Hardware Acceleration of Frequency Domain Scheduler in Wireless Networks

The task of handling the frequency domain scheduler in a 5G wireless network through hardware accelerator is solved, and the problem of extended execution time of frequency domain scheduler in a 5G network is improved, resource allocation efficiency and energy and cost are reduced.

CN111819546BActive Publication Date: 2024-11-22NOKIA TECHNOLOGIES OY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201880090229.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2018-01-11
Publication Date
2024-11-22
Estimated Expiration
2038-01-11

AI Technical Summary

Technical Problem

In 5G wireless networks, frequency domain schedulers have extended execution time for computationally intensive tasks due to large-scale MIMO and short transmission time intervals (TTIs), and existing software schedulers have difficulty efficiently handling multiple sort calls and resource allocations. Causes inefficiency.

Method used

Using hardware accelerator, receives jobs from the processor core by receiving instructions, uses an array stored in memory to determine user resource allocation, and allocate resources based on user metrics and buffer size, and sends the results back to the processor core to realize frequency domain scheduling. hardware acceleration of the device.

Benefits of technology

It reduces the execution time of the scheduler, improves resource allocation efficiency, reduces power consumption and cost, especially in effectively dealing with the computing pressure brought by large-scale MIMO in 5G networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111819546B_ABST
    Figure CN111819546B_ABST
Patent Text Reader

Abstract

A method and apparatus for hardware acceleration of a frequency domain scheduler are provided. An example method includes: receiving instructions from a processor core to process a job, the job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a transmission layer in a transmission time interval; processing a first job by: determining which user of the one or more users is to be allocated on each resource of a corresponding resource in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each resource in the set, and the second array being associated with a buffer size for each of the one or more users; and sending an indication of the determined allocation to the processor core in response to successfully completing the job.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates generally to wireless networks, and more particularly to hardware acceleration of a frequency domain scheduler for wireless networks. Background Art

[0002] This section is intended to provide a background or context for the invention disclosed below. The descriptions herein may include concepts that may be pursued, but are not necessarily concepts that have been previously conceived, implemented, or described. Therefore, unless otherwise expressly noted herein, the content described in this section is not prior art to the descriptions in this application and will not be admitted to be prior art by virtue of being included in this section. Below, after the main part of the detailed description section, abbreviations that can be found in the specification and / or the drawings are defined.

[0003] In wireless networks, the scheduler (also known as MAC packet scheduler or MAC scheduler) is typically run as a software process (or multiple threads) on a central processing unit (CPU) or digital signal processing (DSP) or system-on-chip (SOC) core. We can collectively refer to them as CPU cores, processor cores, or just "cores". The scheduler performs many computationally intensive tasks, including multiple sequencing calls for a large number of user / flow metrics. This problem is further exacerbated by various factors in 5G wireless networks compared to LTE wireless networks, including: larger bandwidth causes the scheduler to allocate more resources; massive MIMO implementations require more synchronized beams to be allocated and higher algorithmic complexity to jointly select multiple beams; and TTI duration is shorter, which shortens the deadline to complete the allocation work. Summary of the invention

[0004] This section is intended to include examples and is not intended to be limiting.

[0005] According to an example of an embodiment, a method is provided, the method comprising: receiving instructions from a first processor core to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer within a transmission time interval; processing the first job by: determining which of one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each of the resources in the set, the second array being associated with a buffer size for each of the one or more users; and sending an indication of the determined allocation of the set of resources to the first processor core in response to successful completion of the first job.

[0006] Additional examples of embodiments include a computer program comprising code for causing the method of the preceding paragraph to be performed when the computer program is executed on a processor. A computer program according to the paragraph, wherein the computer program is a computer program product comprising a computer readable medium carrying computer program code implemented therein for use by a computer.

[0007] According to another example embodiment, a device is provided, the device comprising: a component for receiving instructions from a first processor core to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer within a transmission time interval; a component for processing the first job by: determining which user of the one or more users is to be allocated on each resource of corresponding resources in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each resource in the set, and the second array being associated with a buffer size for each of the one or more users; and a component for sending an indication of the determined allocation of the set of resources to the first processor core in response to successful completion of the first job.

[0008] According to another example embodiment, a device is provided, the device comprising: a hardware accelerator; at least one processor including one or more processing cores; at least one non-volatile memory including computer program code, the at least one memory and the computer program code being configured to, together with the at least one processor and the hardware accelerator, cause the device to at least: receive instructions from a first processor core to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer within a transmission time interval; process the first job by: determining which of the one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in the memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each of the resources in the set, the second array being associated with a buffer size for each of the one or more users; and in response to successfully completing the first job, sending an indication of the determined allocation of the set of resources to the first processor core. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In the attached picture:

[0010] Figure 1 is a block diagram of one possible and non-limiting exemplary system in which exemplary embodiments may be practiced;

[0011] Figure 2 is a simplified process flow diagram of a scheduler according to an exemplary embodiment;

[0012] Figure 3 is a simplified schematic diagram outlining the use of a scheduler accelerator or co-processor with a memory and a processor according to an exemplary embodiment;

[0013] Figure 4 is a message flow diagram for communication between a CPU core and a scheduler accelerator according to an exemplary embodiment; and

[0014] Figure 5 is a logic flow diagram for hardware acceleration of a frequency domain scheduler and illustrates the operation of one or more exemplary methods according to exemplary embodiments, the results of execution of computer program instructions implemented on a computer readable memory, functions performed by logic implemented in hardware, and / or interconnected components for performing functions. DETAILED DESCRIPTION

[0015] The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. All embodiments described in the detailed description are exemplary embodiments, which are provided to enable those skilled in the art to make or use the invention, rather than to limit the scope of the invention defined by the claims.

[0016] The exemplary embodiments herein describe techniques for hardware acceleration of a frequency domain scheduler.After describing a system in which the exemplary embodiments may be used, additional description of these techniques is presented.

[0017] Features as described herein are sometimes referred to in LTE terminology, but it should be noted that these features may be used with other types of systems in the future. For example, unless otherwise noted, references to eNBs (i.e., LTE base stations) also apply to future base stations of wireless networks (such as, for example, base stations in 5G wireless networks referred to as gNBs).

[0018] Steering Figure 1 , which shows a block diagram of one possible and non-limiting exemplary system in which the exemplary embodiments may be practiced. Figure 1In the present invention, a user equipment (UE) 110 wirelessly communicates with a wireless network 100. A UE is a wireless device that can access a wireless network, typically a mobile device. The UE 110 includes one or more processors 120, one or more memories 125, and one or more transceivers 130 interconnected by one or more buses 127. Each of the one or more transceivers 130 includes a receiver Rx 132 and a transmitter Tx 133. The one or more buses 127 can be address, data, or control buses, and can include any interconnection mechanism, such as a series of lines on a motherboard or integrated circuit, optical fiber or other optical communication devices, etc. The one or more transceivers 130 are connected to one or more antennas 128. The one or more memories 125 include computer program code 123. For example, the one or more memories 125 and the computer program code 123 can be configured to, together with the one or more processors 120, cause the user equipment 110 to perform one or more operations described herein. The UE 110 communicates with the gNB 170 via a wireless link 111.

[0019] The gNB / eNB (fifth generation Node B / evolved Node B) 170 is a base station (e.g., for a 5G wireless network and / or an LTE wireless network) that provides access to the wireless network 100 for wireless devices such as UE 110. The gNB 170 includes one or more processors 152 (e.g., which may include various types of processor cores, such as a CPU or DSP or SOC core), one or more memories 155, one or more network interfaces (N / WI / F) 161, and one or more transceivers 160 interconnected by one or more buses 157. Each of the one or more transceivers 160 includes a receiver Rx 162 and a transmitter Tx 163. The one or more transceivers 160 are connected to one or more antennas 158. The one or more memories 155 include computer program code 153. The gNB 170 includes an accelerator module, which includes one or both of the parts 150-1 and / or 150-2, and the accelerator module can be implemented in a variety of ways. The accelerator module may be implemented in hardware as an accelerator module 150-1, such as being implemented as part of one or more processors 152. The accelerator module 150-1 may also be implemented as an integrated circuit, or by other hardware such as a programmable gate array, which may be external to the processor 152, or adjacent to the processor 152 as a co-processor, or the accelerator module 150-1 may even be implemented as a dedicated circuit system within the same chip as the processor 152, or in any other suitable manner. In another example, the accelerator module may be implemented as an accelerator module 150-2, which is implemented as a computer program code 153 and executed by one or more processors 152 (e.g., on one or more dedicated cores within the processor 152, which dedicated cores may be capable of executing certain specialized instructions). For example, the one or more memories 155 and the computer program code 153 are configured to, together with the one or more processors 152, cause the gNB 170 to perform one or more operations described herein. The one or more network interfaces 161 communicate over a network, such as via links 176 and 131. Two or more gNBs 170 communicate using, for example, link 176. Link 176 may be wired or wireless or both, and may implement, for example, an X2 interface.

[0020] The one or more buses 157 may be address, data, or control buses and may include any interconnection mechanism, such as a series of wires on a motherboard or integrated circuit, optical fiber or other optical communication device, a wireless channel, etc. For example, one or more transceivers 160 may be implemented as a remote radio head (RRH) 195, where other elements of the gNB 170 are physically located at a different location from the RRH, and the one or more buses 157 may be partially implemented as a fiber optic cable for connecting the other elements of the gNB 170 to the RRH 195.

[0021] Note that the descriptions herein indicate that a "cell" performs a function, but it should be clear that the gNBs forming the cells will perform the functions. A cell forms part of a gNB. That is, there can be multiple cells per gNB. The term "cell" can refer to the coverage area of ​​a given set of transceivers associated with a gNB, or can refer to a logical portion of a gNB that performs functions related to transmission / reception within that coverage area. For example, for a single gNB carrier frequency and associated bandwidth, there can be three cells, each covering one-third of a 360-degree area, so that the coverage area of ​​a single gNB covers an approximately elliptical or circular shape. Furthermore, each cell can correspond to a single carrier, and a gNB can use multiple carriers. Thus, if there are three 120-degree cells per carrier and there are two carriers, the gNB has a total of 6 cells.

[0022] The wireless network 100 may include one or more network control elements (NCE) 190, which may include MME (mobility management entity) and / or SGW (serving gateway) functionality and which provides connectivity to additional networks such as a telephone network and / or a data communication network (e.g., the Internet). The gNB 170 is coupled to the NCE 190 via a link 131. The link 131 may be implemented as, for example, an S1 interface. The NCE 190 includes one or more processors 175, one or more memories 171, and one or more network interfaces (N / WI / F) 180 interconnected by one or more buses 185. The one or more memories 171 include computer program code 173. The one or more memories 171 and the computer program code 173 are configured to, together with the one or more processors 175, cause the NCE 190 to perform one or more operations.

[0023] Wireless network 100 may implement network virtualization, which is a process of combining hardware and software network resources and network functionality into a single software-based managed entity (virtual network). Network virtualization involves platform virtualization, which is often combined with resource virtualization. Network virtualization is categorized as external (combining many networks or network parts into virtual units) or internal (providing similar network functionality to software containers on a single system). Note that, to some extent, the virtualized entities produced by network virtualization can still be implemented using hardware such as processors 152 or 175 and memories 155 and 171, and such virtualized entities also produce technical effects.

[0024] Computer readable memories 125, 155, and 171 may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory. Computer readable memories 125, 155, and 171 may be components for performing storage functions. Processors 120, 152, and 175 may be of any type suitable for the local technical environment and may include, as non-limiting examples, one or more of: a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), and a processor based on a multi-core processor architecture. Processors 120, 152, and 175 may be components for performing functions such as controlling UE 110, gNB 170, and other functions described herein.

[0025] In general, various embodiments of user equipment 110 may include, but are not limited to, a cellular phone (such as a smart phone), a tablet computer, a personal digital assistant (PDA) with wireless communication capabilities, a portable computer with wireless communication capabilities, an image capture device with wireless communication capabilities (such as a digital camera), a gaming device with wireless communication capabilities, a music storage and playback device with wireless communication capabilities, an Internet device that allows wireless Internet access and browsing, a tablet computer with wireless communication capabilities, and a portable device or terminal containing a combination of such functions. The user equipment may execute one or more service flows simultaneously, for example, these service flows correspond to different applications and have different requirements for throughput or latency or other key performance determination indicators.

[0026] Having thus introduced a suitable but non-limiting technical context for the practice of exemplary embodiments of the present invention, the exemplary embodiments will now be described in more detail.

[0027] Note that the terms "flow" and "user" may be used interchangeably in the following description unless otherwise indicated.

[0028] As mentioned above, schedulers in wireless networks perform many computer-intensive tasks, including multiple sequencing calls for a large number of user / flow metrics. In a gNB, a scheduler instance is typically associated with a cell or carrier and typically performs the task of allocating radio resources between users or flows connected to that cell. Radio resources may include time domain resources, such as, for example, transmission time intervals; or frequency domain resources, sometimes referred to as physical resource blocks (PRBs) or resource block groups (RBGs). The scheduler typically performs the task of determining the allocation or partitioning of radio resources between one or more users or between one or more traffic flows of one or more users. The determination of which users (or flows) are eligible for allocated resources within a given transmission time interval (TTI), sometimes referred to as time domain scheduling or TD scheduling, can be based on various metrics and channel properties of the users or flows, such as a proportional fairness metric and the spectral efficiency of the user's channel.

[0029] The determination of the allocation of specific frequency domain resources to different users / streams (sometimes referred to as frequency domain scheduling or FD scheduling) can be based on various metrics and channel properties of users / streams that can vary from one radio resource to another. For example, wireless channels are typically frequency selective, that is, the channel gain or spectral efficiency of a user on a given frequency domain radio resource can be different from that on a different frequency domain radio resource, and thus the FD scheduling allocation of frequency domain resources can take into account such frequency selectivity differences of different users / streams across different frequency domain resources. This is sometimes referred to as frequency selective scheduling (FSS).

[0030] Furthermore, in multiple-input multiple-output (MIMO, sometimes also referred to as massive MIMO) systems that employ multiple transmit / receive antennas, a scheduler may allocate a given time-domain or frequency-domain resource to multiple users (or streams belonging to different users) simultaneously. This is typically accomplished in MIMO systems by forming different beams toward different users (e.g., by using different sets of antenna weights on multiple antennas) using a technique known as spatial division multiplexing, which allows different transmission beams to be formed toward different users in a manner that reduces overlap between the beams. Thus, the interference experienced by one user (receiving one beam) due to another beam being simultaneously transmitted to a different user is reduced.

[0031] A massive MIMO system may sometimes transmit multiple transmission layers or spatial dimensions. Each MIMO layer may include, for example, a single beam (e.g., the same beam transmitted on all frequency domain resources, where different users / streams are assigned different frequency domain resources, as described above), or may include multiple beams (e.g., different beams generated on frequency domain resources). There may be multiple transmission layers in a cell transmitting simultaneously in a given transmission time interval. The scheduler in a massive MIMO system typically determines the selection of beams and the selection of users for whom specific beams should be used for transmission, as well as the mapping of beams to transmission layers. This selection of beams and layers is sometimes referred to as spatial domain scheduling (SD scheduling) and may be combined with TD or FD scheduling as described above. These aspects of wireless systems and schedulers may be applicable to LTE as well as 5G systems and other systems that may use lower frequency bands (below 6 GHz or below 6 GHz) or higher frequency bands (centimeter waves or millimeter waves, typically above 6 GHz).

[0032] Reference now Figure 2 , which shows a simplified process flow diagram of a scheduler for a wireless network (such as, for example, LTE and 5G wireless networks) according to an example embodiment. In particular, the process flow diagram of the scheduler includes a pre-scheduler component 202; a time domain and spatial domain (TD+SD) component 204; a frequency domain (FD) component 206; and a PDDCH component 208. At the pre-scheduler component 202, all eligible users are determined and passed to the TD+SD scheduling component 204. In each TTI, the TD+SD scheduling component 204 selects a subset of users / streams (and possibly a subset of beams in the case of massive MIMO) and passes this information to the FD scheduler 206 for frequency domain resource allocation for that TTI. In 5G with massive MIMO, multiple spatial dimensions (e.g., beams) may be transmitted simultaneously, and therefore the FD scheduler 206 must perform FD allocations for multiple beams or layers in each TTI. As described above, FD scheduling on a given beam is performed as frequency selective scheduling. This is also expected to occur in 5G wireless networks below 6 GHz. This is a computationally complex operation that requires repeatedly finding, for example, the maximum of some metric M(i, j) for each resource (eg, RBG) j over a given set of streams i.

[0033] One problem with these calculations is that the maximum on each resource j cannot be found in parallel due to the limited size of the buffer. Typically, resources need to be allocated sequentially, one at a time, and then the allocated buffer is decremented. Otherwise, you may encounter a situation where multiple resources are allocated to the same user / stream, but the stream does not have enough data in the buffer to use all the resources, resulting in inefficiency. In addition, this operation must be performed for multiple beams. The example embodiment utilizes a hardware acceleration block to speed up the execution time of the FD scheduler 206 for 5G with massive MIMO. It should also be noted that although the bandwidth in question in LTE may be smaller than in 5G, the same problem applies to massive MIMO in LTE.

[0034] An example of the algorithm used by the FD scheduler is now discussed in more detail. In each TTI, the FD scheduler is provided (by the TD scheduler) with a set of candidate streams, a set of beams (or spatial layers), and a set of frequency domain resources, and the FD scheduler must allocate resources to the streams on different beams.

[0035] The following symbols are used in the following description. For the scheduler algorithm, there are F streams and R frequency domain resources (e.g., RBGs), where stream i has a large amount of data B(i) in its buffer. Each stream i (which may correspond to a user or a bearer) has a scheduling weight W(i). On each transmission layer k, on each frequency resource j, a certain amount of data S(i, j, k) can be transmitted to user / stream i. The scheduling metric for stream i on resource j on transmission layer k is sometimes constructed as M(i, j, k) = W(i) * S(i, j, k). For example, the scheduling weight of a PF (proportional fairness) scheduler may be W(i) = 1 / (R_ave(i)), i.e., the inverse of the average throughput obtained by user i, or alternatively may be a GBR / delay related metric. S(i, j, k) may be a measure of the spectral efficiency of user i on frequency resource j on transmission layer k. As previously described, due to the frequency selectivity of the channel, users may have different spectral efficiencies on each frequency resource j. The scheduling metric for a user is the product of the scheduling weight W(i) and the spectral efficiency S(i, j, k). For simplicity, whenever we work on a transport layer beam idk, the symbol of index k can be removed. For example, S(i, j, k) can be referred to as S(i, j), and M(i, j, k) can be referred to as M(i, j). C(i) represents the number of control channel elements (CCEs) required to send a control indication to user i to convey the resource allocation, and C total Indicates the total number of available CCEs.

[0036] FD Scheduling Algorithm : The exemplary FD scheduling algorithm will iteratively perform the following operations:

[0037] · Select one of the selected transport layers, and perform FD scheduling for the transport layer.

[0038] For each remaining resource j, find the flow F*(j) with the maximum metric on resource j among all flows i that match resource j, where the amount of data remaining in the buffer is non-zero (i.e., B(i)>0), i.e., F*(j)=argmax i M(i, j).

[0039] οAmong all remaining resources j, find j * = argmax j M(F*(j), j). Assign resource j* to flow F*(j*)

[0040] o Decrement the amount of data remaining in the buffer of stream F*(j*) by the amount of data that can be transferred or stream F*(j*) on resource j*:: B(F*(j*))=max(0, B(F*(j*))-S(F*(j*), j*))

[0041] Optionally, a certain number of CCEs are counted C(F*(j*)) for the PDCCH assigned to the user, and the total number of remaining CCEs is decremented, i.e., C total =C total -C(F*(j*))

[0042] o Mark resource j* as assigned, and if the remaining data in the buffer is 0, i.e., if B(F*(j*))=0, mark flow F*(j*) as unsuitable for further allocation in this TTI

[0043] o Repeat the iteration until all data for all flows are assigned, or until all CCEs are exhausted, or there are no more resources on this transport layer

[0044] o Repeat the above process for each transport layer.

[0045] As will be appreciated by those skilled in the art, certain details in the above FD scheduling algorithm have been omitted for simplicity, such as, for example, consideration of new transmissions versus retransmissions. Furthermore, essentially any algorithm may be used for FD scheduling, and the above is only one possible algorithm. It may be noted that these types of FD scheduling algorithms have significant complexity, which increases with the number of users / streams, the number of frequency domain resources, and the number of transmission layers. However, the scheduler has a given amount of time (constrained by the transmission time interval) to perform the task of determining the frequency domain resource allocation. Therefore, it is desirable to find a way to accelerate the execution of such a FD scheduling algorithm, for example with the aid of a suitable hardware accelerator.

[0046] One problem with implementing the FD scheduling algorithm described above is that it requires a series of "max" operations with dependencies between them, rather than a single sort, so a typical sort accelerator that provides a fast implementation of a simple sort is not sufficient. For example, in the first iteration, the scheduler must find the best user on each RBG, then find the best user among each of these best users, and then assign the found RBG to the found user. In order to calculate this using a typical sort accelerator, multiple passes to the accelerator would be required, one array (user metrics on a given RBG) at a time, which requires a lot of overhead because each pass to the accelerator incurs overhead. In addition, the second iteration depends on the outcome of the RBG allocation decision in the first iteration. This means that the user's buffer count needs to be decremented and the CCE may need to be decremented because the first RBG must be deleted from the list before the second iteration begins. Therefore, the FD scheduling algorithm has sequential dependencies and, therefore, it is not possible to perform allocations in parallel on different frequency resources.

[0047] According to an example embodiment, a hardware (HW) accelerator or coprocessor for a wireless MAC scheduler is provided. The HW accelerator (also referred to herein as a scheduler accelerator) may correspond to, for example, Figure 1 The accelerator module in blocks 150-1 and / or 150-2. The HW accelerator may receive as input:

[0048] A first pointer to an array of pointers with a specified length (representing the number of resources), where each pointer in the array of pointers points to an array of measures of spectral efficiency for a set of eligible users / streams for a set of resources. This can be represented as a 2D matrix of size (number of streams)*(number of resources). For example, the first pointer can be a pointer to a one-dimensional array of pointers, where each pointer in the array points to a one-dimensional array of elements, where each element represents a measure of spectral efficiency for one user / stream on one resource, i.e., S(i, j);

[0049] A second pointer to an array of buffer sizes B(i), where the length is equal to the number of streams;

[0050] (optional) a third pointer to an array of scheduling weights W(i), where the length is equal to the number of flows;

[0051] (optional) a fourth pointer to an array of CCEs C(i), where the length of the array is equal to the number of flows (or users);

[0052] a fifth pointer to an array into which the stream assignment decision F*(j) is to be written (or alternatively, a flag indicating that the output should be written to replace an existing array), wherein the length of the array corresponds to the number of resources;

[0053] In this example embodiment, the HW accelerator may then perform the following operations:

[0054] Iteratively execute the above FD scheduling algorithm using the provided metrics, buffer sizes, and CCEs;

[0055] Write the output to the location in memory specified by the fourth pointer (e.g., including the decision of which user is assigned to each resource, the allocated CCEs, the remaining buffer size, etc.), and

[0056] Returns a status value indicating completion or an error condition.

[0057] Reference now Figure 3 , which is a simplified schematic diagram outlining the use of a scheduling accelerator (or coprocessor) with a memory and a processor according to an exemplary embodiment. More specifically, Figure 3 The process flow between the CPU core 302, memory 306 and scheduler accelerator 304 responsible for performing frequency domain scheduling is shown. For example, the process flow diagram can be used in a processor such as Figure 1 The CPU core 302 corresponds to at least one processor 152, the memory 306 corresponds to at least one memory 155, and the scheduler accelerator 304 corresponds to the accelerator module 150-1 and / or 150-2. In step S0, the CPU core 302 selects a transport layer k. In step S1, the CPU core prepares:

[0058] A block of memory containing a two-dimensional array (represented as a pointer in a pointer (Ptr)), called metric_array_ptr in this example;

[0059] ·Buffer size array (in Figure 3 Buf_ptr); and

[0060] (Optional) Control Channel Element (CCE) array (in Figure 3 CCE_ptr), and the total number of control channel elements (Ctotal); and

[0061] A "zeroed" or suitably initialized array into which the output decisions can be written (in Figure 3 is called output_flowid_ptr).

[0062] Each pointer in metric_array_ptr points to an array S(i, j) of per-resource values ​​representing a measure of the spectral efficiency of a user / flow. In step S2, the CPU core 302 passes these input parameters to the scheduler accelerator 304. In step S3, the scheduler accelerator 304 processes the input block, and in step S4, the scheduler accelerator provides a notification to the CPU core about the task completion. The notification may be provided as an event in an event machine (EM) queue, or via another suitable bus, which is represented by the shaded area 303. In step S5, the CPU core 302 selects another transport layer and repeats steps S1-S5 for that transport layer.

[0063] Reference now Figure 4 , which is a message flow diagram for a CPU core and a scheduler accelerator according to an example embodiment. In this example, event machine queues and events are used to facilitate communication between the CPU core and the accelerator. In this example, the CPU core starts the FD scheduler within a given transmission time interval, as shown in 410, and selects a transport layer k at 412. The CPU core then prepares 414 data structures in memory for: flows for each resource (i.e., S(i, j)); buffer sizes (i.e., B(i)); optionally, control channel elements (i.e., C(i)); scheduling weights W(i); and outputs (i.e., F*(j)). In step 414, the CPU core may also indicate the control channel elements C available in memory. totalThe total number of. For example, this can be accomplished by sending one or more signals from the CPU core to allocate data structures in memory. The CPU core sends a message 416 to the EM to enqueue a first event containing indications of various data structures, such as by sending a pointer to the memory location of the data structure. The EM then sends an indication of the next pending task / event 418 (which will be the first event in this example) to the scheduler accelerator, and the scheduler accelerator retrieves 420 data corresponding to the task event from memory. At 422, the scheduler accelerator executes the FD scheduler algorithm to determine the user allocation on each resource based on the specified input. The scheduler accelerator then writes 424 the output data of the allocation decision to memory (e.g., including decisions about which user has been allocated for each resource, the allocated CCEs, the remaining buffer size, etc.), and typically notifies 426 the EM that the task has been completed (or with an indication of an error condition where appropriate) by creating / enqueuing a second event containing a pointer to the output and an error indication. The EM then notifies 428 the CPU core that the task has been completed by distributing the second event to the CPU core, and the CPU core retrieves 430 the allocation decision from memory. At 432, the CPU core updates weights and metrics based on the allocation decision returned by the accelerator, such as scheduling weights W(i) that may be based on a proportional fairness metric if needed. At 434, the CPU core selects the next transport layer k' and repeats the process for that transport layer.

[0064] exist Figure 4 In FIG. 1 , an EM for distributing tasks from a CPU core to a scheduler accelerator is shown, however, in addition to EM, any other suitable method may be used for communication between the CPU core and the scheduler accelerator, such as, for example, via any suitable bus.

[0065] It should also be noted that in the above-described example embodiments, it is described that the CPU core sends a single task (corresponding to a resource allocation on a single transport layer), waits for the task to complete, and then distributes the next job. For example, assume that the FD allocation for one transport layer corresponds to a single beam. In this case, the FD allocation for one beam can be completed before proceeding with the FD allocation for the next beam. This is useful when the selection of the next beam and the user on the beam (or its metric / scheduling weight) depends on the allocation result of the first beam, because it ensures that the same user is not considered on overlapping beams. According to another example embodiment, the CPU core can distribute the next task without waiting for the first task to complete, and continue with other operations to return to process the output of the scheduling accelerator when receiving an appropriate notification that the scheduling accelerator has completed the task. Note that for clarity, as used herein, the terms "task" and "job" are interchangeable.

[0066] According to another example embodiment, multiple CPU cores that feed into the same scheduler accelerator in parallel can be used. For example, two different CPU cores can handle two different cells and simultaneously issue resource allocation requests for their corresponding cells to the accelerator. In some examples, two different cores can even handle two transport layers in the same cell and simultaneously issue resource allocation requests for their corresponding layers to the accelerator. In this case, when the accelerator completes the task, a completion notification must be sent back to the specific initiator. To help the EM route the response back to the correct requesting core, the CPU core can include the core ID and EM queue ID along with the events sent to the accelerator, such as Figure 4 As shown in message 416 in . When the scheduler accelerator completes the job, the scheduler accelerator queues the completion notification event into the correct EM queue and optionally includes the requesting core ID in the event. The EM can distribute the response back to the correct core that requested the event.

[0067] According to an example embodiment, the scheduler accelerator is configured to process multiple jobs / events simultaneously, such as when multiple CPU cores are utilized and multiple jobs are sent to the accelerator from multiple CPU cores simultaneously. For example, the scheduler accelerator can be configured to have a parallel circuit system for executing multiple tasks simultaneously. In such an embodiment, the scheduler accelerator can read multiple jobs from the EM in one batch.

[0068] Figure 5 1 is a logic flow diagram for hardware acceleration of a frequency domain scheduler. The figure also illustrates the operation of one or more exemplary methods according to exemplary embodiments, the results of executing computer program instructions implemented on a computer readable memory, the functions performed by logic implemented in hardware, and / or the interconnected components for performing the functions. For example, accelerator modules 150-1 and / or 150-2 (see Figure 1 ) can include Figure 5 In the multiple blocks, each block is an interconnected component for performing the function in the block. Figure 5 The boxes in are at least partially performed by a base station such as gNB 170, for example under the control of accelerator modules 150-1 and / or 150-2.

[0069] refer to Figure 5, a method is provided, the method comprising: receiving instructions from a first processor core to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval, as shown in box 502; processing the first job by: determining which user of one or more users is to be allocated on each resource of corresponding resources in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each resource in the set, and the second array being associated with a buffer size for each of the one or more users, as shown in box 504; and in response to successfully completing the first job, sending an indication of the determined allocation of the set of resources to the first processor core, as shown in box 506.

[0070] The instruction may include a set of pointers to locations of the first array and the second array in a memory. Determining the allocation of the set of resources may be based on a third array corresponding to a scheduling weight for each of the one or more users, wherein the instruction includes an indication of the location of the third array in the memory. The instruction may include a pointer to the location of a fourth array in the memory, and the determined resource allocation may be written to the memory at the fourth array. Determining the allocation of the set of resources may be based on a fifth array corresponding to a number of control channel elements assigned to each of the one or more users, and the instruction may include an indication of the location of the fifth array in the memory. The length of the first array may represent the number of resources in the set of resources, and the length of the second array may be equal to the number of one or more users. The first array may be an array of pointer values ​​pointing to one or more additional arrays stored in the memory, wherein each of the additional arrays may correspond to a corresponding one of the resources in the set of resources, and each element in each of the additional arrays may correspond to a value of a metric for users on each corresponding resource in the set. The metric may be a spectral efficiency metric. The method may also include: in response to detecting at least one error related to the first job, sending an indication of the at least one error to the first processor core. The method may also include: receiving at least one second instruction to process a second job, the second job being used to determine the allocation of a set of resources corresponding to another transmission layer in a transmission time interval for one or more users. The second job may be received before completing the first job. The second job may be received in response to completing the first job. The second job may be received from a processor core different from the first processor core. The instruction for the first job may be received via an event machine. The first job may be associated with: an identification value indicating that the first processor core corresponds to the first job and a queue positioning value identifying the positioning of the first job in a queue of the event machine. The first job may be processed by a hardware accelerator or a coprocessor. The method may include transmitting data via a first transmission layer based on the determined resource allocation. Each resource in the set of resources may include a frequency domain resource.

[0071] Another example of an embodiment includes a computer program that, when executed on a processor, causes execution of a method, the method comprising: receiving instructions from a first processor core to process a first job, the first job for determining an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer within a transmission time interval; processing the first job by: determining which of the one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each of the resources in the set, the second array being associated with a buffer size for each of the one or more users; and in response to successfully completing the first job, sending an indication of the determined allocation of the set of resources to the first processor core. The computer program may be a computer program product including a computer-readable medium carrying computer program code embodied therein for use by a computer.

[0072] According to another example embodiment, a device is provided, the device comprising: a component for receiving instructions from a first processor core to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval; a component for processing the first job by: determining which user of the one or more users is to be allocated on each resource of corresponding resources in the set based on a first array stored in a memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each resource in the set, and the second array being associated with a buffer size for each of the one or more users; and a component for sending an indication of the determined allocation of the set of resources to the first processor core in response to successful completion of the first job.

[0073] According to another example embodiment, a device is provided, the device comprising: a hardware accelerator; at least one processor including one or more processing cores; at least one non-volatile memory including computer program code, the at least one memory and the computer program code being configured to, together with the at least one processor and the hardware accelerator, cause the device to at least: receive instructions from a first processor core of the at least one processor at the hardware accelerator to process a first job, the first job being used to determine an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval; processing the first job by the hardware accelerator by: determining which of the one or more users is to be allocated on each resource of the corresponding resources in the set based on a first array stored in the memory and a second array stored in the memory, the first array being associated with a metric for each of the one or more users for each resource in the set, the second array being associated with a buffer size for each of the one or more users; and sending an indication of the determined allocation of the set of resources to the first processor core in response to successful completion of the first job.

[0074] The instruction may include: a set of pointers to the locations of the first array and the second array in the memory. The determination of the allocation of the set of resources may be based on a third array corresponding to the scheduling weights of each user in the one or more users, and the instruction may include an indication of the location of the third array in the memory. The instruction may include a pointer to the location of a fourth array in the memory, wherein the determined resource allocation is written to the memory at the fourth array. The determination of the allocation of the set of resources may be based on a fifth array corresponding to the number of control channel elements assigned to each user in the one or more users, and the instruction may include an indication of the location of the fifth array in the memory. The length of the first array may represent the number of resources in the set of resources, and the length of the second array may be equal to the number of one or more users. The first array may be an array of pointer values ​​pointing to one or more additional arrays stored in the memory, wherein each array in the additional arrays corresponds to a corresponding resource in the resources in the set of resources, and each element in each array in the additional arrays corresponds to the value of a metric for users on each corresponding resource in the set. The metric may be a spectral efficiency metric. At least one memory and computer program code may be configured to, together with at least one processor and hardware accelerator, cause the device to at least send an indication of at least one error to the first processor core in response to detecting at least one error related to the first job. At least one memory and computer program code may be configured to, together with at least one processor and hardware accelerator, cause the device to at least receive at least one second instruction to process a second job, the second job being used to determine the allocation of a set of resources corresponding to another transmission layer in a transmission time interval for one or more users. The second job may be received before completing the first job. The second job may be received in response to completing the first job. The second job may be received from a processor core different from the first processor core. The instruction of the first job may be received via an event machine. The first job may be associated with: an identification value indicating that the first processor core corresponds to the first job and a queue positioning value identifying the positioning of the first job in a queue of the event machine. The hardware accelerator may be a separate hardware or coprocessor. At least one memory and computer program code may be configured to, together with at least one processor and hardware accelerator, cause the device to transmit data via the first transmission layer at least based on the determined resource allocation. Each resource in the set of resources may include a frequency domain resource.

[0075] Without in any way limiting the scope, interpretation, or application of the claims appearing below, a technical effect of one or more of the example embodiments disclosed herein is to reduce the execution time of a MAC scheduler by offloading heavy computational tasks from one or more CPU cores to dedicated hardware. One or more CPU core dedicated hardware. Another technical effect of one or more of the example embodiments disclosed herein is to reduce energy and cost by using lower power cores because some of the work associated with the scheduler is performed by dedicated hardware. The example embodiments are beneficial for 5G, where the workload of the scheduler is expanded due to massive MIMO and the execution time is reduced due to smaller TTI durations.

[0076] The embodiments herein may be implemented in software (executed by one or more processors), hardware (e.g., application specific integrated circuits), or a combination of software and hardware. In example embodiments, the software (e.g., application logic, instruction sets) is maintained on any of a variety of conventional computer-readable media. In the context of this document, a "computer-readable medium" may be any medium or component that can contain, store, communicate, propagate, or transport instructions for use by or in connection with an instruction execution system, apparatus, or device, such as a computer, an example of which is provided in Figure 1 The computer-readable medium may include a computer-readable storage medium (e.g., memory 125, 155, 171 or other device), which may be any medium or component that can contain, store, and / or transmit instructions for use by or in connection with an instruction execution system, apparatus, or device, such as a computer. Computer-readable storage media do not include propagating signals.

[0077] If desired, the different functions discussed herein may be performed in different orders and / or concurrently with each other. In addition, if desired, one or more of the above functions may be optional or may be combined.

[0078] Although various aspects of the invention are set out in the independent claims, further aspects of the invention comprise other combinations of features from the described embodiments and / or the dependent claims with features of the independent claims, not just these combinations explicitly listed in the claims.

[0079] It should also be noted herein that although the above describes example embodiments of the invention, these descriptions should not be construed in a limiting sense. Rather, several variations and modifications may be made without departing from the scope of the invention as defined in the appended claims.

[0080] The following abbreviations that may be found in the specification and / or drawings are defined as follows:

[0081] eNB (or eNodeB) Evolved Node B (e.g., LTE base station)

[0082] gNB 5th Generation Node B

[0083] FD Frequency Domain

[0084] HW Hardware

[0085] I / F Interface

[0086] LTE Long Term Evolution

[0087] MME Mobility Management Entity

[0088] NCE Network Control Element

[0089] N / W Network

[0090] RRH Remote Radio Head

[0091] Rx Receiver

[0092] SD Spatial Domain

[0093] SGW Service Gateway

[0094] SOC System on Chip

[0095] TD Time Domain

[0096] Tx: Transmitter

[0097] UE User Equipment (e.g., wireless device, typically a mobile device)

Claims

1. A method for allocating resources, comprising: receiving instructions from a first processor core to process a first job for determining an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval; processing the first job by: determining which of the one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in a memory associated with a metric for each of the one or more users for each of the resources in the set and a second array stored in the memory associated with a buffer size for each of the one or more users; as well as responsive to successful completion of the first job, sending an indication of the determined allocation of the set of resources to the first processor core; The instruction includes: a set of pointers pointing to the locations of the first array and the second array in the memory.

2. According to the method of claim 1, determining the allocation of the set of resources is also based on a third array, the third array corresponding to the scheduling weight of each user of the one or more users, wherein the instruction includes an indication of the location of the third array in the memory. 3 . The method of claim 1 , wherein the instruction comprises a pointer to a location of a fourth array in the memory, wherein the determined resource allocation is written into the memory at the fourth array.

4. The method of claim 1 , wherein determining the allocation of the set of resources is also based on a fifth array corresponding to the number of control channel elements assigned to each of the one or more users, wherein the instruction includes an indication of a location of the fifth array in the memory.

5. The method according to claim 1, wherein: The length of the first array represents the number of resources in the set of resources, and the length of the second array is equal to the number of the one or more users.

6. A method according to claim 1, wherein the first array is an array of pointer values ​​pointing to one or more additional arrays stored in the memory, wherein each of the additional arrays corresponds to a corresponding one of the resources in the set of resources, and each element in each of the additional arrays corresponds to the value of the metric for the user on each corresponding resource in the set.

7. The method of claim 1, wherein the metric is a spectral efficiency metric.

8. The method according to claim 1, further comprising: In response to detecting at least one error associated with the first job, an indication of the at least one error is sent to the first processor core.

9. The method according to claim 1, further comprising: At least one second instruction is received to process a second job for determining, for the one or more users, an allocation of a set of resources corresponding to another transmission layer in the transmission time interval.

10. The method of claim 9, wherein the second job is received before the first job is completed. The method of claim 9 , wherein the second job is received in response to completing the first job.

12. The method of claim 9, wherein the second job is received from a processor core different from the first processor core.

13. The method of claim 1, wherein the instructions for the first job are received via an event machine.

14. The method of claim 13, wherein the first job is associated with an identification value indicating that the first processor core corresponds to the first job and a queue position value identifying a position of the first job in a queue of an event machine.

15. The method of claim 1, wherein the first job is processed by a hardware accelerator or a coprocessor.

16. The method according to claim 1, further comprising: Data is transmitted via the first transmission layer based on the determined resource allocation.

17. The method of claim 1, wherein each resource in the set of resources comprises a frequency domain resource.

18. An apparatus for allocating resources, comprising: means for receiving instructions from a first processor core to process a first job for determining an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval; means for processing the first job by: determining which of the one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in a memory associated with a metric for each of the one or more users for each of the resources in the set and a second array stored in the memory associated with a buffer size for each of the one or more users; as well as means for sending an indication of the determined allocation of the set of resources to the first processor core in response to successful completion of the first job; The instruction includes: a set of pointers pointing to the locations of the first array and the second array in the memory.

19. An apparatus according to claim 18, wherein the determination of the allocation of the set of resources is based on a third array corresponding to a scheduling weight for each of the one or more users, wherein the instruction includes an indication of a location of the third array in the memory.

20. The apparatus of claim 18, wherein the instruction comprises a pointer to a location of a fourth array in the memory, wherein the determined resource allocation is written into the memory at the fourth array.

21. An apparatus according to claim 18, wherein the determination of the allocation of the set of resources is also based on a fifth array corresponding to the number of control channel elements assigned to each of the one or more users, and wherein the instruction includes an indication of the location of the fifth array in the memory.

22. The apparatus of claim 18, wherein: The length of the first array represents the number of resources in the set of resources, and the length of the second array is equal to the number of the one or more users.

23. An apparatus according to claim 18, wherein the first array is an array of pointer values ​​pointing to one or more additional arrays stored in the memory, wherein each of the additional arrays corresponds to a corresponding one of the resources in the set of resources, and each element in each of the additional arrays corresponds to a value of the metric for the user on each corresponding resource in the set.

24. The apparatus of claim 18, wherein the metric is a spectral efficiency metric.

25. The apparatus of claim 18, further comprising: In response to detecting at least one error associated with the first job, means for sending an indication of the at least one error to the first processor core.

26. The apparatus of claim 18, further comprising: A means for receiving at least one second instruction to process a second job for determining, for the one or more users, an allocation of a set of resources corresponding to another transmission layer in the transmission time interval.

27. The apparatus of claim 26, wherein the second job is received prior to completion of the first job.

28. The apparatus of claim 26, wherein the second job is received in response to completion of the first job.

29. The apparatus of claim 26, wherein the second job is received from a processor core different from the first processor core.

30. The apparatus of claim 18, wherein the instructions for the first job are received via an event machine.

31. The apparatus of claim 30, wherein the first job is associated with: an identification value indicating that the first processor core corresponds to the first job and a queue position value identifying a position of the first job in a queue of an event machine.

32. The apparatus of claim 18, wherein the first job is processed by a hardware accelerator or a coprocessor.

33. The apparatus of claim 18, further comprising: Data is transmitted via the first transmission layer based on the determined resource allocation.

34. The apparatus of claim 18, wherein each resource in the set of resources comprises a frequency domain resource.

35. A base station comprising the apparatus according to any one of claims 18 to 34.

36. An apparatus for allocating resources, comprising: Hardware accelerators; at least one processor, the at least one processor comprising one or more processing cores; at least one non-transitory memory, the at least one non-transitory memory comprising computer program code, the at least one memory and the computer program code being configured to, together with the at least one processor and the hardware accelerator, cause the apparatus to at least: receiving, at the hardware accelerator, instructions from a first processor core of the at least one processor to process a first job for determining an allocation of a set of resources to one or more users, the set of resources corresponding to a first transmission layer in a transmission time interval; processing, by the hardware accelerator, the first job by: determining which of the one or more users is to be allocated on each of the corresponding resources in the set based on a first array stored in a memory associated with a metric for each of the one or more users for each of the resources in the set and a second array stored in the memory associated with a buffer size for each of the one or more users; as well as responsive to successful completion of the first job, sending an indication of the determined allocation of the set of resources to the first processor core; The instruction includes: a set of pointers pointing to the locations of the first array and the second array in the memory.

37. An apparatus according to claim 36, wherein the at least one non-volatile memory comprising computer program code, the at least one memory and the computer program code are configured to, together with the at least one processor and the hardware accelerator, cause the apparatus to also perform the method according to any one of claims 2 to 17.

38. A computer program product comprising a program code for causing the execution of the method according to any one of claims 1 to 17.

39. The computer program product of claim 38, wherein the computer program product comprises a computer readable medium carrying computer program code embodied therein for use by a computer.