Method and system for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction

Abstract

The present invention discloses a method and system for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction, which relates to the field of superconducting quantum computing measurement and control technology. The method comprises: a configuration execution circuit decodes a prediction codeword generated on a chip to obtain a parameter group index, the prediction codeword is generated by a pseudo-random prediction engine, the pseudo-random prediction engine uses a seed stack to receive seed data input from an external source, and uses a real-time prediction module to execute a pseudo-random algorithm in real time. The predicted pseudo-random value is used as the prediction codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit according to a set rule; the parameter group index is used to perform a parallel search in a parameter lookup table of the configuration execution circuit, and all parameter data values ​​corresponding to the parameter group index can be output in one clock cycle, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data; the method and system for dynamic parameter configuration are intended to achieve efficient and low-latency quantum bit measurement and control waveform configuration.

Description

Technical Field

[0001] The present invention relates to the field of superconducting quantum computing measurement and control technology, and in particular to a method and system for dynamically configuring quantum bit measurement and control waveform parameters based on pseudo-random prediction. Background Art

[0002] The cross-entropy benchmark (XEB) of qubits is one of the important methods in the field of quantum computing to verify the performance of quantum circuits and quantum supremacy. Its core idea is to measure the similarity between the probability distribution output by the quantum computing system and the theoretical distribution by calculating the cross-entropy benchmark fidelity, and to evaluate whether the computing power of the quantum system exceeds the simulation capability of classical computers. XEB is a typical application example of random circuit sampling (RCS) technology. This type of test requires the quantum computing system to execute a pseudo-random quantum logic gate circuit; in superconducting quantum computing systems, quantum logic gates are achieved by manipulating the state of qubits through microwave measurement and control signals with configurable characteristic information such as shape, length, and tuning increment generated by arbitrary waveform generation devices. Although the traditional arbitrary waveform generator configuration method can meet basic configuration requirements, it still has the following limitations when dealing with the measurement and control waveform generation of complex quantum circuits such as random circuit sampling:

[0003] 1) Insufficient flexibility: Traditional arbitrary waveform generation technology relies on waveform data storage. The overhead of hardware resources limits the dynamic adjustment capability of the waveform, making it difficult to quickly respond to the diverse requirements of complex quantum circuits for waveform shape, length, tuning increment, phase reference, scaling factor, and bias amplitude.

[0004] 2) Limited feature expression: Traditional methods lack the extraction of key waveform features, have limited distinguishability and expression of waveform features, and have redundancy in data configuration and storage, which limits the possibility of dynamic generation of measurement and control waveforms.

[0005] 3) High time cost: The information transmission process of traditional waveform configuration methods is complex and has a low degree of parallelization. The calculation of waveform parameters usually requires the cooperation of the host computer, which has a high configuration time cost and is difficult to adapt to the high real-time waveform configuration requirements of complex quantum algorithms. Summary of the Invention

[0006] Based on the technical problems existing in the background technology, the present invention is based on a method and system for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction, aiming to achieve efficient and low-latency quantum bit measurement and control waveform configuration.

[0007] The method for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction proposed in the present invention includes:

[0008] The configuration execution circuit decodes the predicted codeword generated on the chip to obtain a parameter group index. The predicted codeword is generated by a pseudo-random prediction engine. The pseudo-random prediction engine uses a seed stack to receive external seed data, and uses a real-time prediction module to execute a pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit;

[0009] By using the parameter group index to perform parallel search in the parameter lookup table, all parameter data values ​​corresponding to the parameter group index can be output in one clock cycle, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data value. Each sub-area in the parameter lookup table stores the preloaded waveform parameters.

[0010] Furthermore, when the configuration execution circuit receives the predicted codeword generated by the pseudo-random prediction engine and parses it into a parameter group index, the predicted codeword is parsed using the codeword parsing module, specifically:

[0011] The predicted codeword is parsed into the index of each parameter through the bit field, and the index is used as the parameter group index. In the bit field parsing, the data of different bit fields of the codeword represent the index of different parameter values.

[0012] The parsed parameter group index is cached and sent to the parameter lookup table.

[0013] Furthermore, each sub-region in the parameter lookup table corresponds to a control parameter component, and the parameter value in each sub-region is configured by the host computer and pre-loaded into the parameter lookup table through the bus.

[0014] Furthermore, the pseudo-random prediction engine utilizes a seed stack to receive seed data input from an external source, and utilizes a real-time prediction module to execute a pseudo-random algorithm in real time based on the seed and configured generation coefficients. The real-time prediction module is specifically configured to:

[0015] When the quantum circuit is started, the seed stack receives and caches the seed data input from the outside. The seed data is the initial value of the pseudo-random number generation algorithm.

[0016] The seed stack transfers the cached seed data to the shift register, uses the generation coefficient in the generation coefficient register to filter the shift register value, and then updates the state value of the shift register through an exclusive-OR operation. The updated register state value is input into the index end of the logic gate lookup table.

[0017] The codewords read from the logic gate lookup table are fed back to the codeword cache, which caches the compressed codewords that represent the parameter indices corresponding to the sub-regions in the parameter lookup table.

[0018] The codeword parsing module receives a synchronous trigger signal of the waveform and reads the codeword cache under the action of random prediction enabling, thereby obtaining the predicted codeword.

[0019] Furthermore, when receiving the synchronous trigger signal of the waveform, the codeword parsing module reads the codeword cache under the action of random prediction enable to obtain the predicted codeword. If the random prediction enable fails, the codeword parsing module reads the codeword generated by the on-chip microcontroller as the predicted codeword.

[0020] Furthermore, the shift register value is screened using the generation coefficient in the generation coefficient register, and then the state value of the shift register is updated through an XOR operation, and the updated register state value is input into the index end of the logic gate lookup table, specifically:

[0021] The generation coefficient register selects the value of the corresponding bit of the shift register state value according to the bit with the value 1 in each group of generation coefficients;

[0022] Each group of filtered bits is passed through an XOR operator to generate feedback bits, and the feedback bits are sequentially filled into the leftmost bit of the shift register, and the other bits are shifted right in sequence;

[0023] The high m bits of the shift register are input as pseudo-random prediction results to the index end of the logic gate lookup table, where m is an integer. The logic gate lookup table is the private memory of the pseudo-random prediction engine and is used to store various code words.

[0024] Furthermore, the generation coefficient register and logic gate lookup table in the pseudo-random prediction engine are configured by a host computer.

[0025] Furthermore, each time the pseudo-random prediction engine starts the quantum circuit, the seed stack receives seed data input from the outside, allowing multiple seeds to be input in order;

[0026] In each execution of the quantum circuit, it is only necessary to receive external input seed data once, and predict the configuration parameters of the quantum logic gate in real time based on the pseudo-random algorithm.

[0027] Furthermore, the parameter group index includes a scaling factor index, a tuning increment index, a bias amplitude index, a phase reference index and an envelope information index;

[0028] The parameter lookup table is provided with a tuning increment sub-area, a scaling factor sub-area, a bias amplitude sub-area, a phase reference sub-area, and an envelope information sub-area, and each sub-area is provided with a corresponding lookup table.

[0029] A quantum bit waveform parameter dynamic configuration system based on pseudo-random prediction, including a configuration execution circuit and a pseudo-random prediction engine;

[0030] The pseudo-random prediction engine uses a seed stack to receive seed data input from the outside, and uses a real-time prediction module to execute a pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit and decodes it to obtain a parameter group index.

[0031] The configuration execution circuit uses the parameter group index to perform parallel searches in the parameter lookup table, and outputs all parameter data values ​​corresponding to the parameter group index in one clock cycle, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data value. Each sub-area in the parameter lookup table stores preloaded waveform parameters.

[0032] The advantages of the method and system for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction provided by the present invention are: aiming to achieve efficient and low-latency quantum bit measurement and control waveform configuration, introducing a codeword parsing module controlled by multi-bit codewords and a parameter lookup table using structured lookup table technology on the basis of traditional waveform configuration technology, optimizing the waveform configuration process through codeword programming and lookup table technology, reducing the amount of data and transmission delay of information distributed from the front-end control system to the back-end waveform generation system, and improving the real-time generation capability of dynamic waveforms; at the same time, a pre-pseudo-random prediction engine is integrated into the configuration hardware to realize real-time prediction of random quantum logic gates to meet the high-performance measurement and control requirements of complex quantum computing circuits in quantum computing experiments such as the cross entropy benchmark test (XEB). BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a structural schematic diagram of the present invention;

[0034] Figure 2 Schematic diagram of the waveform configuration circuit structure corresponding to the dynamic configuration method of quantum bit measurement and control waveform parameters;

[0035] Figure 3 This is a schematic diagram of the parsing process of the encoding parsing module;

[0036] Figure 4 Schematic diagram of data retrieval for parameter lookup table;

[0037] Figure 5 This is a structural block diagram of the pseudo-random prediction engine;

[0038] Figure 6 Schematic diagram comparing the dynamic parameter configuration method of this embodiment with the existing classic waveform configuration method. DETAILED DESCRIPTION

[0039] The technical solutions of the present invention are described in detail below through specific embodiments. Numerous specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art may make similar modifications without departing from the scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0040] like Figures 1 to 6 As shown, the method for dynamically configuring the quantum bit waveform parameters of the pseudo-random prediction engine proposed in the present invention includes the following steps:

[0041] Step 1: Configure the execution circuit to decode the predicted codeword generated on the chip to obtain a parameter group index. The predicted codeword is generated by a pseudo-random prediction engine. The pseudo-random prediction engine uses a seed stack to receive external seed data, and uses a real-time prediction module to execute a pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit;

[0042] Step 2: The configuration execution circuit uses the parameter group index to perform parallel search in the parameter lookup table of the configuration execution circuit. All parameter data values ​​corresponding to the parameter group index can be output in one clock cycle, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data value. Each sub-area in the parameter lookup table stores the preloaded waveform parameters.

[0043] This embodiment aims to achieve efficient and low-latency quantum bit measurement and control waveform configuration. Based on traditional waveform configuration technology, it introduces a codeword parsing module controlled by multi-bit codewords and a parameter lookup table using structured lookup table technology. Through codeword programming and lookup table technology, the waveform configuration process is optimized, the data volume and transmission delay of information distributed from the front-end control system to the back-end waveform generation system are reduced, and the real-time generation capability of dynamic waveforms is improved. At the same time, a pre-pseudo-random prediction engine is integrated into the configuration hardware to achieve real-time prediction of random quantum logic gates to meet the high-performance measurement and control requirements of complex quantum computing circuits in quantum computing experiments such as the cross-entropy benchmark test (XEB).

[0044] Specifically, if Figure 2As shown, this embodiment achieves highly time-efficient and low-latency configuration parameter output through the close integration of the codeword parsing module and the parameter lookup table, providing support for efficient, real-time generation of qubit measurement and control waveforms. Furthermore, a hardware pseudo-random prediction engine based on pseudo-random prediction technology enables quantum logic gate prediction during quantum circuit execution, significantly improving the flexibility and real-time performance of measurement and control waveform configuration. This guarantees the execution of large-scale, complex quantum computing circuits and is particularly suitable for providing real-time and efficient hardware support for measurement and control waveform configuration in random circuit sampling experiments such as the Cross Entropy Benchmark (XEB).

[0045] The implementation of the above-mentioned dynamic configuration method of quantum bit waveform parameters relies on the dynamic configuration system of quantum bit waveform parameters, which includes a pseudo-random prediction engine and a configuration execution circuit. The pseudo-random prediction engine includes two key modules: a seed stack and a real-time prediction module, and the configuration execution circuit includes two key modules: a codeword parsing module and a parameter lookup table.

[0046] Among them, the seed stack can receive "seed" data input from the outside, and the seed data can be understood as the initial value of the pseudo-random number generator, that is, the "seed" in the pseudo-random number generation algorithm is an initial value. It is the internal state set when starting the pseudo-random number generation, which determines the starting point of the pseudo-random number sequence. The same seed applied to the same generator will generate the same pseudo-random number sequence, which is often used in scenarios where reproducible results are required; the real-time prediction module executes the pseudo-random algorithm in real time according to the seed and the configured generation coefficient, and the predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next logic gate operation in the quantum circuit; the codeword parsing module is used to receive and parse the predicted codeword that conforms to the coding rules in real time, and provide the parameter group index of the parameter lookup table; the parameter lookup table stores pre-loaded waveform feature information data in different regions (that is, corresponding to the various parameters parsed by the codeword parsing module), and is responsible for quickly and in parallel retrieving and outputting all parameter data values ​​according to the parameter group index, such as the waveform envelope information, bias amplitude, scaling factor and tuning increment, phase reference, etc. This design achieves highly flexible and real-time waveform configuration through the combination of multiple technologies such as feature encoding, lookup table technology, and pseudo-random prediction technology. It is highly compatible with the front-end control system and the back-end waveform generation system, providing a guarantee for the efficient and real-time generation of measurement and control waveforms for complex algorithm superconducting quantum computing.

[0047] In one embodiment, the configuration execution circuit includes a codeword parsing module specifically as follows:

[0048] The key to waveform configuration in this embodiment lies in the codeword representation of key waveform features. The index encoding rules shown in Table 1 are used to implement the combination of multiple waveform feature parameters. The data in different bit fields of the codeword represent the index values ​​of different parameters. The conditional output indicator bit and the waveform output hold indicator bit serve as indicators for special waveform generation modes. The scaling factor index, tuning increment index, offset amplitude index, phase reference index, and envelope information index are parameter lookup table indices for general waveform information in the parameter lookup table. Reserved bits are designed for scalability.

[0049] Table 1 Encoding rules for waveform feature information

[0050]

[0051] In superconducting quantum computing, quantum logic gates executed on qubits are implemented using measurement and control microwave signals emitted by arbitrary waveform generators. Parameters such as the phase value and time integral of the microwave waveform determine the evolution direction and degree of the qubit state. Classic waveform configuration processes are relatively inflexible, requiring host computer software to calculate the waveform's sampled data values ​​at each digital clock edge and serially configure them to the waveform generation circuit via a communication interface. This configuration data has low information density, consumes hardware storage resources, and requires long memory access times during waveform generation. This embodiment proposes a dynamic waveform configuration technique based on parameterized representation of key features. Specifically, a measurement and control waveform (calculated from the parameters mapped by the prediction codewords described in this embodiment) is decomposed into key feature information, including tuning increment, phase reference, scaling factor, offset amplitude, and envelope information. The amount of data required to configure each key feature is significantly less than the number of waveform sampling points. By expressing this information in codewords and dynamically indexing it, a wide variety of flexible feature representations can be achieved.

[0052] In the circuit structure involved in the codeword parsing module of this embodiment, the function of the codeword parser is to parse the codeword into the index of each waveform feature information (i.e., parameter) according to the encoding rules shown in Table 1. The workflow of codeword parsing is to decompress, cache and output the codeword. Figure 3 As shown, the codeword parsing module receives the predicted codeword and the synchronization trigger signal, parses the predicted codeword according to a predetermined format, and obtains a parameter group index, which includes a scaling factor index, a tuning increment index, an offset amplitude index, a phase reference index, and an envelope information index.

[0053] In one embodiment, the parameter lookup table is specifically:

[0054] The circuit structure for realizing a fast response to the parameter group index obtained by the codeword parsing module is a set of structured lookup tables, namely parameter lookup tables, whose circuit structure is as follows: Figure 4 shown.

[0055] The structure of the parameter lookup table is closely linked to the codeword encoding rules of the codeword parsing module. Each encoded waveform feature in the codeword parsing module has an independent sub-region of storage space in the parameter lookup table. The contents of these storage spaces can be written to the system bus and retrieved and read out using parameter group indices (feature parameter index values) from the codeword decoding module. Before the quantum circuit executes, the host computer calculates the values ​​of the possible parameters (scaling factor, phase reference, offset amplitude, tuning increment, and envelope information) based on the circuit characteristics and preloads them into the sub-region corresponding to each parameter in the parameter lookup table via the bus. After the quantum circuit begins execution, the codeword parsing module continuously parses the codeword and inputs the decompressed parameter group index into the parameter lookup table. Each parameter index value in the parameter group index is connected to the index port of its corresponding parameter lookup table, performing parameter retrieval in parallel, and all parameter data values ​​are output in parallel within a single clock cycle. The index retrieval of each sub-area in the parameter lookup table is executed in parallel in the hardware. The parameter lookup table is set with the tuning increment sub-area, scaling factor sub-area, offset amplitude sub-area, phase reference sub-area and envelope information sub-area. Each sub-area has a corresponding lookup table. See Figure 2 That is, the tuning increment sub-area is set with a tuning increment lookup table, the scaling factor sub-area is set with a scaling factor lookup table, the offset amplitude sub-area is set with a offset amplitude lookup table, the phase reference sub-area is set with a phase reference lookup table, and the envelope information sub-area is set with an envelope information lookup table. The output values ​​of the tuning increment sub-area and the phase reference sub-area are the tuning increment and the phase reference, respectively, which are used to control the tuning increment and phase reference of the measurement and control waveform carrier; the output values ​​of the scaling factor sub-area and the offset amplitude sub-area are the scaling degree of the measurement and control waveform and the size of its offset amplitude; the output values ​​of the envelope information sub-area include two pieces of information: the envelope initial address and the number of envelope points. Combined with the envelope data stored in the envelope memory of the waveform generator, the envelope data of the measurement and control waveform can be quickly located.

[0056] Structured lookup table technology utilizes local memory access optimization for parallel lookups. Data reads do not occupy the bus and are therefore not limited by bus memory access bandwidth. Parameter lookup latency can be controlled within a single clock cycle. In contrast, if parameters are stored in hardware global memory, sequential parameter lookups typically require dozens to dozens of clock cycles due to interface bandwidth limitations, communication protocol delays, and other factors. However, after pre-storing data once, the parameter lookup table of this embodiment exhibits a superlinear product-combination relationship between the number of configurable waveform types and the amount of pre-stored information. This allows for fast retrieval speeds within each sub-region of the parameter lookup table and a high degree of parallelism, comprehensively meeting the high real-time and flexibility requirements of quantum computing measurement and control systems.

[0057] In one embodiment, the pseudo-random prediction engine is a hardware acceleration module designed for random circuit sampling tests such as XEB experiments, specifically:

[0058] During the XEB experiment, a large number of random quantum logic gates must be continuously applied to the quantum bits, i.e., random measurement and control microwave waveforms are emitted to the quantum bits. During the experiment, a classical computer simulates this random quantum circuit and compares the simulated data with the actual quantum bit readout data to calibrate the performance of the quantum computing system. Traditional measurement and control waveform configuration does not support pseudo-random prediction. Therefore, to ensure consistency between the simulated and measured circuits, the experiment requires the host to generate a series of random quantum logic gates to be executed using a pseudo-random algorithm, calculate all the corresponding measurement and control waveform data, and load them into the waveform generation circuit. The number of logic gates in the quantum circuit is determined by the experimental objectives and requirements, and each logic gate is controlled by a measurement and control waveform. Some experimental circuits contain dozens to hundreds of logic gate operations, requiring the waveform generator to load dozens to hundreds of measurement and control waveform sampling data. This poses significant challenges to experimental configuration time and waveform generator storage capacity. This embodiment, however, fully utilizes a flexible and efficient waveform configuration method that combines a codeword parsing module with a parameter lookup table, and further develops a hardware-accelerated pseudo-random prediction engine to meet the needs of such experiments.

[0059] The core of the pseudo-random prediction engine consists of a seed stack and a real-time prediction module. It also features a set of coefficient registers and a logic gate lookup table that can be configured by the host computer. The pseudo-random prediction engine uses the seed value and the coefficients in the coefficient registers to execute a linear feedback shift register (LFSR)-based pseudo-random number generation algorithm. The specific steps are as follows (a1) to (a5):

[0060] (a1) Shift register initialization: Before the random circuit sampling experiment begins, the random prediction enable is pulled high, and the seed stack receives one or a series of seed data, which are the initial values ​​of the pseudo-random number generation algorithm. The first seed data is cached and copied to the shift register of the real-time prediction module.

[0061] (a2) Feedback shift calculation: The seed stack transfers the cached seed data to the shift register, uses the generation coefficient in the generation coefficient register to filter the shift register value, and then updates the state value of the shift register through an exclusive-OR operation. The updated register state value is input to the index end of the logic gate lookup table; the upper m bits of the shift register are input to the index end of the logic gate lookup table as the pseudo-random prediction result, where m is an integer.

[0062] That is, the register state value of the shift register is filtered by the generation coefficients in each generation coefficient register using a logical AND operation, and a feedback bit is generated by an exclusive OR operator. The exclusive OR operator is a group of operators that perform a bitwise exclusive OR operation on each filtered shift register state value; the shift register is continuously shifted to the right and the feedback bits are used to fill the empty bits on the left end caused by the shift; a portion of the data at the right end of the shift register is input to the index end of the logic gate lookup table, which is the private memory of the pseudo-random prediction engine and is used to store various different codewords;

[0063] It is understandable that the value of the shift register is equal to the seed value only in the first iteration of the pseudo-random number generation algorithm, and the subsequent feedback bit filling and data shifting operations will change the register value.

[0064] (a3) Sequence output: The upper m bits of the shift register (e.g., m = 16) are entered as pseudo-random prediction results into the index end of the logic gate lookup table, where m is an integer. The logic gate lookup table is a private memory of the pseudo-random prediction engine, used to store various codewords. The pseudo-random prediction results are looked up from the logic gate lookup table and the read codeword is fed back to the codeword cache. The codeword cache caches the compressed codewords that represent the parameter indices corresponding to the sub-regions in the parameter lookup table.

[0065] (a4) When receiving the waveform's synchronous trigger signal, the codeword parsing module reads the codeword cache (i.e., the predicted codeword generated by the pseudo-random prediction engine) under the action of random prediction enable, rather than receiving the codeword from the on-chip microcontroller. Since the codeword is the determining factor of the waveform configuration parameters, and the measurement and control waveform parameters determine the type of quantum logic gate, the above prediction process realizes the prediction of a future random logic gate.

[0066] It should be noted that if the random prediction enable fails, the codeword parsing module will read the codeword cached by the on-chip microcontroller as the predicted codeword instead of obtaining the codeword from the host computer. When obtaining the codeword from the pseudo-random prediction engine or the on-chip microcontroller, the codeword is generated on the chip of the measurement and control system. The generated codeword exists directly in the register of the chip, which is much smaller than the delay of transmitting the codeword from the host computer to the measurement and control system.

[0067] (a5) Seed update: The seed update signal is connected to the seed stack. This signal is usually low (invalid). When it is pulled high, the seed stack pushes out the current seed and copies the next seed in the stack to the shift register to achieve seed update. If the seed update signal is received and the seed stack is empty, the real-time prediction module will stop waveform configuration and report an error.

[0068] With the help of a pseudo-random prediction engine, in random circuit sampling tests such as the XEB experiment, the host computer no longer needs to transfer the waveform data required to measure and control the entire random circuit to the waveform generator's data memory. Instead, it only transmits the seed and generation coefficients to the pseudo-random prediction engine. The waveform configuration circuit then predicts the quantum logic gates to be executed one by one within the hardware and automatically completes the waveform configuration, ultimately realizing random circuit experiments in quantum computing systems. At the same time, the host computer uses the same seed and generation coefficients to simulate the pseudo-random prediction engine's functions through a software algorithm, calculating the quantum circuit configuration being executed on the quantum device, thereby completing the simulation experiment of the quantum circuit on a classical computer. During the experiment, the data required to be transmitted by the host computer does not increase synchronously with the number of logic gates in the circuit. This approach is superior to traditional solutions in which the host computer calculates and transmits the waveform data required for the entire circuit in terms of both configuration time and hardware resources.

[0069] The structure, function, and codeword expression of waveform characteristics of the pseudo-random prediction engine are closely related. Because the waveform parameters corresponding to a quantum logic gate are highly compactly compressed into a codeword, the measurement and control waveforms required for different quantum logic gates can be mapped to corresponding codewords and displayed in a small private memory, thereby realizing random logic gate prediction based on pseudo-random numerical prediction.

[0070] The following will take 3 common waveform configuration scenarios as examples to illustrate the advantages of this embodiment over the prior art in more detail. Figure 6 shown.

[0071] Example 1: Total amount of waveform configuration data;

[0072] This embodiment is a parameter dynamic configuration method that cooperates with an on-chip microcontroller, a codeword parsing module, and a parameter lookup table;

[0073] The quantum logic gate operations that may be required for a quantum circuit to execute Each gate operation is generated by a microwave signal, where They are the total number of scaling factors, the total number of offset amplitudes, the total number of tuning increments, the total number of phase references, and the total number of envelope shape information. Their specific values ​​are determined according to the specific waveform, and the mathematical form of the waveform is as follows:

[0074]

[0075] ;

[0076] in, are the indexes of scaling factor, offset amplitude, phase reference, tuning increment and envelope shape information respectively, For the The envelope shape The sampled data value at the moment; For the first The corrected shape of the envelope is The sampling data at the moment, for The derivative of That is Proportional to The derivative of 、 、 、 Respectively Scaling factor, Bias amplitude, Phase reference and The three tuning increments respectively determine the scaling factor, offset amplitude, phase reference and tuning increment of the measurement and control waveform generated by the arbitrary waveform generator.

[0077] When the existing technology realizes the above experimental requirements, it is usually necessary to calculate the measurement and control waveform required by each quantum logic gate in the quantum circuit into sampling data according to the above formula in the host computer. , and all of them are transmitted from the host computer to the memory in the arbitrary waveform generator through the communication interface to complete the waveform data configuration.

[0078] This embodiment provides a new configuration scheme. In this embodiment, each pair Data and Data (i.e. envelope and correction data) are [ ] is organized into words, and the entire segment is stored in the first Segment address space, the initial address of the envelope is , the number of envelope points is , and Known as the The envelope information of the envelope shape is configured as the first envelope information sub-region in the parameter lookup table through the communication interface before the quantum circuit starts to execute. values. Similarly, 、 、 、 Before the quantum circuit starts executing, it is configured as the first sub-region of the scaling factor in the parameter lookup table. value, bias scaling factor, and the sub-region value, the phase reference sub-region value and the tuning increment sub-region After the quantum circuit starts executing, when the random prediction enable fails, the codeword parsing module reads the codeword cached by the on-chip microcontroller as the predicted codeword instead of reading the codeword from the pseudo-random prediction engine. At the same time, the codeword parsing module and the parameter lookup table operate normally.

[0079] At this time, the on-chip microcontroller sends a trigger pulse to the waveform configuration circuit at a specific moment according to the physical experiment process, and at the same time inputs a code word into the code word decoding module. This trigger is expected to generate a measurement and control waveform. ;

[0080] When the codeword parsing module receives the codeword fed back by the on-chip microcontroller, it intercepts the codeword. , use them as indexes and input them into the index ports of each sub-region of the parameter lookup table at the same time. Each sub-region performs the parameter lookup in parallel, and after one clock cycle, the offset value is , scaling factor , Phase Reference , tuning increment , envelope initial address and envelope points The waveform is output simultaneously to realize the waveform configuration, and the subsequent waveform generation module will calculate and generate the waveform according to these parameters.

[0081] Example 1 mainly highlights the optimization of the total amount of configuration data in this embodiment compared to the traditional waveform configuration method. Example 1 total configuration Different measurement and control waveforms The classic arbitrary waveform generator does not make a targeted decomposition of the waveform characteristics, has no codeword and lookup table structure, and the configuration data is the sampling data of the control waveform. , the total amount of data configured (in words) is obviously data; and in the configuration method provided by the present invention, the data that needs to be loaded into the arbitrary waveform generator through the communication interface is only Segment envelope data and The total amount of data configured (in words) can be expressed as ; Envelope length of the waveform Usually much larger than The total amount of data required to configure the configuration method provided in this embodiment is approximately equal to , compared with the classic waveform configuration method, it reduces about This reduction in configuration data volume stems from the targeted extraction and decomposition of waveform features by the waveform configuration method using codeword parsing and parameter lookup. This not only reduces the time required for waveform configuration, but also reduces the amount of storage resources required by the arbitrary waveform generator. With the same hardware resource consumption, it is possible to configure measurement and control waveforms corresponding to a wider range of quantum logic gates, thus facilitating the generation of measurement and control waveforms for quantum circuits composed of more quantum logic gates.

[0082] Example 2: Dynamic configuration of waveforms controlled by feedback;

[0083] Assume that there is a quantum circuit that needs to perform real-time feedback and modify the measurement and control waveform shape or parameters during the calculation process (such a situation exists in actual quantum computing algorithms, such as quantum error correction scenarios, etc.). Bit, value Dynamically select the measurement and control waveform corresponding to the quantum logic gate to be executed as Which one of is an integer value, The feedback values ​​are The scaling factor and feedback value are The bias amplitude and feedback value are The phase reference and feedback value are taken as The tuning increment and feedback value are The index of the envelope shape information at the time.

[0084] The waveform configuration achieved by existing technologies requires a host computer to calculate the measurement and control waveform sampling data required for the quantum circuit, which makes it difficult to cope with measurement and control waveform adjustments based on feedback values. The only option is to further transmit the feedback data back to the host computer and reconfigure the waveform. However, the communication delay between the host computer and the arbitrary waveform generator is high relative to the lifetime of the quantum bit state, making it difficult for this traditional configuration scheme to correctly handle feedback control requirements.

[0085] The solution of this embodiment is: a parameter dynamic configuration method that cooperates with an on-chip microcontroller, a codeword parsing module, and a parameter lookup table; the configuration process of the waveform of Example 2 is basically the same as that of Example 1, except that: the conditional output bit value in the codeword received by the microcontroller is 1; the codeword parsing module detects that the value of this bit is 1 and will set the feedback value Add to the code word, and then intercept and feedback value according to the bit field Related and other parameters; the subsequent table lookup work is the same as in Example 1;

[0086] Example 2 mainly highlights the dynamic advantages of the waveform configuration method of this embodiment. For general quantum algorithms with feedback control, the waveform configuration based on this method can be combined with the feedback value of the bit state, and the codeword can be dynamically updated to select the measurement and control waveform to be emitted under different conditions, thereby realizing dynamic adjustment of the quantum circuit. The entire process consumes almost no additional logic resources. Only an additional addition is made to the codeword value and the feedback value. The time required for waveform configuration is no different from the sequential output without feedback. In contrast, traditional arbitrary waveform generators that do not have codeword and parameter search mechanisms have poor support for this function. One method is to sacrifice real-time performance and transmit the feedback value to the host computer, which recalculates and reconfigures the waveform data through the communication interface; another method is to increase the sampling data of the sequence controller to select the trigger waveform, but it also requires additional storage resources in the memory to load the sampling data of the measurement and control waveform corresponding to each feedback value.

[0087] The support for dynamic waveform configuration in this embodiment comes from the codeword expression of the key features of the waveform. This codeword expression compresses arbitrarily complex waveform sampling data into one data, so that the measurement and control waveform configuration based on dynamic feedback only adds one step of codeword and feedback value addition operation for this embodiment.

[0088] Example 3: Execute random line situation;

[0089] Perform a random circuit sampling, the quantum logic gates that may be sampled are Types (sampling range, not necessarily all types are sampled, not necessarily without repetition), each type is generated by a specific measurement and control waveform, such as:

[0090]

[0091] ;

[0092] A total of D pseudo-random sampling times, generating D control waveforms (denoted as ) and form a random line, is the sampled control waveform index, that is Respectively The scaling factor, bias amplitude, phase reference, tuning increment, and envelope shape information corresponding to each control waveform are stored. In random sampling circuit experiments, a quantum computing system is used to execute the aforementioned circuit containing a large number of random quantum logic gates, and a host computer performs software simulation. Finally, the data collected after the experiment and simulation are analyzed to calibrate certain properties of the quantum computing system. To correctly simulate the behavior of the quantum computing system, the configuration of the random quantum circuit executed in the quantum computing system and the software simulation must be exactly the same.

[0093] To achieve these experimental requirements, existing technology involves a host computer generating a pseudo-random sequence based on specific rules (such as a linear feedback shift register method) and sampling quantum logic gates according to this sequence to create a complete random quantum circuit. The host computer then calculates all the measurement and control waveform sampling data required for this circuit and transfers it to the memory of the arbitrary waveform generator. This waveform configuration method requires configuration time and hardware resources that scale linearly with the number of logic gates in the circuit.

[0094] The solution of this embodiment: a parameter dynamic configuration method that cooperates with a pseudo-random prediction engine, a codeword parsing module, and a parameter lookup table; the configuration execution circuit's codeword parsing and parameter table lookup process are the same as in Example 1, except that the codeword generation method is changed to real-time generation by the pseudo-random prediction engine.

[0095] The algorithm for generating random quantum logic gates can be briefly understood as a three-level step: 1) Based on the "seed" data, a pseudo-random number generation algorithm is used to generate a pseudo-random number sequence; 2) The pseudo-random number sequence is used as an index to find the corresponding codeword value; 3) The codeword is parsed and the waveform parameters are generated. In the example scenario, the host computer generates a pseudo-random seed and Generate coefficients with the same bit width as the seed , and load them into the seed stack and generated coefficient registers respectively through the communication interface. Figure 5 As shown, the pseudo-random prediction is based on the steps (a1)-(a5) above. Each time a trigger pulse signal is received, a 16-bit pseudo-random prediction value is taken out from the shift register. , which is used as an index to find a code word from the logic gate lookup table ; Finally, the codeword enters the configuration execution circuit to analyze and find the scaling factor , bias amplitude , tuning increment , Phase Reference and envelope shape information , realize the waveform The codeword values ​​in the logic gate lookup table are pre-loaded, and there is an identical backup in the host computer. Obviously, this process can be executed exactly the same by the software algorithm in the host computer, thus ensuring that the host computer software simulation and the quantum computing system experiment have the same configuration.

[0096] The advantage of this embodiment over the prior art is that the generation algorithm of the pseudo-random quantum logic gate is integrated into the hardware, thereby eliminating the need to sample the measurement and control waveform data of all pseudo-random quantum logic gates before waveform configuration. Instead of calculating, transmitting and storing in the envelope memory of the arbitrary waveform generator, only the seed and generation coefficients need to be Similar to the conclusion in Example 1, this change changes the amount of data transmitted from the host computer to the arbitrary waveform generator in the waveform configuration from Reduced to The number of envelope types M of measurement and control waveforms in quantum bit measurement and control is limited, usually much smaller than the number of quantum logic gates to be configured in one circuit, thus reducing the configuration time and storage resources.

[0097] More importantly, the amount of data that existing methods need to transmit and store is positively correlated with the circuit depth D (i.e., the number of quantum logic gates to be executed in the quantum circuit). This means that the storage resources of the arbitrary waveform generator may limit the increase in circuit depth, resulting in poor scalability. However, the amount of configuration data in this embodiment mainly depends on the envelope type and length of the measurement and control waveform. These are limited for typical quantum circuits and can be pre-configured and stored. The circuit depth D is not limited by the waveform configuration. After each synchronous trigger pulse is received, the configuration circuit performs a pseudo-random prediction and maps the predicted value into a codeword, and then configures the corresponding measurement and control waveform until the number of quantum logic gates required for the experiment is reached.

[0098] Therefore, this embodiment compresses information such as the shape, length, tuning increment, phase reference, scaling factor, and bias amplitude of the waveform according to the diverse requirements of quantum bit measurement and control waveforms, ensuring that the measurement and control waveform can be dynamically configured and flexibly adjusted during the execution of the quantum circuit.

[0099] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for dynamic configuration of quantum bit waveform parameters based on pseudo-random prediction, characterized in that: include: The configuration execution circuit decodes the predicted codeword generated on the chip to obtain a parameter group index. The predicted codeword is generated by a pseudo-random prediction engine. The pseudo-random prediction engine uses a seed stack to receive external seed data, and uses a real-time prediction module to execute a pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit; By using the parameter group index to perform parallel search in the parameter lookup table, all parameter data values ​​corresponding to the parameter group index can be output in one clock cycle, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data value. Each sub-area in the parameter lookup table stores the preloaded waveform parameters.

2. The parameter dynamic configuration method according to claim 1, characterized in that: When the configuration execution circuit receives the predicted codeword generated by the pseudo-random prediction engine and parses it into a parameter group index, the predicted codeword is parsed using the codeword parsing module, specifically: The predicted codeword is parsed into the index of each parameter through the bit field, and the index is used as the parameter group index. In the bit field parsing, the data of different bit fields of the codeword represent the index of different parameter values. The parsed parameter group index is cached and sent to the parameter lookup table.

3. The parameter dynamic configuration method according to claim 2, characterized in that: Each sub-region in the parameter lookup table corresponds to a control parameter component, and the parameter value in each sub-region is configured by the host computer and pre-loaded into the parameter lookup table through the bus.

4. The method for dynamic parameter configuration according to claim 2, wherein: The pseudo-random prediction engine uses a seed stack to receive seed data input from the outside, and uses a real-time prediction module to execute the pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The real-time prediction module is specifically used to: When the quantum circuit is started, the seed stack receives and caches the seed data input from the outside. The seed data is the initial value of the pseudo-random number generation algorithm. The seed stack transfers the cached seed data to the shift register, uses the generation coefficient in the generation coefficient register to filter the shift register value, and then updates the state value of the shift register through an exclusive-OR operation. The updated register state value is input into the index end of the logic gate lookup table. The codewords read from the logic gate lookup table are fed back to the codeword cache, which caches the compressed codewords that represent the parameter indices corresponding to the sub-regions in the parameter lookup table. The codeword parsing module receives a synchronous trigger signal of the waveform and reads the codeword cache under the action of random prediction enabling, thereby obtaining the predicted codeword.

5. The method for dynamic parameter configuration according to claim 4, characterized in that: When receiving the synchronous trigger signal of the waveform, the codeword parsing module reads the codeword cache under the action of random prediction enable to obtain the predicted codeword. If the random prediction enable fails, the codeword parsing module reads the codeword generated by the on-chip microcontroller as the predicted codeword.

6. The method for dynamic parameter configuration according to claim 4, characterized in that: The shift register value is filtered using the generation coefficient in the generation coefficient register, and then the state value of the shift register is updated through an XOR operation, and the updated register state value is input into the index end of the logic gate lookup table, specifically: The generation coefficient register selects the value of the corresponding bit of the shift register state value according to the bit with the value 1 in each group of generation coefficients; Each group of filtered bits is passed through an XOR operator to generate feedback bits, and the feedback bits are sequentially filled into the leftmost bit of the shift register, and the other bits are shifted right in sequence; The high m bits of the shift register are input as pseudo-random prediction results to the index end of the logic gate lookup table, where m is an integer. The logic gate lookup table is the private memory of the pseudo-random prediction engine and is used to store various code words.

7. The method for dynamic parameter configuration according to claim 4, characterized in that: The generation coefficient register and logic gate lookup table in the pseudo-random prediction engine are configured by a host computer.

8. The method for dynamic parameter configuration according to claim 1, wherein: Each time the pseudo-random prediction engine starts the quantum circuit, the seed stack receives seed data from the external input, allowing multiple seeds to be input in order; In each execution of the quantum circuit, it is only necessary to receive external input seed data once, and predict the configuration parameters of the quantum logic gate in real time based on the pseudo-random algorithm.

9. The method for dynamic parameter configuration according to claim 1, wherein: The parameter group index includes a scaling factor index, a tuning increment index, a bias amplitude index, a phase reference index and an envelope information index; The parameter lookup table is provided with a tuning increment sub-area, a scaling factor sub-area, a bias amplitude sub-area, a phase reference sub-area, and an envelope information sub-area, and each sub-area is provided with a corresponding lookup table.

10. A quantum bit waveform parameter dynamic configuration system based on pseudo-random prediction, characterized in that: including configuration execution circuitry and a pseudo-random prediction engine; The pseudo-random prediction engine uses a seed stack to receive seed data input from the outside, and uses a real-time prediction module to execute a pseudo-random algorithm in real time according to the seed and the configured generation coefficient. The predicted pseudo-random value indexes the predicted codeword corresponding to the measurement and control waveform of the next quantum logic gate operation in the quantum bit circuit; The configuration execution circuit decodes the predicted codeword to obtain a parameter group index, and uses the parameter group index to perform parallel searches in a parameter lookup table. It takes one clock cycle to output all parameter data values ​​corresponding to the parameter group index, thereby dynamically configuring the quantum bit measurement and control waveform mapped by the parameter data value. Each sub-region in the parameter lookup table stores preloaded waveform parameters.

Images