Resource organization methods, systems, devices, electronic devices, and storage media
The resource configuration method in chiplet integration systems optimizes memory access by prioritizing resource allocation based on access priority, addressing inefficiencies and latency issues in conventional systems.
Patent Information
- Application Number
- KR1020267036126
- Authority / Receiving Office
- KR · KR
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-02-28
- Filing Date
- 2025-09-25
- Publication Date
- 2026-11-16
AI Technical Summary
Conventional chiplet integration systems face challenges in managing memory resource access efficiently, leading to disordered access and increased latency, which affects memory access efficiency and system performance.
A resource configuration method that allocates memory resources based on access priority among function chiplets, including reservation and scheduling of shared resources to ensure orderly management and efficient use.
Improves memory access efficiency by preventing disorder and reducing latency, ensuring stable system operation and maximizing resource utilization.
Smart Images

Figure PCT00006_ABST
Abstract
Description
Technology Field
[0001] Cross-reference of related applications
[0002] This application claims priority to a Chinese patent application filed with the National Intellectual Property Administration of China on February 28, 2025, with application number 202510236629.4 and title of invention "method, system, device, electronic device and storage medium for resource organization," the entire contents of which are incorporated into this application by reference.
[0003] Technology field
[0004] The embodiments of the present disclosure relate to the field of chip technology but are not limited thereto, and relate, for example, to resource configuration methods, systems, devices, electronic devices, and storage media. Background Technology
[0005] Chiplet integration technology adopts innovative design concepts to interconnect via high-speed die-to-die (D2D) interfaces and connects multiple chiplets together using advanced packaging processes, thereby forming large-scale homogeneous or heterogeneous multicore processor systems. The problem to be solved
[0006] The following is an overview of the subject matter described in detail in the main text. This overview is not intended to limit the scope of protection of the claims.
[0007] Embodiments of the present disclosure provide a resource configuration method, system, device, electronic device, and storage medium, and can improve memory access efficiency by rationally configuring memory resources for a functional chiplet. means of solving the problem
[0008] The technical measures adopted in the embodiments of the present disclosure are as follows.
[0009] According to a first aspect provided in an embodiment of the present disclosure, a method for configuring resources is provided, wherein the method is applied to an interconnect chiplet in a chiplet interconnection system, the chiplet interconnection system further comprises a plurality of function chiplets, each function chiplet being provided with a memory resource, and the method comprises the steps of: receiving a resource sharing request transmitted by a first function chiplet, wherein the resource sharing request is used to request the sharing of a memory resource of another function chiplet, and the other function chiplet is a function chiplet among a plurality of function chiplets excluding the first function chiplet; and configuring a memory resource of one or a plurality of target function chiplets for the first function chiplet based on the access priority of the plurality of function chiplets.
[0010] According to the technical means described above, in an embodiment of the present disclosure, after receiving a resource sharing request transmitted by a first function chiplet, an interconnect chiplet may configure the memory resources of another function chiplet for the first function chiplet based on the access priority of the other function chiplet, thereby preventing disorder or randomness when the system configures resources for the function chiplet, reducing time loss due to unreasonable resource configuration, further improving the access efficiency to the system's memory resources, and ensuring the efficient operation of the system.
[0011] In some embodiments, when the used memory resource of the first function chiplet is greater than or equal to the first resource pre-warning threshold and less than the second resource pre-warning threshold, the resource sharing request is used to request that the shared memory resource be reserved in advance, and when the used memory resource of the first function chiplet is greater than or equal to the second resource pre-warning threshold, the resource sharing request is used to request the scheduling of the shared memory resource, wherein the second resource pre-warning threshold is greater than the first resource pre-warning threshold.
[0012] According to the technical means described above, the embodiment of the present disclosure can effectively prevent system performance degradation that occurs when the demand for memory resources surges by reserving shared memory resources in advance for the first function chiplet when the memory resources used by the first function chiplet reach a first resource advance warning threshold. When the memory resources used by the first function chiplet further increase and reach a second resource advance warning threshold, the memory resources can be scheduled for the first function chiplet by promptly responding to the resource sharing request of the first function chiplet. Furthermore, by controlling the reserving and scheduling of memory resources in advance, the response speed of the system can be improved and stable operation of the system can be ensured.
[0013] In some embodiments, when a resource sharing request is used to instruct an interconnect chiplet to reserve memory resources in advance for a first function chiplet, one or more target function chiplets satisfy at least the condition that one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of one or more target function chiplets is greater than 0.
[0014] According to the technical means described above, an embodiment of the present disclosure can rapidly and accurately determine one or more target function chiplets based on the conditions described above.
[0015] In some embodiments, when a resource sharing request is used to instruct an interconnect chiplet to schedule memory resources for a first function chiplet, one or more target function chiplets satisfy at least the condition that all of the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than or equal to the memory resources required for the first function chiplet.
[0016] In some embodiments, the method further includes the step of scheduling all remaining available memory resources of other function chiplets for the first function chiplet when the sum of the remaining available memory resources of other function chiplets is less than the memory resources required for the first function chiplet.
[0017] According to the technical means described above, the embodiment of the present disclosure can maximize the use of existing memory resources by intensively scheduling all remaining available memory resources of other function chiplets, excluding the first function chiplet, to the first function chiplet, as well as improve the resource utilization efficiency of the system and ensure the normal operation of the first function chiplet.
[0018] In some embodiments, the method further comprises the step of transmitting a resource lock message to one or more target function chiplets, wherein the resource lock message is used to indicate that other function chiplets are not allowed to use the remaining available memory resources of the target function chiplet.
[0019] According to the technical means described above, an embodiment of the present disclosure can prevent contention and occupation of such memory resources by other function chiplets by ensuring that a locked memory resource is provided to and used only by a first function chiplet within a specific period by transmitting a resource lock message to one or more target function chiplets.
[0020] In some embodiments, the method further comprises the step of transmitting a resource release message to one or more target function chiplets when it is determined that there is no need to schedule memory resources of one or more target function chiplets for the first function chiplet, wherein the resource release message is used to indicate that the first function chiplet no longer uses memory resources of one or more target function chiplets.
[0021] According to the technical means described above, the embodiment of the present disclosure can release memory resources in a timely manner when it is determined that there is no need to schedule memory resources of one or more target function chiplets for a first function chiplet, thereby preventing idleness and waste of resources, optimizing memory resource usage efficiency, and reducing delays and performance degradation caused by other function chiplets waiting for memory resources.
[0022] According to a second aspect provided in an embodiment of the present disclosure, a resource configuration method is provided, wherein the method is applied to a first function chiplet among a plurality of function chiplets included in a chiplet interconnection system, the chiplet interconnection system further includes an interconnection chiplet, and each of the plurality of function chiplets is provided with a memory resource, and the method comprises the steps of: transmitting a resource sharing request to the interconnection chiplet—the resource sharing request is used to request the sharing of a memory resource of another function chiplet, and the other function chiplet is a function chiplet other than the first function chiplet among the plurality of function chiplets—; receiving a resource configuration message transmitted by the interconnection chiplet—the resource configuration message is used to indicate a memory resource of one or a plurality of target function chiplets among the other function chiplets, and the one or a plurality of target function chiplets is determined according to the access priority of the plurality of function chiplets—; and determining whether to perform data storage using the memory resource of one or a plurality of target function chiplets based on the usage status of the memory resource of the first function chiplet.
[0023] In some embodiments, when the memory resources used by the first function chiplet are greater than or equal to the first resource advance warning threshold and less than the second resource advance warning threshold, the resource sharing request is used to request that the shared memory resources be reserved in advance, and when the memory resources used by the first function chiplet are greater than or equal to the second resource advance warning threshold, the resource sharing request is used to request the scheduling of the shared memory resources.
[0024] Here, the second resource pre-warning threshold is greater than the first resource pre-warning threshold.
[0025] In some embodiments, when a resource sharing request is used to instruct an interconnect chiplet to reserve memory resources in advance for a first function chiplet, one or more target function chiplets satisfy at least the condition that one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of one or more target function chiplets is greater than 0.
[0026] In some embodiments, when a resource sharing request is used to instruct an interconnect chiplet to schedule memory resources for a first function chiplet, one or more target function chiplets satisfy at least the condition that all of the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than or equal to the memory resources required for the first function chiplet.
[0027] In some embodiments, the method further comprises the step of transmitting a resource release notice to an interconnect chiplet, and the resource release notice is used to indicate that there is no need to use the memory resources of one or more target function chiplets.
[0028] In some embodiments, the step of determining whether to perform data storage using the memory resources of one or more target function chiplets based on the usage status of the memory resources of the first function chiplet comprises: a step of determining to perform data storage using the memory resources of one or more target function chiplets when it is determined that the remaining available memory resources of the first function chiplet cannot satisfy the storage requirement; and a step of determining not to perform data storage using the memory resources of one or more target function chiplets when it is determined that the remaining available memory resources of the first function chiplet can satisfy the storage requirement.
[0029] According to a third aspect provided in an embodiment of the present disclosure, a chiplet interconnection system is provided, wherein the chiplet interconnection system comprises an interconnection chiplet and a plurality of function chiplets, each function chiplet being provided with a memory resource, wherein the interconnection chiplet is configured to perform the method of the first aspect and any one possible embodiment thereof described above, and the function chiplet is configured to perform the method of the second aspect and any one possible embodiment thereof described above.
[0030] In some embodiments, each function chiplet includes a D2D interface, and the interconnect chiplet includes a plurality of D2D interfaces, and each D2D interface of the interconnect chiplet is configured to be connected to the D2D interface of the corresponding function chiplet.
[0031] According to a fourth aspect provided in the present disclosure, a resource configuration device is provided, wherein the device is applied to an interconnect chiplet in a chiplet interconnection system, the chiplet interconnection system further comprises a plurality of function chiplets, each function chiplet being provided with a memory resource, the device comprising a communication unit and a configuration unit, the communication unit being configured to receive a resource sharing request transmitted by a first function chiplet, and the configuration unit being configured to configure the memory resources of one or more target function chiplets among other function chiplets for the first function chiplet based on the access priority of the plurality of function chiplets.
[0032] In some embodiments, the configuration is also configured to schedule all remaining available memory resources of other function chiplets for the first function chiplet when the sum of the remaining available memory resources of other function chiplets is less than the memory resources required for the first function chiplet.
[0033] In some embodiments, the communication unit is also configured to transmit a resource lock message to one or more target function chiplets.
[0034] In some embodiments, the communication unit is also configured to transmit a resource release message to one or more target function chiplets when it is determined that there is no need to schedule memory resources of one or more target function chiplets for the first function chiplet.
[0035] According to a fifth aspect provided in an embodiment of the present disclosure, a resource configuration device is provided, wherein the device is applied to a first function chiplet in a chiplet interconnection system, the chiplet interconnection system further comprises a plurality of function chiplets, each function chiplet being provided with a memory resource, the device comprising a communication unit, a processing unit, and a storage unit, wherein the communication unit is configured to transmit a resource sharing request to an interconnection chiplet, and the communication unit is also configured to receive a resource configuration message transmitted by the interconnection chiplet, the processing unit is configured to determine whether to perform data storage using the memory resources of one or a plurality of target function chiplets based on the usage status of the memory resources of the first function chiplet, and the storage unit is configured to store resources.
[0036] In some embodiments, the communication unit is also configured to transmit a resource release notification to the interconnect chiplet when it is determined that there is no need to use the memory resources of one or more target function chiplets.
[0037] In some embodiments, the processing unit is also configured to determine that if it determines that the remaining available memory resources of the first function chiplet cannot satisfy the storage requirement, it determines to perform data storage using the memory resources of one or more target function chiplets, and if it determines that the remaining available memory resources of the first function chiplet can satisfy the storage requirement, it determines not to perform data storage using the memory resources of one or more target function chiplets.
[0038] According to a sixth aspect provided in an embodiment of the present disclosure, an electronic device is provided, wherein the electronic device comprises a chiplet interconnection system of the third aspect described above.
[0039] According to a seventh aspect provided in an embodiment of the present disclosure, an electronic device is provided, wherein the electronic device comprises: a processor; and a memory configured to store executable instructions of the processor, wherein the processor is configured to execute instructions to implement a method of the first aspect and any one possible embodiment thereof described above, or to implement a method of the second aspect and any one possible embodiment thereof described above.
[0040] According to the eighth aspect provided in an embodiment of the present disclosure, a computer-readable storage medium is provided, and when an instruction in the computer-readable storage medium is executed by a processor of an electronic device, the electronic device is enabled to perform the method of the first aspect and any one possible embodiment thereof described above, or to perform the method of the second aspect and any one possible embodiment thereof described above.
[0041] According to the ninth aspect provided in an embodiment of the present disclosure, a computer program product is provided, wherein the computer program product includes computer instructions, and when the computer instructions are executed in an electronic device, the electronic device performs the method of the first aspect and any one possible embodiment thereof described above, or performs the method of the second aspect and any one possible embodiment thereof described above. Effects of the invention
[0042] Accordingly, the above-described technical features of the embodiments of the present disclosure have the following beneficial effects.
[0043] (1) After receiving a resource sharing request transmitted by the first function chiplet, the interconnect chiplet can configure the memory resources of the other function chiplet for the first function chiplet based on the access priority of the other function chiplet. This prevents disorder or randomness when the system configures resources for the function chiplet, reduces time loss due to unreasonable resource configuration, and further improves the access efficiency of the system's memory resources and ensures the efficient operation of the system.
[0044] (2) When the memory resources used by the first function chiplet reach the first resource advance warning threshold, the shared memory resources are reserved in advance for the first function chiplet, thereby effectively preventing system performance degradation that occurs when the demand for memory resources increases rapidly. When the memory resources used by the first function chiplet increase further and reach the second resource advance warning threshold, the memory resources can be scheduled for the first function chiplet by responding quickly to the resource sharing request of the first function chiplet. In addition, by controlling the reservation and scheduling of memory resources in advance, the response speed of the system can be improved and stable operation of the system can be ensured.
[0045] (3) One or more target function chiplets can be determined quickly and accurately.
[0046] (4) By concentrating all remaining available memory resources of other function chiplets, excluding the first function chiplet, on the first function chiplet, not only can existing memory resources be utilized to the maximum extent, but the resource utilization efficiency of the system can also be improved and the normal operation of the first function chiplet can be guaranteed.
[0047] (5) By sending a resource lock message to one or more target function chiplets, the locked memory resource is guaranteed to be provided to and used only by the first function chiplet within a specific period, thereby preventing competition and occupation of this memory resource by other function chiplets.
[0048] (6) When it is determined that there is no need to schedule memory resources of one or more target function chiplets for the first function chiplet, these memory resources can be released in a timely manner, thereby preventing idle and wasted resources and optimizing memory resource usage efficiency, and also reducing delay and performance degradation caused by other function chiplets waiting for memory resources.
[0049] After reading and understanding the drawings and detailed descriptions, you can understand other aspects. Brief explanation of the drawing
[0050] FIG. 1 is a schematic diagram of the structure of a chiplet interconnection system illustrated in an embodiment of the present disclosure. FIG. 2 is a schematic diagram of the structure of another chiplet interconnection system illustrated in an embodiment of the present disclosure. FIG. 3 is a schematic diagram of the structure of an interconnected chiplet illustrated in an embodiment of the present disclosure. FIG. 4 is a schematic diagram of a memory address management element within an interconnect chiplet as illustrated in an embodiment of the present disclosure. FIG. 5 is a schematic diagram of the internal connection structure of an interconnect chiplet illustrated in an embodiment of the present disclosure. FIG. 6 is a schematic flowchart of a resource configuration method illustrated in an embodiment of the present disclosure. FIG. 7 is a schematic flowchart of another resource configuration method illustrated in an embodiment of the present disclosure. FIG. 8 is a schematic diagram of the interaction flow of a resource configuration method illustrated in an embodiment of the present disclosure. FIG. 9 is a block diagram of a resource configuration device illustrated in an embodiment of the present disclosure. FIG. 10 is a block diagram of another resource configuration device illustrated in an embodiment of the present disclosure. FIG. 11 is a block diagram of an electronic device illustrated in an embodiment of the present disclosure. Specific details for implementing the invention
[0051] In order to enable a person skilled in the art to better understand the technical methods of the embodiments of the present disclosure, the technical methods of the embodiments of the present disclosure are described below in a clear and complete manner by combining the drawings.
[0052] It should be noted that terms such as “first,” “second,” etc., in the specification and claims of this disclosure and in the drawings described above are intended to distinguish similar objects and are not necessarily intended to describe a specific order or sequence. Since data used in this manner may be interchangeable where appropriate, it should be understood that the embodiments of this disclosure described herein may be practiced in an order other than the order shown or described herein. The embodiments described in the exemplary embodiments below do not represent all embodiments consistent with this disclosure. On the contrary, they are merely examples of apparatuses and methods consistent with certain aspects of this disclosure as described in detail in the appended claims.
[0053] In the embodiments of the present disclosure, words such as “exemplary,” “for instance,” or “for example” are used to indicate that they are presented as examples, illustrative, or descriptive. Any embodiment or design described as “exemplary,” “for instance,” or “for example” in the embodiments of the present disclosure should not be interpreted as being more desirable or advantageous than other embodiments or designs. Specifically, the use of words such as “exemplary,” “for instance,” or “for example” is intended to present the relevant concepts in a specific manner.
[0054] First, to aid the understanding of those skilled in the art, the technology related to the embodiments of the present disclosure is explained.
[0055] As new application fields such as artificial intelligence, big data, and intelligent driving rapidly advance, the demand for chip computing capabilities in these areas is increasing day by day. These applications require not only the processing of vast amounts of data but also the implementation of complex algorithmic computations and high-speed data exchange, which poses significant challenges to conventional System-on-Chip (SOC) designs. Conventional SOC designs are limited by the physical constraints of single-chip area, power consumption, and manufacturing technology, making them unable to adapt to the development trends of these new application fields.
[0056] To address these challenges, the integrated circuit industry is gradually shifting toward high integration, high performance, low cost, and low power consumption. Against this backdrop, Chiplet integration technology has emerged. By adopting innovative design concepts, Chiplet integration technology interconnects via high-speed D2D interfaces, and connecting multiple chiplets using advanced packaging processes, thereby forming large-scale homogeneous or heterogeneous multicore processor systems. This modular design approach not only enhances chip flexibility but also enables designers to select the optimal combination of chiplets based on specific requirements, thereby achieving an optimal balance between performance and cost.
[0057] However, during the operation of the multicore processor system of the embodiment of the present disclosure, due to the absence of an effective memory resource management mechanism, a problem may arise in which access to local memory resources and remote memory resources becomes disordered when a chiplet accesses memory resources, and furthermore, the delay in accessing memory resources may increase, which may affect the memory access efficiency of the system.
[0058] On the one hand, even if a single chiplet has sufficient local memory resources, a situation may arise where that chiplet accesses the memory resources of other remote chiplets, which further increases memory resource access latency and degrades access efficiency. On the other hand, when a single chiplet needs to access the memory resources of other remote chiplets, a problem of disordered access to remote memory resources occurs, which can affect memory access efficiency. Therefore, determining how to rationally allocate memory resources for a chiplet when it accesses them has become an urgent technical issue that needs to be resolved.
[0059] To solve the aforementioned technical problem, embodiments of the present disclosure provide a resource configuration method. An interconnect chiplet, after receiving a resource sharing request transmitted by a first function chiplet, may configure the memory resources of one or more target function chiplets having the highest access priority among other function chiplets for the first function chiplet based on the access priority of a plurality of function chiplets. This method of configuring memory resources based on access priority enables the system to manage memory resources more effectively and in an orderly manner, reduces time loss due to unreasonable resource configuration, and further improves the access efficiency to the system's memory resources, thereby ensuring the efficient operation of the system.
[0060] The following describes the technical methods of the embodiments of the present disclosure in conjunction with the drawings of the embodiments of the present disclosure. Clearly, the described embodiments are only some of the embodiments of the present disclosure, not all of them.
[0061] FIG. 1 is a schematic diagram of the structure of a chiplet interconnection system (100) illustrated in an embodiment of the present disclosure. As shown in FIG. 1, the chiplet interconnection system (100) includes an interconnection chiplet (101) and a plurality of function chiplets (102). The interconnection chiplet (101) is communicationally connected to each function chiplet (102). Here, each function chiplet (102) is provided with a memory resource. The memory resource may be memory, for example, double data rate (DDR) memory.
[0062] In one example, as illustrated in FIG. 1, the interconnect chiplet (101) may include a plurality of D2D interfaces, and each function chiplet may include a single D2D interface. The interconnect chiplet (101) may be connected to each of the plurality of function chiplets (102) through the plurality of D2D interfaces.
[0063] Here, the interconnect chiplet (101) may be configured to implement the configuration of memory resources between function chiplets (102). The function chiplet (102) processes tasks using local memory resources, and if local memory resources are insufficient, it may access the memory resources of other function chiplets through the interconnect chiplet (101).
[0064] Optionally, embodiments of the present disclosure are not limited to the type of D2D interface protocol. For example, the D2D interface protocol may include, but is not limited to, the universal chiplet interconnect express (UCIE) protocol, the advanced cost-driven chiplet interface (ACC) protocol, or other user-defined protocols.
[0065] Optionally, embodiments of the present disclosure do not specifically limit the number of functional chiplets (102) in the chiplet interconnection system (100) and may include more or fewer functional chiplets (102) than in FIG. 1.
[0066] In another example, as illustrated in FIG. 2, each functional chiplet (102) may include a processor cluster. The processor cluster may include one or more of a central processing unit (CPU), a graphics processing unit (GPU), a neural processing unit (NPU), or an intelligence processing unit (IPU), but is not limited thereto, and embodiments of the present disclosure are not limited thereto.
[0067] Optionally, as shown in FIG. 2, each function chiplet (102) can be connected to a single memory (e.g., DDR memory).
[0068] FIG. 3 is a schematic diagram of the structure of an interconnect chiplet (101) illustrated in an embodiment of the present disclosure, and as shown in FIG. 3, the interconnect chiplet (101) may include a control element, a setting element, a memory address management element, a cache element, and a decoding element.
[0069] The control element is configured to configure memory resources for the function chiplet (102) in response to a resource sharing request transmitted by the function chiplet (102). The control element is also configured to determine the access priority of each function chiplet (102). The control element is also configured to transmit a resource release message to the accessed function chiplet (e.g., target function chiplet) in response to a resource release notice transmitted by the accessing function chiplet (e.g., first function chiplet). Here, the resource release notice and the resource release message are used to indicate that the accessing function chiplet does not need to use the memory resources of the accessed function chiplet.
[0070] The setting element is configured to set the resource scheduling priority of each function chiplet (102). The setting element is also configured to set the accessibility between each function chiplet (102).
[0071] The memory address management element is configured to manage the mapping relationship between the physical address of the memory resource of the function chiplet (102) and the virtual address corresponding to the physical address of the memory resource.
[0072] The cache element is configured to cache access data between function chiplets (102).
[0073] The decoding element is configured to determine the physical address of the memory resource of the accessed function chiplet (e.g., target function chiplet) by decoding the virtual address associated with the accessing function chiplet (e.g., first function chiplet) accessing the memory resource when accessing a memory resource between function chiplets.
[0074] In one example, as illustrated in FIG. 4, the memory address management element may include a shared address mapping unit and a shared address priority unit.
[0075] Here, the shared address mapping unit may include a virtual shared address table of an access-side function chiplet (e.g., a first function chiplet) and a physical shared address table of an access-side function chiplet (e.g., a target function chiplet).
[0076] The physical shared address table is used to store the interface number of the accessed function chiplet and the physical address range number of the memory resource of the accessed function chiplet.
[0077] The virtual shared address table stores the interface number of the access-side function chiplet and the virtual address range number of the memory resource of the accessed-side function chiplet, and is used to implement the mapping relationship between the virtual shared address of the access-side function chiplet and the physical shared address of the accessed-side function chiplet. Here, the address range of the memory resource is the smallest address unit in which the memory resource within the function chiplet can be shared.
[0078] The shared address priority section is configured to manage the access priority of multiple accessed function chiplets.
[0079] FIG. 5 is a schematic diagram of the internal connection structure of an interconnect chiplet (101) illustrated in an embodiment of the present disclosure. As illustrated in FIG. 5, a bidirectional connection, a unidirectional connection, or no connection may be implemented between any two D2D interfaces. For example, a bidirectional connection between an interface (A1) (configured to be connected to a function chiplet (A1)) and an interface (A2) (configured to be connected to a function chiplet (A2)) indicates that the function chiplet (A1) can access the function chiplet (A2) and that the function chiplet (A2) can also access the function chiplet (A1). A unidirectional connection between an interface (A1) and an interface (A2) indicates that the function chiplet (A1) can access the function chiplet (A2) but the function chiplet (A2) cannot access the function chiplet (A1), or that the function chiplet (A1) cannot access the function chiplet (A2) but the function chiplet (A2) can access the function chiplet (A1). The lack of connection between interface (A1) and interface (A2) indicates that the function chiplet (A1) cannot access the function chiplet (A2), and the function chiplet (A2) cannot access the function chiplet (A1).
[0080] The chiplet interconnection system of the embodiments of the present disclosure may be provided in an electronic device. The electronic device may be a terminal device or a server, and the embodiments of the present disclosure are not limited thereto.
[0081] The terminal device of the embodiment of the present disclosure may be a mobile device such as a computer equipped with a mobile terminal device (or referred to as a “cellular” phone) and a portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile device, and these exchange language and / or data with a radio access network (RAN) node. For example, the terminal device may be a vehicle, a mobile phone, a tablet computer, a laptop computer, a palmtop computer, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical surgery, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, a wireless terminal device in a smart home, etc.
[0082] The server of the embodiment of the present disclosure may be a single server, a server cluster, or a cloud server, and the embodiment of the present disclosure is not limited thereto.
[0083] To aid understanding, the resource configuration method provided in the embodiment of the present disclosure is specifically introduced below by combining the drawings.
[0084] Combining FIG. 1, FIG. 6 is a flowchart of a resource configuration method illustrated in an embodiment of the present disclosure, and as shown in FIG. 6, the method includes S601 to S603.
[0085] S601: The first function chiplet transmits a resource sharing request to the interconnect chiplet. In response, the interconnect chiplet receives the resource sharing request from the first function chiplet.
[0086] Here, the first function chiplet may be any of the multiple function chiplets of the chiplet interconnection system of FIG. 1 or FIG. 2 described above. For example, it may be function chiplet (A1). A resource sharing request is used to request a shared memory resource of another function chiplet, and the other function chiplet is a function chiplet other than the first function chiplet among the multiple function chiplets. For example, a resource sharing request may be used to request that a shared memory resource be reserved in advance, or to request the scheduling of a shared memory resource (i.e., a memory resource of another function chiplet).
[0087] In one possible implementation, if the memory resources used by the first function chiplet are greater than or equal to the first resource advance warning threshold and less than the second resource advance warning threshold, the first function chiplet may transmit a resource sharing request used to request the interconnect chiplet to reserve a shared memory resource in advance. In response, the interconnect chiplet may receive the resource sharing request from the first function chiplet, which is used to request the reserve a shared memory resource in advance.
[0088] Optionally, the first resource pre-warning threshold and the second resource pre-warning threshold may be set according to actual requirements. The second resource pre-warning threshold is greater than the first resource pre-warning threshold. For example, the first resource pre-warning threshold may be 80% of the total memory resources of the first function chiplet, and the second resource pre-warning threshold may be 90% of the total memory resources of the first function chiplet. The embodiments of the present disclosure are not specifically limited thereto.
[0089] In another possible implementation, if the memory resources used by the first function chiplet are greater than or equal to the second resource pre-warning threshold, the first function chiplet may transmit a resource sharing request used to request scheduling of shared memory resources to the interconnect chiplet. In response, the interconnect chiplet may receive the resource sharing request used to request scheduling of shared memory resources from the first function chiplet.
[0090] S602: Based on the access priority of multiple function chiplets, the interconnect chiplet configures the memory resources of one or more target function chiplets among the other function chiplets for the first function chiplet.
[0091] Here, the access priority of a function chiplet can be used to indicate the order in which the memory resources of the function chiplet are used. The higher the access priority of a function chiplet, the earlier the order in which the memory resources of that function chiplet are used. For example, if the access priority of function chiplet 1 is higher than the access priority of function chiplet 2, and both the memory resources of function chiplet 1 and function chiplet 2 can be used by function chiplet 3, function chiplet 3 can use the memory resources of function chiplet 1 preferentially.
[0092] Optionally, embodiments of the present disclosure are not limited to the manner of expressing access priority. For example, access priority may be from Grade 1 to Grade N, where N is a positive integer. Access priority may also be from Grade A to Grade Z.
[0093] Specifically, the access priority of the function chiplet can be determined by the following method.
[0094] 1. Transmission delay between the first function chiplet and the target function chiplet.
[0095] Here, the transmission delay between the first function chiplet and the target function chiplet may refer to data transmission time or data transmission delay. The access priority of the target function chiplet may be directly proportional to the transmission delay. For example, the smaller the transmission delay between the first function chiplet and the target function chiplet, the higher the access priority of the target function chiplet.
[0096] 2. Configuration information of the target function chiplet.
[0097] Here, the configuration information of the target function chiplet may include processing performance and may also include memory bandwidth.
[0098] In some examples, interconnect chiplets can determine the access priority of a target function chiplet based on the level of the target function chiplet's processing performance (e.g., data transfer rate) or the size of its memory bandwidth. For example, the higher the processing performance of the target function chiplet, the higher its access priority. Alternatively, the larger the memory bandwidth of the target function chiplet, the higher its access priority.
[0099] In embodiments of the present disclosure, the priority of the function chiplets may be determined in other ways, for example, based on the size of the number of the function chiplets, but is not limited thereto.
[0100] In one possible implementation, after receiving a resource sharing request transmitted by a first function chiplet, the interconnect chiplet may determine one or more target function chiplets with the highest access priority among the other function chiplets based on the access priority of the other function chiplets excluding the first function chiplet among the multiple function chiplets and the remaining available memory resources of the other function chiplets.
[0101] S603: The first function chiplet determines whether to perform data storage using the memory resources of one or more target function chiplets based on the usage status of memory resources.
[0102] In one possible implementation, if the first function chiplet determines that its remaining available memory resources cannot satisfy the storage requirement, it decides to perform data storage using the memory resources of one or more target function chiplets.
[0103] Here, the inability to satisfy the storage requirement may mean that the current remaining available memory resources of the first function chiplet are insufficient to support the storage operation of the first function chiplet.
[0104] In another possible implementation, if the first function chiplet determines that its remaining available memory resources can satisfy the storage requirement, it decides not to perform data storage using the memory resources of one or more target function chiplets.
[0105] Based on the technical means described above, the first function chiplet may transmit a resource sharing request to the interconnect chiplet when local memory resources are insufficient, thereby ensuring that the first function chiplet uses local memory resources preferentially. In response, after receiving the resource sharing request transmitted by the first function chiplet, the interconnect chiplet may configure the memory resources of one or more target function chiplets with the highest access priority among other function chiplets for the first function chiplet, based on the access priority of multiple function chiplets. In this way, by configuring the memory resources of other function chiplets for the first function chiplet according to the access priority of other function chiplets, the interconnect chiplet prevents disorder or randomness in the system's resource configuration, reduces time loss due to unreasonable resource configuration, and further improves the access efficiency to the system's memory resources, thereby ensuring the efficient operation of the system.
[0106] In addition, the first function chiplet can determine whether to use the memory resources of other function chiplets based on the usage status of its own memory resources, thereby ensuring the effective utilization of memory resources and reducing memory resource waste. Furthermore, by resolving the issue of using shared memory resources between function chiplets when the remaining available memory resources of the function chiplets are insufficient, it is possible to ensure the efficient operation of the chiplet interconnection system.
[0107] Below, we introduce the conditions satisfied by one or more target function chiplets when a resource sharing request is used to request that a shared memory resource be reserved in advance, and the conditions satisfied by one or more target function chiplets when a resource sharing request is used to request the scheduling of a shared memory resource.
[0108] In one possible implementation, where a resource sharing request is used to request that a shared memory resource be reserved in advance, one or more target function chiplets satisfy at least the following conditions.
[0109] Condition 1: One or more target function chiplets allow access to the first function chiplet.
[0110] Here, the fact that the target function chiplet allows access to the first function chiplet may mean that a connection relationship exists between the target function chiplet and the first function chiplet, or that the first function chiplet can access the memory resources of the target function chiplet (i.e., there is a unidirectional connection between the first function chiplet and the target function chiplet).
[0111] Condition 2: The sum of the remaining available memory resources of one or more target function chiplets is greater than 0.
[0112] Here, residual available memory resources can also be referred to as residual shared resources. The residual shared resources of a target function chiplet refer to those that, in addition to being available for use by the target function chiplet itself, can also be shared and used by other function chiplets.
[0113] In some examples, when there is one target function chiplet, the remaining available memory resources of that target function chiplet are greater than 0, that is, that the target function chiplet has remaining available memory resources. When there are multiple target function chiplets, for example, three target function chiplets, it can be understood that the sum of the remaining available memory resources of the three target function chiplets is greater than 0, that at least one of the three target function chiplets has remaining available memory resources.
[0114] In another possible implementation, where a resource sharing request is used to request scheduling of a shared memory resource, one or more target function chiplets satisfy at least the following conditions.
[0115] Condition 3: One or more target function chiplets allow access to the first function chiplet.
[0116] Here, regarding the specific meaning of one or more target function chiplets allowing access to the first function chiplet, one may refer to the introduction of the case where the aforementioned resource sharing request is used to instruct the interconnect chiplet to reserve memory resources in advance for the first function chiplet, which is not explained repeatedly here.
[0117] Condition 4: The sum of the remaining available memory resources of one or more target function chiplets is greater than or equal to the memory resources required for the first function chiplet.
[0118] In some examples, when there is one target function chiplet, the remaining available memory resources of said target function chiplet are greater than or equal to the memory resources required for the first function chiplet. When there are multiple target function chiplets, for example, three target function chiplets, the remaining available memory resources of all three target function chiplets are greater than 0, and the sum of the remaining available memory resources of the three target function chiplets is greater than or equal to the memory resources required for the first function chiplet.
[0119] Based on the technical means described above, after receiving a resource sharing request, the interconnect chiplet can quickly and accurately determine one or more target function chiplets according to the conditions satisfied by the target function chiplets corresponding to the resource sharing request in different situations.
[0120] In some embodiments, when a resource sharing request is used to request that a shared memory resource be reserved in advance, configuring the memory resources of one or more target function chiplets for a first function chiplet based on the access priority of the plurality of function chiplets described above may specifically include the following.
[0121] In one possible implementation, the interconnect chiplet reserves memory resources of one or more target function chiplets with the highest priority among the other function chiplets for the first function chiplet based on the access priority of other function chiplets, and transmits a resource reservation message to the first function chiplet, thereby enabling the first function chiplet to use the memory resources of one or more target function chiplets based on the resource reservation message.
[0122] Here, the resource reservation message is used to indicate the memory resources of one or more target function chiplets.
[0123] In some examples, the interconnect chiplet may determine the virtual address of a memory resource based on the physical address of the memory resource of one or more target function chiplets. Then, the interconnect chiplet may transmit the virtual address of the memory resource of one or more target function chiplets to the first function chiplet, thereby enabling the first function chiplet to access the memory resource of one or more target function chiplets based on the virtual address of the memory resource of one or more target function chiplets.
[0124] Based on the technical means described above, the interconnect chiplet reserving memory resources in advance for the first function chiplet is a preventive measure that can effectively prevent system performance degradation that occurs when the demand for memory resources surges.
[0125] In some embodiments, when a resource sharing request is used to request scheduling of a shared memory resource, the above-described interconnect chiplet configuring the memory resources of one or more target function chiplets for a first function chiplet based on the access priority of a plurality of function chiplets may specifically further include the following.
[0126] In one possible implementation, the interconnect chiplet schedules the memory resources of one or more target function chiplets with the highest priority among the other function chiplets for the first function chiplet based on the access priority of the other function chiplets, and transmits a resource scheduling message to the first function chiplet, thereby enabling the first function chiplet to use the memory resources of one or more target function chiplets based on the resource scheduling message.
[0127] Here, the resource scheduling message is intended to represent the memory resources of one or more target function chiplets.
[0128] In some examples, the method of an interconnect chiplet determining a virtual address of a memory resource of one or more target function chiplets based on the physical address of a memory resource of one or more target function chiplets and transmitting the virtual address of a memory resource of one or more target function chiplets to a first function chiplet may refer to the method described above and is not described repeatedly herein.
[0129] Based on the technical means described above, the interconnect chiplet can schedule the memory resources of one or more target function chiplets for the first function chiplet, thereby enabling the first function chiplet to access the memory resources of one or more target function chiplets, and furthermore, can improve the storage performance of the chiplet interconnection system and ensure the efficient operation of the chiplet interconnection system.
[0130] The method provided in an embodiment of the present disclosure further includes scheduling all remaining available memory resources of other function chiplets for the first function chiplet when the sum of the remaining available memory resources of other function chiplets is less than the memory resources required for the first function chiplet.
[0131] Here, another function chiplet allows access to the first function chiplet.
[0132] In one possible implementation, when a resource sharing request is used to request the scheduling of a shared memory resource, the interconnect chiplet can determine the remaining available memory resources of other function chiplets that allow access to the first function chiplet, and if the sum of the remaining available memory resources of other function chiplets is less than the memory resources required by the first function chiplet, the interconnect chiplet can schedule the remaining available memory resources of all other function chiplets for the first function chiplet.
[0133] For example, function chiplet 1, function chiplet 2, and function chiplet 3 all allow access to function chiplet 4, and the sum of the remaining available memory resources of function chiplet 1, function chiplet 2, and function chiplet 3 is less than the memory resources required by function chiplet 4. In this case, the interconnect chiplets can intensively schedule all remaining available memory resources of function chiplet 1, function chiplet 2, and function chiplet 3 to function chiplet 4.
[0134] According to the technical means described above, the interconnect chiplet can intensively schedule all remaining available memory resources of the other function chiplets, excluding the first function chiplet among the plurality of function chiplets, to the first function chiplet, thereby not only improving the resource utilization efficiency of the system by making maximum use of existing memory resources but also ensuring the normal operation of the first function chiplet.
[0135] In some embodiments, when the interconnect chiplet in the above-described S602 configures the memory resources of one or more target function chiplets for the first function chiplet, a resource lock message may be further transmitted to one or more target function chiplets.
[0136] Here, the resource lock message is used to indicate that other function chiplets, excluding the first function chiplet, are not allowed to use the remaining available memory resources of the target function chiplet.
[0137] In one possible implementation, the interconnect chiplet receives a resource sharing request from the first function chiplet, determines one or more target function chiplets with the highest access priority among other function chiplets based on the access priority of multiple function chiplets and the memory resources required by the first function chiplet, and can transmit a resource lock message to the one or more target function chiplets.
[0138] In another possible implementation, the interconnect chiplet can detect the usage status of the memory resources of the first function chiplet. If the memory resources used by the first function chiplet are detected to be greater than the first resource pre-warning threshold, the interconnect chiplet can send a resource lock message to one or more target function chiplets.
[0139] Based on the technical means described above, the interconnect chiplet transmits a resource lock message to one or more target function chiplets, thereby ensuring that the remaining available memory resources of the locked one or more target function chiplets are provided to and used only by the first function chiplet within a specific period, and thus preventing contention and occupation of these memory resources by other function chiplets.
[0140] In some embodiments, when it is determined that there is no need to schedule memory resources of one or more target function chiplets for the first function chiplet, the interconnect chiplet may send a resource release message to one or more target function chiplets.
[0141] Here, the resource release message is used to indicate that the first function chiplet no longer uses the memory resources of one or more target function chiplets. That is, the memory resources of one or more target function chiplets may be used by other function chiplets.
[0142] In one possible implementation, the first function chiplet may transmit a resource release notice to the interconnect chiplet when the used memory resource is less than the first resource pre-warning threshold, and said resource release notice is used to indicate that there is no need to use the memory resources of one or more target function chiplets. In response, the interconnect chiplet may, after receiving the resource release notice transmitted by the first function chiplet, transmit a resource release message to one or more target function chiplets in response to the resource release notice.
[0143] In another possible implementation, the interconnect chiplet can detect the usage status of the memory resources of the first function chiplet and / or the operating status of the first function chiplet. If a situation is detected where the memory resources used by the first function chiplet are less than the first resource pre-warning threshold and / or where the first function chiplet stops operating or its operating status is abnormal, the interconnect chiplet can send a resource release message to one or more target function chiplets.
[0144] Based on the technical means described above, the embodiment of the present disclosure can release the memory resources of the target function chiplets in a timely manner when it is determined that there is no need to schedule the memory resources of one or more target function chiplets for the first function chiplet, thereby preventing idleness and waste of resources and optimizing memory resource usage efficiency. In addition, delays and performance degradation caused by other function chiplets waiting for memory resources can be reduced.
[0145] In some embodiments, when an interconnect chiplet receives a shared resource request transmitted by at least two function chiplets, the resource configuration method provided in the embodiment of the present disclosure may further include S701 to S702, as illustrated in FIG. 7.
[0146] S701: Receives a resource sharing request transmitted by at least two function chiplets.
[0147] S702: According to the resource scheduling priority of each of at least two function chiplets, the memory resources of the other function chiplet are configured for each function chiplet.
[0148] Here, other function chiplets are other function chiplets excluding at least two function chiplets among the plurality of function chiplets.
[0149] Resource scheduling priority can be used to indicate the order in which memory resources are configured for function chiplets. The higher the resource scheduling priority of a function chiplet, the earlier it is scheduled to schedule memory resources for that function chiplet. For example, if the resource scheduling priority of function chiplet 1 is higher than the resource scheduling priority of function chiplet 2, the interconnect chiplet may receive resource sharing requests sent by function chiplet 1 and function chiplet 2 simultaneously, schedule memory resources for function chiplet 1 first, and then schedule memory resources for function chiplet 2.
[0150] In one possible implementation, the interconnect chiplet receives resource sharing requests transmitted by at least two function chiplets, and then stores the resource sharing requests of each function chiplet in a first-in, first-out cache queue according to the resource scheduling priority of each of the at least two function chiplets; that is, it stores the resource sharing requests of the function chiplet with the higher resource scheduling priority first, and the resource sharing requests of the function chiplet with the lower resource scheduling priority later. After that, the interconnect chiplet can sequentially configure memory resources for each function chiplet according to the order of resource sharing requests in the first-in, first-out cache queue. The method by which the interconnect chiplet configures memory resources for each function chiplet may refer to the method by which the interconnect chiplet configures memory resources of one or more target function chiplets for the first function chiplet in the above-described S602, which is not described repeatedly here.
[0151] After scheduling memory resources for a single function chiplet, the interconnect chiplet removes the resource sharing request of that function chiplet from the first-in, first-out cache queue and can respond to the first resource sharing request in the first-in, first-out cache queue until there are no resource sharing requests in the first-in, first-out cache queue.
[0152] Based on the technical means described above, after an interconnect chiplet receives a resource sharing request transmitted by a plurality of function chiplets, the interconnect chiplet can first share memory resources for a function chiplet with a higher resource scheduling priority and then share memory resources for a function chiplet with a lower resource scheduling priority, based on the resource scheduling priority of each function chiplet. This resolves the problem of disordered scheduling and improves the scheduling efficiency of the chiplet interconnection system.
[0153] Hereinafter, a resource configuration method provided in an embodiment of the present disclosure is described in detail, with the interaction between an interconnect chiplet, a first function chiplet, and two target function chiplets as an example, and as illustrated in FIG. 8, the method may specifically include S801 to S819.
[0154] S801: The first function chiplet sends a resource sharing request to the interconnect chiplet.
[0155] Here, regarding resource sharing requests, refer to the introduction in S601 described above, which is not repeated here.
[0156] S802: The interconnect chiplet sends a shared resource detection request to the target function chiplet 1.
[0157] Here, the shared resource detection request is used to detect the memory resource status of the target function chiplet, and the memory resource status includes the remaining available memory resources and used memory resources of the target function chiplet.
[0158] S803: The interconnect chiplet sends a shared resource detection request to the target function chiplet 2.
[0159] S804: Target function chiplet 1 transmits memory resource status to interconnect chiplet.
[0160] S805: Target function chiplet 2 transmits memory resource status to interconnect chiplet.
[0161] S806: The interconnect chiplet determines whether the sum of the remaining available memory resources of target function chiplet 1 and target function chiplet 2 is greater than 0, and if the sum of the remaining available memory resources of target function chiplet 1 and target function chiplet 2 is 0 or less, S807 is performed, and if the sum of the remaining available memory resources of target function chiplet 1 and target function chiplet 2 is greater than 0, S808 is performed.
[0162] In some examples, when there are n target function chiplets, the sum of the remaining available memory resources of the n target function chiplets = remaining available memory resources of target function chiplet 1 + remaining available memory resources of target function chiplet 2 + remaining available memory resources of target function chiplet 3 + … + remaining available memory resources of target function chiplet n.
[0163] In some examples, the remaining available memory resources of the target function chiplet = the total memory resources of the target function chiplet - the used memory resources, or the remaining available memory resources of the target function chiplet = the total memory resources of the target function chiplet * the first resource pre-warning threshold - the used memory resources.
[0164] S807: The interconnect chiplet transmits a no-available-resources-to-the-first-function-chiplet notification.
[0165] Here, the "no available resources" notice is used to indicate that there are no memory resources available to be shared and used by the first function chiplet.
[0166] S808: The interconnect chiplet determines whether the sum of the remaining available memory resources of the target function chiplet 1 and the target function chiplet 2 is greater than or equal to the memory resources required for the first function chiplet, and if the sum of the remaining available memory resources of the target function chiplet 1 and the target function chiplet 2 is less than the memory resources required for the first function chiplet, performs S809 and S811 to S815, and if the sum of the remaining available memory resources of the target function chiplet 1 and the target function chiplet 2 is greater than or equal to the memory resources required for the first function chiplet, performs S810 and S811 to S815.
[0167] S809: The interconnect chiplet configures all remaining available memory resources of target function chiplet 1 and target function chiplet 2 for the first function chiplet.
[0168] S810: The interconnect chiplet configures memory resources for the first function chiplet that are identical to the memory resources required for the first function chiplet. Here, the method by which the interconnect chiplet configures memory resources for the first function chiplet may refer to S602 described above and is not described repeatedly here.
[0169] For example, the memory resources required for the first function chiplet are 10, the remaining available memory resources for the target function chiplet 1 are 5, and the remaining available memory resources for the target function chiplet 2 are 7.
[0170] In this case, the interconnect chiplet may configure five remaining available memory resources of the target function chiplet 1 for the first function chiplet and five remaining available memory resources of the target function chiplet 2. Alternatively, the interconnect chiplet may configure four remaining available memory resources of the target function chiplet 1 for the first function chiplet and six remaining available memory resources of the target function chiplet 2.
[0171] S811: The interconnect chiplet transmits a resource configuration message to the first function chiplet.
[0172] S812: The interconnect chiplet sends a resource lock message to the target function chiplet 1.
[0173] S813: The interconnect chiplet sends a resource lock message to the target function chiplet 2.
[0174] S814: The first function chiplet performs data storage using the memory resources of target function chiplet 1 and target function chiplet 2.
[0175] Here, the priority of the memory resources of the first function chiplet itself is higher than the priority of the memory resources of other function chiplets; that is, the first function chiplet uses its own memory resources first, and then uses the memory resources of other function chiplets.
[0176] S815: The first function chiplet determines whether the memory resource used is greater than or equal to the first resource pre-warning threshold, and if the first function chiplet determines that the memory resource used is greater than or equal to the first resource pre-warning threshold, performs S816, and if the first function chiplet determines that the memory resource used is less than the first resource pre-warning threshold, performs S817 to S819.
[0177] S816: The first function chiplet determines whether to continue requesting the sharing of memory resources, and if the first function chiplet determines to continue requesting the sharing of memory resources, performs S801, and if the first function chiplet determines not to continue requesting the sharing of memory resources, performs S817 to S819.
[0178] S817: The first function chiplet transmits a resource release notice to the interconnect chiplet.
[0179] S818: The interconnect chiplet sends a resource release message to the target function chiplet 1.
[0180] Here, the method of an interconnect chiplet sending a resource release message to a target function chiplet 1 may refer to the method of an interconnect chiplet sending a resource release message to one or more target function chiplets described above, which is not described repeatedly here.
[0181] S819: The interconnect chiplet sends a resource release message to the target function chiplet 2.
[0182] The foregoing description has primarily introduced the methods provided in the embodiments of this disclosure from the perspective of method. To implement the functions described above, resource configuration devices or electronic devices include hardware structures and / or software elements corresponding to performing each function. A person skilled in the art will readily understand that the embodiments of this disclosure may be implemented in hardware or in a combination of hardware and computer software in relation to the parts and algorithm steps of each example described in combination with the embodiments disclosed herein. Whether a function is performed by hardware or by computer software driving the hardware is determined by the specific application and design constraints of the technical method. A person skilled in the art may implement the described functions using different methods for each specific application, but such implementations should not be construed as being outside the scope of the embodiments of this disclosure.
[0183] FIG. 9 is a block diagram of a resource configuration device illustrated in an embodiment of the present disclosure, wherein the device is applied to an interconnect chiplet in a chiplet interconnection system, and the chiplet interconnection system further includes a plurality of function chiplets, each function chiplet being provided with a memory resource. As illustrated in FIG. 9, the resource configuration device includes a communication unit (901) and a configuration unit (902).
[0184] In one possible manner, the communication unit (901) is configured to receive a resource sharing request transmitted by the first function chiplet.
[0185] In one possible manner, the configuration unit (902) is configured to configure the memory resources of one or more target function chiplets for the first function chiplet based on the access priority of the multiple function chiplets.
[0186] In one possible manner, the configuration unit (902) is also configured to schedule all remaining available memory resources of other function chiplets for the first function chiplet when the sum of the remaining available memory resources of other function chiplets is less than the memory resources required for the first function chiplet.
[0187] In one possible manner, the communication unit (901) is also configured to transmit a resource lock message to one or more target function chiplets.
[0188] In one possible manner, the communication unit (901) is also configured to send a resource release message to one or more target function chiplets when it is determined that there is no need to schedule memory resources of one or more target function chiplets for the first function chiplet.
[0189] FIG. 10 is a block diagram of another resource configuration device illustrated in an embodiment of the present disclosure, wherein the device is applied to a first function chiplet in a chiplet interconnection system, and the chiplet interconnection system further includes a plurality of function chiplets, each function chiplet being provided with a memory resource. As illustrated in FIG. 10, the resource configuration device includes a communication unit (1001), a processing unit (1002), and a storage unit (1003).
[0190] In one possible manner, the communication unit (1001) is configured to transmit a resource sharing request to the interconnect chiplet. The communication unit (1001) is also configured to receive a resource configuration message transmitted by the interconnect chiplet.
[0191] In one possible manner, the processing unit (1002) is configured to determine whether to perform data storage using the memory resources of one or more target function chiplets based on the usage status of the memory resources of the first function chiplet.
[0192] In one possible way, the storage unit (1003) is configured to store resources.
[0193] In one possible manner, the communication unit (1001) is also configured to transmit a resource release notification to the interconnect chiplet when it is determined that there is no need to use the memory resources of one or more target function chiplets.
[0194] In one possible manner, the processing unit (1002) is also configured to determine that if it determines that the remaining available memory resources of the first function chiplet cannot satisfy the storage requirement, it will determine to perform data storage using the memory resources of one or more target function chiplets, and if it determines that the remaining available memory resources of the first function chiplet can satisfy the storage requirement, it will determine not to perform data storage using the memory resources of one or more target function chiplets.
[0195] FIG. 11 is a block diagram of an electronic device illustrated in an embodiment of the present disclosure. As shown in FIG. 11, the electronic device includes, but is not limited to, a processor (1101) and a memory (1102).
[0196] Here, the memory (1102) described above is configured to store executable instructions of the processor (1101) described above. It can be understood that the processor (1101) described above is configured to execute instructions to implement the resource configuration method of the embodiment described above.
[0197] It must be explained that a person skilled in the field can understand that the structure of the electronic device shown in FIG. 11 does not constitute a limitation on the electronic device, and that the electronic device may include more or fewer parts than shown in FIG. 11, combine some parts, or have different arrangements of parts.
[0198] The processor (1101) serves as the control center of the electronic device, connecting each part of the entire electronic device using various interfaces and circuits, driving or executing software programs and / or components stored in memory (1102), and calling data stored in memory (1102) to perform various functions of the electronic device and process data, thereby monitoring the entire electronic device. The processor (1101) may include one or more processing units. Optionally, the processor (1101) may integrate an application processor and a modem processor, wherein the application processor primarily processes operating systems, user interfaces, and applications, and the modem processor primarily processes wireless communication. It can be understood that the aforementioned modem processor may not be integrated into the processor (1101).
[0199] The memory (1102) is configured to store software programs and various data. The memory (1102) may primarily include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one functional element (e.g., a decision unit, a processing unit, etc.). Additionally, the memory (1102) may include high-speed random access memory and may include non-volatile memory, such as at least one magnetic disk storage element, a flash memory element, or other non-volatile semiconductor storage element.
[0200] In an exemplary embodiment, a computer-readable storage medium containing instructions, such as a memory (1102) containing instructions, is further provided, and the instructions described above can be executed by a processor (1101) of an electronic device to implement the method of the embodiment described above.
[0201] In actual implementation, the functions of the communication unit (901) and configuration unit (902) of FIG. 9, and the functions of the communication unit (1001), processing unit (1002), and storage unit (1003) of FIG. 10 can all be implemented by the processor (1101) of FIG. 11 calling a computer program stored in memory (1102). The specific execution process can be described by referring to the explanation of the method part of the above-described embodiment, and is not described repeatedly here.
[0202] Optionally, the computer-readable storage medium may be a non-transient computer-readable storage medium, and such non-transient computer-readable storage medium may be, for example, Read-Only Memory (ROM), Random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.
[0203] In an exemplary embodiment, the embodiment of the present disclosure also provides a computer program product comprising one or more instructions, said one or more instructions may be executed by a processor (1101) of an electronic device to complete the method of the embodiment described above.
[0204] It must be explained that when the instructions in the computer-readable storage medium or one or more instructions in the computer program product described above are executed by the processor of an electronic device, they implement each process of the method embodiment described above and can achieve the same technical effect as the method described above, and to avoid duplication, they are not described again here.
[0205] Through the description of the above-described embodiments, those skilled in the art will understand that for the convenience and brevity of the explanation, only the distinction of each functional element described above has been used as an example, and in actual applications, the functions described above can be distributed to different functional elements to be completed as needed; that is, it will be clearly understood that the internal structure of the device can be divided into different functional elements to complete all or part of the functions described above.
[0206] It should be understood that in some of the embodiments provided in this disclosure, the disclosed devices and methods may be implemented in different ways. For example, the device embodiments described above are merely exemplary, and, for example, the distinction between elements or parts is merely a logical functional distinction, and there may be different distinctions in actual implementation, for example, multiple parts or components may be combined or integrated into other devices, and some features may be omitted or not performed. On the other hand, the mutual coupling, direct coupling, or communication connection shown or discussed may be made through some interface, and the indirect coupling or communication connection of devices or parts may be electrical, mechanical, or of other forms.
[0207] The part described as a separated part may or may not be physically separated, and the part indicated as a part may be a single physical part or multiple physical parts, that is, it may be located in one place or distributed in multiple different places. Depending on actual needs, some or all of the parts may be selected to achieve the purpose of the method of the present embodiment.
[0208] Additionally, each functional part in each embodiment of the present disclosure may be integrated into a single processing unit, each part may exist physically separately, or two or more parts may be integrated into a single unit. The integrated parts described above may be implemented in the form of hardware or in the form of software functional parts.
[0209] When implemented in the form of an integrated additional software function and sold or used as an independent product, it may be stored on a single readable storage medium. Based on this understanding, the technical solution of the embodiments of the present disclosure may essentially—that is, a part contributing to the prior art, or all or part of said technical solution—be implemented in the form of a software product, said software product which is stored on a single storage medium and includes a plurality of instructions that cause a single device (which may be a microcontroller, a chip, etc.) or processor to perform all or part of the steps of the method of each embodiment of the present disclosure. The aforementioned storage medium includes various media capable of storing program code, such as a USB memory, an external hard disk, ROM, RAM, a magnetic disk, or an optical disk.
[0210] The above is merely a specific embodiment of the embodiments of the present disclosure, and the scope of protection of the embodiments of the present disclosure is not limited thereto. Any modification or substitution within the technical scope disclosed in the embodiments of the present disclosure shall be included within the scope of protection of the embodiments of the present disclosure. Accordingly, the scope of protection of the embodiments of the present disclosure shall be based on the scope of protection of the claims.
Claims
Claim 1 A method for configuring resources, applied to an interconnect chiplet within a chiplet interconnection system, wherein the chiplet interconnection system further comprises a plurality of function chiplets, each of which is provided with a memory resource, and the method comprises the step of receiving a resource sharing request transmitted by a first function chiplet, wherein the resource sharing request is used to request the sharing of a memory resource of another function chiplet, and the other function chiplet is a function chiplet among the plurality of function chiplets excluding the first function chiplet; and the step of configuring a memory resource of one or more of the other function chiplets for the first function chiplet based on the access priority of the plurality of function chiplets. Claim 2 In claim 1, if the memory resource used by the first function chiplet is greater than or equal to the first resource pre-warning threshold and less than the second resource pre-warning threshold, the resource sharing request is used to request that the shared memory resource be reserved in advance, and if the memory resource used by the first function chiplet is greater than or equal to the second resource pre-warning threshold, the resource sharing request is used to request the scheduling of the shared memory resource, and the second resource pre-warning threshold is greater than the first resource pre-warning threshold, method. Claim 3 In paragraph 2, a method wherein, when the resource sharing request is used to instruct the interconnect chiplet to reserve memory resources in advance for the first function chiplet, the one or more target function chiplets satisfy at least the condition that the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than 0. Claim 4 A method according to paragraph 2, wherein when the resource sharing request is used to instruct the interconnect chiplet to schedule memory resources for the first function chiplet, the one or more target function chiplets satisfy at least the condition that all of the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than or equal to the memory resources required for the first function chiplet. Claim 5 A method according to any one of claims 1 to 4, wherein the method further comprises the step of scheduling all remaining available memory resources of the other function chiplets for the first function chiplet when the sum of the remaining available memory resources of the other function chiplets is less than the memory resources required for the first function chiplet. Claim 6 A method according to any one of claims 1 to 4, wherein the method further comprises the step of transmitting a resource lock message to one or more target function chiplets, wherein the resource lock message is used to indicate that the other function chiplets are not allowed to use the remaining available memory resources of the target function chiplets. Claim 7 In claim 6, the method further comprises the step of transmitting a resource release message to the one or more target function chiplets when it is determined that there is no need to schedule the memory resources of the one or more target function chiplets for the first function chiplet, wherein the resource release message is used to indicate that the first function chiplet no longer uses the memory resources of the one or more target function chiplets. Claim 8 A method for configuring resources, wherein a first function chiplet is applied to a plurality of function chiplets included in a chiplet interconnection system, the chiplet interconnection system further includes an interconnection chiplet, and each of the plurality of function chiplets is provided with a memory resource, and the method comprises the step of transmitting a resource sharing request to the interconnection chiplet - the resource sharing request is used to request the sharing of a memory resource of another function chiplet, and the other function chiplet is a function chiplet among the plurality of function chiplets excluding the first function chiplet -; the step of receiving a resource configuration message transmitted by the interconnection chiplet - the resource configuration message is used to indicate a memory resource of one or more target function chiplets among the other function chiplets, and the one or more target function chiplets are determined according to the access priority of the plurality of function chiplets -; and the step of determining whether to perform data storage using the memory resource of the one or more target function chiplets based on the usage status of the memory resource of the first function chiplet. Claim 9 In claim 8, if the memory resource used by the first function chiplet is greater than or equal to the first resource pre-warning threshold and less than the second resource pre-warning threshold, the resource sharing request is used to request that the shared memory resource be reserved in advance, and if the memory resource used by the first function chiplet is greater than or equal to the second resource pre-warning threshold, the resource sharing request is used to request the scheduling of the shared memory resource, and the second resource pre-warning threshold is greater than the first resource pre-warning threshold, method. Claim 10 In claim 9, where the resource sharing request is used to instruct the interconnect chiplet to reserve memory resources in advance for the first function chiplet, the one or more target function chiplets satisfy at least the condition that the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than 0. Claim 11 In claim 9, when the resource sharing request is used to instruct the interconnect chiplet to schedule memory resources for the first function chiplet, the one or more target function chiplets satisfy at least the condition that all of the one or more target function chiplets allow access to the first function chiplet; and the condition that the sum of the remaining available memory resources of the one or more target function chiplets is greater than or equal to the memory resources required for the first function chiplet. Claim 12 In any one of claims 8 to 11, the method further comprises the step of transmitting a resource release notice to the interconnect chiplet, wherein the resource release notice is used to indicate that there is no need to use the memory resources of the one or more target function chiplets. Claim 13 A method according to any one of claims 8 to 11, wherein the step of determining whether to perform data storage using the memory resources of one or more target function chiplets based on the usage status of the memory resources of the first function chiplet comprises: a step of determining to perform data storage using the memory resources of one or more target function chiplets when it is determined that the remaining available memory resources of the first function chiplet cannot satisfy the storage requirement; and a step of determining not to perform data storage using the memory resources of one or more target function chiplets when it is determined that the remaining available memory resources of the first function chiplet can satisfy the storage requirement. Claim 14 A chiplet interconnection system comprising an interconnection chiplet and a plurality of function chiplets, wherein each function chiplet is provided with a memory resource, wherein the interconnection chiplet is configured to perform a resource configuration method according to claim 1, and the function chiplet is configured to perform a resource configuration method according to claim 8. Claim 15 A system according to claim 14, wherein each of the above-mentioned function chiplets includes a D2D interface, the above-mentioned interconnect chiplets include a plurality of the above-mentioned D2D interfaces, and each of the above-mentioned D2D interfaces of the above-mentioned interconnect chiplets is configured to be connected to the D2D interface of the corresponding above-mentioned function chiplets. Claim 16 A resource configuration device, wherein the device is applied to an interconnect chiplet within a chiplet interconnection system, the chiplet interconnection system further comprises a plurality of function chiplets, each of which is equipped with a memory resource, the device comprises a communication unit and a configuration unit, the communication unit is configured to receive a resource sharing request transmitted by a first function chiplet, wherein the resource sharing request is used to request the sharing of a memory resource of another function chiplet, the other function chiplet is a function chiplet among the plurality of function chiplets excluding the first function chiplet, and the configuration unit is configured to configure the memory resource of one or more target function chiplets among the other function chiplets for the first function chiplet based on the access priority of the plurality of function chiplets. Claim 17 In a resource configuration device, a first function chiplet is applied, wherein the first function chiplet is any function chiplet among a plurality of function chiplets included in a chiplet interconnection system, and the chiplet interconnection system further includes an interconnection chiplet, and each of the function chiplets is equipped with a memory resource, and the device includes a communication unit, a processing unit, and a storage unit, wherein the communication unit is configured to transmit a resource sharing request to the interconnection chiplet, where the resource sharing request is used to request the sharing of a memory resource of another function chiplet, and the other function chiplet is a function chiplet among the plurality of function chiplets excluding the first function chiplet, and the communication unit is also configured to receive a resource configuration message transmitted by the interconnection chiplet, where the resource configuration message is used to indicate a memory resource of one or more target function chiplets among the other function chiplets, and the one or more target function chiplets are determined according to the access priority of the plurality of function chiplets, and the processing unit uses the memory resource of the one or more target function chiplets based on the usage status of the memory resource of the first function chiplet to data A resource configuration device configured to determine whether to perform storage, wherein the storage unit is configured to store resources. Claim 18 An electronic device, wherein the electronic device is equipped with a chiplet interconnection system according to claim 14. Claim 19 A computer-readable storage medium, wherein when a computer execution instruction stored in the computer-readable storage medium is executed by a processor of an electronic device, the electronic device performs a resource configuration method according to claim 1 or a resource configuration method according to claim 8.
