Resource control device, resource control system, and resource control method

The resource control device addresses the challenge of simultaneous abstraction, flexibility, and fairness in FPGA resource management by employing a controller unit, user queues, and scheduler to optimize IP core usage and task prioritization, enhancing resource efficiency.

JP7586200B2Active Publication Date: 2024-11-19NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2022581083
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-02-10
Publication Date
2024-11-19
Estimated Expiration
2041-02-10

AI Technical Summary

Technical Problem

Existing FPGA resource management systems on cloud servers fail to simultaneously satisfy the requirements of abstraction, flexibility, controllability, and fairness, leading to inefficiencies in resource allocation and usage.

Method used

A resource control device with a controller unit, user queues, and a scheduler unit that manages IP cores and task priorities, allowing dynamic resource allocation and multi-stage scheduling to ensure abstraction, flexibility, and fairness among multiple users.

Benefits of technology

The solution enables efficient sharing of FPGA resources among multiple users, improving resource utilization and ensuring fair, flexible, and controllable allocation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007586200000001
    Figure 0007586200000001
  • Figure 0007586200000002
    Figure 0007586200000002
  • Figure 0007586200000003
    Figure 0007586200000003
Patent Text Reader

Abstract

A resource control unit (1) is provided with: a controller unit (2) that sets resources related to IP cores (81 to 83) of an FPGA 8 allowing user programs (3a, 3b) to execute tasks (6a to 6d); a common unit (4) that generates, for the respective user programs (3a, 3b), user queues (5a, 5b) as sets of queues each having a plurality of priorities and stores the tasks (6a to 6d) in the user queues (5a, 5b); and a scheduler unit (7) that selects a task to be executed by any one of the IP cores by multilevel scheduling in the user queues (5a, 5b) and across the user queues (5a, 5b).
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an FPGA resource control device, a resource control system, and a resource control method. [Background technology]

[0002] In recent years, FPGAs (Field-Programmable Gate Arrays) equipped with multiple inference convolutional neural networks per IP core (Intellectual Property Core) have come to be used (see Non-Patent Documents 1 and 2). Such FPGAs can be used for multiple use cases (e.g., pose identification, human detection, object detection, etc.). End users can use the FPGA for each task with a similar execution time order without rewriting the IP core functions. Common processing performed by the FPGA is started when the host CPU program of each end user passes the processing to the FPGA, and each processing in the FPGA is executed as a non-preemptive task, and then the execution result is returned to the CPU (Central Processing Unit). End-user-specific processing that cannot be realized by the common functions of the FPGA is executed within the CPU program. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Xilinx Vitis-AI, [Retrieved February 1, 2021], Internet<URL:https: / / github.com / Xilinx / Vitis-AI> [Non-Patent Document 2] M. Bacis, R. Brondolin and MD Santambrogio, "BlastFunction: an FPGA-as-a-Service system for Accelerated Serverless Computing," 2020 Design, Automation & Test in Europe Conference & Exhibition (DATE), Grenoble, France, 2020, pp. 852-857, doi: 10.23919 / DATE48585.2020.9116333. Summary of the Invention [Problem to be solved by the invention]

[0004] When installing FPGAs in cloud servers to provide services, it is desirable to satisfy the requirements of abstraction, flexibility, control, and fairness. Abstraction means that internal information of the cloud, such as IP core masks, is not exposed to users. Flexibility means that the amount of required resources, such as the number of IP cores in an FPGA, can be dynamically changed from outside the program. Controllability means that each user can set the priority of a task relative to the other tasks of that user. Fairness means that there is no difference between the amount of FPGA resources requested by the user and the execution time actually obtained. Simply running multiple programs for each user as multiple processes poses the problem that it is not possible to simultaneously satisfy the requirements for abstraction, flexibility, control, and fairness. Therefore, an object of the present invention is to appropriately share the functions of an FPGA among multiple users and to improve the resource efficiency of the FPGA. [Means for solving the problem]

[0005] In order to solve the above problems, the resource control device of the present invention includes a controller unit that sets resources related to IP cores of an FPGA on which a program executes a task, a common unit that creates user queues, which are a set of queues each having a plurality of priorities, for each of the programs and stores tasks in the user queues, and a scheduler unit that selects a task to be executed by any of the IP cores by multi-stage scheduling within and between the user queues. The controller unit secures the number of IP cores designated by the program, and includes an IP core usage control unit that, upon receiving a designation for exclusive use of the IP cores, creates and manages a map in which the IP cores are fixedly assigned to each program, and when the total number of IP cores newly designated by the program exceeds the number of IP cores in the FPGA, the IP core usage control unit does not accept this designation. It is characterized by: Other means will be described in the description of the embodiment of the invention. Effect of the Invention

[0006] According to the present invention, it is possible to appropriately share the functions of an FPGA among multiple users and improve the resource efficiency of the FPGA. [Brief description of the drawings]

[0007] [Figure 1] FIG. 2 is a configuration diagram of a resource control device for sharing an accelerator device in the present embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of an operation in a resource control device. [Diagram 3] FIG. 13 is a diagram illustrating an example of an exclusive use operation of an IP core by a resource control device. [Figure 4] FIG. 2 illustrates an example of a resource control device deployed in a user space of a host machine. [Diagram 5] FIG. 1 illustrates an example of a resource control device installed in an OS kernel of a host machine. [Figure 6] FIG. 11 is a diagram illustrating an example of a resource control system in which a controller unit is arranged in a user space of another host machine. [Figure 7] FIG. 1 is a configuration diagram of a resource control device of a comparative example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, a comparative example and an embodiment of the present invention will be described in detail with reference to the drawings. Comparative Example FIG. 7 is a configuration diagram of a resource control device 1G of the comparative example. The resource control device 1G includes an FPGA 8 implemented as hardware, and a CPU (not shown) executes a software program to realize a queue set 5G and a scheduler unit 7G. The resource control device 1G is, for example, a cloud server installed in a data center and providing services to each user via the Internet.

[0009] The FPGA 8 includes multiple IP cores 81 to 83, and executes multiple tasks simultaneously in a non-preemptive manner. In Fig. 7, the IP core 81 is written as "IP core #0", the IP core 82 is written as "IP core #1", and the IP core 83 is written as "IP core #2".

[0010] Queue set 5G includes multiple queues 50, 51 to 5F. The priority of queue 50 is lower than the priority of all other queues, so it is shown as "queue #0" in FIG. 7. The priority of queue 51 is higher than the priority of queue 50 but lower than the priority of all other queues, so it is shown as "queue #1" in FIG. 7. The priority of queue 5F is higher than the priority of all other queues, so it is shown as "queue #15" in FIG. 7.

[0011] The scheduler unit 7G includes a fixed priority scheduler unit 74, which schedules the tasks 6a to 6d stored in the queues 50, 51 to 5F in the order of priority of each queue, and causes the FPGA 8 to execute the tasks.

[0012] The resource control device 1G receives tasks 6a to 6d and the like from a plurality of user programs 3a, 3b, and has the FPGA 8 execute them. To this end, the user programs 3a, 3b are provided with an IP core mask setting unit 31 and a task priority setting unit 32. The task 6a is a task for identifying a person. The tasks 6b, 6c are tasks for identifying a pose. These tasks 6a to 6c are executed by the FPGA 8 in response to instructions from the user program 3a. The task 6d is a task for recognizing an object. The task 6d is executed by the FPGA 8 in response to instructions from the user program 3b.

[0013] The IP core mask setting unit 31 sets which of the IP cores 81 to 83 of the FPGA 8 is to execute a task, and which is not to execute a task. In other words, the core mask for the IP core to be used can be directly specified from the user programs 3a and 3b. Therefore, the internal information of the cloud server is exposed to the user, and there is a drawback in that it lacks abstraction.

[0014] The task priority setting unit 32 sets the priority of a task. The task priority setting unit 32 determines in which of the queues 50, 51 to 5F of the queue set 5G a task is to be stored. The task priority setting unit 32 allows each user to set the priority of a certain task relative to other tasks of the user. Specifically, the tasks 6b and 6c for identifying a pose can be executed before the task 6a for identifying a person. However, since the amount of FPGA 8 resources to be used is determined inside the user programs 3a and 3b, it cannot be dynamically changed from outside the user programs 3a and 3b, and there is a drawback in that it lacks flexibility.

[0015] The fixed priority scheduler unit 74 simply takes out tasks stored in the queue with the highest priority and assigns them to IP cores. Therefore, the user cannot specify the amount of resources he or she wants to use for a task. This can lead to a discrepancy between the amount of FPGA 8 resources required by the user program and the expected execution time that is actually obtained, which can cause a lack of fairness.

[0016] In other words, simply running multiple user programs as multiple processes may not simultaneously satisfy the requirements for abstraction, flexibility, control, and fairness.

[0017] <<Present Embodiment>> FIG. 1 is a configuration diagram of a resource control device 1 for sharing an accelerator device in this embodiment. The resource control device 1 includes an FPGA 8 implemented as hardware, and a CPU (not shown) executes a software program to embody a controller unit 2, a common unit 4, user queues 5a and 5b, and a scheduler unit 7. This resource control device 1 is, for example, installed in a data center, and is a cloud server that provides services to each user via the Internet.

[0018] The controller unit 2 includes a command reception unit 21, a user queue management unit 22, and an IP core usage control unit 23. The controller unit 2 has a function related to IP core setting, and sets resources related to the IP cores 81 to 83 of the FPGA 8 on which the program executes a task. The controller unit 2 checks the availability of the IP cores, specifies an IP core mask internally, and sets it in the scheduler unit 7. Therefore, information inside the cloud server is not exposed to the user programs 3a, 3b, etc. Furthermore, by providing the controller unit 2 with the command reception unit 21, resources can be dynamically controlled, providing flexibility.

[0019] The command reception unit 21 dynamically receives resource control commands from the user from outside the program. The resource control commands are commands that describe the number of IP cores to be used, whether the IP cores are to be used exclusively, etc. If the resource control command is not received, the command reception unit 21 notifies the user of that fact.

[0020] Each time a user program 3a, 3b, etc. is started, the user queue management unit 22 notifies the user queue creation unit 41 of the common unit 4 of the creation of a user queue for this program.

[0021] The IP core usage control unit 23 manages the occupied / free state of the IP cores 81-83 of the FPGA 8 in the physical host, secures the number of IP cores specified by the command reception unit 21, and creates and manages a map in which the IP cores are assigned in a fixed manner so as to be exclusive to each user as necessary. The IP core usage control unit 23 also notifies the scheduler unit 7 of the assignment information every time it updates the assignment information in which any of the IP cores 81-83 is assigned to a task. If the number of free IP cores is insufficient for the task that the user program is to execute, the IP core usage control unit 23 notifies the command reception unit 21 that the designation cannot be accepted. Then, the command reception unit 21 notifies the user that the resource control command from the user was not accepted.

[0022] The common unit 4 includes a user queue creation unit 41 and a user queue distribution unit 42. The common unit 4 creates a user queue, which is a set of queues each having a plurality of priorities, for each program, and stores tasks in the user queues. The user queue creation unit 41 receives information about available user queues from the controller unit 2, and creates a user queue for a new program every time the program is deployed and started.

[0023] When the user queue allocator 42 receives a task from a program, it selects a user queue corresponding to a user identifier assigned to the program, and stores the task in the queue of the corresponding priority based on the priority of the task in the user program 3a, 3b. The user programs 3a, 3b each include a task priority setting unit 32, which sets a priority for each task.

[0024] The task priority setting section 32 of the user program 3a sets the priority level #0 for the tasks 6a and 6b and hands them over to the common section 4, and sets the priority level #1 for the task 6c and hands it over to the common section 4. The task priority setting unit 32 of the user program 3b sets priority #1 for task 6d and passes it to the common unit 4. Tasks 6a and 6b are stored in queue 50 of user queue 5a, and task 6c is stored in queue 51 of user queue 5a. Task 6d is stored in queue 51 of user queue 5b.

[0025] The scheduler unit 7 includes an inter-user queue scheduler unit 71, an intra-user queue scheduler unit 72, and an IP core mask setting unit 73. The scheduler unit 7 selects a task to be executed by any of the IP cores 81 to 83 by multi-stage scheduling within the user queue and between the user queues. The inter-user queue scheduler unit 71 selects the user queues 5a and 5b from which a task is to be taken out by a fair algorithm such as round robin. These user queues 5a and 5b are a set of queues 50 and 51 with multiple priorities. The user queues 5a and 5b are a queue set with 16 levels of priority, but only two of them are shown in the figure.

[0026] The intra-user queue scheduler unit 72 selects a task to be executed according to an algorithm that takes priority into consideration, such as taking out a task from the queue with the highest priority among the user queues selected by the inter-user queue scheduler unit 71. The intra-user queue scheduler unit 72 schedules user tasks in the user queues 5a and 5b independently of the inter-user queue scheduler unit 71, thereby enabling priority control of each task. In other words, the intra-user queue scheduler unit 72 and the inter-user queue scheduler unit 71 realize the controllability of the resource control device 1.

[0027] The IP core mask setting unit 73 receives information from the controller unit 2, sets an IP core mask for each task, and controls so that unspecified IP cores are not used. Here, the IP core mask refers to the designation of an IP core for a task.

[0028] The common unit 4 prepares a plurality of independent user queues 5a, 5b each including a queue of each priority. Furthermore, the scheduler unit 7 includes an inter-user queue scheduler unit 71 that determines which user queue 5a, 5b to select. The common unit 4 ensures fairness in resource allocation of the FPGA 8 by scheduling in multiple stages using a priority control algorithm of the intra-user queue scheduler unit 72 and an algorithm of the inter-user queue scheduler unit 71 in combination.

[0029] FIG. 2 is a diagram showing an example of the operation of the resource control device 1. As shown in FIG. The controller unit 2 in FIG. 2 is notified in sequence of resource control instructions 20a to 20c. First, a resource control command 20a is notified to the controller unit 2. The resource control command 20a states that this is a deployment request for user program A (user program 3a) and that the number of IP cores to be used is two. By notifying the controller unit 2 of the resource control command 20a together with the user program 3a, the user program 3a is deployed and its execution starts. The controller unit 2 manages the mapping between the two IP cores in the FPGA 8 and the user program 3a. At this time, the two IP cores in the FPGA 8 are assigned to executing the task of the user program 3a.

[0030] Next, the controller unit 2 is notified of a resource control command 20b. The resource control command 20b describes that this is a deployment request for the user program B (user program 3b) and that the number of IP cores to be used is one. Even if the user program 3a is already being executed, by notifying the controller unit 2 of the resource control command 20b together with the user program 3b, the user program 3b is deployed and its execution starts. The controller unit 2 manages the mapping between one IP core in the FPGA 8 and the user program 3b.

[0031] At this time, two IP cores in the FPGA 8 are assigned to execute a task of the user program 3a, and the remaining IP core is assigned to execute a task of the user program 3b.

[0032] Finally, the controller unit 2 is notified of a resource control command 20c. The resource control command 20c is a command related to a user program C (not shown). The user programs 3a and 3b are already being executed, and all IP cores in the FPGA 8 are assigned to these user programs 3a and 3b. Since the deployment request exceeds the resource capacity, the controller unit 2 notifies the user of this fact and does not deploy the user program C.

[0033] When the user programs 3a and 3b are deployed, the in-user-queue scheduler unit 72 of the scheduler unit 7 adjusts the ratio of the execution times of the user programs 3a and 3b to 2:1 using an algorithm such as round robin. This ratio of execution times is equal to the ratio of the numbers of IP cores in the resource control instructions 20a and 20b. This allows the controller unit 2 to fairly allocate the IP cores of the FPGA 8 to the user programs 3a and 3b.

[0034] FIG. 3 is a diagram showing an example of an exclusive use operation of an IP core by the resource control device 1. In FIG. The controller unit 2 in FIG. 3 is notified of resource control instructions 20a and 20b. The resource control instruction 20a describes that this is a deployment request related to the user program A (user program 3a), that the number of IP cores to be used is two, and that the IP cores will be used exclusively. When the user notifies the controller unit 2 of the user program 3a together with the resource control command 20a, the user program 3a is deployed and its execution is started. The controller unit 2 manages two IP cores in the FPGA 8, and manages the exclusive mapping between the user program 3a and the IP cores. At this time, the two IP cores in the FPGA 8 are exclusively assigned to the task execution of the user program 3a.

[0035] The resource control instruction 20b describes that this is a deployment request for user program B (user program 3b), that one IP core is to be used, and that the IP core is to be used exclusively. Even if the user program 3a is already being executed, the user can notify the controller unit 2 of the user program 3b together with a resource control command 20b, whereby the user program 3b is deployed and its execution is started. The controller unit 2 manages the mapping between the IP cores in the FPGA 8 and the user programs 3a and 3b. At this time, two IP cores in the FPGA 8 are exclusively assigned to execute the task of the user program 3a, and the remaining IP core is exclusively assigned to execute the task of the user program 3b.

[0036] FIG. 4 is a diagram showing an example of the resource control device 1 deployed in the user space of the host machine 1B. The host machine 1B includes a CPU 93 and an FPGA 8 as a hardware layer, and has an OS (Operating System) 92 installed therein. In the user space of the host machine 1B, a controller unit 2 and an FPGA library 91 are embodied, and user programs 3a and 3b are deployed therein.

[0037] The FPGA library 91 includes a multiqueue 5 and a scheduler unit 7, and operates as the above-mentioned resource control device 1 in combination with the controller unit 2. A new user queue is generated in the multiqueue 5 every time a user program is deployed. The scheduler unit 7 includes sections corresponding to the inter-user queue scheduler unit 71, intra-user queue scheduler unit 72, and IP core mask setting unit 73 shown in FIG. 1.

[0038] The controller unit 2 sends commands to the multiqueue 5 and the scheduler unit 7. The FPGA library 91 causes the FPGA 8 and the IP cores 81 to 83 to execute tasks via an FPGA driver 94 installed in an OS 92.

[0039] FIG. 5 is a diagram showing an example of the resource control device 1 deployed in the kernel space of the OS 92 of the host machine 1C. The host machine 1C includes a CPU 93 and an FPGA 8 as a hardware layer, and has an OS 92 installed therein. A controller unit 2, a CPU scheduler 921, and an FPGA driver 94 are installed in the kernel space of the OS 92 of the host machine 1B. An FPGA library 91 and user programs 3a and 3b are deployed in the user space of the host machine 1C.

[0040] The controller unit 2 includes a CPU control unit 24, a device control unit 25, a GPU (Graphic Processing Unit) control unit 26, and an FPGA control unit 27. The CPU control unit 24 is a part that controls the cores 931 to 932 configured in the CPU 93, and sends instructions to the CPU scheduler 921.

[0041] The GPU control unit 26 is a unit that controls a GPU (not shown). The FPGA control unit 27 is a unit that controls the FPGA 8, and includes units corresponding to the command reception unit 21, the user queue management unit 22, and the usage IP core control unit 23 shown in FIG.

[0042] The FPGA driver 94 includes a multiqueue 5 and a scheduler unit 7. The multiqueue 5 and the scheduler unit 7 are controlled by the FPGA control unit 27. A new user queue is generated in the multiqueue 5 every time a new user program is deployed. The scheduler unit 7 includes parts corresponding to the inter-user queue scheduler unit 71, the intra-user queue scheduler unit 72, and the IP core mask setting unit 73 shown in FIG. 1.

[0043] FIG. 6 is a diagram showing an example of a resource control system in which a controller unit 2 is arranged in a user space of another host machine 1D. The resource control system shown in Fig. 6 includes a host machine 1E and a host machine 1D equipped with a controller unit 2. The host machine 1D includes a CPU 93 as a hardware layer, and an OS 92 is installed. The controller unit 2 is embodied in the user space of the host machine 1D. This controller unit 2 has the same functions as the controller unit 2 shown in Fig. 1.

[0044] The host machine 1E includes a CPU 93 and an FPGA 8 as a hardware layer, and has an OS 92 installed. An FPGA library 91 is implemented in the user space of the host machine 1E, and user programs 3a and 3b are deployed therein. This FPGA library 91 has the same functions as the FPGA library 91 shown in FIG.

[0045] The FPGA library 91 includes a multiqueue 5 and a scheduler unit 7, and operates as the above-mentioned resource control device 1 in combination with the controller unit 2 of the host machine 1D. Every time a user program is deployed, a new user queue corresponding to the user program is generated in the multiqueue 5. The scheduler unit 7 includes sections equivalent to the inter-user queue scheduler unit 71, intra-user queue scheduler unit 72, and IP core mask setting unit 73 shown in FIG. 1.

[0046] The controller unit 2 sends commands to the multiqueue 5 and the scheduler unit 7. The FPGA library 91 causes the FPGA 8 and the IP cores 81 to 83 to execute tasks via an FPGA driver 94 installed in an OS 92.

[0047] "effect" The effects of the resource control device, the resource control system, and the resource control method according to the present invention will be described below. .

[0048] 《Claim 1》 A controller that configures resources related to IP cores in an FPGA for executing tasks by a program; a common unit that creates a user queue, which is a set of queues each having a plurality of priorities, for each of the programs and stores tasks in the user queues; a scheduler unit that selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; A resource control device comprising:

[0049] This allows the FPGA functions to be shared among multiple users in an optimal manner, improving the resource efficiency of the FPGA.

[0050] 《Claim 2》 The scheduler unit a user queue scheduler unit that selects a user queue from which a task is to be taken out; a user queue intra-scheduler unit for taking out a task from a queue having a highest priority among the queues in which tasks are registered, within the user queues selected by the inter-user queue scheduler unit; 2. The resource control device according to claim 1, further comprising:

[0051] This enables multi-stage scheduling between and within multiple users, improving the resource efficiency of FPGAs.

[0052] 《Claim 3》 The scheduler unit The method further includes an IP core mask setting unit that controls each task so as not to use an IP core that is not specified for that task. 2. The resource control device according to claim 1 .

[0053] This enables multi-stage scheduling between and within multiple users, improving the resource efficiency of FPGAs.

[0054] 《Claim 4》 The controller unit a control unit for controlling an IP core to be used, which secures a number of IP cores designated by the program and, upon receiving a designation for exclusive use of the IP cores, creates and manages a map in which the IP cores are fixedly assigned to each program; The IP core control unit includes: If the total number of IP cores newly specified by the program exceeds the number of IP cores in the FPGA, the specification is not accepted. 2. The resource control device according to claim 1 .

[0055] This allows the FPGA functions to be shared appropriately among multiple users. 《Claim 5》 The common part is A user queue creating unit is provided for creating a user queue for a new program each time the program is started. 2. The resource control device according to claim 1 .

[0056] This allows multiple users to share FPGA resources fairly.

[0057] 《Claim 6》 The common part is When a task is received from the program, a user queue related to the program is selected based on an identifier, and the task is registered in the user queue based on a task priority. 2. The resource control device according to claim 1 .

[0058] This allows multiple users to share FPGA resources fairly.

[0059] 《Claim 7》 A controller that configures resources related to IP cores in an FPGA for executing tasks by a program; a common unit that creates a user queue, which is a set of queues each having a plurality of priorities, for each of the programs and stores tasks in the user queues; a scheduler unit that selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; A resource control system comprising:

[0060] This allows the FPGA functions to be shared among multiple users in an optimal manner, improving the resource efficiency of the FPGA.

[0061] 《Claim 8》 A step in which a controller unit sets resources related to an IP core of an FPGA in which a program executes a task; a common unit creating a user queue, which is a set of queues each having a plurality of priorities, for each of the programs, and storing tasks in the user queues; A scheduler unit selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; 2. A resource control method comprising the steps of:

[0062] This allows the FPGA functions to be shared among multiple users in an optimal manner, improving the resource efficiency of the FPGA. [Explanation of symbols]

[0063] 1,1G Resource Control Unit 1B, 1C, 1D, 1E host machines 2. Controller 20a~20c Resource control instructions 21 Command Reception Section 22 User Queue Management Unit 23 IP core control section 24 CPU control unit 25 Device control section 26 GPU control unit 27 FPGA control unit 3a User Program 3b User Program 31 IP core mask setting section 32 Task Priority Setting Unit 4 Common areas 41 User Queue Creation Section 42 User queue distribution unit 5 Multi-Queue 5a,5b User Queue 5G Queue Set 50,51,5F Queue Tasks 6a-6d 7 Scheduler 7G Scheduler 71 User inter-queue scheduler 72 User Queue Scheduler 73 IP core mask setting section 74 Fixed Priority Scheduler 8 FPGA 81~83 IP Core 91 FPGA Library 92 OS 921 CPU Scheduler 93 CPU 931~933 Core 94 FPGA Drivers

Claims

1. A controller that configures resources related to IP cores of an FPGA on which a program executes a task; a common unit that creates a user queue, which is a set of queues each having a plurality of priorities, for each of the programs and stores tasks in the user queues; a scheduler unit that selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; Equipped with The controller unit a usage IP core control unit that secures the number of IP cores designated by the program and, upon receiving a designation for exclusive use of the IP cores, creates and manages a map in which the IP cores are fixedly assigned to each program; The IP core control unit includes: If the total number of IP cores newly specified by the program exceeds the number of IP cores in the FPGA, the specification is not accepted. A resource control device comprising:

2. The scheduler unit a user queue scheduler unit that selects a user queue from which a task is to be taken out; a user queue intra-scheduler unit for taking out a task from a queue having a highest priority among the queues in which tasks are registered, within the user queues selected by the inter-user queue scheduler unit; The resource control device according to claim 1, further comprising:

3. The scheduler unit The method further includes an IP core mask setting unit that controls so as not to use IP cores that are not specified for each task.

2. The resource control device according to claim 1 .

4. The common part is A user queue creating unit is provided for creating a user queue for a new program each time the program is started.

2. The resource control device according to claim 1 .

5. The common part is When a task is received from the program, a user queue related to the program is selected based on an identifier, and the task is registered in the user queue based on a task priority.

2. The resource control device according to claim 1 .

6. A controller that configures resources related to IP cores of an FPGA on which a program executes a task; a common unit that creates a user queue, which is a set of queues each having a plurality of priorities, for each of the programs and stores tasks in the user queues; a scheduler unit that selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; Equipped with The controller unit a usage IP core control unit that secures the number of IP cores designated by the program and, upon receiving a designation for exclusive use of the IP cores, creates and manages a map in which the IP cores are fixedly assigned to each program; The IP core control unit includes: If the total number of IP cores newly specified by the program exceeds the number of IP cores in the FPGA, the specification is not accepted. A resource control system comprising:

7. A step in which a controller unit sets resources related to an IP core of an FPGA on which a program executes a task; a common unit creating a user queue, which is a set of queues each having a plurality of priorities, for each of the programs, and storing tasks in the user queues; A scheduler unit selects a task to be executed by any one of the IP cores by multi-stage scheduling within the user queue and between the user queues; a step in which an IP core control unit of the controller unit secures the number of IP cores designated by the program, and when a designation of exclusive use of the IP cores is received, creates and manages a map in which IP cores are fixedly assigned to each program; a step of the IP core control unit not accepting the designation if the total number of IP cores newly designated by the program exceeds the number of IP cores in the FPGA; A resource control method comprising:

Citation Information

Patent Citations

  • Message controller

    JP2000155693A

  • Semiconductor integrated circuit

    JP2007004340A

  • Data distribution apparatus and data distribution method

    JP2015130135A

  • System and method for supporting charging for use of accelerator part

    JP2019082819A