CN114298892A - A cache module and system for distributed processing unit - Google Patents

Abstract

The application provides a cache module and a system applied to a distributed processing unit, wherein the cache module comprises: the device comprises a control module, a storage module and a sniffing module; the control module is used for receiving the data request and processing the data request; the storage module is used for caching data; and the sniffing module is used for receiving and executing the data consistency instruction. The cache module provided by the application ensures the adaptability with the MESI data consistency protocol and the variation thereof through the control module and the sniffing module, and can effectively improve the efficiency of data caching in parallel computing.

Description

A cache module and system for distributed processing units

technical field

The present application relates to the technical field of image processing, and in particular, to a cache module and system applied to distributed processing units.

Background technique

In the process of executing instructions by the computer processing unit, in order to speed up the reading and writing of data involved, the cache will be used as the location for reading and writing data, and the data in the cache will be updated to the main memory after the operation is completed. middle. In the scenario of multi-core central processing unit (CPU) and graphics processing unit (Graphics Processing Unit, GPU), multiple processing units will have their own caches, and when they operate in parallel, the same data in main memory It may be read and modified by different processing units, and a certain mechanism is required to ensure the consistency of the data between the main memory and each cache during the reading and writing process.

In the field of graphics processing and computing, rendering, texture filtering, etc. have high requirements on computing performance and throughput, and a cache solution suitable for distributed processing units is required.

SUMMARY OF THE INVENTION

In order to solve one of the above technical defects, the present application provides a cache module and system applied to a distributed processing unit.

A first aspect of the present application provides a cache module applied to a distributed processing unit, the cache module comprising: a control module, a storage module and a sniffing module;

The control module is used to receive and process the data request;

The storage module is used to cache data;

The sniffing module is used for receiving and executing data consistency instructions.

Optionally, the control module is configured to receive a data request from the processing unit, and determine whether the requested data is stored in the storage module according to the data request.

Optionally, the control module is configured to return a hit signal after judging that the request data is stored in the storage module, read the request data from the storage module, and process the request data.

Optionally, the control module is configured to modify the status identifier of the request data.

Optionally, the control module is configured to acquire the request data from other cache modules after judging that the request data is not stored in the storage module.

Optionally, the control module is configured to read the request data from the external main memory after judging that the request data is not stored in the storage module and the request data is not obtained from other cache modules.

Optionally, the control module is configured to store the request data in the storage module.

Optionally, the data consistency instruction includes one or more of the following: an instruction to copy and send the data in the storage module to other cache modules, and mark the copy of the data in the storage module as unavailable The instruction to write the data in the storage module back to the main memory.

Optionally, copies of data that have been modified in other cache modules are marked as unavailable.

Optionally, the processing unit includes a plurality of modules;

The control module is used for receiving data requests from multiple modules in the processing unit, arbitrating the priority of each module, and processing each data request according to the arbitration result.

Optionally, the sniffing module is configured to copy the requested data in the storage module according to the instruction to copy and send the data in the storage module to other cache modules, and write the copied data. into the storage module in the other cache modules.

In a second aspect of the present application, a cache system applied to a distributed processing unit is provided, the cache system includes: a general control module and a plurality of cache modules;

The general control module is used to control the consistency of data transmission between the cache modules;

The cache module is the cache module described in the first aspect above.

Optionally, the master control module is used to receive a data request from any cache module, generate an instruction to copy and send the data in the storage module to other cache modules, and copy and send the storage module. Instructions for data in a module to other cache modules are sent to other cache modules.

The present application provides a cache module and system applied to a distributed processing unit. The cache module includes: a control module, a storage module and a sniffing module; a control module for receiving and processing data requests; storage The module is used to cache data; the sniffing module is used to receive and execute data consistency instructions. The cache module provided by the present application ensures the adaptability with the MESI data consistency protocol and its variants through the control module and the sniffing module, and can effectively improve the efficiency of data cache during parallel computing.

Description of drawings

The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The schematic embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the attached image:

1 is a schematic structural diagram of a cache module applied to a distributed processing unit according to an embodiment of the present application;

2 is a schematic structural diagram of another cache module applied to a distributed processing unit according to an embodiment of the present application;

3 is a schematic structural diagram of a cache system applied to a distributed processing unit according to an embodiment of the present application;

4 is a schematic structural diagram of another cache system applied to a distributed processing unit provided by an embodiment of the present application;

5 is a schematic structural diagram of another cache system applied to a distributed processing unit provided by an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a general control module according to an embodiment of the present application.

Detailed ways

In order to make the technical solutions and advantages of the embodiments of the present application more clear, the exemplary embodiments of the present application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and Not all embodiments are exhaustive. It should be noted that the embodiments in the present application and the features of the embodiments may be combined with each other in the case of no conflict.

In the process of realizing this application, the inventor found that in the scenario of multi-core CPU and GPU, multiple processing units will have their own caches, and when they operate in parallel, the same data in the main memory may be processed by different processing units. Reading and modifying requires a certain mechanism to ensure the consistency of data between the main memory and each cache during the reading and writing process. In the field of graphics processing and computing, rendering, texture filtering, etc. have high requirements on computing performance and throughput, and a cache solution suitable for distributed processing units is required.

In view of the above problems, the embodiment of the present application provides a cache module and system applied to a distributed processing unit, the cache module includes: a control module, a storage module and a sniffing module; a control module for receiving data requests And process the data request; the storage module is used to cache data; the sniffing module is used to receive and execute data consistency instructions. The cache module provided by the present application ensures the adaptability with the MESI data consistency protocol and its variants through the control module and the sniffing module, and can effectively improve the efficiency of data cache during parallel computing.

Referring to FIG. 1 , the cache module provided in this embodiment includes: a control module, a storage module, and a sniffing module.

1. Control module

The control module is used to receive and process the data request.

The data request is sent by the processing unit.

If the processing unit includes multiple modules. Then each module may send a data request, then the control module is used for receiving data requests from multiple modules in the processing unit, arbitrating the priority of each module, and processing each data request according to the arbitration result.

For example, the data requests sent by each module are processed in order from high to low according to the priority obtained by arbitration.

Referring to FIG. 2 , the control module is specifically configured to receive a data request from the processing unit, and determine whether the requested data is stored in the storage module according to the data request.

In addition, after judging that the request data is stored in the storage module, the control module will return a hit signal, read the request data from the storage module, and process the request data. After judging that the request data is not stored in the storage module, obtain the request data from other cache modules (for example, send an instruction to copy and send the data in the storage module to other cache modules to other cache modules, and other cache modules based on the copy and Instructions that send data in a memory module to other cache modules send request data). After judging that the request data is not stored in the storage module and the request data is not obtained from other cache modules, the request data is read from the external main memory.

In the case where the request data is not stored in the storage module, if the main control module obtains the request data (which can be obtained from other cache modules or from an external main memory), the request data is stored in the storage module.

In addition, the control module is also used to modify the status flag of the request data.

2. Storage module

A storage module for caching data.

3. Sniffing module

The sniffing module is used to receive and execute data consistency commands.

Wherein, the data consistency instruction includes but is not limited to one or more of the following: an instruction to copy and send data in the storage module to other cache modules, an instruction to mark the copy of the data in the storage module as an unavailable instruction, to store the An instruction to write data in the module back to main memory.

Mark copies of data that have been modified in other cache modules as unavailable. That is, when a piece of data is modified in other cache modules, the sniffing module of this cache module receives an instruction to mark the copy of the data in the storage module as unavailable.

If the sniffing module receives an instruction to copy and send the data in the storage module to other cache modules, the sniffing module will copy the request in the storage module according to the instruction to copy and send the data in the storage module to other cache modules data, and the copied data is written to storage modules in other cache modules.

A cache module applied to a distributed processing unit provided by this embodiment includes: a control module, a storage module and a sniffing module; a control module for receiving and processing data requests; a storage module for caching Data; sniffing module for receiving and executing data consistency commands. The cache module provided in this embodiment ensures the adaptability with the MESI data consistency protocol and its variants through the control module and the sniffing module, and can effectively improve the efficiency of data cache during parallel computing.

Referring to FIG. 3 , this embodiment provides a cache system applied to a distributed processing unit, where the cache system includes: a general control module and a plurality of cache modules.

1. Master control module

The master control module is used to control the consistency of data transmission among the cache modules.

Specifically, the master control module is used to receive a data request from any cache module, generate an instruction to copy and send the data in the storage module to other cache modules, and copy and send the data in the storage module to other cache modules. instructions are sent to other cache modules.

2. Cache module

Each cache module is a cache module as shown in FIG. 1 .

For example, any cache module includes: a control module, a storage module, and a sniffing module.

1) Control module

The control module is used to receive and process the data request.

The data request is sent by the processing unit.

If the processing unit includes multiple modules. Then each module may send a data request, then the control module is used for receiving data requests from multiple modules in the processing unit, arbitrating the priority of each module, and processing each data request according to the arbitration result.

For example, the data requests sent by each module are processed in order from high to low according to the priority obtained by arbitration.

The control module is specifically configured to receive a data request from the processing unit, and determine whether the requested data is stored in the storage module according to the data request.

In addition, after judging that the request data is stored in the storage module, the control module will return a hit signal, read the request data from the storage module, and process the request data. After judging that the requested data is not stored in the storage module, obtain the requested data from other cache modules (for example, send an instruction to copy and send the data in the storage module to other cache modules to other cache modules, and other cache modules based on the copy and Instructions that send data in a memory module to other cache modules send request data). After judging that the request data is not stored in the storage module and the request data is not obtained from other cache modules, the request data is read from the external main memory.

In the case where the request data is not stored in the storage module, if the main control module obtains the request data (which can be obtained from other cache modules or from an external main memory), the request data is stored in the storage module.

In addition, the control module is also used to modify the status flag of the request data.

2) Storage module

A storage module for caching data.

3) Sniffing module

The sniffing module is used to receive and execute data consistency commands.

Wherein, the data consistency instruction includes but is not limited to one or more of the following: an instruction to copy and send data in the storage module to other cache modules, an instruction to mark the copy of the data in the storage module as an unavailable instruction, to store the An instruction to write data in the module back to main memory.

Mark copies of data that have been modified in other cache modules as unavailable. That is, when a piece of data is modified in other cache modules, the sniffing module of this cache module receives an instruction to mark the copy of the data in the storage module as unavailable.

If the sniffing module receives an instruction to copy and send the data in the storage module to other cache modules, the sniffing module will copy the request in the storage module according to the instruction to copy and send the data in the storage module to other cache modules data, and the copied data is written to storage modules in other cache modules.

The cache module is composed of a control module, a storage module, and a sniffing module, and each distributed processing unit contains one of the cache modules. Multiple distributed processing units are connected to the same general control module to realize data sharing among multiple cache modules.

The data request from the processing unit is first sent to the control module, and the control module judges whether the requested data is available in the cache and performs different operations according to the result: if the data is available, a hit signal is returned, and the storage module is instructed to read the data; If the data is not available, the data is requested from the main memory or the cache of other units, and the instruction is read out after the data is obtained. The sniffing module is used for receiving instructions from the master control module, so as to realize data sharing among multiple cache modules and ensure data consistency. Specifically, the instructions received by the sniffing module mainly include: copying the data requested by other cache modules in this cache module and sending it to the requester, when other cache modules modify a certain data, copy the data in this module Mark a copy of the data as unavailable, write some data back to main memory, etc. The storage module acts as a storage unit for cached data.

The service objects of the cache module include other modules within the processing unit that need to read and write data and the cache modules of other processing units. Data transmission is performed between the cache modules through the sniffing module under the control of the general control module.

Other modules inside the processing unit have different priorities for reading and writing cached data according to their functions. Therefore, the control module needs to arbitrate the requests of these modules. After the request with the highest priority passes the arbitration, the control module is based on the type of request. and whether the requested data is available in the storage module to modify the status flag of the data, and at the same time, the storage module reads out the corresponding cached data according to the data address contained in the request and returns a data available signal. If the requested data does not exist in the storage module, the control module returns a suspend signal, and sends a data request to other connected cache modules through the master control module.

For example, when the data to be read by the cache module A has been cached in the cache module B, the data request sent by the cache module A is sent to all the cache modules connected to the cache module A through the general control module. After receiving the request, the control module in module B checks whether the data does not exist or is occupied. If the data is available, it returns a signal to the general control module, and the sniffing module in the cache module B will cache the data requested by the cache module A. Data is written to the storage module of cache module A. If there is no available request data in all the cache modules connected to the cache module A, the cache module A reads the requested data from the external main memory and caches it in the storage module of A.

Taking four connected cache modules as an example, their structural relationship and signals with the master control module and the main memory are shown in Figure 4. The cache module sends the state of each cached data (including available, shared, modified, unavailable, etc.) and data requests, and receive return signals and data request signals from the master control module. At the same time, each cache module is connected with the main memory through a direct memory access (DMA) module, so as to realize the functions of reading data from the main memory and writing the data in the cache into the main memory. In addition, under the control of the master control module, each cache module can send cache data to the data sharing link through its own sniffing module for use by other cache modules in need.

Inside the cache module, the structural relationship and signals of its main modules are shown in Figure 5. The cached data requests from the processing unit or the master control module enter the control module in turn after arbitration, and the control module judges whether the data is available and requests the data. The party sends a feedback signal. If the data is available, the control module will also send the address of the data in the storage module, and the data requester can read and write the storage module through the address. If the data is not available, the control module sends a request to the master control module, and the data is finally copied to the storage module by the direct memory access (DMA) module receiving the data sent by the sniffing module from the main memory or other cache modules. The sniffing module receives the data sniffing instruction from the master control module, reads the corresponding data in the storage module and outputs it to other cache modules that request data, and at the same time, the data can also be written into the main memory.

In addition, the structure of the master control module is shown in Figure 6. After the data requests from multiple cache modules are arbitrated, the highest priority data search request is input into the sniffing module of each cache module, and the request corresponds to The data flag is pending until at least one cache module completes the transfer of the data. Before that, the same data request will not be preferentially processed by the arbiter due to the flag being pending.

The cache system applied to the distributed processing unit provided by this implementation realizes data sharing among multiple cache modules. Its general control module, cache control module and sniffing module take into account the adaptability with commonly used cache coherence protocols and several variants thereof, which can effectively improve the efficiency of data cache when multiple processing units perform parallel computing.

The cache system applied to the distributed processing unit provided by this embodiment is composed of a general control module and a plurality of cache modules, wherein each cache module includes: a control module, a storage module and a sniffing module; a control module, It is used to receive and process data requests; the storage module is used to cache data; the sniffing module is used to receive and execute data consistency instructions. Through the master control module, the control module and the sniffing module, the adaptability with the MESI data consistency protocol and its variants is ensured, and the efficiency of data cache can be effectively improved in parallel computing.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein. The solutions in the embodiments of the present application may be implemented in various computer languages, for example, the object-oriented programming language Java and the literal translation scripting language JavaScript, and the like.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

While the preferred embodiments of the present application have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be construed to include the preferred embodiment and all changes and modifications that fall within the scope of this application.

Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims (13)

1. a cache module applied to a distributed processing unit, wherein the cache module comprises: a control module, a storage module and a sniffing module; The control module is used to receive and process the data request; The storage module is used to cache data; The sniffing module is used for receiving and executing data consistency instructions. 2 . The cache module according to claim 1 , wherein the control module is configured to receive a data request from a processing unit, and determine whether the requested data is stored in the storage module according to the data request. 3 . 3. The high-speed cache module according to claim 2, wherein the control module is used to return a hit signal after judging that the requested data is stored in the storage module, and read from the storage module the request data, and process the request data. 4 . The cache module according to claim 3 , wherein the control module is configured to modify the status identifier of the request data. 5 . 5 . The cache module according to claim 2 , wherein the control module is configured to acquire the request data from other cache modules after judging that the request data is not stored in the storage module. 6 . 6. The cache module according to claim 5, wherein the control module is used for judging that the request data is not stored in the storage module, and after the request data is not obtained from other cache modules , read the request data from external main memory. 7. The cache module according to claim 5 or 6, wherein the control module is configured to store the request data in the storage module. 8. The cache module according to claim 1, wherein the data consistency instruction comprises one or more of the following: an instruction to copy and send data in the storage module to other cache modules, An instruction to mark the copy of the data in the storage module as unavailable, and an instruction to write the data in the storage module back to main memory. 9. The cache module of claim 8, wherein copies of data that have been modified in other cache modules are marked as unavailable. 10. The cache module of claim 1, wherein the processing unit comprises a plurality of modules; The control module is used for receiving data requests from multiple modules in the processing unit, arbitrating the priority of each module, and processing each data request according to the arbitration result. 11. The cache module according to claim 3, wherein the sniffing module is configured to copy the storage module according to the instruction of copying and sending data in the storage module to other cache modules the requested data in the module, and the copied data is written into the storage module in the other cache module. 12. A cache system applied to a distributed processing unit, wherein the cache system comprises: a general control module and a plurality of cache modules; The general control module is used to control the consistency of data transmission between the cache modules; The cache module is the cache module according to any one of claims 1-11. 13. The cache system according to claim 12, wherein the master control module is used to receive the data request of any cache module, generate copy and send the data in the storage module to other caches The instruction of the module sends the instruction of copying and sending the data in the storage module to other cache modules to other cache modules.

Images