CN117544433A - Core particle, low power consumption control method, chip and computer equipment - Google Patents
Core particle, low power consumption control method, chip and computer equipment Download PDFInfo
- Publication number
- CN117544433A CN117544433A CN202311389522.0A CN202311389522A CN117544433A CN 117544433 A CN117544433 A CN 117544433A CN 202311389522 A CN202311389522 A CN 202311389522A CN 117544433 A CN117544433 A CN 117544433A
- Authority
- CN
- China
- Prior art keywords
- link layer
- core particle
- core
- data
- low
- 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.)
- Pending
Links
- 239000007771 core particle Substances 0.000 title claims abstract description 294
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000005540 biological transmission Effects 0.000 claims abstract description 302
- 230000004044 response Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 16
- 239000010410 layer Substances 0.000 description 517
- 238000010586 diagram Methods 0.000 description 12
- 238000004891 communication Methods 0.000 description 9
- 230000003993 interaction Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000012792 core layer Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000008187 granular material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L12/00—Data switching networks
- H04L12/02—Details
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L12/00—Data switching networks
- H04L12/02—Details
- H04L12/10—Current supply arrangements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
- H04L69/30—Definitions, standards or architectural aspects of layered protocol stacks
- H04L69/32—Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level
- H04L69/322—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions
- H04L69/323—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions in the physical layer [OSI layer 1]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L69/00—Network arrangements, protocols or services independent of the application payload and not provided for in the other groups of this subclass
- H04L69/30—Definitions, standards or architectural aspects of layered protocol stacks
- H04L69/32—Architecture of open systems interconnection [OSI] 7-layer type protocol stacks, e.g. the interfaces between the data link level and the physical level
- H04L69/322—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions
- H04L69/324—Intralayer communication protocols among peer entities or protocol data unit [PDU] definitions in the data link layer [OSI layer 2], e.g. HDLC
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Dc Digital Transmission (AREA)
Abstract
本申请实施例提供一种芯粒、低功耗控制方法、芯片及计算机设备,所述芯粒与互联的对端芯粒之间传输数据流,所述芯粒包括:芯粒互联接口;所述芯粒互联接口包括:链路层和物理层;所述链路层,用于在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路,以使得所述链路层进入低功耗状态;其中,所述物理层维持工作状态不变。本申请实施例可以在芯粒互联场景下,降低芯粒互联的功耗。
Embodiments of the present application provide a core particle, a low-power control method, a chip, and a computer device. The core particle transmits a data stream with an interconnected counterpart core particle. The core particle includes: a core particle interconnection interface; The core-to-chip interconnection interface includes: a link layer and a physical layer; the link layer is used to close the data transmission path of the link layer during the transmission idle time of the data stream, so that the link layer Enter a low power consumption state; wherein the physical layer maintains its working state unchanged. The embodiments of this application can reduce the power consumption of core-to-die interconnection in the core-to-die interconnection scenario.
Description
技术领域Technical field
本申请实施例涉及芯片技术领域,具体涉及一种芯粒、低功耗控制方法、芯片及计算机设备。The embodiments of the present application relate to the field of chip technology, and specifically relate to a core chip, a low-power control method, a chip, and a computer device.
背景技术Background technique
芯粒是能够实现一定功能且含有互联接口的单元芯片,多个芯粒通过互联接口进行互联可以形成SOC(System On Chip,片上系统)等芯片。芯粒互联是指在两个芯粒的互联接口存在物理连接的基础上,使用互联协议来协调调度两个芯粒之间的通信,从而实现两个芯粒之间的互连互通。A chip is a unit chip that can achieve certain functions and contains an interconnection interface. Multiple cores are interconnected through the interconnection interface to form a SOC (System On Chip) and other chips. Core-to-core interconnection refers to using an interconnection protocol to coordinate and schedule the communication between two cores on the basis of the physical connection between the interconnection interfaces of two cores, thereby achieving interconnection and interoperability between two cores.
在芯粒互联场景下,芯粒之间的数据传输和处理需要消耗较大功耗,因此如何提供技术方案,以降低芯粒互联的功耗,成为了本领域技术人员亟需解决的技术问题。In the scenario of chip interconnection, data transmission and processing between chips consume a lot of power. Therefore, how to provide technical solutions to reduce the power consumption of chip interconnection has become an urgent technical problem that technicians in the field need to solve. .
发明内容Contents of the invention
有鉴于此,本申请实施例提供一种芯粒、低功耗控制方法、芯片及计算机设备,以在芯粒互联场景下,降低芯粒互联的功耗。In view of this, embodiments of the present application provide a chip, a low-power control method, a chip, and a computer device to reduce the power consumption of core-to-chip interconnection in a core-to-die interconnection scenario.
第一方面,本申请实施例提供一种芯粒,所述芯粒与互联的对端芯粒之间传输数据流,所述芯粒包括:芯粒互联接口;所述芯粒互联接口包括:链路层和物理层;In a first aspect, embodiments of the present application provide a core particle that transmits data streams between the core particle and an interconnected counterpart core particle. The core particle includes: a core particle interconnection interface; the core particle interconnection interface includes: Link layer and physical layer;
所述链路层,用于在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路,以使得所述链路层进入低功耗状态;The link layer is configured to close the data transmission path of the link layer during the idle time of transmission of the data stream, so that the link layer enters a low power consumption state;
其中,所述物理层维持工作状态不变。Wherein, the physical layer maintains its working state unchanged.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述芯粒向所述对端芯粒发送数据流;所述链路层,用于在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路,包括:Optionally, transmitting a data stream between the core particle and an interconnected opposite end core particle includes: the core particle sends a data stream to the opposite end core particle; and the link layer is configured to During the transmission idle time, close the data transmission path of the link layer, including:
在所述传输空闲时间到来时,获取低功耗进入请求;When the transmission idle time arrives, obtain a low-power entry request;
响应于所述低功耗进入请求,向所述对端芯粒发送进入低功耗信号,并关闭所述链路层的数据发送通路;In response to the low-power entry request, send a low-power entry signal to the opposite end chip, and close the data transmission path of the link layer;
其中,所述进入低功耗信号用于控制所述对端芯粒的链路层关闭数据接收通路,以使得所述对端芯粒的链路层进入低功耗状态。Wherein, the entering low power consumption signal is used to control the link layer of the opposite end core chip to close the data receiving path, so that the link layer of the opposite end core chip enters a low power consumption state.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述对端芯粒向所述芯粒发送数据流;所述链路层,用于在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路包括:Optionally, transmitting a data stream between the core particle and an interconnected opposite end core particle includes: the opposite end core particle sends a data stream to the core particle; and the link layer is configured to The transmission idle time, closing the data transmission path of the link layer includes:
在所述传输空闲时间到来时,获取所述对端芯粒的链路层所发送的进入低功耗信号;When the transmission idle time arrives, obtain the low-power consumption signal sent by the link layer of the opposite end core;
响应于所述进入低功耗信号,关闭链路层的数据接收通路。In response to the entering low power consumption signal, the data receiving path of the link layer is closed.
可选的,所述链路层通过关闭数据传输通路的工作时钟,以关闭所述链路层的数据传输通路。Optionally, the link layer closes the data transmission path of the link layer by turning off the working clock of the data transmission path.
可选的,所述链路层,还用于在所述数据流的传输工作时间,开启所述链路层的数据传输通路,以使得所述链路层退出低功耗状态。Optionally, the link layer is also configured to open a data transmission path of the link layer during the transmission working time of the data stream, so that the link layer exits the low power consumption state.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述芯粒向所述对端芯粒发送数据流;所述链路层,用于在所述数据流的传输工作时间,开启所述链路层的数据传输通路,包括:Optionally, transmitting a data stream between the core particle and an interconnected opposite end core particle includes: the core particle sends a data stream to the opposite end core particle; and the link layer is configured to The transmission working time is to open the data transmission path of the link layer, including:
在所述传输工作时间到来时,开启所述链路层的数据发送通路,并向所述对端芯粒发送退出低功耗信号;When the transmission working time arrives, open the data transmission path of the link layer, and send an exit low-power consumption signal to the opposite end core;
其中,所述退出低功耗信号用于控制所述对端芯粒的链路层重启所述数据流的传输。Wherein, the exit low power consumption signal is used to control the link layer of the opposite end core chip to restart the transmission of the data stream.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述对端芯粒向所述芯粒发送数据流;所述链路层,用于在所述数据流的传输工作时间,开启所述链路层的数据传输通路,包括:Optionally, transmitting a data stream between the core particle and an interconnected opposite end core particle includes: the opposite end core particle sends a data stream to the core particle; and the link layer is configured to The transmission working time is to open the data transmission path of the link layer, including:
在所述传输工作时间到来时,开启链路层的数据接收通路;When the transmission working time arrives, open the data receiving path of the link layer;
获取所述对端芯粒的链路层所发送的退出低功耗信号,并响应于所述退出低功耗信号,重启所述数据流的传输。Obtain the low-power exit signal sent by the link layer of the opposite end core particle, and restart the transmission of the data stream in response to the low-power exit signal.
可选的,所述链路层通过开启数据传输通路的工作时钟,以开启所述链路层的数据传输通路。Optionally, the link layer starts the data transmission path of the link layer by turning on the working clock of the data transmission path.
可选的,所述链路层以链路层的工作时钟作为驱动时钟,关闭所述链路层的数据传输通路,或者开启所述链路层的数据传输通路。Optionally, the link layer uses the working clock of the link layer as a driving clock to close the data transmission path of the link layer, or open the data transmission path of the link layer.
可选的,所述传输空闲时间的时间长度为所述链路层的预设数量的工作时钟周期。Optionally, the length of the transmission idle time is a preset number of working clock cycles of the link layer.
可选的,所述链路层包括:低功耗控制逻辑;Optionally, the link layer includes: low power consumption control logic;
所述低功耗控制逻辑,用于执行所述在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路,以使得所述链路层进入低功耗状态的步骤;以及执行所述在所述数据流的传输工作时间,开启所述链路层的数据传输通路,以使得所述链路层退出低功耗状态的步骤。The low-power control logic is used to perform the step of closing the data transmission path of the link layer during the transmission idle time of the data stream, so that the link layer enters a low-power state; and Execute the step of opening the data transmission path of the link layer during the transmission working time of the data stream, so that the link layer exits the low power consumption state.
第二方面,本申请实施例提供一种低功耗控制方法,应用于如上述第一方面所述的芯粒,所述芯粒与互联的对端芯粒之间传输数据流;所述方法包括:In the second aspect, embodiments of the present application provide a low power consumption control method, which is applied to the core particle as described in the first aspect, and transmits data streams between the core particle and the interconnected counterpart core particle; the method include:
在数据流的传输空闲时间,关闭链路层的数据传输通路,以使得链路层进入低功耗状态;During the idle time of data flow transmission, close the data transmission path of the link layer so that the link layer enters a low power consumption state;
以及,维持物理层的工作状态不变。And, maintain the working status of the physical layer unchanged.
可选的,还包括:Optional, also includes:
在数据流的传输工作时间,开启链路层的数据传输通路,以使得链路层退出低功耗状态。During the transmission working time of the data flow, the data transmission path of the link layer is opened so that the link layer exits the low power consumption state.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述芯粒向所述对端芯粒发送数据流;所述在数据流的传输空闲时间,关闭链路层的数据传输通路,包括:Optionally, transmitting the data stream between the core particle and the interconnected counterpart core particle includes: the core particle sends the data stream to the counterpart core particle; and during the idle time of the data stream transmission, the link is closed. Layer data transmission path, including:
在所述传输空闲时间到来时,获取低功耗进入请求;When the transmission idle time arrives, obtain a low-power entry request;
响应于所述低功耗进入请求,向所述对端芯粒发送进入低功耗信号,并关闭链路层的数据发送通路;其中,所述进入低功耗信号用于控制所述对端芯粒的链路层关闭数据接收通路,以使得所述对端芯粒的链路层进入低功耗状态;In response to the low-power entry request, an entry low-power signal is sent to the opposite end core and the data transmission path of the link layer is closed; wherein the entry low-power signal is used to control the opposite end. The link layer of the core particle closes the data receiving path, so that the link layer of the opposite end core particle enters a low power consumption state;
所述在数据流的传输工作时间,开启链路层的数据传输通路,包括:The process of opening the link layer data transmission path during the transmission working time of the data stream includes:
在所述传输工作时间到来时,开启链路层的数据发送通路,并向所述对端芯粒发送退出低功耗信号;其中,所述退出低功耗信号用于控制所述对端芯粒的链路层重启所述数据流的传输。When the transmission working time arrives, the data transmission path of the link layer is opened, and an exit low-power consumption signal is sent to the opposite end core; wherein the exit low-power consumption signal is used to control the opposite end core. The granular link layer restarts the transmission of the data stream.
可选的,所述关闭链路层的数据发送通路包括:以链路层的工作时钟周期作为驱动时钟,关闭链路层的数据发送通路的工作时钟;Optionally, closing the data transmission path of the link layer includes: using the working clock cycle of the link layer as the driving clock, closing the working clock of the data sending path of the link layer;
所述开启链路层的数据发送通路包括:以链路层的工作时钟周期作为驱动时钟,开启链路层的数据发送通路的工作时钟。The step of opening the data transmission path of the link layer includes: using the working clock cycle of the link layer as a driving clock, opening the working clock of the data sending path of the link layer.
可选的,所述芯粒与互联的对端芯粒之间传输数据流包括:所述对端芯粒向所述芯粒发送数据流;所述在数据流的传输空闲时间,关闭链路层的数据传输通路,包括:Optionally, transmitting the data stream between the core particle and the interconnected opposite end core particle includes: the opposite end core particle sends the data stream to the core particle; and during the idle time of the data stream transmission, the link is closed. Layer data transmission path, including:
在传输空闲时间到来时,获取所述对端芯粒的链路层所发送的进入低功耗信号;When the transmission idle time arrives, obtain the low-power consumption signal sent by the link layer of the opposite end core;
响应于所述进入低功耗信号,关闭链路层的数据接收通路;In response to the entering low-power signal, close the data receiving path of the link layer;
所述在数据流的传输工作时间,开启链路层的数据传输通路,包括:The process of opening the link layer data transmission path during the transmission working time of the data stream includes:
在所述传输工作时间到来时,开启链路层的数据接收通路;When the transmission working time arrives, open the data receiving path of the link layer;
获取所述对端芯粒发送的退出低功耗信号,并响应于所述退出低功耗信号,重启所述数据流的传输。Obtain the low-power exit signal sent by the opposite end core chip, and in response to the low-power exit signal, restart the transmission of the data stream.
可选的,所述关闭链路层的数据接收通路包括:以链路层的工作时钟周期作为驱动时钟,关闭链路层的数据接收通路的工作时钟;Optionally, closing the data receiving path of the link layer includes: using the working clock cycle of the link layer as the driving clock, closing the working clock of the data receiving path of the link layer;
开启链路层的数据接收通路包括:以链路层的工作时钟周期作为驱动时钟,开启链路层的数据接收通路的工作时钟。Turning on the data receiving path of the link layer includes: using the working clock cycle of the link layer as the driving clock, turning on the working clock of the data receiving path of the link layer.
第三方面,本申请实施例提供一种芯片,包括互联的多个芯粒,所述芯粒如上述第一方面所述的芯粒。In a third aspect, embodiments of the present application provide a chip including a plurality of interconnected core particles, and the core particles are as described in the first aspect.
第四方面,本申请实施例一种计算机设备,包括如上述第三方面所述的芯片。In a fourth aspect, an embodiment of the present application provides a computer device, including the chip described in the third aspect.
本申请实施例提供的芯粒可以用于芯粒互联场景,在芯粒互联场景下,芯粒与互联的对端芯粒之间传输数据流。基于此,本申请实施例提供的芯粒可以包括芯粒互联接口,所述芯粒互联接口可以包括:链路层和物理层;所述链路层,用于在所述数据流的传输空闲时间,关闭所述链路层的数据传输通路,以使得所述链路层进入低功耗状态;其中,所述物理层维持工作状态不变。可见,本申请实施例提供的芯粒可以在数据流的传输空闲时间到来时,关闭芯粒的链路层的数据传输通路,从而使得芯粒的链路层的数据传输通路不工作,芯粒的链路层进入低功耗状态,降低芯粒的功耗消耗。同时,芯粒的物理层作为链路层的下一分层,由于芯粒的链路层的数据传输通路不传输数据,因此芯粒的物理层可以在链路层进入低功耗状态时,停止数据的传输工作,从而芯粒的物理层可以维持工作状态不变,避免物理层进行状态切换时的物理层重新训练过程,进而减少物理层的功耗消耗,并且增加数据流传输所能使用的时间(增加的时间与省去的物理层重新训练的时间相对应),提升数据传输效率。The core particles provided in the embodiments of the present application can be used in core particle interconnection scenarios. In the core particle interconnection scenario, data streams are transmitted between core particles and interconnected peer core particles. Based on this, the core particles provided in the embodiments of the present application may include a core particle interconnection interface, and the core particle interconnection interface may include: a link layer and a physical layer; the link layer is used to transmit the data stream when the data flow is idle. time, the data transmission path of the link layer is closed, so that the link layer enters a low power consumption state; wherein the physical layer maintains the working state unchanged. It can be seen that the core particle provided by the embodiment of the present application can close the data transmission path of the link layer of the core particle when the transmission idle time of the data stream arrives, so that the data transmission path of the link layer of the core particle does not work, and the core particle The link layer enters a low-power state to reduce the power consumption of the core. At the same time, the physical layer of the core particle serves as the next layer of the link layer. Since the data transmission path of the link layer of the core particle does not transmit data, the physical layer of the core particle can enter the low power consumption state when the link layer enters the low power consumption state. Stop the data transmission work, so that the physical layer of the core particle can maintain the working state unchanged, avoiding the physical layer retraining process when the physical layer performs state switching, thereby reducing the power consumption of the physical layer, and increasing the use of data stream transmission time (the added time corresponds to the saved physical layer retraining time) to improve data transmission efficiency.
可见,本申请实施例可以在芯粒与对端芯粒的数据流的传输空闲时间,通过控制链路层进入低功耗状态,而物理层维持工作状态不变,实现芯粒的低功耗控制,降低芯粒的功耗消耗;并且避免在物理层进行低功耗控制所带来的状态切换的重新训练过程,省去物理层的重新训练带来的功耗和时间占用,提升了数据流传输所能使用的时间。因此本申请实施例提供的方案能够在芯粒互联场景下,降低芯粒互联的功耗,并且通过避免物理层的重新训练,提升数据流传输所能使用的时间,提升数据传输效率。It can be seen that the embodiment of the present application can control the link layer to enter a low-power state during the idle time of the data stream transmission between the core and the peer core, while the physical layer maintains the working state unchanged, thereby achieving low power consumption of the core. control, reducing the power consumption of the core; and avoiding the retraining process of state switching caused by low-power control at the physical layer, eliminating the power consumption and time occupation caused by retraining of the physical layer, and improving data The time available for streaming. Therefore, the solution provided by the embodiments of this application can reduce the power consumption of core-to-die interconnection in the core-to-die interconnection scenario, and by avoiding retraining of the physical layer, it can increase the time available for data stream transmission and improve data transmission efficiency.
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是芯粒互联的可选结构示意图。Figure 1 is a schematic diagram of an optional structure of core-to-chip interconnection.
图2是芯粒互联协议在芯粒互联接口的分层的可选结构示意图。Figure 2 is a schematic diagram of an optional layered structure of the core-to-chip interconnection protocol in the core-to-chip interconnection interface.
图3是芯粒互联的另一可选结构示意图。Figure 3 is a schematic diagram of another optional structure of core-chip interconnection.
图4是芯粒互联场景下,低功耗控制方法的可选流程示意图。Figure 4 is a schematic flow diagram of an optional low-power control method in a chip-to-chip interconnection scenario.
图5是本申请实施例提供的芯粒互联的可选结构示意图。Figure 5 is a schematic diagram of an optional structure of core-to-chip interconnection provided by an embodiment of the present application.
图6是本申请实施例提供的低功耗控制方法的可选流程示意图。Figure 6 is an optional flow diagram of the low power consumption control method provided by the embodiment of the present application.
图7是本申请实施例提供的低功耗控制方法中对应进入低功耗状态的可选流程示意图。FIG. 7 is a schematic diagram of an optional flow corresponding to entering a low power consumption state in the low power consumption control method provided by the embodiment of the present application.
图8是本申请实施例提供的低功耗控制方法中对应退出低功耗状态的可选流程示意图。FIG. 8 is a schematic diagram of an optional flow corresponding to exiting the low power consumption state in the low power consumption control method 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.
为便于理解芯粒互联,以芯粒1和芯粒2为例,图1示例性的示出了芯粒互联的可选结构示意图。如图1所示,芯粒内部可以包括芯粒互联接口以及多个功能模块,其中,芯粒互联接口通过系统总线,与多个功能模块相连接。而在芯粒外部,芯粒1和芯粒2可以通过芯粒互联接口实现通信互联,传输数据流。针对任一芯粒,芯粒的对端芯粒可以是与该芯粒相通信的另一芯粒。In order to facilitate understanding of core-particle interconnection, taking core particle 1 and core particle 2 as an example, Figure 1 exemplarily shows an optional structural schematic diagram of core-particle interconnection. As shown in Figure 1, the core chip may include a core chip interconnection interface and multiple functional modules. The core chip interconnection interface is connected to multiple functional modules through a system bus. Outside the core, core 1 and core 2 can communicate and interconnect through the core interconnect interface to transmit data streams. For any core particle, the opposite core particle of the core particle may be another core particle in communication with the core particle.
需要说明的是,通过芯粒互联接口互联的两个芯粒可以是任意类型的芯粒,例如,通过互联接口进行互联的两个功能芯粒、通过互联接口进行互联的功能芯粒与接口芯粒等。It should be noted that the two core particles interconnected through the core particle interconnection interface can be any type of core particles, for example, two functional core particles interconnected through the interconnection interface, a functional core particle and an interface core interconnected through the interconnection interface. Granules etc.
基于芯粒的类型不同,芯粒中实现芯粒功能的功能模块的形式也可能不同。例如,功能芯粒可以包括处理器核芯粒(比如CPU核芯粒),处理器核芯粒中的功能模块可以是处理器核和高速缓存(Cache)集成的模块等。又例如,接口芯粒可以是对接网络通信、USB(Universal Serial Bus、通用串行总线)通信、PCIE(Peripheral ComponentInterconnect Express,高速串行计算机扩展总线标准)通信等通信方式的接口模块所在的芯粒。Based on the different types of core particles, the forms of functional modules in the core particles that implement core particle functions may also be different. For example, the functional core may include a processor core (such as a CPU core), and the functional module in the processor core may be a module integrated with a processor core and a cache (Cache), etc. For another example, the interface chip can be the chip where the interface module is located for network communication, USB (Universal Serial Bus, Universal Serial Bus) communication, PCIE (Peripheral Component Interconnect Express, high-speed serial computer expansion bus standard) communication and other communication methods. .
芯粒在通过芯粒互联接口进行互联时,依据的是芯粒互联协议,其中,芯粒互联协议可以定义互联的芯粒进行数据传输的分层机制。芯粒互联协议在芯粒的分层由上至下可以包括:应用层、协议层、链路层和物理层。需要说明的是,应用层、协议层、链路层和物理层是基于芯粒互联协议的分层在芯粒中实现的用于芯粒互联的电路模块;也就是说,应用层、协议层、链路层和物理层可以是电路模块的形式,例如,应用层电路模块(简称应用层)、协议层电路模块(简称协议层)、链路层电路模块(简称链路层)、物理层电路模块(简称物理层)。在可能的实现中,链路层和物理层可以部署于芯粒的互联接口,作为互联接口中的电路模块;应用层和协议层可以部署于芯粒的互联接口外部,并设置为链路层的上层电路模块。When core particles are interconnected through the core particle interconnection interface, they are based on the core particle interconnection protocol. Among them, the core particle interconnection protocol can define the hierarchical mechanism for data transmission between interconnected core particles. The core layer of the core interconnect protocol can include: application layer, protocol layer, link layer and physical layer from top to bottom. It should be noted that the application layer, protocol layer, link layer and physical layer are circuit modules implemented in the core based on the layering of the core-to-chip interconnection protocol for core-to-chip interconnection; that is to say, the application layer, protocol layer , the link layer and the physical layer can be in the form of circuit modules, for example, application layer circuit module (referred to as application layer), protocol layer circuit module (referred to as protocol layer), link layer circuit module (referred to as link layer), physical layer Circuit module (referred to as physical layer). In a possible implementation, the link layer and physical layer can be deployed on the interconnection interface of the core particle as a circuit module in the interconnection interface; the application layer and protocol layer can be deployed outside the interconnection interface of the core particle and set as the link layer. upper circuit module.
作为可选实现,链路层和物理层可以部署于芯粒的芯粒互联接口。在一个示例中,图2示出了芯粒互联协议在芯粒互联接口的分层的可选结构示意图,如图2所示,芯粒互联协议在芯粒互联接口的分层可以包括链路层(例如,图2中芯粒1的链路层11、芯粒2的链路层21),和物理层(例如,图2中芯粒1的物理层12、芯粒2的物理层22)。As an optional implementation, the link layer and physical layer can be deployed on the core interconnect interface of the core. In one example, Figure 2 shows an optional structural diagram of the layering of the core interconnection protocol at the core interconnection interface. As shown in Figure 2, the layering of the core interconnection protocol at the core interconnection interface may include links. layer (for example, the link layer 11 of core particle 1 and the link layer 21 of core particle 2 in Figure 2), and the physical layer (for example, the physical layer 12 of core particle 1 and the physical layer 22 of core particle 2 in Figure 2 ).
在芯粒中,链路层用于为芯粒间的通信提供可靠性保障,包括但不限于:纠错、校验、重传机制等。In core particles, the link layer is used to provide reliability guarantee for communication between core particles, including but not limited to: error correction, verification, retransmission mechanism, etc.
在芯粒中,物理层用于提供芯粒进行数据传输的物理传输通路(简称通路)。In the core particle, the physical layer is used to provide the physical transmission path (referred to as the path) for the core particle to transmit data.
在芯粒互联协议的分层中,链路层与物理层可以通过相适配的接口连接。In the layering of the core-to-chip interconnection protocol, the link layer and the physical layer can be connected through adapted interfaces.
需要进一步说明的是,芯粒互联协议的分层中还可以包括应用层和协议层。应用层为互联协议的顶部分层,软件可以运行于应用层。例如,软件可以由芯粒的处理器硬件执行并运行于应用层,处理器硬件比如芯粒的处理器核等。协议层位于应用层之下,用于支持芯粒承载的业务所对应的业务协议,与芯粒承载的业务的类型相关,并且,协议层也可以支持用户自定义协议。It should be further explained that the layering of the core-to-chip interconnection protocol may also include an application layer and a protocol layer. The application layer is the top layer of the Internet protocol, and software can run on the application layer. For example, the software may be executed by the core's processor hardware, such as the core's processor core, etc., and run at the application layer. The protocol layer is located below the application layer and is used to support the business protocols corresponding to the services carried by the core. It is related to the type of services carried by the core. Moreover, the protocol layer can also support user-defined protocols.
在芯粒发送数据的方向,芯粒的应用层的数据通过协议层接收,协议层将数据发送至链路层(例如协议层可以将数据数据切分为子数据,并传输给链路层),链路层通过相应的接口将数据分发到物理层传输。In the direction of the core particle sending data, the core particle's application layer data is received through the protocol layer, and the protocol layer sends the data to the link layer (for example, the protocol layer can divide the data data into sub-data and transmit it to the link layer) , the link layer distributes data to the physical layer for transmission through the corresponding interface.
芯粒互联的目的之一是实现芯粒之间的通信,即实现芯粒之间的数据传输。例如芯粒之间传输的数据为数据流时,需要实现芯粒之间的数据流传输。基于图2所示芯粒互联协议的分层,对芯粒1的数据发送方向进行介绍。在芯粒1的内部,链路层11获取协议层发送的数据流,链路层11通过链路层的数据传输通路将数据流传输至物理层12,芯粒1的物理层12通过数据传输通路将数据流传输至对端芯粒2的物理层22。也就是说,在芯粒1的数据发送方向上,数据流是在芯粒的链路层11和物理层12进行逐层传递,并通过物理层12将数据流传输给对端的芯粒2。One of the purposes of core-particle interconnection is to realize communication between cores, that is, to realize data transmission between cores. For example, when the data transmitted between core particles is a data stream, data stream transmission between core particles needs to be implemented. Based on the layering of the core-to-chip interconnection protocol shown in Figure 2, the data sending direction of core 1 is introduced. Inside the core 1, the link layer 11 obtains the data stream sent by the protocol layer. The link layer 11 transmits the data stream to the physical layer 12 through the data transmission path of the link layer. The physical layer 12 of the core 1 transmits the data through the data transmission path. The path transmits the data stream to the physical layer 22 of the peer core 2 . That is to say, in the data sending direction of core 1, the data stream is transmitted layer by layer in the link layer 11 and physical layer 12 of the core, and the data stream is transmitted to the opposite end core 2 through the physical layer 12.
对芯粒2的数据接收方向进行介绍。芯粒2的物理层22通过数据传输通路,接收芯粒1的物理层12发送的数据流;进而,在芯粒2内部,物理层22通过物理层的数据传输通路将数据流传输至链路层21。也就是说,在芯粒2的数据接收方向上,芯粒2的物理层22接收对端的芯粒1的物理层12发送的数据流后,数据流是在芯粒2的物理层22、链路层21进行逐层传递。The data receiving direction of core 2 is introduced. The physical layer 22 of the core particle 2 receives the data stream sent by the physical layer 12 of the core particle 1 through the data transmission path; further, inside the core particle 2, the physical layer 22 transmits the data stream to the link through the data transmission path of the physical layer. Layer 21. That is to say, in the data receiving direction of core 2, after the physical layer 22 of core 2 receives the data stream sent by the physical layer 12 of core 1 at the opposite end, the data flow is transmitted between the physical layer 22 of core 2 and the link. Road layer 21 performs layer-by-layer transmission.
可见,芯粒的数据传输分为发送数据和接收数据,即芯粒可以作为发送方,向对端芯粒发送数据信息,也可以作为接收方,接收对端芯粒发送的数据信息。进而,在芯粒的数据传输分为发送数据和接收数据的情况下,芯粒的数据传输通路可以分为数据发送通路和数据接收通路。参照图3所示的芯粒互联的另一可选结构示意图,芯粒的链路层可以包括数据发送通路(Tx Data Signal)和数据接收通路(Rx Data Signal);其中,链路层的数据发送通路例如图中芯粒1的链路层11的数据发送通路111,芯粒2的链路层21的数据发送通路212;链路层的数据接收通路例如图中芯粒1的链路层11的数据接收通路112,芯粒2的链路层21的数据接收通路211。芯粒的物理层可以包括数据发送通路(Tx Data Lane)和数据接收通路(Rx Data Lane);物理层的数据发送通路例如图中芯粒1的物理层12的数据发送通路121,芯粒2的物理层22的数据发送通路222;芯粒的物理层的数据接收通路例如图中芯粒1的物理层12的数据接收通路122,芯粒2的物理层22的数据接收通路221。It can be seen that the data transmission of the core particle is divided into sending data and receiving data, that is, the core particle can be a sender, sending data information to the opposite end core particle, or it can be a receiver, receiving data information sent by the opposite end core particle. Furthermore, when the data transmission of the core particle is divided into sending data and receiving data, the data transmission path of the core particle can be divided into a data sending path and a data receiving path. Referring to another optional structural diagram of core-chip interconnection shown in Figure 3, the link layer of the core chip may include a data transmission path (Tx Data Signal) and a data reception path (Rx Data Signal); where, the link layer data The transmission path is, for example, the data transmission path 111 of the link layer 11 of the core 1 in the figure, and the data transmission path 212 of the link layer 21 of the core 2; the data reception path of the link layer is, for example, the link layer of the core 1 in the figure. The data receiving path 112 of 11, and the data receiving path 211 of the link layer 21 of core 2. The physical layer of the core particle may include a data transmission path (Tx Data Lane) and a data reception path (Rx Data Lane); the data transmission path of the physical layer is, for example, the data transmission path 121 of the physical layer 12 of the core particle 1 in the figure, and the data transmission path of the core particle 2. The data sending path 222 of the physical layer 22 of the core chip; the data receiving path of the physical layer of the core chip are, for example, the data receiving path 122 of the physical layer 12 of the core chip 1 in the figure, and the data receiving path 221 of the physical layer 22 of the core chip 2.
需要说明的是,Tx表示发送,Rx表示接收。作为发送方的芯粒(例如芯粒1)通过数据发送通路,能够将数据传输至作为接收方的芯粒(例如芯粒2)的数据接收通路,从而实现芯粒的互联通信。It should be noted that Tx means sending and Rx means receiving. The core particle as the sender (for example, core particle 1) can transmit data to the data reception path of the core particle as the receiver (for example, core particle 2) through the data sending path, thereby realizing the interconnected communication of the core particles.
需要进一步说明的是,不同数据传输通路的数据,由于传输路径的不同和物理层处理的差异,因此数据到达接收方的延迟不同,从而不同数据传输通路的数据需要对齐,例如:芯粒1中链路层11的数据发送通路111需要与物理层12的数据发送通路121对齐,链路层11的数据接收通路112需要与物理层12的数据接收通路122对齐;芯粒2中链路层21的数据接收通路211需要与物理层22的数据接收通路221对齐,链路层21的数据发送通路212需要与物理层22的数据发送通路222对齐。在芯粒互联场景下,芯粒互联接口的链路层能够控制物理层的训练,以完成数据传输通路的对齐处理,使得数据能够准确地传输和接收,并找到正确的采样时间点。It needs to be further explained that due to different transmission paths and differences in physical layer processing, the data in different data transmission paths have different delays in reaching the receiver, so the data in different data transmission paths need to be aligned, for example: in Core 1 The data sending path 111 of the link layer 11 needs to be aligned with the data sending path 121 of the physical layer 12, and the data receiving path 112 of the link layer 11 needs to be aligned with the data receiving path 122 of the physical layer 12; the link layer 21 in the core 2 The data receiving path 211 needs to be aligned with the data receiving path 221 of the physical layer 22, and the data sending path 212 of the link layer 21 needs to be aligned with the data sending path 222 of the physical layer 22. In the core-to-chip interconnection scenario, the link layer of the core-to-grain interconnection interface can control the training of the physical layer to complete the alignment of the data transmission path, so that data can be accurately transmitted and received, and the correct sampling time point can be found.
在芯粒中上层业务模块(例如协议层),存在业务工作状态和业务空闲状态,因此可以在业务工作状态和业务空闲状态下,通过低功耗状态的退出和进入,来实现控制芯粒的功耗消耗。例如,当没有数据时,上层业务模块处于业务空闲状态,数据传输通路传输的数据流中的数据为空数据,上层业务模块可以控制芯粒互联接口进入低功耗状态,以关闭物理层的数据传输功能,降低功耗。当有数据时,上层业务模块进入业务工作状态,可以控制芯粒互联接口退出低功耗状态,恢复正常的数据传输。In the upper-layer business module of the core (such as the protocol layer), there is a business working state and a business idle state. Therefore, the core can be controlled by exiting and entering the low-power state in the business working state and business idle state. Power consumption. For example, when there is no data, the upper-layer business module is in the idle state, and the data in the data stream transmitted by the data transmission channel is empty data. The upper-layer business module can control the chip interconnection interface to enter a low-power state to turn off the data of the physical layer. transmission function to reduce power consumption. When there is data, the upper-layer business module enters the business working state and can control the chip interconnection interface to exit the low-power state and resume normal data transmission.
以芯粒互联的芯粒1为发送方,芯粒2为接收方为例,芯粒1可以发起低功耗进入过程,以与芯粒2同步进入低功耗状态;当需要退出低功耗时,芯粒1可以发起低功耗退出过程,以与芯粒2同步退出低功耗状态。在可选实现中,以通过控制物理层实现低功耗状态的进入和退出为例,图4示例性的示出了芯粒互联场景下,低功耗控制方法的可选流程示意图。Taking core 1 as the sender and core 2 as the receiver in the core-to-chip interconnection as an example, core 1 can initiate a low-power entry process to enter the low-power state synchronously with core 2; when it is necessary to exit low-power When, Core 1 can initiate a low-power exit process to exit the low-power state synchronously with Core 2. In an optional implementation, taking the entry and exit of a low-power state by controlling the physical layer as an example, Figure 4 illustrates an optional flow chart of a low-power control method in a core-to-chip interconnection scenario.
如图4所示,在进入低功耗状态时,芯粒1的链路层11向芯粒2的链路层21发送进入低功耗握手信号,以请求芯粒2进入低功耗状态。As shown in Figure 4, when entering the low power consumption state, the link layer 11 of the core 1 sends an entry low power handshake signal to the link layer 21 of the core 2 to request the core 2 to enter the low power state.
当芯粒2接收到该进入低功耗握手信号,若芯粒2支持进入低功耗状态,则响应该进入低功耗握手信号,芯粒2的链路层21向芯粒1的链路层11发送进入低功耗握手信号,从而芯粒1和芯粒2握手成功,完成进入低功耗状态的交互。When core 2 receives the low-power handshake signal, if core 2 supports entering the low-power state, it responds to the low-power handshake signal, and the link layer 21 of core 2 establishes a link to core 1. Layer 11 sends the low-power handshake signal, so that the handshake between core 1 and core 2 is successful, completing the interaction to enter the low-power state.
在完成进入低功耗状态的交互时,芯粒1的链路层11控制物理层12进入低功耗状态;同时,芯粒2的链路层21控制物理层22进入低功耗状态。When the interaction to enter the low-power state is completed, the link layer 11 of the core 1 controls the physical layer 12 to enter the low-power state; at the same time, the link layer 21 of the core 2 controls the physical layer 22 to enter the low-power state.
芯粒的链路层控制物理层进入低功耗状态,能够关闭物理层的数据传输通路。例如,芯粒1作为发送方,芯粒1的物理层12关闭数据发送通路,停止向芯粒2的物理层21发送新的数据;芯粒2作为接收方,芯粒2的物理层22关闭数据接收通路,停止接收芯粒1的物理层11发送的数据。The link layer of the core particle controls the physical layer to enter a low power consumption state and can close the data transmission path of the physical layer. For example, core 1 serves as the sender, and the physical layer 12 of core 1 closes the data transmission path and stops sending new data to the physical layer 21 of core 2; core 2 serves as the receiver, and the physical layer 22 of core 2 closes The data receiving path stops receiving data sent by the physical layer 11 of the core 1.
由于进入低功耗状态时,物理层关闭了数据传输通路,从而在退出低功耗状态时,链路层需要控制物理层重新训练,调整数据传输通路,以完成传输通道的数据对齐等处理,使得链路层传输的信号能够被物理层准确接收和传输。因此,在退出低功耗状态时,芯粒1的链路层11控制物理层12重新训练。同时,芯粒2的链路层21控制物理层22重新训练。Since the physical layer closes the data transmission path when entering the low-power state, when exiting the low-power state, the link layer needs to control the physical layer to retrain and adjust the data transmission path to complete the data alignment of the transmission channel. This enables signals transmitted by the link layer to be accurately received and transmitted by the physical layer. Therefore, when exiting the low power consumption state, the link layer 11 of the core 1 controls the physical layer 12 to retrain. At the same time, the link layer 21 of the core 2 controls the physical layer 22 to retrain.
在物理层重新训练后,芯粒1的链路层11向芯粒2的链路层21发送退出低功耗握手信号。After the physical layer is retrained, the link layer 11 of core 1 sends an exit low-power handshake signal to the link layer 21 of core 2.
当芯粒2接收到芯粒1发送的退出低功耗握手信号,则响应该退出低功耗握手信号,芯粒2的链路层21向芯粒1的链路层11发送退出低功耗握手信号,完成与芯粒1退出低功耗状态的交互,芯粒1与芯粒2恢复正常的数据流传输。When core 2 receives the exit low-power handshake signal sent by core 1, in response to the exit low-power handshake signal, the link layer 21 of core 2 sends an exit low-power handshake signal to the link layer 11 of core 1. The handshake signal completes the interaction with Core 1 to exit the low-power state, and Core 1 and Core 2 resume normal data flow transmission.
需要说明的是,物理层训练的过程较为繁琐,包括数据通路的对齐、时钟到数据的对齐、数据保序等,导致其训练所需的时间较长,一般经过较长的微秒(us)级时间,才能够达到物理层的重新锁定。因此,通过物理层进入低功耗状态和退出低功耗状态,实现芯粒的低功耗控制的方案,由于退出低功耗状态时需要进行物理层的重新训练,这增加了芯粒低功耗控制的进出时间和操作复杂度,不利于芯粒互联的功耗降低。It should be noted that the physical layer training process is relatively cumbersome, including data path alignment, clock-to-data alignment, data order preservation, etc., resulting in a longer training time, which generally takes longer microseconds (us). level of time to achieve re-locking of the physical layer. Therefore, through the physical layer entering and exiting the low-power state, the low-power control scheme of the core is implemented. Since the physical layer needs to be retrained when exiting the low-power state, this increases the cost of low-power control of the core. The entry and exit time and operation complexity of power consumption control are not conducive to reducing the power consumption of chip interconnection.
有鉴于此,本申请实施例提供芯粒互联场景下改进的低功耗控制方案,使得低功耗控制基于链路层实现,无需改变物理层的工作状态。In view of this, embodiments of the present application provide an improved low-power control solution in a core-to-chip interconnection scenario, so that low-power control is implemented based on the link layer without changing the working state of the physical layer.
基于上述思路,作为可选实现,本申请实施例可以在数据流的传输空闲时间到来时,关闭芯粒的链路层的数据传输通路,使得芯粒的链路层的数据传输通路不工作,芯粒的链路层进入低功耗状态,降低芯粒的功耗消耗。同时,芯粒的物理层作为链路层的下一分层,由于芯粒的链路层的数据传输通路不传输数据,因此芯粒的物理层可以在链路层进入低功耗状态时,停止数据的传输工作,从而芯粒的物理层可以维持工作状态不变,避免物理层进行状态切换时的物理层重新训练过程,进而减少物理层的功耗消耗,并且增加数据流传输所能使用的时间(增加的时间与省去的物理层重新训练的时间相对应),提升数据传输效率。Based on the above ideas, as an optional implementation, the embodiment of the present application can close the data transmission path of the link layer of the core particle when the transmission idle time of the data stream arrives, so that the data transmission path of the link layer of the core particle does not work. The link layer of the core particle enters a low power consumption state, reducing the power consumption of the core particle. At the same time, the physical layer of the core particle serves as the next layer of the link layer. Since the data transmission path of the link layer of the core particle does not transmit data, the physical layer of the core particle can enter the low power consumption state when the link layer enters the low power consumption state. Stop the data transmission work, so that the physical layer of the core particle can maintain the working state unchanged, avoiding the physical layer retraining process when the physical layer performs state switching, thereby reducing the power consumption of the physical layer, and increasing the use of data stream transmission time (the added time corresponds to the saved physical layer retraining time) to improve data transmission efficiency.
为实现上述目的,作为可选实现,本申请实施例可以对链路层进行改进,使得数据流的传输空闲时间到来时,关闭链路层的数据传输通路,以及在数据流的传输工作时间到来时,开启链路层的数据传输通路。其中,图5示例性的示出了本申请实施例的芯粒互联的可选结构示意图。如图5所示,在芯粒互联场景下,芯粒1与互联的芯粒2之间传输数据流。In order to achieve the above purpose, as an optional implementation, embodiments of the present application can improve the link layer, so that when the idle time for the transmission of the data stream arrives, the data transmission path of the link layer is closed, and when the transmission working time of the data stream arrives, When , open the data transmission channel of the link layer. Among them, FIG. 5 exemplarily shows an optional structural schematic diagram of core-chip interconnection according to the embodiment of the present application. As shown in Figure 5, in the core-to-core interconnection scenario, data streams are transmitted between core 1 and interconnected core 2.
其中,针对任一芯粒(例如芯粒1或芯粒2),芯粒可以包括芯粒互联接口,芯粒互联接口可以包括链路层和物理层,本申请实施例的链路层可以用于在数据流的传输空闲时间,关闭所述链路层的数据传输通路,以使得所述链路层进入低功耗状态。其中,芯粒的物理层维持工作状态不变。Among them, for any core particle (for example, core particle 1 or core particle 2), the core particle may include a core particle interconnection interface, and the core particle interconnection interface may include a link layer and a physical layer. The link layer in the embodiment of the present application may be During the transmission idle time of the data stream, the data transmission path of the link layer is closed, so that the link layer enters a low power consumption state. Among them, the physical layer of the core particle maintains its working state unchanged.
所述数据流的传输空闲时间指的是链路层的上层业务模块,在业务空闲状态传输空数据的时间。The transmission idle time of the data flow refers to the time during which the upper-layer service module of the link layer transmits empty data in the service idle state.
在芯粒的数据传输分为发送数据和接收数据情况下,所述链路层的数据传输通路可以包括数据发送通路和数据接收通路。When the data transmission of the core particle is divided into sending data and receiving data, the data transmission path of the link layer may include a data sending path and a data receiving path.
在一些实施例中,对应进入低功耗控制状态时,链路层的数据传输通路的关闭,本申请实施例中的链路层还用于在数据流的传输工作时间,开启所述链路层的数据传输通路,以使得所述链路层退出低功耗状态。In some embodiments, corresponding to the closing of the data transmission path of the link layer when entering the low-power control state, the link layer in the embodiment of the present application is also used to open the link during the transmission working time of the data stream. layer data transmission path, so that the link layer exits the low power consumption state.
为便于理解本申请实施例基于链路层进行低功耗控制的方式,图6示例性的示出了本申请实施例的低功耗控制方法的可选流程示意图。图6仅以一个芯粒为例进行示出。如图6所示,应用于芯粒的低功耗控制方法可以包括:In order to facilitate understanding of the way in which low power consumption control is performed based on the link layer in this embodiment of the present application, FIG. 6 exemplarily shows an optional flowchart of the low power consumption control method in this embodiment of the present application. Figure 6 shows only one core particle as an example. As shown in Figure 6, low-power control methods applied to core chips can include:
步骤S100,链路层在数据流的传输空闲时间,关闭所述链路层的数据传输通路。Step S100: The link layer closes the data transmission path of the link layer during the idle time of the data stream transmission.
步骤S200,链路层在数据流的传输工作时间,开启所述链路层的数据传输通路。Step S200: The link layer opens the data transmission path of the link layer during the transmission working time of the data stream.
对应参照图5,以芯粒互联的芯粒1为发送方,芯粒2为接收方为例,对低功耗控制过程进行说明。Correspondingly referring to Figure 5, taking core 1 as the sender and core 2 as the receiver of the core-to-chip interconnection as an example, the low power consumption control process will be described.
在数据流的传输空闲时间,如图5所示,芯粒1的链路层11关闭其数据发送通路111,从而,芯粒1的链路层11进入低功耗状态,而芯粒1的物理层12的工作状态维持不变,芯粒1的物理层12的数据发送通路121仍为通道对齐状态,并且由于芯粒1的链路层11的数据发送通路111不传输数据,芯粒1的物理层12可以在链路层11进入低功耗状态时,停止数据发送通路121的数据发送工作。During the idle time of data stream transmission, as shown in Figure 5, the link layer 11 of core 1 closes its data sending path 111. As a result, the link layer 11 of core 1 enters a low power consumption state, while the link layer 11 of core 1 The working state of the physical layer 12 remains unchanged. The data transmission path 121 of the physical layer 12 of the core 1 is still in the channel alignment state. Since the data transmission path 111 of the link layer 11 of the core 1 does not transmit data, the core 1 The physical layer 12 can stop the data sending work of the data sending path 121 when the link layer 11 enters the low power consumption state.
作为接收方的芯粒2可以与作为发送发的芯粒1同步进入低功耗状态。在一个可选示例中,芯粒1的链路层11可以在数据流的传输空闲时间,向芯粒2的链路层21发送对应进入低功耗状态的信号,以使芯粒2的链路层21关闭其数据接收通路211,从而,芯粒2的链路层21进入低功耗状态,而芯粒2的物理层22的工作状态维持不变,芯粒2的物理层22的数据接收通路221仍为通道对齐状态,并且由于芯粒2的链路层21的数据接收通路211不接收数据,芯粒2的物理层22可以在链路层21进入低功耗状态时,停止数据接收通路221的数据接收工作,降低芯粒的功耗消耗。Core 2 as the receiver can enter the low power consumption state synchronously with core 1 as the sender. In an optional example, the link layer 11 of core 1 can send a signal corresponding to entering the low power consumption state to the link layer 21 of core 2 during the idle time of data stream transmission, so that the link layer of core 2 The link layer 21 closes its data receiving path 211, thus the link layer 21 of the core 2 enters a low power consumption state, while the working state of the physical layer 22 of the core 2 remains unchanged, and the data of the physical layer 22 of the core 2 remains unchanged. The receiving path 221 is still in the channel aligned state, and since the data receiving path 211 of the link layer 21 of the core 2 does not receive data, the physical layer 22 of the core 2 can stop data when the link layer 21 enters the low power consumption state. The data reception work of the receiving path 221 reduces the power consumption of the chip.
在数据流的传输工作时间,如图5所示,芯粒1的链路层11会开启其数据发送通路111,从而,芯粒1的链路层11退出低功耗状态。而且,为使数据流的传输重启,芯粒1的链路层11可以通过数据发送通路111向芯粒2的链路层21发送退出低功耗信号。During the transmission working time of the data stream, as shown in Figure 5, the link layer 11 of the core 1 will open its data transmission path 111, so that the link layer 11 of the core 1 exits the low power consumption state. Furthermore, in order to restart the transmission of the data stream, the link layer 11 of the core 1 may send an exit low-power consumption signal to the link layer 21 of the core 2 through the data transmission path 111 .
作为接收方的芯粒2与作为发送发的芯粒1同步退出低功耗状态,基于芯粒1的链路层11在数据流的传输工作时间,开启其数据发送通路111,芯粒2的链路层21可以在数据流的传输工作时间,开启其数据接收通路211,从而,芯粒2的链路层21退出低功耗状态。并且,芯粒2的链路层21的数据接收通路211可以对应获取芯粒1的链路层11发送的退出低功耗信号,以重启数据流的传输。Core 2 as the receiver and core 1 as the sender exit the low-power state synchronously. The link layer 11 based on core 1 opens its data transmission path 111 during the transmission working time of the data stream. Core 2's The link layer 21 can open its data receiving path 211 during the transmission working time of the data stream, so that the link layer 21 of the core 2 exits the low power consumption state. Furthermore, the data receiving path 211 of the link layer 21 of the core 2 can correspondingly obtain the exit low power consumption signal sent by the link layer 11 of the core 1 to restart the transmission of the data stream.
需要说明的是,图5中以芯粒2为发送方,芯粒1为接收方实现低功耗控制的过程可以同理参照上述内容,本申请实施例对此不做过多赘述。It should be noted that in Figure 5 , with core 2 as the sender and core 1 as the receiver, the process of implementing low power consumption control can refer to the above content in the same way, and the embodiments of this application will not go into details.
可以看出,本申请实施例提供的芯粒可以在数据流的传输空闲时间到来时,关闭芯粒的链路层的数据传输通路,从而使得芯粒的链路层的数据传输通路不工作,芯粒的链路层进入低功耗状态,降低芯粒的功耗消耗。同时,芯粒的物理层作为链路层的下一分层,由于芯粒的链路层的数据传输通路不传输数据,因此芯粒的物理层可以在链路层进入低功耗状态时,停止数据的传输工作,从而芯粒的物理层可以维持工作状态不变,避免物理层进行状态切换时的物理层重新训练过程,进而减少物理层的功耗消耗,并且增加数据流传输所能使用的时间(增加的时间与省去的物理层重新训练的时间相对应),提升数据传输效率。It can be seen that the core particle provided by the embodiment of the present application can close the data transmission path of the link layer of the core particle when the transmission idle time of the data stream arrives, so that the data transmission path of the link layer of the core particle does not work. The link layer of the core particle enters a low power consumption state, reducing the power consumption of the core particle. At the same time, the physical layer of the core particle serves as the next layer of the link layer. Since the data transmission path of the link layer of the core particle does not transmit data, the physical layer of the core particle can enter the low power consumption state when the link layer enters the low power consumption state. Stop the data transmission work, so that the physical layer of the core particle can maintain the working state unchanged, avoiding the physical layer retraining process when the physical layer performs state switching, thereby reducing the power consumption of the physical layer, and increasing the use of data stream transmission time (the added time corresponds to the saved physical layer retraining time) to improve data transmission efficiency.
可见,本申请实施例在芯粒与对端芯粒的数据流的传输空闲时间,通过控制链路层进入低功耗状态,而物理层维持工作状态不变,实现芯粒的低功耗控制,降低芯粒的功耗消耗;并且避免在物理层进行低功耗控制所带来的状态切换的重新训练过程,省去物理层的重新训练带来的功耗和时间占用,提升了数据流传输所能使用的时间。因此本申请实施例提供的方案能够在芯粒互联场景下,降低芯粒互联的功耗,并且通过避免物理层的重新训练,提升数据流传输所能使用的时间,提升数据传输效率。It can be seen that the embodiment of the present application controls the link layer to enter a low-power state during the idle time of the data stream transmission between the core and the peer core, while the physical layer maintains the working state unchanged, thereby achieving low-power control of the core. , reduce the power consumption of the core; and avoid the retraining process of state switching caused by low-power control at the physical layer, eliminating the power consumption and time occupation caused by retraining of the physical layer, and improving the data flow The time available for transmission. Therefore, the solution provided by the embodiments of this application can reduce the power consumption of core-to-die interconnection in the core-to-die interconnection scenario, and by avoiding the retraining of the physical layer, it can increase the time available for data stream transmission and improve data transmission efficiency.
在一个可选实现中,为实现链路层在数据流的传输空闲时间,关闭其数据传输通路,本申请实施例的链路层可以设置低功耗控制逻辑(例如图5中芯粒1中的低功耗控制逻辑13、芯粒2中的低功耗控制逻辑23)。上述芯粒的链路层进入和退出低功耗状态的功能可以由芯粒的链路层设置的低功耗控制逻辑实现。低功耗控制逻辑是本申请实施例在芯粒的链路层中设置的逻辑电路,所述低功耗控制逻辑可以用于在数据流的传输空闲时间,控制链路层关闭所述链路层的数据传输通路,以及在数据流的传输工作时间,控制链路层开启所述链路层的数据传输通路。从而,基于低功耗控制逻辑的控制,链路层能够实现低功耗状态的进入和退出。In an optional implementation, in order to realize that the link layer closes its data transmission path during the idle time of data stream transmission, the link layer in the embodiment of the present application can set low-power control logic (for example, in core 1 in Figure 5 Low power consumption control logic 13, low power consumption control logic 23 in core 2). The function of the link layer of the core particle entering and exiting the low-power state can be implemented by the low-power control logic set by the link layer of the core particle. The low-power control logic is a logic circuit set in the link layer of the core in the embodiment of the present application. The low-power control logic can be used to control the link layer to close the link during the idle time of data stream transmission. The data transmission path of the layer, and during the transmission working time of the data flow, the link layer is controlled to open the data transmission path of the link layer. Therefore, based on the control of low-power control logic, the link layer can realize entry and exit of low-power states.
可以理解的是,在链路层的低功耗控制,基于链路层的低功耗控制逻辑实现的情况下,由于芯粒的链路层在进入低功耗状态时,芯粒的物理层的工作状态维持不变,从而在链路层开启其数据传输通路后,链路层通过数据传输通路能够将数据流传输至物理层的数据传输通路,避免芯粒的物理层进行状态切换时的物理层重新训练过程,减少了物理层重新训练的功耗消耗和相应时间。It can be understood that when the low power consumption control of the link layer is implemented based on the low power consumption control logic of the link layer, when the link layer of the core particle enters the low power consumption state, the physical layer of the core particle The working state remains unchanged, so that after the link layer opens its data transmission path, the link layer can transmit the data stream to the data transmission path of the physical layer through the data transmission path, avoiding the problem when the physical layer of the core chip performs state switching. The physical layer retraining process reduces the power consumption and response time of physical layer retraining.
下面基于上述所述的芯粒,在芯粒互联场景下,对本申请实施例的低功耗控制方案进行详细描述。图7示例性的示出了本申请实施例中低功耗控制方法中对应进入低功耗状态的可选流程示意图。Based on the above-mentioned core chips, the low power consumption control scheme of the embodiment of the present application is described in detail in the core chip interconnection scenario. FIG. 7 exemplarily shows an optional flow chart corresponding to entering the low power consumption state in the low power consumption control method in the embodiment of the present application.
基于芯粒互联场景下,数据传输方向的特点,芯粒可以向对端芯粒发送数据流,对端芯粒也可以向芯粒发送数据流。Based on the characteristics of the data transmission direction in the core-particle interconnection scenario, the core particle can send data streams to the opposite end core particle, and the opposite end core particle can also send data streams to the core particle.
作为一种可选实现,以图7所示芯粒1的角度出发,芯粒2作为芯粒1的对端芯粒。在芯粒与互联的对端芯粒之间传输数据流为芯粒1向对端芯粒2发送数据流的情况下,结合图7所示,芯粒1的链路层11可以执行以下步骤。As an optional implementation, starting from the angle of core particle 1 shown in FIG. 7 , core particle 2 serves as the opposite end core particle of core particle 1 . When the data stream transmitted between a core particle and the interconnected peer core particle is that core particle 1 sends a data stream to the peer core particle 2, as shown in Figure 7, the link layer 11 of core particle 1 can perform the following steps .
步骤S101,在传输空闲时间到来时,获取低功耗进入请求。Step S101: When the transmission idle time arrives, obtain a low power consumption entry request.
所述低功耗进入请求可以是位于链路层11之上的协议层发送至链路层11,以指示链路层11进入低功耗状态。The low power consumption entry request may be sent by a protocol layer located above the link layer 11 to the link layer 11 to instruct the link layer 11 to enter a low power consumption state.
步骤S102,响应所述低功耗进入请求,向对端芯粒发送进入低功耗信号。Step S102: In response to the low power consumption entry request, send a low power consumption entry signal to the opposite end chip.
其中,芯粒1的链路层11发送的进入低功耗信号用于控制对端芯粒2的链路层21关闭数据接收通路,以使得对端芯粒2的链路层21进入低功耗状态。在可选实现中,芯粒1的链路层11通过数据发送通路向对端芯粒2的链路层21发送进入低功耗信号。Among them, the low-power consumption signal sent by the link layer 11 of the core particle 1 is used to control the link layer 21 of the opposite end core particle 2 to close the data receiving path, so that the link layer 21 of the opposite end core particle 2 enters low power consumption. consumption status. In an optional implementation, the link layer 11 of the core chip 1 sends the low-power consumption signal to the link layer 21 of the opposite core chip 2 through the data transmission path.
在一个实现示例中,芯粒的链路层中设置的低功耗控制逻辑可以在接收到芯粒的协议层发送的低功耗进入请求后,向对端的芯粒发送进入低功耗信号。例如,芯粒1的链路层11中设置的低功耗控制逻辑13可以在接收到芯粒1的协议层发送的低功耗进入请求后,向对端的芯粒2发送进入低功耗信号。In an implementation example, the low-power control logic set in the link layer of the core can send an entry low-power signal to the opposite core after receiving the low-power entry request sent by the protocol layer of the core. For example, the low-power control logic 13 set in the link layer 11 of the core 1 can send an entry low-power signal to the opposite core 2 after receiving the low-power entry request sent by the protocol layer of the core 1. .
在链路层11向对端芯粒发送进入低功耗信号的同时,链路层11可以执行步骤S103,关闭数据发送通路。While the link layer 11 sends the low-power consumption signal to the opposite end core chip, the link layer 11 may perform step S103 to close the data transmission path.
在一些实施例中,芯粒的链路层关闭数据传输通路可以是基于链路层的工作时钟实现。在一个示例中,芯粒1的链路层11可以通过关闭数据传输通路的工作时钟,以关闭链路层11的数据传输通路。在一个可选实现中,芯粒的链路层中设置的低功耗控制逻辑,可以向芯粒的链路层的数据传输通路发送工作时钟关闭通知,以使得芯粒的链路层关闭数据传输通路。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以向芯粒1的链路层11的数据传输通路发送工作时钟关闭通知,以使得芯粒1的链路层11关闭数据传输通路。In some embodiments, closing the data transmission path at the link layer of the core may be implemented based on the working clock of the link layer. In one example, the link layer 11 of the core 1 can close the data transmission path of the link layer 11 by turning off the working clock of the data transmission path. In an optional implementation, the low-power control logic set in the link layer of the core particle can send a working clock shutdown notification to the data transmission path of the link layer of the core particle, so that the link layer of the core particle closes the data transmission path. For example, the low-power control logic 13 provided in the link layer 11 of the core 1 can send a working clock shutdown notification to the data transmission path of the link layer 11 of the core 1, so that the link layer 11 of the core 1 Close the data transmission path.
在芯粒发送数据流给对端芯粒的情况下,芯粒的链路层中设置的低功耗控制逻辑,可以在获取芯粒的协议层发送的低功耗进入请求后,向对端的芯粒发送进入低功耗信号,并且向芯粒的链路层的数据传输通路发送工作时钟关闭通知,以使得芯粒的链路层关闭数据传输通路。例如在图5中,在芯粒1发送数据流时,芯粒1的链路层11通过关闭数据发送通路111的工作时钟,实现关闭链路层11的数据发送通路111。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以在获取到芯粒1的协议层发送的低功耗进入请求后,向对端的芯粒2发送进入低功耗信号,并且向芯粒1的链路层11的发送通路111发送工作时钟关闭通知,以使得芯粒1的链路层11关闭数据发送通路111。When the core particle sends a data stream to the opposite end core particle, the low-power control logic set in the link layer of the core particle can send the low-power entry request to the opposite end after obtaining the low-power entry request sent by the protocol layer of the core particle. The core particle sends an entry low-power signal, and sends a working clock shutdown notification to the data transmission path of the link layer of the core particle, so that the link layer of the core particle closes the data transmission path. For example, in FIG. 5 , when the core particle 1 transmits a data stream, the link layer 11 of the core particle 1 closes the data transmission path 111 of the link layer 11 by turning off the working clock of the data transmission path 111 . For example, the low-power control logic 13 set in the link layer 11 of the core 1 can send the low-power entry request to the opposite core 2 after obtaining the low-power entry request sent by the protocol layer of the core 1. signal, and sends a working clock shutdown notification to the transmission path 111 of the link layer 11 of the core particle 1, so that the link layer 11 of the core particle 1 closes the data transmission path 111.
在一些实施例中,基于链路层通过关闭数据传输通路的工作时钟,实现关闭数据传输通路的情况下,所述链路层可以是以链路层的工作时钟作为驱动,关闭链路层的数据传输通路。例如,芯粒的链路层中设置的低功耗控制逻辑,可以是以链路层的工作时钟作为驱动,向链路层的数据传输通路发送工作时钟关闭通知。例如芯粒1的链路层11可以是以链路层11的工作时钟为驱动,关闭的数据发送通路111。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以在获取到芯粒1的协议层发送的低功耗进入请求后,以链路层11的工作时钟为驱动,向芯粒1的链路层11的发送通路111发送工作时钟关闭通知。In some embodiments, when the data transmission path is closed by closing the working clock of the data transmission path based on the link layer, the link layer may use the working clock of the link layer as a driver to close the link layer. Data transmission path. For example, the low-power control logic set in the link layer of the core chip can be driven by the working clock of the link layer and send a working clock shutdown notification to the data transmission path of the link layer. For example, the link layer 11 of the core 1 may be a closed data transmission path 111 driven by the working clock of the link layer 11 . For example, the low-power control logic 13 set in the link layer 11 of the core 1 can be driven by the working clock of the link layer 11 after obtaining the low-power entry request sent by the protocol layer of the core 1. The working clock shutdown notification is sent to the transmission path 111 of the link layer 11 of the core particle 1 .
作为另一种可选实现,以图7所示的芯粒2的角度出发,芯粒1作为芯粒2的对端芯粒。在芯粒与互联的对端芯粒之间传输数据流为芯粒1向芯粒2发送数据流的情况下,结合图7所示,基于对端的芯粒1在传输空闲时间到来时,向芯粒2发送进入低功耗信号,从而,芯粒2的链路层21可以执行以下步骤。As another optional implementation, starting from the perspective of core particle 2 shown in FIG. 7 , core particle 1 serves as the opposite end core particle of core particle 2 . When the transmission data stream between a core particle and the interconnected peer core particle is that core particle 1 sends a data stream to core particle 2, as shown in Figure 7, when the transmission idle time arrives, core particle 1 based on the peer end sends a data stream to core particle 2. Core 2 sends an incoming low power signal so that the link layer 21 of Core 2 can perform the following steps.
步骤S104,在传输空闲时间到来时,获取所述对端芯粒的链路层所发送的进入低功耗信号。Step S104: When the transmission idle time arrives, obtain the low-power consumption signal sent by the link layer of the opposite end core chip.
步骤S105,响应于所述低功耗进入信号,关闭数据接收通路。Step S105: In response to the low power consumption entry signal, the data receiving path is closed.
需要说明的是,基于芯粒互联场景下,数据流在两个芯粒之间同时流动的数据传输特点,链路层21关闭数据接收通路可以是与链路层11关闭数据发送通路同时执行的,以实现作为接收方的对端芯粒2和作为发送方的芯粒1同步进入低功耗状态。It should be noted that, based on the data transmission characteristics of simultaneous flow of data streams between two cores in a core-to-core interconnection scenario, the link layer 21 closing the data receiving path can be executed at the same time as the link layer 11 closing the data sending path. , to realize that the opposite end core particle 2 as the receiver and the core particle 1 as the sender enter the low power consumption state simultaneously.
在一个可选实现中,芯粒的链路层中设置的低功耗控制逻辑,可以向芯粒的链路层的数据传输通路发送工作时钟关闭通知,以使得芯粒的链路层关闭数据传输通路。比如,芯粒2的链路层21中设置的低功耗控制逻辑23,可以向芯粒2的链路层21的数据传输通路发送工作时钟关闭通知,以使得芯粒2的链路层21关闭数据传输通路。In an optional implementation, the low-power control logic set in the link layer of the core particle can send a working clock shutdown notification to the data transmission path of the link layer of the core particle, so that the link layer of the core particle closes the data transmission path. For example, the low-power control logic 23 provided in the link layer 21 of the core 2 can send a working clock shutdown notification to the data transmission path of the link layer 21 of the core 2, so that the link layer 21 of the core 2 Close the data transmission path.
在一个示例中,芯粒中设置的低功耗控制逻辑,可以获取对端芯粒发送的进入低功耗信号,从而芯粒中的低功耗控制逻辑可以向芯粒的链路层的数据传输通路发送工作时钟关闭通知,以使得芯粒的链路层关闭数据传输通路。比如,以芯粒2接收芯粒1发送的数据流为例,芯粒2的链路层21设置的低功耗控制逻辑23可以接收,对端的芯粒1发送的进入低功耗信号,从而向芯粒2的链路层21的数据接收通路发送工作时钟关闭通知,以使得芯粒2的链路层21关闭数据接收通路。In one example, the low-power control logic set in the core can obtain the incoming low-power signal sent by the opposite end core, so that the low-power control logic in the core can send data to the link layer of the core. The transmission path sends a working clock shutdown notification, so that the link layer of the core particle closes the data transmission path. For example, taking core 2 to receive the data stream sent by core 1 as an example, the low-power control logic 23 set up in the link layer 21 of core 2 can receive the low-power signal sent by core 1 at the opposite end, so that The working clock shutdown notification is sent to the data receiving path of the link layer 21 of the core particle 2, so that the link layer 21 of the core particle 2 closes the data receiving path.
在一些实施例中,基于链路层通过关闭数据传输通路的工作时钟,实现关闭数据传输通路的情况下,所述链路层可以是以链路层的工作时钟作为驱动,关闭链路层的数据传输通路。例如,芯粒的链路层中设置的低功耗控制逻辑,可以是以链路层的工作时钟作为驱动,向链路层的数据传输通路发送工作时钟关闭通知。例如芯粒2的链路层21可以是以链路层21的工作时钟为驱动,关闭的数据接收通路211。比如,芯粒2的链路层21中设置的低功耗控制逻辑23,可以在获取到对端芯粒1发送的进入低功耗信号后,以链路层21的工作时钟为驱动,向芯粒2的链路层21的数据接收通路211发送工作时钟关闭通知。In some embodiments, when the data transmission path is closed by closing the working clock of the data transmission path based on the link layer, the link layer may use the working clock of the link layer as a driver to close the link layer. Data transmission path. For example, the low-power control logic set in the link layer of the core chip can be driven by the working clock of the link layer and send a working clock shutdown notification to the data transmission path of the link layer. For example, the link layer 21 of the core 2 may be a closed data receiving path 211 driven by the working clock of the link layer 21 . For example, the low-power control logic 23 set in the link layer 21 of the core particle 2 can, after acquiring the low-power consumption signal sent by the opposite end core particle 1, drive the working clock of the link layer 21 to The data receiving path 211 of the link layer 21 of the core 2 sends a working clock shutdown notification.
在一些实施例中,基于链路层以链路层的工作时钟作为驱动时钟,关闭所述链路层的数据传输通路,从而链路层进入低功耗状态的速度能够与链路层的工作时钟周期相对应,为此,本申请实施例中所述传输空闲时间的时间长度可以为链路层的预设数量的工作时钟周期,以实现链路层进入低功耗状态的时长与数据流的传输空闲时长相匹配,降低芯粒的功耗。In some embodiments, based on the link layer, the working clock of the link layer is used as the driving clock to close the data transmission path of the link layer, so that the speed at which the link layer enters the low power consumption state can be consistent with the working speed of the link layer. Corresponding to the clock cycle, for this reason, the length of the transmission idle time described in the embodiment of the present application can be a preset number of working clock cycles of the link layer, so as to realize that the length of the link layer entering the low power consumption state is consistent with the data flow. The transmission idle time is matched to reduce the power consumption of the chip.
在一些实施例中,对应链路层进入低功耗状态,在传输工作时间到来时,链路层需要退出低功耗状态。图8示例性的示出了本申请实施例中低功耗控制方法中对应退出低功耗状态的可选流程示意图。In some embodiments, the corresponding link layer enters the low power consumption state, and when the transmission working time arrives, the link layer needs to exit the low power consumption state. FIG. 8 exemplarily shows an optional flow chart corresponding to exiting the low power consumption state in the low power consumption control method in the embodiment of the present application.
作为一种可选实现,以图8所示芯粒1的角度出发,芯粒2作为芯粒1的对端芯粒。在芯粒与互联的对端芯粒之间传输的数据流为:芯粒1向对端的芯粒2发送数据流的情况下,芯粒1的链路层11可以执行以下步骤。As an optional implementation, starting from the angle of core particle 1 shown in FIG. 8 , core particle 2 serves as the opposite end core particle of core particle 1 . When the data stream transmitted between a core particle and an interconnected peer core particle is: core particle 1 sends a data stream to the peer core particle 2, the link layer 11 of core particle 1 can perform the following steps.
步骤S201,在传输工作时间到来时,开启数据发送通路。Step S201: When the transmission working time arrives, the data transmission path is opened.
作为可选实现,芯粒1的链路层11可以是基于协议层的低功耗退出请求,开启数据发送通路。在一个具体示例中,协议层向芯粒1的链路层11发送低功耗退出请求,以指示传输工作时间到来,从而链路层开启数据发送通路。As an optional implementation, the link layer 11 of the core 1 can open the data transmission path based on the low-power exit request of the protocol layer. In a specific example, the protocol layer sends a low-power exit request to the link layer 11 of the core 1 to indicate the arrival of the transmission working time, so that the link layer opens the data transmission path.
作为另一可选实现,可以预先设定传输空闲时间的时间长度为链路层的预设数量的工作时钟周期,从而在芯粒的链路层进入低功耗状态之后,经过预设数量的链路层的工作时钟周期,可以确认数据流的传输工作时间到来,进而芯粒的链路层可以自动开启数据传输通路。在一个可选实现中,芯粒的链路层进入低功耗状态之后,如果计时达到预设数量的链路层的工作时钟周期,则芯粒的链路层中设置的低功耗控制逻辑,可以控制芯粒的链路层中的数据传输通路进行开启。例如,芯粒的链路层中设置的低功耗控制逻辑,可以向芯粒的链路层的数据传输通路,发送工作时钟开启通知,以控制芯粒的链路层中的数据传输通路进行开启。As another optional implementation, the length of the transmission idle time can be preset to a preset number of working clock cycles of the link layer, so that after the link layer of the core enters the low power consumption state, after a preset number of The working clock cycle of the link layer can confirm the arrival of the transmission working time of the data stream, and then the link layer of the core can automatically open the data transmission path. In an optional implementation, after the link layer of the core particle enters the low-power state, if the timing reaches a preset number of link layer working clock cycles, the low-power control logic set in the link layer of the core particle , the data transmission path in the link layer of the core particle can be controlled to be opened. For example, the low-power control logic set in the link layer of the core can send a working clock start notification to the data transmission path of the link layer of the core to control the data transmission path in the link layer of the core. Turn on.
例如,芯粒1的链路层11可以开启数据发送通路111,从而通过链路层11开启数据发送通路,芯粒1中协议层的数据流能够传输至链路层,从而传输至物理层12。For example, the link layer 11 of core 1 can open the data transmission path 111, thereby opening the data transmission path through the link layer 11, and the data flow of the protocol layer in core 1 can be transmitted to the link layer, thereby transmitting to the physical layer 12 .
在链路层11开启数据发送通路的同时,链路层11可以执行步骤S202,向对端芯粒发送退出低功耗信号。While the link layer 11 opens the data transmission path, the link layer 11 may perform step S202 and send an exit low-power consumption signal to the opposite end core.
其中,芯粒1的链路层11发送的退出低功耗信号控制对端芯粒2的链路层21重启数据流的传输。Among them, the exit low-power consumption signal sent by the link layer 11 of the core particle 1 controls the link layer 21 of the opposite end core particle 2 to restart the transmission of the data stream.
在一些实施例中,芯粒的链路层开启数据传输通路可以是基于链路层的工作时钟实现。在一个示例中,芯粒1的链路层11可以通过开启数据传输通路的工作时钟,以开启链路层11的数据传输通路。在一个可选实现中,芯粒的链路层中设置的低功耗控制逻辑,可以向芯粒的链路层的数据传输通路发送工作时钟开启通知,以使得芯粒的链路层开启数据传输通路。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以向芯粒1的链路层11的数据传输通路发送工作时钟开启通知,以使得芯粒1的链路层11开启数据传输通路。In some embodiments, opening the data transmission path at the link layer of the core may be implemented based on the working clock of the link layer. In one example, the link layer 11 of the core 1 can open the data transmission path of the link layer 11 by turning on the working clock of the data transmission path. In an optional implementation, the low-power control logic set in the link layer of the core particle can send a working clock start notification to the data transmission path of the link layer of the core particle, so that the link layer of the core particle starts data transmission path. For example, the low-power control logic 13 provided in the link layer 11 of the core 1 can send a working clock start notification to the data transmission path of the link layer 11 of the core 1, so that the link layer 11 of the core 1 Open the data transmission path.
在芯粒发送数据流给对端芯粒的情况下,芯粒的链路层中设置的低功耗控制逻辑,可以在获取芯粒的协议层发送的低功耗退出请求后,向芯粒的链路层的数据传输通路发送工作时钟开启通知,以使得芯粒的链路层开启数据传输通路。例如在图5中,在芯粒1发送数据流时,芯粒1的链路层11通过开启数据发送通路111的工作时钟,实现开启链路层11的数据发送通路111。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以在获取到芯粒1的协议层发送的低功耗退出请求后,向芯粒1的链路层11的发送通路111发送工作时钟开启通知,以使得芯粒1的链路层11开启数据发送通路111。When the core particle sends a data stream to the peer core particle, the low-power control logic set in the link layer of the core particle can send a low-power exit request to the core particle after obtaining the low-power exit request sent by the protocol layer of the core particle. The data transmission path of the link layer sends a working clock start notification, so that the link layer of the core particle opens the data transmission path. For example, in FIG. 5 , when the core particle 1 sends a data stream, the link layer 11 of the core particle 1 realizes opening the data transmission path 111 of the link layer 11 by turning on the working clock of the data transmission path 111 . For example, the low-power control logic 13 set in the link layer 11 of the core 1 can send a low-power exit request to the link layer 11 of the core 1 after obtaining the low-power exit request sent by the protocol layer of the core 1. The path 111 sends a working clock start notification, so that the link layer 11 of the core 1 opens the data sending path 111 .
在一些实施例中,基于链路层通过开启数据传输通路的工作时钟,实现开启数据传输通路的情况下,所述链路层可以是以链路层的工作时钟作为驱动,开启链路层的数据传输通路。例如,芯粒的链路层中设置的低功耗控制逻辑,可以是以链路层的工作时钟作为驱动,向链路层的数据传输通路发送工作时钟开启通知。例如芯粒1的链路层11可以是以链路层11的工作时钟为驱动,开启的数据发送通路111。比如,芯粒1的链路层11中设置的低功耗控制逻辑13,可以在获取到芯粒1的协议层发送的低功耗退出请求后,以链路层11的工作时钟为驱动,向芯粒1的链路层11的发送通路111发送工作时钟开启通知。In some embodiments, when the data transmission path is opened based on the link layer by turning on the working clock of the data transmission path, the link layer can use the working clock of the link layer as a driver to turn on the link layer. Data transmission path. For example, the low-power control logic set in the link layer of the core chip can be driven by the working clock of the link layer and send a working clock start notification to the data transmission path of the link layer. For example, the link layer 11 of the core 1 may be a data transmission path 111 that is driven by the working clock of the link layer 11 . For example, the low-power control logic 13 set in the link layer 11 of the core 1 can be driven by the working clock of the link layer 11 after obtaining the low-power exit request sent by the protocol layer of the core 1. The working clock start notification is sent to the transmission path 111 of the link layer 11 of the core particle 1 .
作为另一种可选实现,以图8所示芯粒2的角度出发,芯粒1作为芯粒2的对端芯粒。在芯粒与互联的对端芯粒之间传输数据流为对端芯粒1向芯粒2发送数据流的情况下,基于对端的芯粒1在传输工作时间到来时,链路层开启数据发送通路,芯粒2的链路层21可以执行以下步骤。As another optional implementation, starting from the angle of core particle 2 shown in FIG. 8 , core particle 1 serves as the opposite core particle of core particle 2 . When the data flow transmitted between the core particle and the interconnected peer core particle is that the peer core particle 1 sends a data stream to core particle 2, when the transmission working time of the peer core particle 1 comes, the link layer starts the data flow. For the transmit path, the link layer 21 of core 2 can perform the following steps.
步骤S203,在传输工作时间到来时,开启数据接收通路。Step S203: When the transmission working time arrives, the data receiving path is opened.
作为一种可选实现,在预先设定传输空闲时间的时间长度为链路层的预设数量的工作时钟周期情况下,当数据流传输空闲对应的预设数量的链路层的工作时钟周期到达,指示传输工作时间到来,芯粒2的链路层21自动开启数据传输通路,例如芯粒2的链路层21开启数据接收通路211。从而,基于芯粒2的链路层21开启数据接收通路,芯粒2的链路层21能够接收芯粒1的链路层11传输的信息。As an optional implementation, when the length of the transmission idle time is preset to a preset number of working clock cycles of the link layer, when the data stream transmission is idle for a preset number of working clock cycles of the link layer, arrival, indicating that the transmission working time has arrived, the link layer 21 of the core 2 automatically opens the data transmission path, for example, the link layer 21 of the core 2 opens the data receiving path 211. Therefore, the data reception path is opened based on the link layer 21 of the core 2, and the link layer 21 of the core 2 can receive the information transmitted by the link layer 11 of the core 1.
在一个可选实现中,芯粒的链路层中设置的低功耗控制逻辑,可以向芯粒的链路层的数据传输通路发送工作时钟开启通知,以使得芯粒的链路层开启数据传输通路。比如,芯粒2的链路层21中设置的低功耗控制逻辑23,可以向芯粒2的链路层21的数据传输通路发送工作时钟开启通知,以使得芯粒2的链路层21开启数据传输通路。In an optional implementation, the low-power control logic set in the link layer of the core particle can send a working clock start notification to the data transmission path of the link layer of the core particle, so that the link layer of the core particle starts data transmission path. For example, the low-power control logic 23 provided in the link layer 21 of the core 2 can send a working clock start notification to the data transmission path of the link layer 21 of the core 2, so that the link layer 21 of the core 2 Open the data transmission path.
在一些实施例中,基于链路层通过开启数据传输通路的工作时钟,实现开启数据传输通路的情况下,所述链路层可以是以链路层的工作时钟作为驱动,开启链路层的数据传输通路。例如,芯粒的链路层中设置的低功耗控制逻辑,可以是以链路层的工作时钟作为驱动,向链路层的数据传输通路发送工作时钟开启通知。例如芯粒2的链路层21可以是以链路层21的工作时钟为驱动,开启的数据接收通路211。比如,芯粒2的链路层21中设置的低功耗控制逻辑23,可以在预设数量的工作时钟周期到达后,以链路层21的工作时钟为驱动,向芯粒2的链路层21的数据接收通路211发送工作时钟开启通知。In some embodiments, when the data transmission path is opened based on the link layer by turning on the working clock of the data transmission path, the link layer can use the working clock of the link layer as a driver to turn on the link layer. Data transmission path. For example, the low-power control logic set in the link layer of the core chip can be driven by the working clock of the link layer and send a working clock start notification to the data transmission path of the link layer. For example, the link layer 21 of the core 2 may be a data receiving path 211 that is driven by the working clock of the link layer 21 . For example, the low-power control logic 23 set in the link layer 21 of the core 2 can be driven by the working clock of the link layer 21 to the link of the core 2 after the preset number of working clock cycles arrives. The data receiving path 211 of layer 21 sends the working clock start notification.
步骤S204,获取对端芯粒的链路层所发送的退出低功耗信号,并响应于所述退出低功耗信号,重启数据流的传输。Step S204: Obtain the exit low-power consumption signal sent by the link layer of the opposite end core chip, and in response to the exit low-power consumption signal, restart the transmission of the data stream.
基于芯粒1的链路层11的数据发送通路开启,和芯粒2的链路层21的数据接收通路的开启,作为发送方的芯粒1与作为接收方的芯粒2之间可以恢复正常数据流传输。Based on the opening of the data sending path of the link layer 11 of the core particle 1 and the opening of the data receiving path of the link layer 21 of the core particle 2, the relationship between the core particle 1 as the sender and the core particle 2 as the receiver can be restored. Normal data streaming.
可以看出,本申请实施例的芯粒,在芯粒与对端芯粒的数据流的传输空闲时间,通过控制链路层进入低功耗状态,而物理层维持工作状态不变,实现芯粒的低功耗控制,降低芯粒的功耗消耗;并且避免在物理层进行低功耗控制所带来的状态切换的重新训练过程,省去物理层的重新训练带来的功耗和时间占用,提升了数据流传输所能使用的时间。因此,本申请实施例提供的方案能够在芯粒互联场景下,降低芯粒互联的功耗,并且通过避免物理层的重新训练,提升数据流传输所能使用的时间,提升数据传输效率。It can be seen that the core particle in the embodiment of the present application enters a low power consumption state by controlling the link layer during the idle time of the data stream transmission between the core particle and the opposite end core particle, while the physical layer maintains the working state unchanged, realizing that the core particle Low-power control of the chip reduces the power consumption of the chip; and avoids the retraining process of state switching caused by low-power control at the physical layer, saving the power consumption and time caused by retraining of the physical layer. Occupation increases the time available for data stream transmission. Therefore, the solution provided by the embodiments of this application can reduce the power consumption of core-to-die interconnection in the core-to-die interconnection scenario, and by avoiding the retraining of the physical layer, the time available for data stream transmission can be increased, and the data transmission efficiency can be improved.
本申请实施例还提供一种芯片,包括互联的多个芯粒,所述芯粒可以如本申请实施例提供的芯粒。Embodiments of the present application also provide a chip, including a plurality of interconnected core particles. The core particles may be the core particles provided in the embodiments of the present application.
本申请实施例还提供一种计算机设备,例如服务器设备或者终端设备,该计算机设备可以包括如本申请实施例提供的上述芯片。An embodiment of the present application also provides a computer device, such as a server device or a terminal device. The computer device may include the above-mentioned chip provided in the embodiment of the present application.
上文描述了本申请实施例提供的多个实施例方案,各实施例方案介绍的各可选方式可在不冲突的情况下相互结合、交叉引用,从而延伸出多种可能的实施例方案,这些均可认为是本申请实施例披露、公开的实施例方案。The above describes multiple embodiment solutions provided by the embodiments of the present application. The optional methods introduced in each embodiment solution can be combined and cross-referenced with each other without conflict, thereby extending a variety of possible embodiment solutions. These can be considered as embodiments disclosed and disclosed in the embodiments of this application.
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 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 (19)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311389522.0A CN117544433A (en) | 2023-10-24 | 2023-10-24 | Core particle, low power consumption control method, chip and computer equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311389522.0A CN117544433A (en) | 2023-10-24 | 2023-10-24 | Core particle, low power consumption control method, chip and computer equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117544433A true CN117544433A (en) | 2024-02-09 |
Family
ID=89792806
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311389522.0A Pending CN117544433A (en) | 2023-10-24 | 2023-10-24 | Core particle, low power consumption control method, chip and computer equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117544433A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117834755A (en) * | 2024-03-04 | 2024-04-05 | 中国人民解放军国防科技大学 | Interface circuit and chip between protocol layer and adapter layer for chiplet interconnection interface |
-
2023
- 2023-10-24 CN CN202311389522.0A patent/CN117544433A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117834755A (en) * | 2024-03-04 | 2024-04-05 | 中国人民解放军国防科技大学 | Interface circuit and chip between protocol layer and adapter layer for chiplet interconnection interface |
CN117834755B (en) * | 2024-03-04 | 2024-05-10 | 中国人民解放军国防科技大学 | Interface circuit and chip between protocol layer and adapter layer for chiplet interconnection interface |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106970886B (en) | Controlling a physical link of a first protocol using an extended functionality architecture of a second protocol | |
US9830292B2 (en) | Architected protocol for changing link operating mode | |
JP6243469B2 (en) | Provision of low power physical unit to load / store communication protocol | |
US9223735B2 (en) | Providing a consolidated sideband communication channel between devices | |
US8638783B2 (en) | Optimized link training and management mechanism | |
US7793030B2 (en) | Association of multiple PCI express links with a single PCI express port | |
KR101812835B1 (en) | Serial interface | |
US10579549B2 (en) | Staggered transmissions on a multi-drop half-duplex bus | |
EP4323880B1 (en) | I2c bus architecture using shared clock and dedicated data lines | |
CN117544433A (en) | Core particle, low power consumption control method, chip and computer equipment | |
CN109388597A (en) | A kind of data interactive method and device based on FPGA | |
US20200201804A1 (en) | I3c device timing adjustment to accelerate in-band interrupts | |
CN101449254B (en) | Flow control for universal serial bus (USB) | |
CN100547960C (en) | Asymmetric high-speed half-duplex communication system and communication method | |
US8913527B2 (en) | Multiple die communication system | |
US20200233829A1 (en) | Multi-lane system power management interface | |
CN102508809A (en) | Device for converting from field bus to PC104 bus and data transmission method | |
CN202563496U (en) | Conversion device from field bus to personal computer (PC) 104 bus | |
US20180181533A1 (en) | Ending write data transfer in i3c hdr-ddr mode |
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 |