trasyn: TensoR-based Arbitrary unitary SYNthesis

This repo contains the code and data for the paper Reducing T Gates with Unitary Synthesis. We are actively preparing our method to be a user-friendly package.

Reproduce Results

Install the dependencies in requirements.txt. Download the gridsynth binary and put it in src/. Optionally, install cupy to enable GPU acceleration. For the best performance, please export CUPY_ACCELERATORS=cub if cub is not the default.

• src/unique_matrices.py generates the unique matrices and corresponding gate sequences in assets/.
• src/benchmark_random_unitary.py synthesizes random single-qubit unitaries. The random unitaries are in data/random_1q/, and the results are in data/random_unitary/.
• data/qasm/ contains raw qasm files we perform our circuit experiments on. src/generate_qaoa.py generates the QAOA qasm files with the gate ordering that maximizes rotation merging.
• src/count_nontrivial_rotations.py transpiles the raw qasm files with in qiskit using 16 settings as desribed in the paper and counts the number of nontrivial rotations in the transpiled circuits for each setting. The transpiled circuits with the least rotations for the $R_z$ and $U3$ intermediate representations are stored in data/qasm_transpiled_rz/ and data/qasm_transpiled_u3/, respectively.
• src/benchmark_circuits.py synthesizes the $R_z$ or $U3$ in the transpiled circuits. The results and the synthesized qasm files are in data/circuit_benchmark/.
• src/pyzx_opt.py optimizes the synthesized circuits with pyzx. The data after optimization is updated in data/circuit_benchmark/.
• src/1q_errors_tradeoff.py computes the process fidelity of synthesized single-qubit $R_z$ gates under varying logical error rates and synthesis error thresholds to study their tradeoff. The results are in data/logical_error/.

Reproduce Figures and Tables

Install matplotlib, pandas, and scipy. Run src/plot_performance.ipynb.