Reorder gates to enable further cancellations
- Rotation Merging: Rz(θ₁)·Rz(θ₂) = Rz(θ₁ + θ₂)
- Constant Folding: Simplify known-value operations
Qubit Routing
- SABRE Algorithm: State-of-the-art routing for NISQ devices
- Multiple Topologies: Linear, grid, heavy-hex support
- Minimizes SWAP overhead for hardware constraints
Testing
- 340 Unit Tests: Comprehensive coverage across parser, IR, passes, and routing
- GoogleTest framework with CI integration
Cross-Layer Analysis Pipeline
| Stage | Tool | Purpose |
|---|---|---|
| Input | VQE Circuit | From quantum-vqe |
| 1 | quantum-circuit-optimizer | Measure: depth, gate count |
| 2 | cuda-quantum-simulator | Validate: correctness |
| 3 | QubitPulseOpt | Measure: pulse fidelity |
| Output | Analysis | Correlate gate metrics with pulse fidelity |
Technology Stack
- Language: C++17
- Build: CMake
- Parser: Hand-written recursive descent (OpenQASM 3.0)
- IR Design: DAG-based (LLVM-inspired)
- Routing: SABRE algorithm
- Testing: GoogleTest (340 tests)
Links
Why Build From Scratch?
Building a quantum compiler from scratch (rather than using Qiskit/TKET) demonstrates understanding of compiler infrastructure fundamentals. More importantly, owning the implementation allows precise control over optimization decisions—essential for studying cross-layer fidelity effects.
This project directly supports my PhD application narrative: "I build the software control stack to make noisy quantum hardware perform like a reliable computer."