Request for Benchmark Data: Tracking Performance of Grover Benchmark on CPU and GPU

[Kihang-Youn]

Could you kindly provide benchmark data for reference? For instance, it would be highly beneficial if there is a way to track the latest performance metrics for both the CPU and GPU versions of the Grover Benchmark. I’ve searched through related papers, but finding detailed or up-to-date data has been quite challenging.

Having such data readily available would not only help in better understanding the current capabilities but also contribute positively to future optimizations and research directions. Your assistance in this matter would be greatly appreciated.

[TysonRayJones]

Hi Kihang-Youn,

That's a great idea. We're in the process of overhauling QuEST on the v4 branch which includes a new unit-testing system. This will eventually include tests to detect performance regressions. It seems natural there to profile big integration tests like Grover's.

Of course, comparing benchmarking data requires running on the same hardware which greatly affects full-state simulation runtime, which is usually bottlenecked by memory bandwidth. As such, it's usually easier to just run QuEST on the hardware from which one is producing under benchmarks to compare. Ergo fixed data mightn't be so useful, but it's certainly worthwhile preparing scripts for quick testing on candidate hardware.

I'll keep this discussion open as a reminder, and as a space for more thoughts!