Granted Patent B2

US 12,386,672 · App. 17/648,065 · Granted Aug 12, 2025

Inventors: Rômulo Teixeira de Abreu Pinho (Rio de Janeiro, BR); Victor Fong (Medford, MA); Kenneth Durazzo (Morgan Hill, CA)

Assignee: Dell Products L.P.

G06F9/505G06F9/4887G06N10/60

Abstract

Claims (39)

1. A method for optimizing execution of hybrid quantum-classical computing workloads, comprising:

receiving a computation workflow defined by a graph that includes a group of computing nodes, which comprise quantum computing nodes and classic computing nodes, and edges representing interdependencies between the computing nodes, wherein each node comprises an estimation of an execution time for each resource allocation configuration;

receiving a catalogue of quantum computing instances that are available in a hybrid classic-quantum computation infrastructure, wherein the catalogue identifies all quantum computing instances and classic computing instances in the hybrid classic-quantum computation infrastructure;

transforming the graph to create a first graph transformation,

wherein each of the quantum computing nodes is assigned a respective candidate resource allocation that identifies candidate resources operable to execute a respective quantum algorithm associated with that quantum computing node, wherein the transforming is performed using information from the catalogue, and

wherein the transforming comprises:

allocating to each quantum computing node the respective candidate resource based on one or more given quantum circuit parameters, and based on a service level agreement estimation for execution of a function that is associated with that quantum computing node;

determining whether a quantum computing instance in the respective candidate resource is a simulation engine or a quantum computing device; and

based on the determination, obtaining a service level agreement from a classic computing device or from the quantum computing device;

optimizing the computation workflow by selecting, for each of the quantum computing nodes, a configuration of resources from the candidate resource allocation based on all execution times, wherein the optimizing comprises transforming the first graph transformation to create a second graph transformation, and the second graph transformation comprises the selected configuration of resources for each node, which meets the service level agreement; and

executing, by classic computing instances or quantum computing instances of resources in the selected configuration based on the optimization, each of the quantum computing nodes in the second graph transformation.

2. The method as recited in claim 1 , wherein the second graph transformation is output to an execution module for execution in the hybrid classic-quantum computation infrastructure.

3. The method as recited in claim 1 , wherein the catalogue identifies all quantum computing instances in the hybrid classic-quantum computation infrastructure.

4. The method as recited in claim 1 , wherein quantum circuit parameters specify, for each node, aspects of a quantum circuit associated with a respective quantum algorithm implemented in that node, and the quantum circuit parameters comprise a number of qubits of the quantum circuit, and a circuit depth of the quantum circuit.

5. The method as recited in claim 1 , wherein the selecting of resources is performed based on experimental observations about resource consumption by the respective quantum algorithms associated with each quantum computing node.

6. The method as recited in claim 1 , wherein the second graph transformation is based on requirements specified in a service level agreement, and the second graph transformation identifies an overall hybrid classic-quantum configuration that is executable to satisfy the service level agreement.

7. The method as recited in claim 1 , wherein the second graph transformation defines an overall algorithm which, when executed, performs one or more functions according to a service level agreement.

8. The method as recited in claim 1 , wherein the selected resources specified in the second graph transformation are operable to execute respective quantum algorithms associated with each quantum computing node so that all execution times of the quantum algorithms aggregated together define an overall execution time that is within a time limit specified by a service level agreement.

9. The method as recited in claim 1 , wherein the workflow graph further comprises classic computing nodes, and the optimizing further comprises selecting, for each of the classic computing nodes, a resource from a respective candidate resource allocation associated with that classic computing node.

10. A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations for optimizing execution of hybrid quantum0classical computing workloads, the operations comprising:

receiving a computation workflow defined by a graph that includes a group of computing nodes, which comprise quantum computing nodes and classic computing nodes, and edges representing interdependencies between the computing nodes, wherein each node comprises an estimation of an execution time for each resource allocation configuration;

receiving a catalogue of quantum computing instances that are available in a hybrid classic-quantum computation infrastructure, wherein the catalogue identifies all quantum computing instances and classic computing instances in the hybrid classic-quantum computation infrastructure;

transforming the graph to create a first graph transformation,

wherein each of the quantum computing nodes is assigned a respective candidate resource allocation that identifies candidate resources operable to execute a respective quantum algorithm associated with that quantum computing node, wherein the transforming is performed using information from the catalogue, and

wherein the transforming comprises:

allocating to each quantum computing node the respective candidate resource based on one or more given quantum circuit parameters, and based on a service level agreement estimation for execution of a function that is associated with that quantum computing node;

determining whether a quantum computing instance in the respective candidate resource is a simulation engine or a quantum computing device; and

based on the determination, obtaining a service level agreement from a classic computing device or from the quantum computing device;

optimizing the computation workflow by selecting, for each of the quantum computing nodes, a configuration of resources from the candidate resource allocation based on all execution times, wherein the optimizing comprises transforming the first graph transformation to create a second graph transformation, and the second graph transformation comprises the selected configuration of resources for each node, which meets the service level agreement; and

executing, by classic computing instances or quantum computing instances of resources in the selected configuration based on the optimization, each of the quantum computing nodes in the second graph transformation.

11. The non-transitory storage medium as recited in claim 10 , wherein the second graph transformation is output to an execution module for execution in the hybrid classic-quantum computation infrastructure.

12. The non-transitory storage medium as recited in claim 10 , wherein the catalogue identifies all quantum computing instances in the hybrid classic-quantum computation infrastructure.

13. The non-transitory storage medium as recited in claim 10 ,

wherein quantum circuit parameters specify, for each node, aspects of a quantum circuit associated with a respective quantum algorithm implemented in that node, and the quantum circuit parameters comprise a number of qubits of the quantum circuit, and a circuit depth of the quantum circuit.

14. The non-transitory storage medium as recited in claim 10 , wherein the selecting of resources is performed based on experimental observations about resource consumption by the respective quantum algorithms associated with each quantum computing node.

15. The non-transitory storage medium as recited in claim 10 , wherein the second graph transformation is based on requirements specified in a service level agreement, and the second graph transformation identifies an overall hybrid classic-quantum configuration that is executable to satisfy the service level agreement.

16. The non-transitory storage medium as recited in claim 10 , wherein the second graph transformation defines an overall algorithm which, when executed, performs one or more functions according to a service level agreement.

17. The non-transitory storage medium as recited in claim 10 , wherein the selected resources specified in the second graph transformation are operable to execute respective quantum algorithms associated with each quantum computing node so that all execution times of the quantum algorithms aggregated together define an overall execution time that is within a time limit specified by a service level agreement.

18. The non-transitory storage medium as recited in claim 10 , wherein the workflow graph further comprises classic computing nodes, and the optimizing further comprises selecting, for each of the classic computing nodes, a resource from a respective candidate resource allocation associated with that classic computing node.

Assignments (1)

Continuity (1)

References Cited (33)