Among Ang Li’s many other projects at Pacific Northwest National Laboratory, he and his colleagues also have developed a quantum circuit simulator.
Quantum machines are arguably the computers of the future. They use the properties of quantum physics to store and manipulate data, letting them perform much faster than conventional machines. Standard computers store data in strings of zeros and ones. Quantum computers use qubits, which represent the state of being one and zero at the same time. This lets quantum computers simultaneously compute billions of possibilities and errors, making it a useful tool for multiple disciplines.
The simulator Li and his PNNL colleagues developed uses GPU/CPU clusters to ensure that the quantum computers actually work before they’re deployed. “Quantum simulation is a difficult thing for classical computers because of all the states that a quantum state can represent,” Li notes.
To make such predictions, he used a system called a density matrix to simulate quantum circuits, and presented a talk on the method this past November at SC20, the annual international high-performance computing conference. Quantum computers rely on quantum circuits, which change based on operations that modify a qubit’s state. Density matrices contain all the information about a particular open quantum state and thus are more reliable simulations of future quantum machines. Li’s team ran the density matrix on linked GPUs and found it works 10 times faster than existing simulators.
Moving forward, Li and his PNNL colleagues will try to support different high-level systems that use GPUs and possibly integrate these simulations with high-level quantum programming frameworks and languages, such as Microsoft Q#/QDK, IBM Qiskit and Google Cirq.