IBM recently posted on its blog that the IBM Quantum team has made great progress along its development roadmap in the past year. In the development process, the improvement of quantum computing performance is particularly important. Among them, the three key indicators to measure the performance of quantum computing include scale, quality and speed.A few days ago, on the basis of the original index, IBM proposed a speed measurement index-circuit layer operations per second, also known as CLOPS.
The following is the full text of the blog:
The IBM Quantum team has made great achievements in the past year. We have made considerable progress along our development roadmap and plan to implement frictionless quantum computing on more than 1,000 systems by the end of 2023. We expect this to be a turning point in the development of this field.
Our guiding principle and goal for advancing quantum computing systems is to increase the amount of useful work that these systems can complete-in short, the performance of quantum computing. Performance has three key attributes: scale, quality, and speed. The introduction of quantum computers into an organization’s computing workflow needs to promote the development of these three areas.
Three key indicators to measure the performance of quantum computing: scale, quality, and speed (Source: IBM)
For scale, we measure progress by the number of qubits in the system. We released a 65-qubit processor last year, and we are expected to deliver a 127-qubit Eagle processor this year, thanks to continuous innovation in expansion technology.
In terms of quality, we use the previously defined quantum volume and provide today’s users with multiple systems with 128 quantum volumes; speed is an often-discussed attribute that has a real impact on the workload of the application. But a suitable system-agnostic metric has not yet appeared, which can capture the complete dependency between the hardware and software executed by the circuit.
Today, we want to propose an indicator, which we call the number of operations per second of the circuit layer, or CLOPS.
Improve the practicality of quantum computing through CLOPS
With the development of quantum computing and the beginning of solving practical problems, we must pay more attention to how much work a quantum computing system can do in a given unit of time. We expect that the actual workload involves the interaction between quantum and classical-a complete program will call the quantum processor as an accelerator for certain tasks; an algorithm will require multiple calls to the quantum processor. Therefore, the runtime system that allows efficient quantum classical communication is essential for achieving high performance. We have embedded this runtime interaction into the CLOPS benchmark proposal.
CLOPS is an index related to the speed at which a quantum processor executes a circuit—specifically, this index measures the speed at which the processor can execute the same type of parameterized model circuit layer used to measure quantum volume.
Increasing the speed of quantum processors is essential to support recent algorithms based on variational methods, which require thousands of iterations. The improved qubit bit gate time allows us to greatly expand the scope of current quantum systems, and allows us to go one step further than classical computing hardware.
——Pranav Gokhale, founder and CEO of Super.tech
Run-time architecture and compilation phase. The circuit mode of Quantum Volume benchmark test and its offline compilation. (Source: IBM)
Quantum circuit is the basic calculation unit of quantum computer, just like the logic circuit of classical calculation. Benchmarking requires the execution of many instances of the model circuit and different parameters generated at runtime. Each part of this hardware-software stack contributes to CLOPS, including the repetition rate of the quantum processor, the speed of gate operation, the time required to compile at runtime, generate classical control instructions, and the data transfer rate between all units.
To achieve the highest performance quantum computing system, we need to completely rethink the architecture that manages the operation of quantum computing programs. This is why we introduced Qiskit Runtime.
Qiskit Runtime is a portable, safe, and containerized architecture that runs quantum programs on classical computing units tightly integrated with quantum processors. At the same time, Qiskit Runtime allows quantum computers to become part of any computing environment to speed up calculations, processing tasks, and data transmission to the quantum processing unit to maximize efficiency.
Today, our fastest system can perform 1,400 circuit-level operations per second.
Improvements in quantum hardware will reduce circuit delay time and idle time between successive circuits; further improvements in the runtime architecture will reduce the initialization time of data loading and improve runtime compilation.
Superconducting qubits are the natural choice for high-performance quantum computing
Our goal is to develop practical quantum computing, and we believe that our superconducting qubit system provides the best opportunity to promote the popularization of quantum computing. Other quantum architectures can achieve high performance in some (but not all) aspects of scale, quality, and speed. For example, trapped ions have shown the ability to achieve high quantum volume, but face challenges in solving speed, while spin qubits can achieve high speeds, but have so far faced challenges in the ability to push mass or scale.
We expect that when it comes to performance improvements in scalability, quality, and speed, superconducting qubits will provide the greatest opportunity to achieve sustained growth in these three areas.
In fact, we have already seen the benefits of speed in real scientific demonstrations. In 2017, our modelling of the lithium hydride molecule required running 4.8 billion quantum circuits, which would have taken months to years on our previous commercial stack. But now, with Qiskit Runtime and other improvements, we can increase the calculation speed by 120 times.
If we want to accelerate the application of quantum computers, we need to focus on the useful work that quantum computers can do, and we need to continuously improve in all three key performance areas:
·We are committed to realizing our development roadmap, as evidenced by Falcon (27), Hummingbird (65) and Eagle (127).
In terms of quality, we are actively promoting core research to improve the basic coherence and gate error of superconducting qubits.
· We have now introduced CLOPS to improve these three performance indicators.
We believe that through our choice of superconducting qubit architecture and the introduction of Qiskit Runtime, as we continue to execute our development roadmap, we will achieve greater performance improvements in the near future.