CN115061789B - Transient current control method and related devices - Google Patents

Transient current control method and related devices Download PDF

Info

Publication number
CN115061789B
CN115061789B CN202210648644.6A CN202210648644A CN115061789B CN 115061789 B CN115061789 B CN 115061789B CN 202210648644 A CN202210648644 A CN 202210648644A CN 115061789 B CN115061789 B CN 115061789B
Authority
CN
China
Prior art keywords
data processing
time interval
processing unit
calling
extended
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210648644.6A
Other languages
Chinese (zh)
Other versions
CN115061789A (en
Inventor
倪怡芳
王斌
陈俊
潘于
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hygon Information Technology Co Ltd
Original Assignee
Hygon Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hygon Information Technology Co Ltd filed Critical Hygon Information Technology Co Ltd
Priority to CN202210648644.6A priority Critical patent/CN115061789B/en
Publication of CN115061789A publication Critical patent/CN115061789A/en
Application granted granted Critical
Publication of CN115061789B publication Critical patent/CN115061789B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00Arrangements for program control, e.g. control units
    • G06F9/06Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46Multiprogramming arrangements
    • G06F9/48Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806Task transfer initiation or dispatching
    • G06F9/4843Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G06F9/4881Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Exchange Systems With Centralized Control (AREA)

Abstract

The embodiment of the application provides a transient current control method and a related device, wherein the transient current control method comprises the following steps: receiving a data processing unit call request; determining an extended calling time interval between each data processing unit which needs to be called currently according to the data processing unit calling request, wherein the extended calling time interval is larger than a standard calling time interval; and sending a call request to each data processing unit in turn according to the call prolonging time interval. The embodiment of the application can ensure the performance of the integrated circuit while realizing the control of the transient current.

Description

瞬态电流控制方法及相关装置Transient current control method and related devices

技术领域Technical field

本申请实施例涉及计算机技术领域,尤其涉及一种瞬态电流控制方法及相关装置。Embodiments of the present application relate to the field of computer technology, and in particular, to a transient current control method and related devices.

背景技术Background technique

随着复杂的半导体技术的发展,集成电路的研究、发展也逐步深入,超大规模集成电路的需求也越来越多。With the development of complex semiconductor technology, the research and development of integrated circuits have gradually deepened, and the demand for very large-scale integrated circuits has also increased.

超大规模集成电路所包含的晶体管数量非常大,随之而来的,超大规模集成电路的功耗就会越来越大,其中,瞬态电流(DI/DT)过大所带来的超大规模集成电路中的很多负载过载运行而造成的这些负载的损耗例如发热等问题,并且瞬态电流过大还会造成供电电压瞬间下降,过低的供电电压会造成超大规模集成电路的功能错误,带来严重的后果。The number of transistors contained in VLSI circuits is very large. As a result, the power consumption of VLSI circuits will become larger and larger. Among them, the VLSI caused by excessive transient current (DI/DT) The overload operation of many loads in integrated circuits causes problems such as heat generation, etc., and excessive transient current will also cause an instant drop in the supply voltage. An excessively low supply voltage will cause functional errors in very large scale integrated circuits, leading to There will be serious consequences.

为此,在通用图形处理器(GPGPU)设计中,通常使用降低输入频率的方法以使瞬态电流下降;或者设计整体超大规模集成电路时增设硬件设备,以控制瞬态电流来稳定整体负载的正常工作。For this reason, in the design of general-purpose graphics processing units (GPGPU), the method of reducing the input frequency is usually used to reduce the transient current; or when designing the overall VLSI circuit, hardware devices are added to control the transient current to stabilize the overall load. normal work.

但是,使用降频的方法不仅会影响超大规模集成电路的性能,还会提高对于频率的控制要求;而增加硬件设备则不仅会因器件的增多影响电路电流的稳定性,还会增加封测的难度。However, using the frequency reduction method will not only affect the performance of VLSI, but also increase the frequency control requirements; and adding hardware equipment will not only affect the stability of the circuit current due to the increase in devices, but also increase the packaging and testing requirements. Difficulty.

因此,如何在实现对瞬态电流的控制的同时,保证集成电路的性能,就成为亟需解决的技术问题。Therefore, how to control the transient current while ensuring the performance of the integrated circuit has become an urgent technical problem that needs to be solved.

发明内容Contents of the invention

本申请实施例解决的技术问题是如何在实现对瞬态电流的控制的同时,保证集成电路的性能。The technical problem solved by the embodiments of the present application is how to control the transient current while ensuring the performance of the integrated circuit.

为解决上述问题,本申请实施例提供一种瞬态电流控制方法及相关装置,包括:In order to solve the above problems, embodiments of the present application provide a transient current control method and related devices, including:

第一方面,本申请实施例提供一种瞬态电流控制方法,适用于集成电路的资源分配模块,所述方法包括:In a first aspect, embodiments of the present application provide a transient current control method, which is suitable for a resource allocation module of an integrated circuit. The method includes:

接收数据处理单元调用请求;Receive data processing unit call request;

根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔;According to the data processing unit call request, determine the extended call time interval between each data processing unit that currently needs to be called, and the extended call time interval is greater than the standard call time interval;

根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。According to the extended calling time interval, calling requests are sent to each of the data processing units in sequence.

第二方面,本申请实施例提供一种瞬态电流控制装置,所述装置包括:In a second aspect, embodiments of the present application provide a transient current control device, which includes:

接收模块,适于接收数据处理单元调用请求;The receiving module is suitable for receiving data processing unit call requests;

时间间隔确定模块,适于根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔;A time interval determination module, adapted to determine an extended call time interval between each data processing unit that currently needs to be called based on the data processing unit call request, where the extended call time interval is greater than the standard call time interval;

发送模块,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。The sending module is adapted to send calling requests to each of the data processing units in sequence according to the extended calling time interval.

第三方面,本申请实施例还提供一种集成电路,以实现如第一方面所述的瞬态电流控制方法。In a third aspect, embodiments of the present application further provide an integrated circuit to implement the transient current control method described in the first aspect.

第四方面,本申请实施例还提供一种电子设备,包括如第四方面所述的集成电路。In a fourth aspect, embodiments of the present application further provide an electronic device, including the integrated circuit described in the fourth aspect.

与现有技术相比,本申请实施例的技术方案具有以下优点:Compared with the existing technology, the technical solutions of the embodiments of the present application have the following advantages:

本申请实施例提供的瞬态电流控制方法,资源分配模块接收数据处理单元调用请求,确定数据处理单元调用请求中所需要调用的各个数据处理单元的调用时间,通过将各个数据处理单元的之间间隔的调用时间延长,以完成数据处理单元的逐个调用。这样,可以提高将需要调用的各个数据处理单元的加载启动时间相互错开的概率,以降低正在调用的数据处理单元从启动到达到稳定工作状态的期间内同时加载启动其他数据处理单元的可能性,从而减少多个数据处理单元同时启动时瞬态电流瞬间增大而破坏集成电路芯片的性能的情况的发生,降低负载的损耗,提高集成电路内部器件的性能,延长集成电路的使用寿命。In the transient current control method provided by the embodiment of the present application, the resource allocation module receives the data processing unit call request, determines the calling time of each data processing unit that needs to be called in the data processing unit call request, and divides the time between each data processing unit. The interval call time is extended to complete the call-by-call of the data processing unit. In this way, the probability of staggering the loading and starting times of each data processing unit that needs to be called can be increased to reduce the possibility of loading and starting other data processing units at the same time during the period from startup to stable working state of the data processing unit being called. This reduces the occurrence of instantaneous increase in transient current that damages the performance of the integrated circuit chip when multiple data processing units are started at the same time, reduces the loss of the load, improves the performance of the internal components of the integrated circuit, and extends the service life of the integrated circuit.

附图说明Description of drawings

为了更清楚地说明本申请实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据提供的附图获得其他的附图。In order to explain the embodiments of the present application or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only This is an embodiment of the present application. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without exerting creative efforts.

图1为GPGPU的内部基本架构示意图。Figure 1 is a schematic diagram of the internal basic architecture of GPGPU.

图2为GPGPU的数据处理单元加载任务时与时间的关系示意图。Figure 2 is a schematic diagram of the relationship between the GPGPU data processing unit loading tasks and time.

图3为数据处理单元加载任务时基本的瞬态电流的变化示意图。Figure 3 is a schematic diagram of the basic transient current changes when the data processing unit loads a task.

图4为本申请实施例提供的瞬态电流控制方法的一流程图。FIG. 4 is a flow chart of a transient current control method provided by an embodiment of the present application.

图5为本申请实施例提供的瞬态电流控制方法的数据处理单元加载任务时与时间的关系示意图。FIG. 5 is a schematic diagram of the relationship between task loading and time in the data processing unit of the transient current control method provided by the embodiment of the present application.

图6为本申请实施例提供的瞬态电流控制方法下数据处理单元加载任务时的瞬态电流的变化示意图。FIG. 6 is a schematic diagram of the change of the transient current when the data processing unit loads a task under the transient current control method provided by the embodiment of the present application.

图7为本申请实施例提供的瞬态电流控制方法的另一流程图。FIG. 7 is another flowchart of the transient current control method provided by the embodiment of the present application.

图8为本申请实施例提供的瞬态电流控制方法的又一流程图。FIG. 8 is another flow chart of the transient current control method provided by the embodiment of the present application.

图9为本申请实施例所提供的瞬态电流控制装置的可选结构框图。Figure 9 is an optional structural block diagram of the transient current control device provided by the embodiment of the present application.

具体实施方式Detailed ways

下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

图1示例性的示出了GPGPU的内部基本架构示意图,如图1所示,该基本架构可以包括:资源分配模块01和数据处理单元02。Figure 1 exemplarily shows a schematic diagram of the internal basic architecture of a GPGPU. As shown in Figure 1, the basic architecture may include: a resource allocation module 01 and a data processing unit 02.

在超大规模集成电路中,尤其在通用图形处理器(GPGPU)中,数据处理单元(DU)02是构建整个GPGPU的基础模块,是最小的独立功能单元。In very large-scale integrated circuits, especially in general-purpose graphics processing units (GPGPU), the data processing unit (DU) 02 is the basic module that builds the entire GPGPU and is the smallest independent functional unit.

资源分配模块01负责向数据处理单元02分配运算任务。数据处理单元02根据资源分配模块01分配的运算任务,完成取指、译码、执行和写回等执行运算任务的流程。The resource allocation module 01 is responsible for allocating computing tasks to the data processing unit 02. The data processing unit 02 completes the process of executing computing tasks such as fetching, decoding, executing and writing back according to the computing tasks assigned by the resource allocation module 01 .

其中,资源分配模块01分配的运算任务来自于经处理程序发出的运算任务请求,当处理程序发现集成电路需要进行运算任务时,向资源分配模块01发送运算任务的请求,资源分配模块01根据接收到的运算任务请求,向数据处理单元02分配运算任务。Among them, the computing tasks allocated by the resource allocation module 01 come from the computing task requests issued by the processing program. When the processing program finds that the integrated circuit needs to perform computing tasks, it sends the computing task request to the resource allocation module 01. The resource allocation module 01 receives the Upon receipt of the computing task request, the computing task is assigned to the data processing unit 02.

数据处理单元02接收到运算任务时,以一定的时间(窗口期)逐渐启动,执行运算任务。When the data processing unit 02 receives the computing task, it gradually starts to execute the computing task within a certain time (window period).

容易理解的是,在窗口期的时间段内,数据处理单元02启动,会产生较大的瞬态电流,而瞬态电流造成的损耗会占据绝大多数的GPGPU的功耗,当多个数据处理单元02同时进行处理运算任务时,就会出现一个窗口期内同时有两个或多个数据处理单元02在启动,即不同的数据处理单元02的窗口期可能重叠,各个数据处理单元02的窗口期重叠会造成瞬态电流进一步加大,不仅对GPGPU的功耗影响很大,会造成超大规模集成电路的功能错误,从而影响GPGPU芯片的性能。It is easy to understand that during the window period, when the data processing unit 02 is started, a large transient current will be generated, and the loss caused by the transient current will account for the vast majority of the power consumption of the GPGPU. When multiple data When the processing units 02 perform processing tasks at the same time, there will be two or more data processing units 02 starting at the same time within a window period, that is, the window periods of different data processing units 02 may overlap, and the window periods of each data processing unit 02 may overlap. Overlapping window periods will cause the transient current to further increase, which will not only have a great impact on the power consumption of the GPGPU, but also cause functional errors in the VLSI circuit, thereby affecting the performance of the GPGPU chip.

其中,窗口期表示的是一个数据处理单元02初始启动时刻到运行稳定时刻之间的时间间隔。一个数据处理单元02在一个窗口期内被运算任务触发启动并逐渐运行至稳定。The window period represents the time interval between the initial startup time of a data processing unit 02 and the stable operation time. A data processing unit 02 is triggered to start by a computing task within a window period and gradually runs until stable.

具体的,请参考图2,图2示例性的示出了GPGPU的数据处理单元加载任务时与时间的关系示意图。Specifically, please refer to FIG. 2 , which is a schematic diagram illustrating the relationship between the time when the data processing unit of the GPGPU loads a task and time.

为了方便对各个数据处理单元DU在加载任务时的具体情况,在图2中仅展示了部分数据处理单元DU的工作情况进行解释说明。In order to facilitate the explanation of the specific conditions of each data processing unit DU when loading tasks, only the working conditions of some data processing units DU are shown in Figure 2.

如图中所示,DU0在时刻5处接收到资源分配模块02分配的运算任务,根据运算任务DU0被触发启动,在窗口期的时间段内DU0开始逐渐运行,直到时刻9满负荷运算,然后稳定运行。在DU0被触发启动后时间间隔T后,DU1在时刻6被分配到运算任务开始启动运行,直到时刻10处达到满负荷稳定运行。同样的DU2和DU3都相应的在接收到分配的运算任务后在各自的窗口期的时间段内,逐渐运行直到达到满负荷运算,稳定运行。As shown in the figure, DU0 receives the computing task assigned by the resource allocation module 02 at time 5, and is triggered to start according to the computing task DU0. During the window period, DU0 begins to run gradually until time 9 is fully loaded, and then Stable operation. After the time interval T after DU0 is triggered to start, DU1 is assigned to the computing task at time 6 and starts running until it reaches full load and stable operation at time 10. Similarly, DU2 and DU3 will gradually run within their respective window periods after receiving the assigned computing tasks until they reach full load computing and operate stably.

可以看到,在一个窗口期的时间间隔内,即图2中DU0触发时刻5到DU0运行平稳的时刻9的时间段内,同时运行了四个数据处理单元02,包括:DU0、DU1、DU2和DU3,这样在一个窗口期的时间间隔范围内,就会包括有四个数据处理单元02同时运行时产生的瞬态电流,并且由于不同的数据处理单元02之间的时间间隔在所有的数据处理单元02执行运算任务所需的总时间范围内的占比是非常小的,因此多个数据处理单元02执行运算任务可以视为是同时执行的,那么多个数据处理单元02产生的瞬态电流的累加也就可以看做是同一时刻产生的瞬态电流,这样就会造成在某一时刻瞬态电流的急剧增大。It can be seen that within the time interval of a window period, that is, the time period from the trigger time 5 of DU0 to the stable time 9 of DU0 in Figure 2, four data processing units 02 are running at the same time, including: DU0, DU1, and DU2 and DU3, so that within the time interval of a window period, the transient current generated when four data processing units 02 are running at the same time will be included, and due to the time intervals between different data processing units 02, all data The proportion of the total time required by the processing unit 02 to execute the computing task is very small. Therefore, the execution of the computing tasks by multiple data processing units 02 can be regarded as being executed simultaneously. Then the transient state generated by the multiple data processing units 02 The accumulation of current can also be regarded as the transient current generated at the same time, which will cause a sharp increase in the transient current at a certain moment.

具体的,请参考图3,图3示例性的示出了数据处理单元加载任务时基本的瞬态电流的变化示意图。Specifically, please refer to FIG. 3 , which exemplarily shows a schematic diagram of basic transient current changes when the data processing unit loads a task.

如图中所示,当集成电路需要进行任务运算时,资源分配模块01分配运算任务,多个数据处理单元02被运算任务触发启动,在时刻1340处就会产生巨大的瞬态电流变化,该瞬态电流即为多个数据处理单元02执行运算任务时,加载任务的瞬态电流的累加。As shown in the figure, when the integrated circuit needs to perform task calculations, the resource allocation module 01 allocates the calculation tasks, and multiple data processing units 02 are triggered by the calculation tasks. At time 1340, a huge transient current change will occur. The transient current is the accumulation of transient currents loading tasks when multiple data processing units 02 execute computing tasks.

瞬态电流过大一方面会引起集成电路上的很多负载过载运行,造成这些负载的损耗,例如集成电路内部环境过热等问题;另一方面,瞬态电流过大还会造成供电电压瞬间下降,如果电压低于了一些阈值电压,则会造成集成电路内部器件的功能错误,带来严重的后果,因此控制瞬态电流就是集成电路中一个非常关键的议题。On the one hand, excessive transient current will cause overload operation of many loads on the integrated circuit, resulting in losses of these loads, such as overheating of the internal environment of the integrated circuit. On the other hand, excessive transient current will also cause an instant drop in the supply voltage. If the voltage is lower than some threshold voltage, it will cause functional errors of the internal components of the integrated circuit, bringing serious consequences. Therefore, controlling transient current is a very critical issue in integrated circuits.

为了解决前述问题,本申请实施例提供了一种瞬态电流控制方法,可以在实现对瞬态电流的控制的同时,保证集成电路的性能。In order to solve the aforementioned problems, embodiments of the present application provide a transient current control method, which can control the transient current while ensuring the performance of the integrated circuit.

为了说明本申请实施例的实施方式,请参考图4-图6,图4示出了本申请实施例提供的瞬态电流控制方法的一流程图;图5示例性的示出了本申请实施例提供的瞬态电流控制方法的数据处理单元加载任务时与时间的关系示意图;In order to illustrate the implementation of the embodiment of the present application, please refer to Figures 4-6. Figure 4 shows a flow chart of the transient current control method provided by the embodiment of the present application; Figure 5 exemplarily shows the implementation of the present application. A schematic diagram of the relationship between the data processing unit loading tasks and time for the transient current control method provided in the example;

图6示例性的示出了本申请实施例提供的瞬态电流控制方法下数据处理单元加载任务时的瞬态电流的变化示意图。FIG. 6 is a schematic diagram illustrating the change of the transient current when the data processing unit loads a task under the transient current control method provided by the embodiment of the present application.

如图4所示,本申请实施例所提供的瞬态电流控制方法,适用于集成电路的资源分配模块,可以包括如下步骤:As shown in Figure 4, the transient current control method provided by the embodiment of the present application is suitable for the resource allocation module of the integrated circuit and may include the following steps:

在步骤S10中,接收数据处理单元调用请求。In step S10, a data processing unit calling request is received.

容易理解的是,数据处理单元调用请求是由处理程序通过处理器发送。It is easy to understand that the data processing unit call request is sent by the handler through the processor.

需要说明的是,当多个数据处理单元调用请求时,资源分配模块一个一个接收,本申请以当前接收的数据处理单元调用请求为基础进行说明。It should be noted that when multiple data processing unit call requests are received, the resource allocation module receives them one by one. This application explains based on the currently received data processing unit call request.

在步骤S11中,根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔。In step S11, based on the data processing unit calling request, an extended calling time interval between each data processing unit that currently needs to be called is determined, and the extended calling time interval is greater than the standard calling time interval.

基于数据处理单元调用请求确定需要调用的数据处理单元02,容易理解的是,一个数据处理单元调用请求根据需要完成的运算任务需求需要调用多个数据处理单元02。The data processing unit 02 that needs to be called is determined based on the data processing unit call request. It is easy to understand that one data processing unit call request needs to call multiple data processing units 02 according to the computing task requirements that need to be completed.

各个数据处理单元02需要依次加载,完成调用,因此各个数据处理单元02之间会存在调用时间间隔,因此,需要确定调用时间间隔,为了进行瞬态电流的控制,调用时间间隔为延长调用时间间隔。Each data processing unit 02 needs to be loaded in sequence to complete the call, so there will be a call time interval between each data processing unit 02. Therefore, the call time interval needs to be determined. In order to control the transient current, the call time interval is to extend the call time interval. .

需要说明的是,标准调用时间间隔为在进行集成电路设计时所确定的,各个数据处理单元02逐个加载执行运算任务时,相邻数据处理单元02触发的时间间隔。It should be noted that the standard calling time interval is determined during integrated circuit design. When each data processing unit 02 loads and performs computing tasks one by one, the time interval between adjacent data processing units 02 is triggered.

延长调用时间间隔大于标准调用时间间隔,具体可以根据需要确定为标准调用时间间隔的整数倍,比如:2倍、3倍等等。The extended calling time interval is greater than the standard calling time interval. Specifically, it can be determined as an integer multiple of the standard calling time interval as needed, such as: 2 times, 3 times, etc.

这样,仅通过很简单的运算,就可以得到延长调用时间间隔。In this way, the calling time interval can be extended by just a very simple operation.

在一种具体实施方式中,延长调用时间间隔包括预设延长调用时间间隔,即提前确定的延长调用时间间隔,只要确定需要进行调用时间间隔的延长时,就可以预设延长调用时间间隔,从而很迅速地实现延长调用时间间隔的确定。In a specific implementation, extending the calling time interval includes preset extending the calling time interval, that is, extending the calling time interval determined in advance. As long as it is determined that the calling time interval needs to be extended, the calling time interval can be preset to be extended, so that Determination of extended call intervals is achieved very quickly.

预设时可以根据标准调用时间间隔,增加标准调用时间间隔的整数倍来获得预设延长调用时间间隔,这样,能够确保在一个延长调用时间间隔内,仅有一个数据处理单元02被触发,从而减缓瞬态电流的上升趋势,保证集成电路的内部器件的正常工作。When presetting, you can increase the integer multiple of the standard calling time interval according to the standard calling time interval to obtain the preset extended calling time interval. In this way, it can be ensured that only one data processing unit 02 is triggered within an extended calling time interval. Slow down the rising trend of transient current and ensure the normal operation of the internal devices of the integrated circuit.

当然,在其他具体实施方式中,延长调用时间间隔还包括非预设延长调用时间间隔,根据需要临时进行确定,比如可以根据请求的不同,确定不同的延长调用时间间隔,比如软件程序根据接收到的数据处理单元调用请求的数量自行设置。Of course, in other specific implementations, the extended call time interval also includes non-preset extended call time intervals, which are temporarily determined as needed. For example, different extended call time intervals can be determined based on different requests. For example, the software program can determine different extended call time intervals based on the received The number of data processing unit call requests is set by itself.

在步骤S12中,根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。In step S12, calling requests are sent to each of the data processing units in sequence according to the extended calling time interval.

延长调用时间间隔为标准调用时间间隔的基础上进一步延长,将下一个数据处理单元02的原始的触发时间延后执行,即确保下一个数据处理单元(图5中所示的DU1)被触发启动时,前一个数据处理单元(图5中所示的DU0)已经在该延长调用时间间隔内基本达到稳定运行,从而达到仅存在一个或者较少量的数据处理单元02在同时进行加载启动。Extend the calling time interval to the standard calling time interval and further extend the original triggering time of the next data processing unit 02, that is, ensure that the next data processing unit (DU1 shown in Figure 5) is triggered and started. At that time, the previous data processing unit (DU0 shown in Figure 5) has basically reached stable operation within the extended calling time interval, so that there is only one or a smaller number of data processing units 02 loading and starting at the same time.

当延长调用时间间隔等于或大于一个数据处理单元02的加载启动时间时,在延长调用时间间隔内,仅有一个数据处理单元02的加载启动;当延长调用时间间隔大于标准调用时间间隔但小于一个数据处理单元02的加载启动时间时,在延长调用时间间隔内,同时进行加载启动的数据处理单元02的数量大于1,但小于标准调用时间间隔时的数量。When the extended call time interval is equal to or greater than the load start time of one data processing unit 02, there is only one load start of the data processing unit 02 within the extended call time interval; when the extended call time interval is greater than the standard call time interval but less than one During the loading start time of data processing unit 02, within the extended call time interval, the number of data processing units 02 that load and start at the same time is greater than 1, but less than the number at the standard call time interval.

通过将各个数据处理单元02的触发运行时间相互错开,以使在一个延长调用时间间隔内仅包括数量较少的正在启动的数据处理单元02,这样,在一个延长调用时间间隔内产生的瞬态电流仅为数量较少的数据处理单元02产生,各个数据处理单元02根据延长调用时间间隔逐次执行,瞬态电流的变化趋势就会缓慢,且不会变化很大。By staggering the triggering runtimes of the various data processing units 02 so that only a smaller number of starting data processing units 02 are included in an extended call time interval, the transients generated during an extended call time interval are The current is only generated by a small number of data processing units 02, and each data processing unit 02 is executed one after another according to the extended calling time interval. The changing trend of the transient current will be slow and will not change greatly.

如图5中所示,图中ST即为延长调用时间间隔,同样的,延长调用时间间隔在多个数据处理单元02同时工作时所需的总的时间范围内占比非常小。As shown in Figure 5, ST in the figure refers to the extended call time interval. Similarly, the extended call time interval accounts for a very small proportion in the total time range required when multiple data processing units 02 work simultaneously.

具体的,数据处理单元02工作时,由于ST时间段大于标准调用时间间隔,在一个ST时间段内,加载启动的数据处理单元02的数量较少,产生的瞬态电流即为较少的数据处理单元02产生,当需要调用多个数据处理单元02工作时,就可以减缓由于多个数据处理单元02在同一窗口期内同时交互加载启动而产生巨大的瞬态电流的情况,降低功耗,同时保证集成电路的性能。Specifically, when the data processing unit 02 is working, since the ST time period is greater than the standard calling time interval, within an ST time period, the number of loaded and started data processing units 02 is small, and the transient current generated is less data. The processing unit 02 is generated. When multiple data processing units 02 need to be called to work, the huge transient current generated due to the simultaneous interactive loading and startup of multiple data processing units 02 within the same window period can be mitigated, and power consumption can be reduced. At the same time, the performance of the integrated circuit is guaranteed.

在一种实施例中,请参考图6,图6为本申请实施例提供的瞬态电流控制方法下数据处理单元加载任务时的瞬态电流的变化示意图。In one embodiment, please refer to FIG. 6 , which is a schematic diagram of the change of the transient current when the data processing unit loads a task under the transient current control method provided by the embodiment of the present application.

如图中所示,相同的数据处理单元02执行运算任务的情况下,由于本申请实施例提供的瞬态电流控制方法中延长了相邻数据处理单元02启动的时间间隔,从而当数据处理单元02被运算任务触发加载启动时,本申请实施例提供的瞬态电流控制方法中所产生的瞬态电流是缓慢上升的。这样,缓解了由于多个数据处理单元02被运算任务触发启动时,由于时间间隔较短所产生的巨大的瞬态电流的情况,从而可以给每一个执行运算任务的数据处理单元02的加载启动提供稳定的工作环境,保证数据处理单元02可以正常工作。As shown in the figure, when the same data processing unit 02 performs a computing task, since the transient current control method provided by the embodiment of the present application extends the time interval between the startup of adjacent data processing units 02, when the data processing unit 02 02 When the load is triggered by a computing task, the transient current generated in the transient current control method provided by the embodiment of the present application rises slowly. In this way, the situation of huge transient currents generated due to the short time interval when multiple data processing units 02 are triggered and started by the computing task is alleviated, thereby providing a load startup for each data processing unit 02 that performs the computing task. A stable working environment ensures that the data processing unit 02 can work normally.

因此,本申请实施例所提供的瞬态电流控制方法,通过延长数据处理单元02执行运算任务时的时间间隔,提高将需要调用的各个数据处理单元的加载启动时间相互错开的概率,以降低正在调用的数据处理单元从启动到达到稳定工作状态的期间内同时加载启动其他数据处理单元的可能性,从而减少多个数据处理单元同时启动时瞬态电流瞬间增大而破坏集成电路芯片的性能的情况的发生,降低负载的损耗,提高集成电路内部器件的性能,延长集成电路的使用寿命。Therefore, the transient current control method provided by the embodiment of the present application increases the probability of staggering the loading startup time of each data processing unit that needs to be called by extending the time interval when the data processing unit 02 performs the computing task, so as to reduce the risk of the data processing unit being called. The called data processing unit has the possibility of loading and starting other data processing units at the same time from startup to reaching a stable working state, thereby reducing the possibility that the transient current will increase instantaneously when multiple data processing units are started at the same time and damage the performance of the integrated circuit chip. situation occurs, reduce the loss of the load, improve the performance of the internal components of the integrated circuit, and extend the service life of the integrated circuit.

为了资源分配模块更方便的管理数据处理单元的启动时间,在一些实施方式中,请参考图7,图7示例性的示出了本申请实施例提供的瞬态电流控制方法的另一流程图。In order for the resource allocation module to more conveniently manage the startup time of the data processing unit, in some embodiments, please refer to FIG. 7 , which exemplarily shows another flow chart of the transient current control method provided by the embodiment of the present application. .

如图中所示,为了确定延长调用时间间隔,本申请实施例提供的瞬态电流控制方法,还可以包括以下步骤:As shown in the figure, in order to determine the extension of the calling time interval, the transient current control method provided by the embodiment of the present application may also include the following steps:

在步骤S110中,根据所述数据处理单元调用请求确定当前需要调用的数据处理单元的数量。In step S110, the number of data processing units that currently need to be called is determined according to the data processing unit calling request.

一个数据处理单元调用请求,可能仅调用一个数据处理单元DU执行运算任务的请求,也可能是调用2个数据处理单元02或者更多的数据处理单元02来执行运算任务,因此,需要首先基于数据处理单元调用请求确定需要调用的数据处理单元02的数量。A data processing unit call request may only call one data processing unit DU to perform a computing task, or it may call two data processing units 02 or more data processing units 02 to perform a computing task. Therefore, it needs to be based on the data first. The processing unit call request determines the number of data processing units 02 that need to be called.

在步骤S111中,判断所述数据处理单元的数量是否大于预定数量,当所述数据处理单元的数量大于预定数量时,执行步骤S112,否则执行步骤S113。In step S111, it is determined whether the number of the data processing units is greater than a predetermined number. When the number of the data processing units is greater than the predetermined number, step S112 is executed; otherwise, step S113 is executed.

基于确定的数据处理单元02的数量,判断是否大于预定数量,如果大于,那么执行步骤S112,进行各个数据处理单元之间的调用时间间隔的延长,防止同时处于启动加载的数据处理单元过多;否则,执行步骤S13确定各个数据处理单元之间的标准调用时间间隔。Based on the determined number of data processing units 02, determine whether it is greater than the predetermined number. If it is greater, then execute step S112 to extend the calling time interval between each data processing unit to prevent too many data processing units that are starting to load at the same time; Otherwise, step S13 is executed to determine the standard calling time interval between each data processing unit.

需要说明的是,所述预定数量为同时处于启动加载状态的数据处理单元02数量的允许值。It should be noted that the predetermined number is an allowable value for the number of data processing units 02 that are in the startup loading state at the same time.

在确定延长时间间隔之前对需要调用的数据处理单元02进行判断,根据判断结果来准确的设置延长调用时间间隔,这样,可以更灵活的控制对数据处理单元02的调用。Before determining the extended time interval, the data processing unit 02 that needs to be called is judged, and the extended call time interval is accurately set based on the judgment result. In this way, the call to the data processing unit 02 can be more flexibly controlled.

如果预定数量过少,使得仅需要调用较少的数据处理单元时,也进行调用时间间隔的延长,一方面对于瞬态电流的控制量有限,另一方面,还会延长启动时间,为了充分体现对瞬态电流的控制效果,在一种具体实施方式中,所述预定数量可以为至少16个,比如:20个、80个等等,当根据运算任务的数量确定数据处理单元调用请求只需要调用少于预定数量的数据处理单元02时,无需进行调用时间间隔的延长。If the predetermined number is too small, so that only a small number of data processing units need to be called, the calling time interval will also be extended. On the one hand, the control amount of the transient current is limited, and on the other hand, the startup time will be extended. In order to fully reflect For the control effect of transient current, in a specific implementation, the predetermined number can be at least 16, such as: 20, 80, etc., when the data processing unit call request is determined based on the number of computing tasks, only When less than a predetermined number of data processing units 02 are called, there is no need to extend the calling time interval.

在步骤S112中,确定各个数据处理单元之间的延长调用时间间隔。In step S112, the extended call time interval between the respective data processing units is determined.

当所述数据处理单元的数量大于预定数量时,确定各个数据处理单元之间调用时间间隔为延长调用时间间隔。When the number of the data processing units is greater than the predetermined number, the calling time interval between each data processing unit is determined to be the extended calling time interval.

当然,后续进行各个数据处理单元的调用时,就根据延长调用时间间隔发送调用请求。Of course, when subsequent calls are made to each data processing unit, the call request will be sent according to the extended call time interval.

在步骤S113中,确定各个数据处理单元之间的标准调用时间间隔。In step S113, the standard calling time interval between the respective data processing units is determined.

当所述数据处理单元的数量大于预定数量时,确定各个数据处理单元之间的调用时间间隔为标准调用时间间隔。When the number of the data processing units is greater than the predetermined number, the calling time interval between the respective data processing units is determined to be the standard calling time interval.

当然,后续进行各个数据处理单元的调用时,就根据标准调用时间间隔发送调用请求。Of course, when subsequent calls are made to each data processing unit, the call request will be sent according to the standard calling time interval.

当接收到的数据处理单元调用请求为调用小于或等于预定数量的数据处理单元02时,就不会发生在原始的一个窗口期内出现瞬态电流叠加的情况,此时,资源分配模块02就可以不需要设置延长调用时间间隔,直接向数据处理单元02分配运算任务。When the received data processing unit call request is to call less than or equal to the predetermined number of data processing units 02, transient current superposition will not occur within the original window period. At this time, the resource allocation module 02 There is no need to set the extended calling time interval, and the computing tasks can be directly assigned to the data processing unit 02.

当接收到的数据处理单元调用请求为调用大于预定数量的数据处理单元02时,进行延长调用时间间隔的设置,将相邻两个数据处理单元02的触发启动时间错开。When the received data processing unit calling request is to call more than a predetermined number of data processing units 02, the calling time interval is set to extend the triggering start time of two adjacent data processing units 02.

从而,当数据处理单元调用请求需要资源分配模块01调用大量的数据处理单元02时,实施本申请实施例提供的瞬态电流控制方法,进入瞬态电流控制模式,错开需要调用的各个数据处理单元02的调用时间,减缓瞬态电流的上升。当仅需要较少的数据处理单元02就可以时,按照器件本身调用的方式,即由资源分配模块01直接向数据处理单元02发送调用请求,不需经过本申请实施例提供的瞬态电流控制方法,进入瞬态电流控制模式,然后再想数据处理单元02发送调用请求。可以更加灵活的控制调用请求的发送。Therefore, when the data processing unit call request requires the resource allocation module 01 to call a large number of data processing units 02, the transient current control method provided by the embodiment of the present application is implemented, the transient current control mode is entered, and each data processing unit that needs to be called is staggered The calling time of 02 slows down the rise of transient current. When only a small number of data processing units 02 are needed, the calling method of the device itself is used, that is, the resource allocation module 01 directly sends a calling request to the data processing unit 02 without going through the transient current control provided by the embodiment of the present application. method, enter the transient current control mode, and then send a call request to the data processing unit 02. The sending of call requests can be controlled more flexibly.

这样,可以在需要进行瞬态电流控制的情况下,再进行调用时间间隔的延长,在保证控制效果的同时,减少不必要的控制,实现更加灵活的控制数据处理单元02执行运算任务。In this way, when transient current control is required, the calling time interval can be extended to ensure the control effect while reducing unnecessary control and achieving more flexible control of the data processing unit 02 to perform computing tasks.

当然,在其他实施例中,也可以不论所需的数据处理单元02数量的多少,都实施本申请实施例提供的瞬态电流控制方法,进入瞬态电流控制模式,实现数据处理单元的调度,这样节省了判断的时间同时也可以实现瞬态电流的控制。Of course, in other embodiments, regardless of the number of data processing units 02 required, the transient current control method provided by the embodiment of the present application can also be implemented, enter the transient current control mode, and realize the scheduling of the data processing units. This saves judgment time and also enables control of transient current.

在一些实施例中,为了提高控制的方便性,并降低硬件结构的调整难度,可以利用延时阻塞软件根据所述数据处理单元调用请求,确定各个数据处理单元之间的延长调用时间间隔。In some embodiments, in order to improve the convenience of control and reduce the difficulty of adjusting the hardware structure, delay blocking software can be used to determine the extended call time interval between each data processing unit according to the data processing unit call request.

利用软件程序来实现延长调用时间间隔的调整,基于数据处理单元调用请求,延时阻塞软件确定延长调用时间间隔,进而控制每间隔延长调用时间间隔时,向数据处理单元发送调用请求,这样,可以在不增加硬件设施的情况下,更加方便、灵活的实现延长调用时间间隔的设置。Use software programs to adjust the call interval extension. Based on the call request of the data processing unit, the delay blocking software determines the call interval extension, and then controls the call request to be sent to the data processing unit when the call interval is extended every interval. In this way, Without adding hardware facilities, it is more convenient and flexible to extend the call interval setting.

当然,在其他实施例中还可以改变集成电路的硬件设施,增设延时使用的硬件器件,为集成电路器件工作时提供延长调用时间间隔。Of course, in other embodiments, the hardware facilities of the integrated circuit can also be changed, and hardware devices used for delay can be added to provide an extended call time interval for the integrated circuit device when it is working.

这样,可以预留出足够的时间,保证在同时处于加载启动状态的数据处理单元的数量较少,从而减缓瞬态电流所带来的冲击,保护器件的正常工作。In this way, enough time can be reserved to ensure that the number of data processing units in the loading and startup state at the same time is small, thereby slowing down the impact of transient current and protecting the normal operation of the device.

在一些实施方式中,为了实现根据延长调用时间间隔向各个所述数据处理单元发送调用请求的方便性,请参考图8,图8示例性的示出了本申请实施例提供的瞬态电流控制方法调用请求发送步骤的一流程图。In some embodiments, in order to achieve the convenience of sending call requests to each of the data processing units according to the extended call time interval, please refer to Figure 8. Figure 8 exemplarily shows the transient current control provided by the embodiment of the present application. A flowchart of the steps for sending a method call request.

如图中所示,本申请实施例提供的瞬态电流控制方法,调用请求发送步骤可以包括:As shown in the figure, in the transient current control method provided by the embodiment of the present application, the call request sending step may include:

在步骤S120中,向各个所述数据处理单元中的前一数据处理单元发送所述调用请求,并开始进行时间间隔计时,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求。In step S120, send the call request to the previous data processing unit in each of the data processing units, start the time interval timing, and stop sending the call to the later data processing unit in each of the data processing units. ask.

当完成前一数据处理单元的调用请求的发送后,需要确定后一数据处理单元的调用请求的发送时机,因此开始进行时间间隔计时,当然,还需要停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求。After completing the sending of the call request of the previous data processing unit, it is necessary to determine the timing of sending the call request of the subsequent data processing unit, so start the time interval timing. Of course, it is also necessary to stop sending the call request to each of the subsequent data processing units. The data processing unit sends the calling request.

具体地,可以通过资源分配模块中已有的计时器实现计时。Specifically, timing can be implemented through an existing timer in the resource allocation module.

在步骤S121中,判断所述时间间隔计时是否等于所述延长调用时间间隔,判断结果为是执行步骤S122,否则执行步骤S123。In step S121, it is determined whether the time interval timing is equal to the extended calling time interval. If the determination result is yes, step S122 is executed; otherwise, step S123 is executed.

在步骤S122中,向各个所述数据处理单元中后一数据处理单元发送所述调用请求,并执行步骤S124。In step S122, the calling request is sent to the later data processing unit in each of the data processing units, and step S124 is executed.

当时间间隔计时等于所述延长调用时间间隔时,向各个所述数据处理单元中后一数据处理单元发送所述调用请求。When the time interval is equal to the extended calling time interval, the calling request is sent to the later data processing unit in each of the data processing units.

在步骤S123中,不向各个所述数据处理单元中后一数据处理单元发送所述调用请求。In step S123, the calling request is not sent to the later data processing unit among the respective data processing units.

当时间间隔计时不等于所述延长调用时间间隔时,不向各个所述数据处理单元中后一数据处理单元发送所述调用请求。When the time interval timing is not equal to the extended calling time interval, the calling request is not sent to the later data processing unit in each of the data processing units.

当然,还需继续等待和判断,直至时间间隔计时等于所述延长调用时间间隔,执行步骤S122。Of course, it is necessary to continue to wait and judge until the time interval is equal to the extended calling time interval, and step S122 is executed.

在步骤S124中,判断是否完成各个所述数据处理单元的调用请求的发送,若是,执行步骤S126,若否,执行步骤S125。In step S124, it is determined whether the sending of the call request of each data processing unit is completed. If yes, step S126 is executed. If not, step S125 is executed.

在步骤S125中,将所述后一数据处理单元作为新的所述前一数据处理单元,并执行步骤120。In step S125, the latter data processing unit is regarded as the new previous data processing unit, and step 120 is executed.

在步骤S126中,结束。In step S126, it ends.

这样,就可以很方便地实现将各个数据处理单元的执行时间错开,使得一个延长调用时间间隔内,仅有较少数量的数据处理单元加载启动,控制瞬态电流的变化趋势为缓慢上升,保护集成电路内部器件的性能。In this way, it is easy to stagger the execution time of each data processing unit, so that within an extended call time interval, only a small number of data processing units are loaded and started, and the changing trend of the transient current is controlled to rise slowly, protecting The performance of the internal components of an integrated circuit.

为了方便实现控制向数据处理单元发送调用请求,在一种具体实施方式中,可以利用调用请求发送寄存器:In order to facilitate the control of sending call requests to the data processing unit, in a specific implementation, the call request sending register can be used:

为此,开始进行时间间隔计时,可以通过设置调用请求发送寄存器的状态为阻塞状态,实现停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求,而当所述时间间隔计时等于所述延长调用时间间隔时,可以通过设置调用请求发送寄存器的状态为非阻塞状态,就可以实现向各个所述数据处理单元中后一数据处理单元发送所述调用请求。To this end, start the time interval timing. You can stop sending the call request to the later data processing unit in each of the data processing units by setting the state of the call request sending register to the blocking state. When the time interval counts, When equal to the extended calling time interval, the calling request can be sent to the latter data processing unit among each of the data processing units by setting the state of the calling request sending register to a non-blocking state.

这样,通过设置调用请求发送寄存器的状态来实现数据处理单元调用请求的逐次发送,不会影响集成电路内部电路结构的设计,还可以辅助资源分配模块01的运算任务的分配,保证了集成电路内部器件的性能。In this way, by setting the state of the call request sending register to realize the sequential sending of the call request of the data processing unit, it will not affect the design of the internal circuit structure of the integrated circuit, and can also assist in the allocation of computing tasks of the resource allocation module 01, ensuring that the internal device performance.

为了解决前述问题,本申请实施例还提供一种瞬态电流控制装置,该装置可以认为是实现本申请实施例提供的瞬态电流控制方法所需设置的功能模块。下文描述的装置内容可与上文描述的方法内容相互对应参照。In order to solve the aforementioned problems, embodiments of the present application also provide a transient current control device, which can be considered as a functional module required to implement the transient current control method provided by the embodiments of the present application. The device content described below may be mutually referenced with the method content described above.

作为一种可选实现中,图9示出了本申请实施例所提供的瞬态电流控制装置的可选结构框图。As an optional implementation, FIG. 9 shows an optional structural block diagram of the transient current control device provided by the embodiment of the present application.

如图9所示,该瞬态电流控制装置可以包括:As shown in Figure 9, the transient current control device may include:

接收模块900,适于接收数据处理单元调用请求;The receiving module 900 is adapted to receive a data processing unit call request;

时间间隔确定模块901,适于根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔;The time interval determination module 901 is adapted to determine the extended call time interval between each data processing unit that currently needs to be called based on the data processing unit call request, and the extended call time interval is greater than the standard call time interval;

发送模块902,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。The sending module 902 is adapted to send calling requests to each of the data processing units in sequence according to the extended calling time interval.

可以提高将需要调用的各个数据处理单元的加载启动时间相互错开的概率,以降低正在调用的数据处理单元从启动到达到稳定工作状态的期间内同时加载启动其他数据处理单元的可能性,从而减少多个数据处理单元同时启动时瞬态电流瞬间增大而破坏集成电路芯片的性能的情况的发生,降低负载的损耗,提高集成电路内部器件的性能,延长集成电路的使用寿命。The probability of staggering the loading and start-up times of each data processing unit that needs to be called can be increased to reduce the possibility that other data processing units are loaded and started at the same time during the period from startup to stable working state of the data processing unit being called, thereby reducing When multiple data processing units are started at the same time, the transient current increases instantaneously and destroys the performance of the integrated circuit chip. It reduces the loss of the load, improves the performance of the internal components of the integrated circuit, and extends the service life of the integrated circuit.

在一种实施方式中,所述时间间隔确定模块901适于根据所述数据处理单元调用请求,确定当前需要调用的各个运算单元模之间的延长调用时间间隔,包括:In one implementation, the time interval determination module 901 is adapted to determine, based on the data processing unit call request, the extended call time interval between each operating unit module that currently needs to be called, including:

根据所述数据处理单元调用请求确定当前需要调用的数据处理单元的数量;当所述数据处理单元的数量大于预定数量时,确定各个数据处理单元之间的延长调用时间间隔。The number of data processing units that currently need to be called is determined according to the data processing unit calling request; when the number of data processing units is greater than a predetermined number, an extended calling time interval between each data processing unit is determined.

在GPGPU进行运算任务的计算时,资源分配模块02接收到的数据处理单元调用请求中可以包含仅调用一个数据处理单元执行运算任务的请求也可以是调用2个数据处理单元或者更多的数据处理单元来执行运算任务,因此,可以在确定延长时间间隔之前对需要调用的数据处理单元进行判断,根据判断结果来准确的设置延长调用时间间隔,可以更灵活的控制数据处理单元单元的调用。When the GPGPU is calculating a computing task, the data processing unit call request received by the resource allocation module 02 may include a request to call only one data processing unit to perform the computing task, or may be a request to call two data processing units or more data processing units. Therefore, the data processing unit that needs to be called can be judged before determining the extended time interval, and the extended calling time interval can be accurately set based on the judgment result, which can more flexibly control the calling of the data processing unit unit.

其中,各个数据处理单元调用请求的接收方式为同时接收。Among them, the call requests of each data processing unit are received simultaneously.

当然,在一些实施例中,所述时间间隔确定模块901中获得的延长调用时间间隔可以是利用延时阻塞软件根据所述数据处理单元调用请求,确定各个数据处理单元之间的延长调用时间间隔。Of course, in some embodiments, the extended call time interval obtained in the time interval determination module 901 can be determined by using delay blocking software according to the data processing unit call request to determine the extended call time interval between each data processing unit. .

利用软件程序来实现延长调用时间间隔的调整,可以在不增加硬件设施的情况下,更加灵活的实现调用时间的设置。Using software programs to adjust the extension of the call time interval can achieve a more flexible setting of the call time without adding hardware facilities.

在另一些实施方式中,所述延长调用时间间隔还可以包括预设延长调用时间间隔。In other embodiments, the extended calling time interval may also include a preset extended calling time interval.

即延长调用时间间隔除了软件程序根据接收到的数据处理单元调用请求的数量自行设置之外,还可以是人为提前预设的,预设时可以根据标准调用时间间隔,增加标准调用时间间隔的整数倍来获得预设延长调用时间间隔,这样,能够确保在一个延长调用时间间隔内,仅包括1个数据处理单元在逐渐执行运算任务至稳定状态,从而减缓瞬态电流的上升趋势,保证集成电路的内部器件的正常工作。That is, in addition to being set by the software program according to the number of call requests received by the data processing unit, the extended calling time interval can also be preset manually in advance. When preset, the standard calling time interval can be increased by an integer of the standard calling time interval. times to obtain the preset extended calling time interval. This ensures that within an extended calling time interval, only one data processing unit is gradually executing computing tasks to a stable state, thereby slowing down the rising trend of transient current and ensuring that the integrated circuit the normal operation of the internal devices.

需要说明的是,延长调用时间间隔可以是标准调用时间间隔的整数倍。这样,可以预留出足够的时间,保证在第一个数据处理单元运行达到稳定的期间内,另一个数据处理单元是未被触发启动的状态,从而减缓瞬态电流所带来的冲击,保护器件的正常工作。It should be noted that the extended calling time interval can be an integer multiple of the standard calling time interval. In this way, enough time can be reserved to ensure that during the period when the operation of the first data processing unit reaches stability, the other data processing unit is not triggered to start, thereby mitigating the impact of transient current and protecting the normal operation of the device.

所述发送模块902还适于:The sending module 902 is also suitable for:

向各个所述数据处理单元中的前一数据处理单元发送所述调用请求,并开始进行时间间隔计时;Send the call request to the previous data processing unit in each of the data processing units, and start time interval timing;

当所述时间间隔计时等于所述延长调用时间间隔时,向各个所述数据处理单元中后一数据处理单元发送所述调用请求,并将所述后一数据处理单元作为新的所述前一数据处理单元,直至完成各个所述数据处理单元的调用请求的发送。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the later data processing unit in each of the data processing units, and the later data processing unit is used as the new previous one. The data processing unit waits until the sending of the call request of each data processing unit is completed.

这样就可以将各个数据处理单元的执行时间错开,使得一个延长调用时间间隔内,只允许一个数据处理单元执行至稳定,控制瞬态电流的变化趋势为缓慢上升,保护集成电路内部器件的性能。In this way, the execution time of each data processing unit can be staggered, so that within an extended call time interval, only one data processing unit is allowed to execute until stable, and the changing trend of the transient current is controlled to rise slowly, protecting the performance of the internal devices of the integrated circuit.

为了根据延长调用时间间隔实现各个数据处理单元调用请求的发送,在一种具体实施方式中,所述发送模块902,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求,包括:In order to implement the sending of call requests to each data processing unit according to the extended calling time interval, in a specific implementation, the sending module 902 is adapted to send calls to each of the data processing units in sequence according to the extended calling time interval. Requests include:

通过设置调用请求发送寄存器的状态为阻塞状态,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求;Stop sending the call request to the latter data processing unit in each of the data processing units by setting the state of the call request sending register to the blocking state;

当所述时间间隔计时等于所述延长调用时间间隔时,通过设置调用请求发送寄存器的状态为非阻塞状态,向各个所述数据处理单元中后一数据处理单元发送所述调用请求。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the latter data processing unit in each of the data processing units by setting the state of the calling request sending register to a non-blocking state.

这样,通过设置寄存器的状态来实现数据处理单元调用请求的逐次发送,不会影响集成电路内部电路结构的设计,还可以辅助资源分配模块01的运算任务的分配,保证了集成电路内部器件的性能。In this way, by setting the state of the register, the data processing unit call request is sent one after another, which will not affect the design of the internal circuit structure of the integrated circuit, and can also assist in the allocation of computing tasks of the resource allocation module 01, ensuring the performance of the internal devices of the integrated circuit. .

本申请实施例还提供一种集成电路,执行本申请实施例上述提供的瞬态电流控制方法。An embodiment of the present application also provides an integrated circuit that executes the transient current control method provided above in the embodiment of the present application.

本申请实施例还提供一种电子设备,包括如上述提供的集成电路。An embodiment of the present application also provides an electronic device, including the integrated circuit provided above.

虽然本申请实施例披露如上,但本申请并非限定于此。任何本领域技术人员,在不脱离本申请的精神和范围内,均可作各种更动与修改,因此本申请的保护范围应当以权利要求所限定的范围为准。Although the embodiments of the present application are disclosed as above, the present application is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims (18)

1.一种瞬态电流控制方法,其特征在于,适用于集成电路的资源分配模块,包括:1. A transient current control method, characterized in that it is suitable for a resource allocation module of an integrated circuit, including: 接收数据处理单元调用请求;Receive data processing unit call request; 根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔,且小于一个数据处理单元的加载启动时间,所述标准调用时间间隔为在进行集成电路设计时所确定的,各个数据处理单元逐个加载执行运算任务时,相邻数据处理单元触发的时间间隔;According to the data processing unit call request, the extended call time interval between each data processing unit that currently needs to be called is determined. The extended call time interval is greater than the standard call time interval and less than the loading startup time of one data processing unit, so The above-mentioned standard calling time interval is determined during the design of integrated circuits. When each data processing unit loads and performs computing tasks one by one, the time interval between adjacent data processing units is triggered; 根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。According to the extended calling time interval, calling requests are sent to each of the data processing units in sequence. 2.如权利要求1所述的瞬态电流控制方法,其特征在于,所述根据所述数据处理单元调用请求确定各个数据处理单元之间的延长调用时间间隔的步骤包括:2. The transient current control method according to claim 1, wherein the step of determining the extended call time interval between each data processing unit according to the data processing unit call request includes: 根据所述数据处理单元调用请求确定当前需要调用的数据处理单元的数量;Determine the number of data processing units that currently need to be called according to the data processing unit calling request; 当所述数据处理单元的数量大于预定数量时,确定各个数据处理单元之间的延长调用时间间隔。When the number of data processing units is greater than a predetermined number, an extended calling time interval between respective data processing units is determined. 3.如权利要求1所述的瞬态电流控制方法,其特征在于,所述根据数据处理单元调用请求确定各个数据处理单元之间的延长调用时间间隔的步骤还包括:3. The transient current control method according to claim 1, wherein the step of determining the extended calling time interval between each data processing unit according to the data processing unit calling request further includes: 利用延时阻塞软件根据所述数据处理单元调用请求,确定各个数据处理单元之间的延长调用时间间隔。Delay blocking software is used to determine the extended calling time interval between each data processing unit according to the data processing unit calling request. 4.如权利要求2所述的瞬态电流控制方法,其特征在于,所述预定数量包括至少16个。4. The transient current control method according to claim 2, wherein the predetermined number includes at least 16. 5.如权利要求1所述的瞬态电流控制方法,其特征在于,所述延长调用时间间隔包括所述标准调用时间间隔的整数倍。5. The transient current control method according to claim 1, wherein the extended calling time interval includes an integer multiple of the standard calling time interval. 6.如权利要求1所述的瞬态电流控制方法,其特征在于,所述延长调用时间间隔包括预设延长调用时间间隔。6. The transient current control method according to claim 1, wherein the extended calling time interval includes a preset extended calling time interval. 7.如权利要求1所述的瞬态电流控制方法,其特征在于,所述根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求的步骤包括:7. The transient current control method according to claim 1, wherein the step of sending call requests to each of the data processing units in sequence according to the extended call time interval includes: 向各个所述数据处理单元中的前一数据处理单元发送所述调用请求,并开始进行时间间隔计时,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求;Send the call request to the previous data processing unit in each of the data processing units, start timing the time interval, and stop sending the call request to the subsequent data processing unit in each of the data processing units; 当所述时间间隔计时等于所述延长调用时间间隔时,向各个所述数据处理单元中后一数据处理单元发送所述调用请求,并将所述后一数据处理单元作为新的所述前一数据处理单元,直至完成各个所述数据处理单元的调用请求的发送。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the later data processing unit in each of the data processing units, and the later data processing unit is used as the new previous one. The data processing unit waits until the sending of the call request of each data processing unit is completed. 8.如权利要求7所述的瞬态电流控制方法,其特征在于,所述停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求的步骤包括:8. The transient current control method according to claim 7, wherein the step of stopping sending the call request to the latter data processing unit in each of the data processing units includes: 通过设置调用请求发送寄存器的状态为阻塞状态,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求;Stop sending the call request to the latter data processing unit in each of the data processing units by setting the state of the call request sending register to the blocking state; 所述当所述时间间隔计时等于所述延长调用时间间隔时,向各个所述数据处理单元中后一数据处理单元发送所述调用请求的步骤包括:When the time interval timing is equal to the extended calling time interval, the step of sending the calling request to the later data processing unit in each of the data processing units includes: 当所述时间间隔计时等于所述延长调用时间间隔时,通过设置调用请求发送寄存器的状态为非阻塞状态,向各个所述数据处理单元中后一数据处理单元发送所述调用请求。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the latter data processing unit in each of the data processing units by setting the state of the calling request sending register to a non-blocking state. 9.一种瞬态电流控制装置,其特征在于,包括:9. A transient current control device, characterized by comprising: 接收模块,适于接收数据处理单元调用请求;The receiving module is suitable for receiving data processing unit call requests; 时间间隔确定模块,适于根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,所述延长调用时间间隔大于标准调用时间间隔,且小于一个数据处理单元的加载启动时间,所述标准调用时间间隔为在进行集成电路设计时所确定的,各个数据处理单元逐个加载执行运算任务时,相邻数据处理单元触发的时间间隔;The time interval determination module is adapted to determine the extended calling time interval between each data processing unit that currently needs to be called according to the data processing unit calling request. The extended calling time interval is greater than the standard calling time interval and less than one data processing unit. The loading start time of the unit. The standard calling time interval is determined during integrated circuit design. When each data processing unit loads and performs computing tasks one by one, the time interval between adjacent data processing units is triggered; 发送模块,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求。The sending module is adapted to send calling requests to each of the data processing units in sequence according to the extended calling time interval. 10.如权利要求9所述的瞬态电流控制装置,其特征在于,所述时间间隔确定模块,适于根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,包括:10. The transient current control device according to claim 9, wherein the time interval determination module is adapted to determine the extension between each data processing unit that currently needs to be called according to the data processing unit call request. Calling interval, including: 根据所述数据处理单元调用请求确定当前需要调用的数据处理单元的数量;Determine the number of data processing units that currently need to be called according to the data processing unit calling request; 当所述数据处理单元的数量大于预定数量时,确定各个数据处理单元之间的延长调用时间间隔。When the number of data processing units is greater than a predetermined number, an extended calling time interval between respective data processing units is determined. 11.如权利要求9所述的瞬态电流控制装置,其特征在于,所述时间间隔确定模块,适于根据所述数据处理单元调用请求,确定当前需要调用的各个数据处理单元之间的延长调用时间间隔,包括:11. The transient current control device according to claim 9, wherein the time interval determination module is adapted to determine the extension between each data processing unit that currently needs to be called according to the data processing unit call request. Calling interval, including: 利用延时阻塞软件根据所述数据处理单元调用请求,确定各个数据处理单元之间的延长调用时间间隔。Delay blocking software is used to determine the extended calling time interval between each data processing unit according to the data processing unit calling request. 12.如权利要求10所述的瞬态电流控制装置,其特征在于,所述预定数量包括至少16个。12. The transient current control device of claim 10, wherein the predetermined number includes at least 16. 13.如权利要求9所述的瞬态电流控制装置,其特征在于,所述延长调用时间间隔包括所述标准调用时间间隔的整数倍。13. The transient current control device according to claim 9, wherein the extended calling time interval includes an integer multiple of the standard calling time interval. 14.如权利要求9所述的瞬态电流控制装置,其特征在于,所述延长调用时间间隔包括预设延长调用时间间隔。14. The transient current control device according to claim 9, wherein the extended calling time interval includes a preset extended calling time interval. 15.如权利要求9所述的瞬态电流控制装置,其特征在于,所述发送模块,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求,包括:15. The transient current control device according to claim 9, wherein the sending module is adapted to send calling requests to each of the data processing units in sequence according to the extended calling time interval, including: 向各个所述数据处理单元中的前一数据处理单元发送所述调用请求,并开始进行时间间隔计时,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求;Send the call request to the previous data processing unit in each of the data processing units, start timing the time interval, and stop sending the call request to the subsequent data processing unit in each of the data processing units; 当所述时间间隔计时等于所述延长调用时间间隔时,向各个所述数据处理单元中后一数据处理单元发送所述调用请求,并将所述后一数据处理单元作为新的所述前一数据处理单元,直至完成各个所述数据处理单元的调用请求的发送。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the later data processing unit in each of the data processing units, and the later data processing unit is used as the new previous one. The data processing unit waits until the sending of the call request of each data processing unit is completed. 16.如权利要求15所述的瞬态电流控制装置,其特征在于,所述发送模块,适于根据所述延长调用时间间隔,依次向各个所述数据处理单元发送调用请求,包括:16. The transient current control device according to claim 15, wherein the sending module is adapted to send calling requests to each of the data processing units in sequence according to the extended calling time interval, including: 通过设置调用请求发送寄存器的状态为阻塞状态,停止向各个所述数据处理单元中后一数据处理单元发送所述调用请求;Stop sending the call request to the latter data processing unit in each of the data processing units by setting the state of the call request sending register to the blocking state; 当所述时间间隔计时等于所述延长调用时间间隔时,通过设置调用请求发送寄存器的状态为非阻塞状态,向各个所述数据处理单元中后一数据处理单元发送所述调用请求。When the time interval timing is equal to the extended calling time interval, the calling request is sent to the latter data processing unit in each of the data processing units by setting the state of the calling request sending register to a non-blocking state. 17.一种集成电路,其特征在于,执行如权利要求1-8任一项所述的瞬态电流控制方法。17. An integrated circuit, characterized in that it performs the transient current control method according to any one of claims 1-8. 18.一种电子设备,其特征在于,包括如权利要求17所述的集成电路。18. An electronic device, characterized by comprising the integrated circuit according to claim 17.
CN202210648644.6A 2022-06-09 2022-06-09 Transient current control method and related devices Active CN115061789B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210648644.6A CN115061789B (en) 2022-06-09 2022-06-09 Transient current control method and related devices

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210648644.6A CN115061789B (en) 2022-06-09 2022-06-09 Transient current control method and related devices

Publications (2)

Publication Number Publication Date
CN115061789A CN115061789A (en) 2022-09-16
CN115061789B true CN115061789B (en) 2024-02-09

Family

ID=83199639

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210648644.6A Active CN115061789B (en) 2022-06-09 2022-06-09 Transient current control method and related devices

Country Status (1)

Country Link
CN (1) CN115061789B (en)

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102929714A (en) * 2012-10-19 2013-02-13 国电南京自动化股份有限公司 uC/OS-II-based hardware task manager
CN103645792A (en) * 2013-11-30 2014-03-19 无锡中星微电子有限公司 Power management unit
CN107368172A (en) * 2017-07-27 2017-11-21 郑州云海信息技术有限公司 A kind of storage server powering method, apparatus and system
CN108400587A (en) * 2018-03-05 2018-08-14 中国商用飞机有限责任公司北京民用飞机技术研究中心 A kind of load management method and system of the more electric systems of civil aircraft
CN108963996A (en) * 2018-07-24 2018-12-07 郑州云海信息技术有限公司 A kind of method and device reducing transient state inrush current
CN109491485A (en) * 2018-10-09 2019-03-19 郑州云海信息技术有限公司 A kind of system and method for high-performance server accelerator card off-peak electrification
CN110618742A (en) * 2019-08-20 2019-12-27 苏州浪潮智能科技有限公司 PDB board and working method thereof
CN111290787A (en) * 2019-06-19 2020-06-16 锐迪科(重庆)微电子科技有限公司 Arithmetic device and arithmetic method
CN113054835A (en) * 2021-03-05 2021-06-29 山东英信计算机技术有限公司 Delay starting method, system and related device of circuit module
CN113534689A (en) * 2020-04-13 2021-10-22 北京四维图新科技股份有限公司 Delayed start control method and device
CN114242138A (en) * 2021-12-01 2022-03-25 海光信息技术股份有限公司 Time delay controller, memory controller and time sequence control method
CN114443137A (en) * 2020-10-30 2022-05-06 中科寒武纪科技股份有限公司 An integrated computing device, chip, board, device and computing method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11507310B2 (en) * 2019-09-02 2022-11-22 SK Hynix Inc. Memory controller and operating method thereof

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102929714A (en) * 2012-10-19 2013-02-13 国电南京自动化股份有限公司 uC/OS-II-based hardware task manager
CN103645792A (en) * 2013-11-30 2014-03-19 无锡中星微电子有限公司 Power management unit
CN107368172A (en) * 2017-07-27 2017-11-21 郑州云海信息技术有限公司 A kind of storage server powering method, apparatus and system
CN108400587A (en) * 2018-03-05 2018-08-14 中国商用飞机有限责任公司北京民用飞机技术研究中心 A kind of load management method and system of the more electric systems of civil aircraft
CN108963996A (en) * 2018-07-24 2018-12-07 郑州云海信息技术有限公司 A kind of method and device reducing transient state inrush current
CN109491485A (en) * 2018-10-09 2019-03-19 郑州云海信息技术有限公司 A kind of system and method for high-performance server accelerator card off-peak electrification
CN111290787A (en) * 2019-06-19 2020-06-16 锐迪科(重庆)微电子科技有限公司 Arithmetic device and arithmetic method
CN110618742A (en) * 2019-08-20 2019-12-27 苏州浪潮智能科技有限公司 PDB board and working method thereof
CN113534689A (en) * 2020-04-13 2021-10-22 北京四维图新科技股份有限公司 Delayed start control method and device
CN114443137A (en) * 2020-10-30 2022-05-06 中科寒武纪科技股份有限公司 An integrated computing device, chip, board, device and computing method
CN113054835A (en) * 2021-03-05 2021-06-29 山东英信计算机技术有限公司 Delay starting method, system and related device of circuit module
CN114242138A (en) * 2021-12-01 2022-03-25 海光信息技术股份有限公司 Time delay controller, memory controller and time sequence control method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
James Herr ; Alison Steer ; .浪涌抑制器保护电子产品免受高压瞬态影响.世界电子元器件.2007,(第12期),全文. *
Jimena L. Bastos.A design paradigm for integrated protection of shipboard power systems.《Proceedings of the 2007 Summer Computer Simulation Conference》.2007,第1-10页. *
总体电路瞬态浪涌抑制;梁君;杨友超;赵岩;;计算机测量与控制(04);全文 *
李少东.隔离开关的智能组件设计.《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》.2016,全文. *

Also Published As

Publication number Publication date
CN115061789A (en) 2022-09-16

Similar Documents

Publication Publication Date Title
US10649935B2 (en) Deferred inter-processor interrupts
US8566826B2 (en) System and method for synchronizing transient resource usage between virtual machines in a hypervisor environment
US6298448B1 (en) Apparatus and method for automatic CPU speed control based on application-specific criteria
US8122269B2 (en) Regulating power consumption in a multi-core processor by dynamically distributing power and processing requests by a managing core to a configuration of processing cores
CN102081555B (en) Method and device for adjusting clock interruption period
EP2911055A1 (en) Parallel computing device
US12032989B2 (en) Timer processing method, apparatus, electronic device and computer storage medium
US7451454B2 (en) Event handling mechanism
KR102789082B1 (en) Dynamic interrupt rate control in computing systems
CN115061789B (en) Transient current control method and related devices
EP1393175A2 (en) A resource management method
JP2021521541A (en) Implementation of processing quality assurance of central processing unit when processing accelerator requests
CN108292236B (en) Information processing method and device
CN118051299A (en) Cloud platform virtual machine performance regulation and control method, system, equipment and medium
WO2024021485A1 (en) Processor performance adjustment method and apparatus, electronic device and storage medium
CN114564289A (en) Task switching method, device and electronic device
CN107480071B (en) Cache data migration method and device
US20100229010A1 (en) Computer system, method for controlling the same, and program
US20240385871A1 (en) Communications across privilege domains within a central processing unit core
US20090271594A1 (en) Semiconductor integrated circuit, semiconductor integrated circuit control device, load distribution method, load distribution program, and electronic device
CN118585314A (en) A stack space task allocation method using dynamic management strategy
CN119938278A (en) A task scheduling processing method, system, device and medium
KR101061133B1 (en) Interrupt handling system
CN117632425A (en) Task allocation method and device, electronic equipment and storage medium
CN119415224A (en) A program execution method, device, equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant