TriQ is the backend compiler for the Scaffold quantum programming language. TriQ takes two inputs: 1) a gate sequence produced by ScaffCC and 2) qubit connectivity and calibration data for the target machine. It compiles the program gate sequence by choosing a good initial placement of the program qubits on the hardware qubits, reducing communication, and by applying gate optimization techniques.
TriQ generates optimized quantum assembly code for both superconducting and ion trap quantum computers. We support the 14 and 5-qubit superconducting devices from IBM (IBMQ14, IBMQ5), the 16-qubit superconducting systems from Rigetti (Aspen1, Aspen3) and 5-qubit trapped ion system from University of Maryland.
Dependencies:
To install TriQ,
make
To run a program prog_name.scaffold with TriQ, first exract an intermediate representation using ScaffCC.
./scaffold.sh -b prog_name.scaffold
Run TriQ on the intermediate representation
python ir2dag.py prog_name.qasm prog_name.in
./triq prog_name.in <output_file_name> <target_backend_name>
For example, to compile programs for IBMQ5
./triq prog_name.in prog_name.out ibmqx5 0
For each backend, TriQ assumes that the error data for single qubit, two-qubit and readout operations are specified in the config directory. For example, for IBMQ5, the Single qubit fidelity data should be specified in ibmqx5_S.rlb. Each line in this file specifies a qubit number and the corresponding fidelity. Two-qubit gate fidelity is specified in ibmqx5_T.rlb and readout fidelity is specified in ibmqx5_M.rlb.
If you use TriQ in your work, we would appreciate it if you cite our paper:
Prakash Murali, Jonathan M. Baker, Ali Javadi Abhari, Frederic T. Chong, Margaret Martonosi. "Noise-Adaptive Compiler Mappings for Noisy Intermediate-Scale Quantum Computers", International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), 2019.
Please reach out to Prakash Murali ([email protected]) for any questions or clarifications.
More compiler options and helper methods to extract reliablity data automatically from vendor APIs