Birds of a Feather Archive

Quantum Jumps: From NISQ to FTQC to Hybrid HPC


Authors: Kevin Kissell (Alice & Bob), Dave Bacon (Google Quantum), Katherine Klymko (Lawrence Berkeley National Laboratory (LBNL)), Matthew Otten (University of Wisconsin-Madison), Aaron Lott (HPE; University of Maryland, College Park)

Abstract: Quantum computing is experiencing accelerating transitions, from noisy intermediate-scale quantum (NISQ) devices to fault-tolerant quantum computers, and from quantum computers being objects of research to their becoming scientific computation tools. Quantum programming habits and assumptions must be revisited, and for the HPC community, this is more than an incremental hardware upgrade; it affects entire stacks. This session will explore how compilation workflows, execution models, and algorithmic strategies are expected to change to incorporate error correction and integrate with production HPC workflows. Thought leaders from the quantum industry, HPC, and academic research will exchange perspectives and engage the audience in discussion.

Long Description: The past decade of experimentation on NISQ quantum computers has brought us a wealth of knowledge about quantum systems and quantum algorithms. And one of the key learnings is that fault-tolerant error-corrected quantum computers will be necessary to realize the potential of quantum computing to solve otherwise intractable problems in science and mathematics. But what exactly does this transition to FTQC entail? And how will it affect the way HPC users and developers interact with quantum processors?

Fault-tolerant quantum computers are more than just NISQ devices with better qubits. Instead of computing on noisy physical qubits directly, these systems use error correction codes to implement logical qubits; robust, encoded units of quantum information that allow vastly deeper and wider quantum circuit programs to be run, achieving scales at which there is practical quantum advantage over “classical” computing solutions. The underlying physical qubits are designed to maximize FTQC performance and efficiency, and not for individual physical qubit longevities and physical gate fidelities.

The resource requirements of fault-tolerant quantum computing are both quantitatively and qualitatively different from NISQ. Regardless of the physical modality chosen for quantum information, FTQC requires literally orders of magnitude more resources per logical qubit and per logical operation than NISQ, with the trade-offs varying with the chosen modality. While the compilation of quantum algorithms for a NISQ machine might focus on minimizing the number of 2-qubit CNOT gates, due to their higher sensitivity to noise, compiling the same algorithms for an FTQC machine may focus instead on minimizing the number of non-Clifford gates, due to their slower execution and their consumption of specialized resources (“magic” states). The sheer scale required to implement enough logical qubits to solve meaningful problems means that some form of distributed quantum computing will be required. While NISQ machines tend to be topologically uniform and homogeneous, arrays of superconducting/semiconductor structures or of atoms suspended in a vacuum. FTQC systems can require a “magic state economy” to support non-Clifford gates, and a “Bell pair economy” to enable quantum operations distributed across multiple substrates or vacuum columns. Non-trivial computer architectural thinking becomes a requirement for quantum processor design and deployment.

As quantum processors become true processing elements for research computing, their integration with existing scientific computing facilities becomes more interesting and demanding. In most prior work on NISQ machines in HPC environments, the algorithms and applications had almost complete decoupling of quantum and “classical” HPC phases. FTQC creates a wider range of potential computational pipelines, and designs for more closely coupled hybrid quantum HPC systems; hybrid quantum supercomputers are being proposed and evaluated. Closer coupling potentially allows us to circle back, and potentially use classical HPC resources, not only to run the classical parts of hybrid applications, but to support the operation of the FTQC elements.

An appreciation of the rich and subtle design space opening before us is multidisciplinary and requires a breadth of points of view. We intend to bring those together in this session and explore them with the audience.


Back to Birds of a Feather Archive Listing