EP 404: Quantum at PSC sponsored by HPC

One of the things I liked about how you were talking about that is you're talking about the integration of HPC and quantum.

They're not really different things.

They're two different vectors maybe for approaching similar types of problems.

But one thing we find in our research is that quantum hasn't, for the most part, been about, oh, I'm going to try to do the same things faster or larger scale, but rather that it's unlocking the approaches to new types of problems that you might be able to address

yes absolutely that's a great point and that's really what we're trying to stress uh to our community especially with with the upcoming deployment that we have um is that at this point you're not really looking into uh advantage over methods and approaches that you already have running on classical systems.

You're really looking at unlocking different types of ways of doing things.

And then eventually when the technology gets there, when these quantum machines are scaled up and become fault tolerant and so on, then you're really ready to squeeze the advantage that that you should be getting there but it's really about trying to understand what can we do that is different than what we're already doing not in a way of like um getting advantage in the short term, but more like how can I think differently about problems and how can I use quantum computing or should I say quantum information processing power to tackle those problems?

What role does your HPC experience play then in terms of your quantum strategy? Where are those overlapping?

Yeah, that's another great question.

So I should say that HPC is central to us, right? We're basically a national HPC center.

That's how we work.

AI has been, over the last several years, has been playing a larger role.

People call it HPC AI.

But in the end, it's these large scale production level systems that we maintain and operate.

And the idea here is quantum has been coming into this ecosystem more and more, right? Over the last two or three years, the community has the quantum computing community really shifted their understanding of quantum computers as standalone machines.

So maybe five or seven years ago, that was a more predominant vision in the community where, oh, these are going to be monolithic systems and so on and so forth.

For people working with high-performance computing systems, we kind of knew that that would never be the case, right? We know that for certain other, quote-unquote, accelerators or coprocessors, you need a lot of other infrastructure that is not related to that.

You can think about GPUs, graphic processing units.

They're really important for AI workloads these days, but a GPU by itself doesn't do anything, right? So the same for quantum computing here.

The strategy here is really to try to bring the quantum processing unit into this ecosystem and then leverage that as a coprocessor.

Yeah, is that really the way that you're viewing quantum then, primarily as a co-processor within a larger ecosystem? And does it take part of the workflow then where you're integrating HPC techniques, AI techniques, quantum techniques as part of a total workflow? Precisely, yes.

That's really the approach that we're taking and that's really what we think is the right thing to do, right? Instead of focusing on potential workloads that could run entirely on a quantum computer, even if you have like a big fault-tolerant computer, the most interesting use cases are the hybrid workflows.

And you can think about hybrid across different dimensions, right? Not only the classical and quantum dimension, you have something running on classical hardware like a CPU or a GPU+.

parts of it running on the QPU.

It's also about applications and what can you do leveraging what is already very efficient, right? So we have a lot of simulation methods that fully include quantum correlations that are very efficient.