IBM has made significant strides in developing its quantum software development kit, Qiskit, to improve performance and efficiency. The company’s efforts are crucial as slow circuit building and transpilation can deter researchers from studying problems on real quantum systems. IBM’s team has refactored much of the code into the Rust programming language, which includes security and memory management features to build higher-quality software.
Additionally, they have developed an algorithm called LightSABRE, an enhancement of the SABRE algorithm that maps and routes circuits onto physical devices while minimizing gates and circuit layers. This upgrade has delivered an 18.9% decrease in costly SWAP gates for the same circuits versus SABRE. The Qiskit Transpiler Service, which uses AI-powered transpiler passes, has also demonstrated a consistent ability to produce shallower, higher-quality circuits than the standard Qiskit transpiler.
These advancements solidify Qiskit’s role as a trusted benchmarking suite for quantum SDKs and make it an attractive option for researchers and developers interested in performance.
The results show that Qiskit, an SDK developed by IBM, outperforms other SDKs in terms of speed and circuit quality. Specifically, Qiskit is 13 times faster at transpiling circuits and produces circuits with 24% fewer two-qubit gates than the next best-performing SDK, TKET.
The article highlights the importance of performance in quantum computing, as slow circuit building and manipulation can deter researchers from studying problems on real quantum systems. The authors also emphasize that they are not experts in other quantum SDKs and have made Benchpress open-source to allow others to review and run the tests.
The article goes on to discuss new features and upgrades to Qiskit that contribute to its high performance, including refactoring code into the Rust programming language, which provides security and memory management features. The authors also mention an upcoming paper detailing an algorithm called LightSABRE, which is an enhancement of the SABRE algorithm used for circuit transpilation.
Furthermore, the article highlights the benefits of using the Qiskit Transpiler Service (QTS), which demonstrated a consistent ability to produce shallower and higher-quality circuits than the standard Qiskit transpiler. The authors also discuss the importance of building a performant software stack that can run circuits on real hardware, representing different quantum computing architectures.
Finally, the article mentions IBM’s efforts to track metrics such as CLOPS (circuit layer operations per second) and error per layered gate to benchmark Qiskit’s ability to run circuits on quantum hardware. The authors also mention exploring ways to augment these metrics to measure primitive quality using mirror circuits and other measures.
External Link: Click Here For More
