Quantum resource estimation is a benchmarking method used to estimate and compare the resources required to run quantum algorithms, and to track progress toward fault-tolerant quantum computing. It allows direct comparison between platforms by determining the number of physical qubits and time needed to run a specific quantum computing application under a given set of architectural assumptions. Quantum resource estimations combine a wide variety of factors that impact quantum computation – across hardware and software – thereby leveling the playing field between different qubit modalities and system architectures, and providing a metric to gauge the impact of improvements in hardware, software, and algorithms.
What are the types of quantum resource estimation (QRE)?
A full quantum computing stack involves complex, interdependent levels of hardware, software, and algorithms, making it challenging to compare the anticipated performance of different modalities or systems. Some or all of these factors may be considered in each quantum resource estimation, depending on the desired accuracy and effort, so it’s important to understand the differences.
Logical QRE is the least accurate method of estimation. It assumes “perfect” (error-free) qubits, providing a high-level estimate of the number of qubits and operations required by an algorithm. It tends to consider only the most time-consuming type of gates (such as Toffoli or T gates) and ignores all other operations. This first-order approach is useful for comparing potential applications, but it does not take the qubit type, architecture, or compilation strategy into consideration.
Physical QRE is an intermediate approach that includes error correction. It factors in the encoding of physical qubits into logical qubits for better physical resource estimations. It also includes the decomposition of logical operations into basic operations on physical qubits to arrive at a more realistic run time. Physical quantum resource estimation requires more upfront preparation work and algorithm optimization than logical QRE – mapping the algorithm to a quantum circuit, careful selection of error correction code parameters, etc. – but provides more meaningful values as a result of the additional information.
Hardware-Aware QRE is even more advanced and resource-intensive. In addition to accounting for concrete implementation of an algorithm and error correction parameters, this approach takes into consideration each system’s particular architecture, qubit topology and performance, parallelization, realistic noise models, and other factors. This allows head-to-head comparisons of required physical qubit resources and run times across quantum computing platforms.
Distributed QRE: Next-generation quantum resource estimation
Most QRE tools in use today assume that all qubits will operate within a single module, yet they predict a need for tens of thousands to millions of qubits for important algorithms like the factorization of large numbers used in encryption. Networking multiple modules together via quantum interconnects – an approach known as distributed quantum computing – is now being recognized as a more practical and expedient path to implementing applications that require many thousands to millions of qubits.
Distributed quantum computing can be challenging and expensive to implement if not integrated into the design from the start. It relies on quantum interconnects with performance comparable to intra‑module operations. In many architectures, such interconnects are difficult and inefficient to add after the fact due to transduction losses, and in some cases are fundamentally incompatible with the underlying hardware.
This gives architectures with a built-in quantum interconnect a distinct advantage over monolithic systems, as they can both scale up and scale out by linking optimally sized, high-performance qubit processors via quantum interconnects. This is the basis of Photonic’s Entanglement First™ Architecture, which features an efficient, native optical interconnect at telecom wavelengths, compatible with existing fibre optic infrastructure.
The need to include distributed costs in quantum resource estimation for comparative and roadmap purposes cannot be overlooked. Resource estimations that account for full system costs allow for more accurate assessments of each architecture’s path to commercial value.
Related content: To learn how Photonic constructed a new class of QLDPC error correction codes that unlock faster and more efficient quantum computation compared to surface codes, read our white paper: Launching SHYPS.