Patent Application
20250225424
The exemplary embodiment relates to quantum computing and finds particular application in connection with a method and system for optimally sharing the computing resources among various users simultaneously.
Current quantum computers work as individual computing systems, connected classically to a control system that controls the operations on qubits (or quantum bits). The number of qubits is important for quantum computing as it is one measure of how much potential power the quantum computer has. Because of the challenging hardware requirements for building quantum computers, it is particularly challenging to build single quantum computers with many qubits, and it is estimated that the qubit limit for a single quantum computer is roughly on the order of 10,000. In order to perform quantum algorithms, logical qubits are needed, where each logical qubit requires potentially hundreds or thousands of physical qubits to implement, greatly decreasing the power of the quantum computer. There is a further demand that thousands of logical qubits are needed before the true power of quantum computing becomes apparent.
In order to scale quantum computers up and overcome the challenges, one approach is to take a collection of smaller scale quantum computers and network them together via classical and quantum communication channels, along with the classical control system; this forms a distributed quantum computer. In order to perform the same quantum algorithms, one would perform on a single quantum computer over a distributed system of quantum computers, additional tasks need to be performed beforehand. For example, additional operations need to be injected into the algorithm steps to produce a logically equivalent version for the distributed system. These steps include communication steps, transmitting both entanglement and classical information, and logical quantum control operations. The additional operations injected to the algorithm steps require coordination and synchronization between the distributed system, which have very strict requirements. The algorithm in some cases is disassembled and reassembled, combining measurement outputs from various times and qubits across the system. For additional speedup and resource utilization, the algorithm structure can also be modified such that a parallel execution of the algorithm is possible. The system can also include operations to gain this speed-up, allocating resources over the distributed system.
Further, because it is assumed that quantum computers will be primarily accessed via cloud environments, it will be necessary to optimally share the computing resources amongst various users simultaneously. A secondary problem of how to optimally and securely allocate quantum computing resources over a distributed quantum computing system arises. Therefore, there is a need for a system that allows as many users as possible access to the system simultaneously.
Various details of the present disclosure are hereinafter summarized to provide a basic understanding. This summary is not an extensive overview of the disclosure and is intended neither to identify certain elements of the disclosure, nor to delineate the scope thereof. Rather, the primary purpose of the summary is to present certain concepts of the disclosure in a simplified form prior to the more detailed description that is presented hereinafter.
In accordance with one aspect of the presently described embodiments, a system for distributed quantum computing execution is provided. The system comprises a processor and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the processor configured to receive a plurality of quantum programs from one or more users, establish a communication connection with at least one distributed quantum computer, discovering all nodes of the distributed quantum computer, wherein each node comprises at least one quantum processor, process the plurality of quantum programs to create a remapped set of distributed quantum programs, determine an allocation of available quantum resources needed for the set of quantum programs, and convert the remapped set of quantum programs into at least one set of machine level instructions for each node in the distributed quantum computer. The system further includes a processor configured to distribute the at least one set of machine instructions to each of the nodes, wherein each node executes instructions based at least in part on the allocation determined by the processor, monitor the performance of the distributed quantum computer, collect all output from the executed instructions, retrieve the collection of output, convert the output, and send the converted output to at least one user.
In accordance with another aspect of the presently described embodiments, a method for distributed quantum computing execution is provided. The method comprises receiving a plurality of quantum programs from one or more users, establishing a communication connection with at least one distributed quantum computer, discovering all nodes of the distributed quantum computer, wherein each node comprises at least one quantum processor, processing the plurality of quantum programs to create a remapped set of distributed quantum programs, determining an allocation of available quantum resources needed for the set of quantum programs, and converting the remapped set of quantum programs into at least one set of machine level instructions for each node in the distributed quantum computer. The method further includes distributing the at least one set of machine instructions to each of the nodes, wherein each node executes instructions based at least in part on the allocation determined by the processor, monitoring the performance of the distributed quantum computer, collecting all output from the executed instructions, retrieving the collection of output, converting the output, and sending the converted output to at least one user.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, the topology of the distributed quantum computer is provided by at least one of user input or an automatic network discovery protocol.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, remapping further comprises determining quality of available quantum resources within the distributed quantum computer.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, the plurality or collection of quantum programs may be transpiled to remove quantum operations requiring greater than 2 qubits.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, the at least one memory and the computer program code are further configured to, with the processor, cause the system to ensure all of the quantum operations are executable, wherein ensuring all of the quantum operations are executable comprises one or more of the following: remapping any non-executable gates to an equivalent gate that is executable, or signaling an execution error.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, resource allocation is further comprised of finding qubits in the system to run the distributed quantum algorithm, determining quality of quantum resources, ordering qubits by their quality, designating qubits for computation and reserving the remaining qubits for communication, and determining an allocation of quantum resources based at least in part on the designation for computation and the remapped distributed set of quantum programs.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, resource allocation is further comprised of a machine learning system trained to optimize the allocation of qubits.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, resource allocation is optimized by a reinforcement learning system, the reinforcement learning system is comprised of receiving information on quality of execution and current state of the quantum resources, training a machine learning network to optimize entanglement and communication resources, generating a resource allocation policy based at least in part on the optimization of entanglement and communication resources, and setting a value for feedback based on the performance of the policy to use for the next iteration.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, additional optimization can be performed to minimize noise by choosing the best qubits based on at least one of coherence time or gate fidelity for that qubit.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, the at least one memory and the computer program code are further configured to, with the processor, cause the system to perform a synchronization of nodes involved in the computation and calibrate the nodes of the distributed quantum computer.
Optionally, and in accordance with any of the above aspects of the exemplary embodiment, setting instruction durations based in part on the physical aspects of entanglement generation.
The following description and drawings set forth certain illustrative implementations of the disclosure in detail, which are indicative of several exemplary ways in which the various principles of the disclosure may be carried out. The illustrative examples, however, are not exhaustive of the many possible embodiments of the disclosure. Other objects, advantages and novel features of the disclosure will be set forth in the following detailed description of the disclosure when considered in conjunction with the drawings, in which:
According to the presently described embodiments, in at least one form, a high-level compiler takes a user algorithm (or programs) in the form of a monolithic quantum circuit and a designated resource allocation. It then performs data and communication qubit allocation with consideration of qubit quality and communication resource minimization, placing qubits accordingly. Once allocated, the monolithic input is remapped to the distributed allocation, filling any missing communication steps as needed. The compiler inserts entanglement generation steps when needed such that entanglement resources are available with high probability when they are needed. Next, the high-level compiler injects any error correction steps. Then, a low-level compiler takes the equivalent circuit and converts it to machine level instructions with time-stamped operations (or at least grouped operations). These instructions are generated for each quantum processor (also referred to as a node) of a quantum computer used in the algorithm execution. These individual instructions are then passed to each respective node in the distributed quantum computer and executed.
Referring now to the drawings wherein the showings are for purposes of illustrating the exemplary embodiments only and not for purposes of limiting the claimed subject matter,
The distributed compiler orchestrator (DCO) or orchestrator 102 illustrated in
In general, the system 100 operates based on the input of a collection of user programs, or circuits, 106 to, for example, an input interface 112 and instructions provided in or to the system 100 for distributed quantum computing execution. This system is also configured to optionally monitor and control a DQC or distributed quantum computer 200.
The DCO or orchestrator 102 includes one or more communication interfaces (I/O), such as network interfaces 112 for communicating with external devices, such as a distributed quantum controller 104, a DQC or distributed quantum computer 200, and any external systems or sources such as those including datasets (e.g., artificial intelligence datasets) and other resources (e.g., software routines, processors and/or memories) that may be used by the system 100 to implement the presently described embodiments. The various hardware components such as processor 110, and memory 108 may all be connected by a bus 114.
With continued reference to
Software modules that may be stored in, for example, memory 108 are intended to encompass any collection or set of instructions executable by the system 100 so as to configure the system to perform the task that is the intent of the software. The term “software” as used herein is intended to encompass such instructions stored in storage medium such as RAM, a hard disk, optical disk, or so forth, and is also intended to encompass so-called “firmware” that is software stored on a ROM or so forth. Such software may be organized in various ways, and may include software components organized as libraries, Internet-based programs stored on a remote server or so forth, source code, interpretive code, object code, directly executable code, and so forth. It is contemplated that the software may invoke system-level code or calls to other software residing on the server or other location to perform certain functions.
The communication interfaces 112 may include, for example, a modem, a router, a cable, and/or Ethernet port, etc.
The distributed quantum controller or DQOS 104 illustrated in
In general, the system 100 operates based on the input of a collection of user programs, or circuits, 106 to, for example, an input interface 112 and instructions provided in or to the system 100 for distributed quantum computing execution. This system is also configured to optionally monitor and control a DQC or distributed quantum computer 200.
The distributed quantum controller 104 includes one or more communication interfaces (I/O), such as network interfaces 120 for communicating with external devices, such as a DQO 102, a DQC or distributed quantum computer 200, and any external systems or sources such as those including datasets (e.g., artificial intelligence datasets) and other resources (e.g., software routines, processors and/or memories) that may be used by the system 100 to implement the presently described embodiments. The various hardware components such as processor 118, and memory 116 may all be connected by a bus 124.
With continued reference to
Software modules that may be stored in, for example, memory 116 are intended to encompass any collection or set of instructions executable by the system 100 so as to configure the system to perform the task that is the intent of the software. The term “software” as used herein is intended to encompass such instructions stored in storage medium such as RAM, a hard disk, optical disk, or so forth, and is also intended to encompass so-called “firmware” that is software stored on a ROM or so forth. Such software may be organized in various ways, and may include software components organized as libraries, Internet-based programs stored on a remote server or so forth, source code, interpretive code, object code, directly executable code, and so forth. It is contemplated that the software may invoke system-level code or calls to other software residing on the server or other location to perform certain functions.
The communication interfaces 120 may include, for example, a modem, a router, a cable, and/or Ethernet port, etc.
As will be appreciated, while the DCO or orchestrator 102 and distributed quantum controller or DQOS 104 are illustrated by way of example, the system 100 may be hosted by fewer or more linked computing devices. Each computing device may include, for example, a server computer, desktop, laptop, or tablet computer, smart phone device or any other computing device capable of implementing the method described herein.
As mentioned, the DCO or orchestrator 102 receives a plurality or collection of quantum programs 106 from one or more users. For example, embodiments are contemplated where the quantum programs 106 can be a quantum circuit or a series of logical quantum instructions (e.g., in the form of a programming language designed for describing quantum circuits) within the DCO or orchestrator 102. In other embodiments, quantum circuits and instructions may be represented by visual programming.
Of course, as noted herein, according to the presently described embodiments, in at least one form, the DCO or orchestrator 102 establishes a communication connection with at least one DQC or distributed quantum computer 200. Through the communication connection the DCO or orchestrator 102 discovers all quantum processors or nodes of the DQC or distributed quantum computer 200. Further the DCO or orchestrator 102 may configure the processor 110 to discover the topology of the DQC or distributed quantum computer 200. For example, the physical arrangement, quantity of qubits, quality of qubits, and type of quantum processors or nodes available in the DQC or distributed quantum computer 200 are recognized in the discovery routine.
It should be appreciated that, in at least one implementation of the presently described embodiments, the system 100 utilizes the DCO or orchestrator 102 to perform high-level compiling which includes instructions to remap the plurality of quantum program by processing the plurality of quantum program to create a set of distributed quantum programs. It should also be appreciated that the remapping process could require a variety of different conversion processes. In one form, the remapping of the plurality of quantum programs to create a set of distributed quantum programs involves converting the plurality of quantum programs to a layer array before remapping to a set of distributed quantum programs. In another form, the plurality of quantum programs may need to be transpiled before converting the plurality of quantum programs to a layer array. Such a form may also involve labeling qubits to minimize non-local gates based, at least in part, on the converted layer array and network topology.
The high-level compiling includes, for example, minimizing the use of communication where possible. Such high-level compiling will, in at least one form, also include a resource allocation in addition to the remapping. This resource allocation may involve determining an allocation of available quantum resources needed for the set of quantum programs. For example, in at least one implementation of the presently described embodiments, the DCO or orchestrator 102 to performing high-level compiling will determine the quality of quantum resources, order qubits by quality, order operational qubits, assign qubits to each quantum processor, and determine the allocation of quantum resources. In some implementations, determining the allocation of quantum resources may involve allocating data and communication qubits and preparing an allocation table.
Thus, in accordance with the presently described embodiments, high-level compiling creates a set of distributed quantum programs that is further processed by low-level compiling to convert the remapped set of quantum programs into at least one set of machine level instructions for each quantum processor or node in the distributed quantum network. The instructions are defined as the instruction to a given node and the time in which it needs to be executed. To produce the instruction set, the system can use information, such as gate times and communication latency, to produce accurate timing. For example, the DCO or orchestrator 102 may perform low-level compiling to generate instructions and an instruction distribution. In another embodiment, the DCO or orchestrator 102 may perform low-level compiling to generate instructions that involve utilizing the allocation table created by performing high-level compiling and the DQC or distributed quantum computer 200 structure to convert the allocation to precise mid-level instructions. In some embodiments, instructions are prepared as a table before converting to precise mid-level instructions. In other embodiments, offsetting the next operations may be performed after preparing the instructions as a table.
According to the presently described embodiments, low-level compiling also involves generating an instruction distribution from the precise mid-level instructions. For example, generating an instruction distribution would include extracting the precise mid-level instructions for each quantum processor or node from the table, converting the instructions into individual tables, remapping high-level instructions, and packaging tables for distribution. In some embodiments, sending at least one set of machine level instructions to the distributed quantum controller 104 is contemplated.
It should also be appreciated that the distributed quantum controller 104 can cause the DQC or distributed quantum computer 200 to execute quantum instructions. According to the presently described embodiments, in one form, the distributed quantum controller 104 receives at least one set of machine level instructions from the DCO or orchestrator 102. In one example, the distributed quantum controller 104 calibrates the nodes of the distributed quantum computer, ensures synchronization between the nodes, distributes at least one set of machine instructions to each of the nodes, wherein each node executes instructions based at least in part on the allocation determined by the DCO or orchestrator 102. Further, the distributed quantum controller 104 monitors the performance of the distributed quantum network, receives output from each node, collects all output from the executed instructions, and sends the collection of outputs to the DCO or orchestrator 102. In the environment where the distributed quantum controller 104 sends the collection of outputs to the DCO or orchestrator 102, the DCO or orchestrator 102 converts the outputs and sends the converted outputs to at least one user.
Turning to
In another embodiment, DQC or distributed quantum computer 200 can be located physically in different locations (such that the network layer is needed). It should be appreciated that each QPU 204 includes a classical network interface 208 that allows the QPU to receive and transmit messages to the network. The messages are sent to the distributed quantum controller 104 for processing.
The distributed quantum controller 104 processes the messages and sends instructions to a classical controller 206 (e.g., a field programmable gate array or FPGA). The classical controller triggers the quantum network interface (QNI) 216 to generate, transmit, or receive entanglement resources to or from other QPUs. The network interface 216 and the quantum hardware 210 interact such that the QNI 216 can transduce flying qubits to matter qubits if needed, and store them in the quantum hardware in, for example, a quantum memory. Flying qubits can also directly interact with matter qubits without storage, still an interaction initiated by interacting the network interface with the quantum hardware. It should be appreciated that the components can all send messages to each other, layer by layer, for status polling, algorithm outputs, or other reasons.
It should be appreciated that the quantum hardware 210 may include quantum memory 212, at least one quantum transistor 214, and a source of quantum entanglement 218. In some embodiments the quantum entanglement source is local. The quantum entanglement source may communicate directly with the quantum hardware 210, through the quantum network interface 216 or a combination of both. In other embodiments, the quantum entanglement source is centrally located.
It should be appreciated that the quantum entanglement source 218 generates the entangled qubits needed for the algorithm. For example, this can be a spontaneous parametric down conversion or a quantum dot-based quantum entanglement source. It should also be appreciated that spontaneous parametric down conversion and quantum dot-based quantum entanglement sources are merely example quantum entanglement sources. Other entanglement sources may be implemented.
In other embodiments, the quantum hardware would be anything that is capable of representing a quantum bit or qubit, such as superconducting qubits, ion-based qubits, photonic based qubits, or molecule-based qubits. In other embodiments quantum hardware can also be a quantum transducer, or quantum memories to store qubits, such as entangled qubits, or anything that can convert signal frequencies for quantum processing.
Now, with reference to
In
The method 300 further includes the step of high-level compiling or HLC (at 302). In one form, the DCO (or orchestrator) 102 receives (at 306) a plurality or collection of quantum programs 106.
High-level compiling (or HLC) (at 302) generally ensures that all of the quantum operations are executable. For this, the high-level compiling (or HLC) (at 302) first remaps (at 304) any non-executable gates to an equivalent gate that is executable (or signals an execution error). In some embodiments, the plurality or collection of quantum programs 106 may be transpiled (at 312) to remove quantum operations that require greater than 2 qubits. It will be appreciated that this method can remap quantum programs 106 comprising more than 2-qubit gates.
In order to find any quantum operations that need to be remapped, the circuit can be converted into a layer structure (at 316). Using the layer array and the network topology received (at 314) by the DCO (or orchestrator) 102, the next step of remapping involves labeling (at 318) qubits to minimize non-local gates. For each layer, any 2-qubit gate that has each qubit physically separated between devices, the operation needs to be remapped to an equivalent distributed version (at 320). This process involves converting the monolithic circuit to a distributed one.
In other embodiments, the circuit can be remapped to a hyper-graph structure to perform optimization for minimizing communication resources. This can be done before or after step 318. From here, from the output of the optimizers, the allocated qubit resources are labeled (at 318) such that communication resources and number of operations are minimized.
With reference to
With reference to
With reference to
Once generated (at 362), the machine instructions are packaged up and sent
Next, the DCO or orchestrator 102 sends all of the individual instructions to the distributed quantum controller 104 (also referred to as the distributed quantum operating system or DQOS).
In some embodiments, with reference to
With reference to
The method illustrated in
In some embodiments, the optimization method is configured to produce as few communication steps as possible, trading off circuit width and depth for higher algorithm fidelity and fewer communication steps. In other embodiments, there may be a second layer of transpiling in a lower layer that can modify the circuit according to the available resources at that time of execution in case the resources are not free at that time (due to scheduling).
The system is configured to allow the number of retries to vary during error correction, because entanglement generation is not always deterministic. The compiler allows for flexibility here and can be set to allow for more retries at the cost of algorithm fidelity.
It should be appreciated that once instructions are sent (at 382) to the individual QPUs 204 (or before, order is not important here), they perform a synchronization (at 380) through the centralized controller 206 for the QPUs 204 involved in the computation. A round of calibration (at 378) can also be done at this stage to ensure the devices are working correctly. In some embodiments, the quantum processors or QPUs 204 can potentially receive multiple quantum program instructions at the same time but with non-overlapping resources. Some operations could therefore have the same timestamp.
In another embodiment the method includes the distributed quantum controller or DQOS 104 sending the instructions to a classical controller 206 (e.g., a field programmable gate array or FPGA). The distributed quantum controller or DQOS 104 configures a classical controller to trigger (at 384) the quantum network interface (QNI) 216 to generate, transmit, or receive entanglement resources to or from other QPUs 204. The network interface 216 and the quantum hardware 210 interact such that the QNI 216 can transduce flying qubits to matter qubits if needed, and store them in the quantum hardware in, for example, a quantum memory. Flying qubits can also directly interact with matter qubits without storage, still an interaction initiated by interacting the network interface with the quantum hardware. It should be appreciated that the components can all send messages to each other, layer by layer, for status polling, algorithm outputs, or other reasons.
It will also be appreciated that, in some embodiments, after the distributed quantum computer 200 receives (at 382) the instructions, the distributed quantum controller or DQOS 104 configures a classical controller to cause each QPU 204 to acknowledge that they have received instructions. Then each QPU 204 is triggered to start its clock, process individual instructions, and open the quantum network interface. Outputs are gathered and processed. Then the results are sent back to the DCO or orchestrator 102.
In some embodiments, a machine learning system is trained to optimize the allocation of qubits during resource allocations. For example, in some embodiments this may be a reinforcement learning network. With reference to
The exemplary system may thus include one or more computing devices, each of which may be a PC, such as a desktop, a laptop, server computer, microprocessor, cellular telephone, tablet computer, combination thereof, or other computing device capable of executing instructions for performing the exemplary method.
The methods described herein and illustrated in the figures may be at least partially implemented in a computer program product or products that may be executed on a computer or computers. The computer program product(s) may comprise a non-transitory computer-readable recording medium on which a control program is recorded (stored), such as a disk, hard drive, or the like. Common forms of non-transitory computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tape, or any other magnetic storage medium, CD-ROM, DVD, or any other optical medium, a RAM, a PROM, an EPROM, a FLASH-EPROM, or other memory chip or cartridge, or any other non-transitory medium from which a computer can read and use.
Alternatively, the methods may be implemented in transitory media, such as a transmittable carrier wave in which the control program is embodied as a data signal using transmission media, such as acoustic or light waves, such as those generated during radio wave and infrared data communications, and the like.
The exemplary methods may be implemented on one or more general purpose computers, special purpose computers, a programmed microprocessor or microcontroller and peripheral integrated circuit elements, an ASIC or other integrated circuit, a digital signal processor, a hardwired electronic or logic circuit such as a discrete element circuit, a programmable logic device such as a PLD, PLA, FPGA, Graphical card CPU (GPU), or PAL, or the like.
As will be appreciated, the steps of the methods need not all proceed in the order illustrated and fewer, more, or different steps may be performed.
The exemplary embodiment has been described with reference to the preferred embodiments. Obviously, modifications and alterations will occur to others upon reading and understanding the preceding detailed description. It is intended that the exemplary embodiment be construed as including all such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
