CN109298839A - Storage controller, storage device, system and method based on PIS - Google Patents

Storage controller, storage device, system and method based on PIS Download PDF

Info

Publication number
CN109298839A
CN109298839A CN201811261719.5A CN201811261719A CN109298839A CN 109298839 A CN109298839 A CN 109298839A CN 201811261719 A CN201811261719 A CN 201811261719A CN 109298839 A CN109298839 A CN 109298839A
Authority
CN
China
Prior art keywords
data
host
controller
storage
sent
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811261719.5A
Other languages
Chinese (zh)
Inventor
杨庆
李卫军
陈祥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Dapu Microelectronics Co Ltd
Original Assignee
Shenzhen Dapu Microelectronics Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Dapu Microelectronics Co Ltd filed Critical Shenzhen Dapu Microelectronics Co Ltd
Priority to CN201811261719.5A priority Critical patent/CN109298839A/en
Publication of CN109298839A publication Critical patent/CN109298839A/en
Priority to PCT/CN2019/095197 priority patent/WO2020082813A1/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601Interfaces specially adapted for storage systems
    • G06F3/0602Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
    • G06F3/061Improving I/O performance
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601Interfaces specially adapted for storage systems
    • G06F3/0628Interfaces specially adapted for storage systems making use of a particular technique
    • G06F3/0655Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices
    • G06F3/0658Controller construction arrangements
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601Interfaces specially adapted for storage systems
    • G06F3/0668Interfaces specially adapted for storage systems adopting a particular infrastructure
    • G06F3/0671In-line storage system
    • G06F3/0683Plurality of storage devices
    • G06F3/0688Non-volatile semiconductor memory arrays

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Advance Control (AREA)
  • Memory System (AREA)

Abstract

The present invention relates to a kind of storage controllers based on PIS, storage device, system and method, including processor, for controlling the operation of storage device, and can send the first pre-processing instruction to data processing unit according to the operational order that host side is sent;Host-side interface connects the storage controller and the host side by the host-side interface;Flash controller carries out data transmission for the storage controller with flash array;Data processing unit, the first pre-processing instruction for being sent according to processor, pre-processes the data stored in flash array, and pretreated data are sent to host side by host-side interface;Above-mentioned host-side interface, flash controller, data processing unit are connect with processor, and the data processing unit is connect with the host-side interface and the flash controller respectively.Inventive memory device can storing data and data calculate, data bus resource occupancy is effectively reduced.

Description

Storage controller, storage device, system and method based on PIS
Technical field
The present invention relates to a kind of storage controllers based on PIS, storage device, system and method.
Background technique
As shown in Figure 1, the CPU and memory of traditional Feng's Neumann computer are independently arranged, led to by bus Letter.As shown in Fig. 2, traditional memory (such as SSD) includes controller and multiple flash, controller includes processor, flash Instruction is transmitted to processor by interface by controller, internal bus, buffer and interface etc., host, and processor passes through internal total Instruction is distributed to flash controller and buffer by line, and instruction is transmitted to flash by internal bus by flash controller, Flash is received and is read and write data from flash particle after instruction, then by internal bus data is transmitted to buffer, by buffer Host is transferred data to by interface, or interface is directly transferred data to by Flash controller, then passes through interface Transfer data to host.Traditional memory further includes ECC, and the data for reading and writing flash controller carry out error correction, so After be then forwarded to flash controller.
The CPU and memory of traditional computer are independent two disparate modules, and computer is responsible for calculating, and memory is responsible for Storage, when carrying out big data analysis, mass data first will be again and again from the memory that memory is transmitted to CPU, then carries out Data Analysis Services.The fast development of growth and artificial intelligence recently as big data business, data volume is increasing, needle To the data-intensive applications of these data, the amount of analysis of CPU is also increasing, if according to traditional memory and CPU The mode being independently arranged with memory, it is necessary to which the bus of more high bandwidth is just able to achieve increasing data volume from memory Be transmitted to CPU, but the bus of traditional computer is STD bus, bandwidth be it is fixed, therefore, in Data-intensive computing In artificial intelligence application, the bandwidth of standard does not catch up with data volume and amount of analysis, it is impossible to meet demand.In addition, due to data Measure it is increasing, if data are transmitted by traditional STD bus, by the energy of drain bus significantly.Moreover, because data It measures increasing, the big quantity space of CPU will be occupied, influence the speed of service of CPU.
Summary of the invention
Goal of the invention of the invention is to provide a kind of storage controller based on PIS, storage device, system and side Method, storage device not only can store data, but also can carry out data calculating, and the resources occupation rate of data/address bus is effectively reduced, into And improve the overall performance of computing system.
Based on the same inventive concept, there are four independent technical solutions for present invention tool:
1, a kind of storage controller based on PIS, comprising:
Processor, for controlling the operation of storage device, and can be according to the operational order that host side is sent to data processing Unit sends the first pre-processing instruction;
Host-side interface connects the storage controller and the host side by the host-side interface;
Flash controller carries out data transmission for the storage controller with flash array;
Data processing unit, the first pre-processing instruction for being sent according to processor, to the number stored in flash array According to being pre-processed, pretreated data are sent to host side by host-side interface;
Above-mentioned host-side interface, flash controller, data processing unit are connect with processor, the data processing unit It is connect respectively with the host-side interface and the flash controller.
Further, the processor can be sent according to the operational order that the host side is sent to the flash controller Storing data to be pre-treated is sent to by the second pre-processing instruction, the flash controller according to second pre-processing instruction The data processing unit.
Further, the flash controller includes Data Management Unit, and the Data Management Unit will be for that will store number According to being encoded and carry out transfer management;The flash controller is according to the second pre-processing instruction, by storage number to be pre-treated According to after Data Management Unit coding and transfer management, it is sent to the data processing unit.
Further, the host-side interface, flash controller, data processing unit are connected by AXI bus and processor It connects, the data processing unit is connected respectively at the host-side interface with the flash controller by AXI bus.
2, a kind of storage device, including the above-mentioned storage controller based on PIS.
3, a kind of storage system, host side connect one or more storage devices, described storage device packet by data/address bus Include the above-mentioned storage controller based on PIS.
4, a kind of data preprocessing method of storage device, the operational order that storage device receiving host end is sent, judgement Whether operational order includes pre-processing instruction, if it is, after storage device pre-processes storing data, then will pretreatment Storing data afterwards is sent to host side;If it is not, then storing data is directly sent to host side by storage device.
The invention has the benefit that
Inventive memory device controller (DPU) not only can control the operation of storage device, but also can carry out to the data of storage Calculation processing, that is, storage device not only can store data, but also can carry out data calculating, existing architectural framework be breached, in number According to computing capability is introduced in storage processor, realize PIS (Processing in Storage).Processor combination number of the present invention Firmware is handled according to processing unit or processor integrated data, so that completing in the nearest place of range data to storing data Pretreatment, and treated data are sent into host, greatly reduce the load and energy consumption of the buses such as PCIe, while releasable CPU frees the CPU of host side from heavy repeated labor, improves the CPU speed of service, improves big data processing With analysis ability, and the resources occupation rate of I/O bus is reduced, greatlys improve the overall performance of computing system, can sufficiently meet Magnanimity, heterogeneous, non-structured big data, cloud storage requirement.
As shown in fig. 6,2.1GB data input keyword " Chinese " search that will be downloaded from wiki (wikipedia) In operation test, 9.2s is needed using the computer processing time of normal memory, and occupy CPU up to 55%;And pass through this hair Bright DPU (Nidal ADA Technology), the processing time only needs 6s, and only takes up CPU 5% or so.As shown in fig. 7, passing through Data volume after the computer disposal of DPU of the present invention is only that treated the 3% of data volume for common computer, is greatly reduced It is transferred to the data volume of host, therefore, greatly reduces CPU occupation proportion, while greatly reducing bus load and power consumption.
Detailed description of the invention
Fig. 1 is the data communication schematic diagram of existing host side and memory;
Fig. 2 is the structure principle chart of existing memory;
Fig. 3 is the structure principle chart of inventive memory device;
Fig. 4 is the system architecture diagram of storage system of the present invention;
Fig. 5 is the data preprocessing method flow chart of inventive memory device;
Fig. 6 is the host side CPU accounting comparison signal of inventive memory device controller and existing storage controller Figure;
Fig. 7 is the host side CPU processing data volume pair of inventive memory device controller and existing storage controller Compare schematic diagram.
Specific embodiment
The present invention is described in detail for each embodiment shown in reference to the accompanying drawing, but it should be stated that, these Embodiment is not limitation of the present invention, those of ordinary skill in the art according to these embodiments made by function, method, Or equivalent transformation or substitution in structure, all belong to the scope of protection of the present invention within.
Embodiment one:
Storage controller based on PIS:
As shown in figure 3, the storage controller (DPU) based on PIS includes:
Processor, for controlling the operation of storage device, and can be according to the operational order that host side is sent to data processing Unit sends the first pre-processing instruction;
Host-side interface connects storage controller and host side by host-side interface;The interface of DPU uses PCIe Interface or NVMe interface etc..
Flash controller (falsh controller) carries out data biography for storage controller and flash array (flash) It is defeated;
Data processing unit (ADA), the first pre-processing instruction for being sent according to processor, to being stored in flash array Data pre-processed, pretreated data are sent to host side by host-side interface;
Above-mentioned host-side interface, flash controller, data processing unit are connect with processor, data processing unit difference It is connect with host-side interface and flash controller.
When it is implemented, processor is connect by internal bus with flash controller, ADA and host-side interface, flash Controller also passes through internal bus and connect with flash, and ADA also passes through internal bus and connect with interface.The internal bus of DPU uses AXI (Advanced eXtensible Interface) bus, has the characteristics that high-performance, high bandwidth, low latency, can meet Performance needed for DPU internal transmission.
Processor can be central processing unit (Central Processing Unit, CPU), network processing unit (network processor, NP), general processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit (Application-Specific Integrated Circuit, ASIC), field programmable gate Array (Field Programmable Gate Array, FPGA) or other programmable logic device, transistor logic device Part, hardware component or any combination thereof.In the case where processor is a CPU, which can be monokaryon CPU, can also be with It is multi-core CPU.
Data processing unit includes text search unit and/or Data Format Transform unit, and text search unit is for real When full-text search is carried out to data stream, Data Format Transform unit is for be converted into storing data can computational format.Text is searched Rope (GREP), Data Format Transform (SCANF), classification (Sorting), compression, decompression etc..GREP and SCANF is big data Using widest two kinds of Data-intensive computings in the middle.Since storage calculates the scalability of PIS framework in fusion chip, after It is continuous to realize computing function in other disks according to different industries to the specific requirements of calculating.Data Format Transform (SCANF) Be one by storing data by certain readable format (such as ASCII, Unicode) be converted to can computational format (such as binary system) mistake Journey.Steps are as follows for traditional data calculating: (1) initial data being read into host memory from storage device;(2) host CPU Initial data in memory is formatted, and is saved in another position of memory;(3) after application program is to conversion Data are operated.Clearly as CPU and caching participate in a large amount of simple and duplicate data exchanges and calculating, it is above-mentioned The efficiency of data usage mode is very low, and the load of rambus and I/O bus is very heavy.Many numerical value are calculated and are appointed For business, the work of above-mentioned Data Format Transform is necessary and occupies for up to 60% execution time.
Flash controller includes Data Management Unit (ADM), and Data Management Unit is for being encoded simultaneously storing data Carry out transfer management;Flash controller is according to third pre-processing instruction, by storing data to be pre-treated through Data Management Unit After coding and transfer management, it is sent to data processing unit.
When the data that host needs to transfer flash are analyzed, processor is sent an instruction to by host-side interface, After processor receives operational order, judge in received operational order whether to include pre-processing instruction, such as include, processor is by the One pre-processing instruction is sent to ADA and the second pre-processing instruction is sent to flash controller, and flash controller receives second Data are read and write from flash particle after pre-processing instruction, then required data are pre-processed by ADA, after pretreatment, will be handled Data afterwards are sent to interface by internal bus, send the data to host by host interface, so that host is further located Reason.When flash controller includes Data Management Unit (ADM), from flash after flash controller the second pre-processing instruction of reception Particle reads and writes data, then by ADM data are encoded and carried out with the transfer management of data, then again by ADA to required Data are pre-processed, and after pretreatment, treated data by internal bus are sent to interface, are sent out data by interface Host is given, so that host is further processed.
Further, operational order has identification code, and identification code includes pre-processing instruction, passes through identification code and storage device In pre-stored identification code be compared, compare successfully, then start pretreatment, compare failure, then do not start pretreatment, directly Send the data to host.
Embodiment two:
Storage controller based on PIS:
Embodiment two and the difference of embodiment one essentially consist in, in embodiment two, ADA and processor as integral structure, That is ADA is poured into a software form in the firmware of processor, is further reduced the setting of DPU internal bus pin.That is, processor packet Data processing firmware is included, data processing firmware is for pre-processing storing data.Remaining working principle of embodiment two is same Embodiment one.
Embodiment three:
Storage device:
Described storage device includes the storage controller (DPU) based on PIS of embodiment one or embodiment two.
Example IV:
Storage system:
As shown in figure 4, host side connects multiple storage devices by data/address bus, described storage device include embodiment one or The storage controller (DPU) based on PIS of embodiment two.Storage device (memory) is set on the slot of host, is led to It crosses and multiple storage devices is set, mass data can be handled.
Embodiment five:
The data preprocessing method of storage device:
As shown in figure 5, whether the operational order (data command) that storage device receiving host end is sent, judge operational order Comprising pre-processing instruction, if it is, after storage device pre-processes storing data, then by pretreated storing data It is sent to host side;If it is not, then storing data is directly sent to host side by storage device.
Pretreatment include in real time to data stream carry out full-text search, by storing data be converted into can computational format, classification (Sorting), compress, decompress etc..
Judge that operational order whether include the method for pre-processing instruction includes: that host sends and refers to the operation of identification code It enabling, identification code includes pre-processing instruction, it is compared by identification code with pre-stored identification code in storage storage device, than To success, then start pretreatment, compares failure, then do not start pretreatment, directly send the data to host.
It should be appreciated that device and method disclosed in the above embodiment provided herein, can also pass through it Its mode realizes that the apparatus embodiments described above are merely exemplary.The foregoing is merely specific realities of the invention Mode is applied, but scope of protection of the present invention is not limited thereto, anyone skilled in the art takes off in the present invention It in the technical scope of dew, can easily think of the change or the replacement, should be covered by the protection scope of the present invention.Therefore, this hair Bright protection scope should be subject to the protection scope in claims.

Claims (7)

1. a kind of storage controller based on PIS characterized by comprising
Processor, for controlling the operation of storage device, and can be according to the operational order that host side is sent to data processing unit Send the first pre-processing instruction;
Host-side interface connects the storage controller and the host side by the host-side interface;
Flash controller carries out data transmission for the storage controller with flash array;
Data processing unit, the first pre-processing instruction for being sent according to processor, to the data stored in flash array into Pretreated data are sent to host side by host-side interface by row pretreatment;
Above-mentioned host-side interface, flash controller, data processing unit are connect with processor, the data processing unit difference It is connect with the host-side interface and the flash controller.
2. the storage controller according to claim 1 based on PIS, it is characterised in that: the processor can basis The operational order that the host side is sent sends the second pre-processing instruction to the flash controller, the flash controller according to Storing data to be pre-treated is sent to the data processing unit by second pre-processing instruction.
3. the storage controller according to claim 2 based on PIS, it is characterised in that: the flash controller packet Data Management Unit is included, the Data Management Unit by storing data for being encoded and carrying out transfer management;The flash memory Controller is according to the second pre-processing instruction, by storing data to be pre-treated after Data Management Unit coding and transfer management, It is sent to the data processing unit.
4. the storage controller according to claim 1 based on PIS, it is characterised in that: the host-side interface, sudden strain of a muscle Memory controller, data processing unit are connect by AXI bus with processor, and the data processing unit is distinguished by AXI bus It is connected in the host-side interface with the flash controller.
5. a kind of storage device, it is characterised in that: including described in Claims 1-4 any one based on the storage device of PIS Controller.
6. a kind of storage system, it is characterised in that: host side connects one or more storage devices, the storage by data/address bus Device includes the storage controller described in Claims 1-4 any one based on PIS.
7. a kind of data preprocessing method of storage device, it is characterised in that: the operation that storage device receiving host end is sent refers to It enables, judges whether operational order includes pre-processing instruction, if it is, after storage device pre-processes storing data, then Pretreated storing data is sent to host side;If it is not, then storing data is directly sent to host side by storage device.
CN201811261719.5A 2018-10-26 2018-10-26 Storage controller, storage device, system and method based on PIS Pending CN109298839A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201811261719.5A CN109298839A (en) 2018-10-26 2018-10-26 Storage controller, storage device, system and method based on PIS
PCT/CN2019/095197 WO2020082813A1 (en) 2018-10-26 2019-07-09 Pis-based memory device controller, memory device, system, and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811261719.5A CN109298839A (en) 2018-10-26 2018-10-26 Storage controller, storage device, system and method based on PIS

Publications (1)

Publication Number Publication Date
CN109298839A true CN109298839A (en) 2019-02-01

Family

ID=65158987

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811261719.5A Pending CN109298839A (en) 2018-10-26 2018-10-26 Storage controller, storage device, system and method based on PIS

Country Status (2)

Country Link
CN (1) CN109298839A (en)
WO (1) WO2020082813A1 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020082813A1 (en) * 2018-10-26 2020-04-30 深圳大普微电子科技有限公司 Pis-based memory device controller, memory device, system, and method
CN111949211A (en) * 2020-07-10 2020-11-17 深圳宏芯宇电子股份有限公司 Storage device and storage control method
CN112181293A (en) * 2020-09-17 2021-01-05 深圳大普微电子科技有限公司 Solid state disk controller, solid state disk, storage system and data processing method
WO2023060891A1 (en) * 2021-10-14 2023-04-20 华为技术有限公司 Data processing method, apparatus and system, and device, storage medium and program product
WO2023072048A1 (en) * 2021-10-27 2023-05-04 华为技术有限公司 Network storage method, storage system, data processing unit, and computer system

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102760045B (en) * 2011-04-29 2015-07-08 无锡江南计算技术研究所 Intelligent storage device and data processing method thereof
JP5853899B2 (en) * 2012-03-23 2016-02-09 ソニー株式会社 Storage control device, storage device, information processing system, and processing method therefor
CN104834484B (en) * 2015-05-11 2018-10-23 上海新储集成电路有限公司 Data processing system based on embedded programmable logic array and processing method
CN107870878A (en) * 2017-10-31 2018-04-03 深圳清华大学研究院 Storage system, terminal and computer installation
CN209044575U (en) * 2018-10-26 2019-06-28 深圳大普微电子科技有限公司 Storage controller, storage device and system based on PIS
CN109298839A (en) * 2018-10-26 2019-02-01 深圳大普微电子科技有限公司 Storage controller, storage device, system and method based on PIS

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020082813A1 (en) * 2018-10-26 2020-04-30 深圳大普微电子科技有限公司 Pis-based memory device controller, memory device, system, and method
CN111949211A (en) * 2020-07-10 2020-11-17 深圳宏芯宇电子股份有限公司 Storage device and storage control method
US11609713B2 (en) 2020-07-10 2023-03-21 Hosin Global Electronics Co., Ltd Storage device and storage control method
CN112181293A (en) * 2020-09-17 2021-01-05 深圳大普微电子科技有限公司 Solid state disk controller, solid state disk, storage system and data processing method
CN112181293B (en) * 2020-09-17 2023-08-22 深圳大普微电子科技有限公司 Solid state disk controller, solid state disk, storage system and data processing method
WO2023060891A1 (en) * 2021-10-14 2023-04-20 华为技术有限公司 Data processing method, apparatus and system, and device, storage medium and program product
WO2023072048A1 (en) * 2021-10-27 2023-05-04 华为技术有限公司 Network storage method, storage system, data processing unit, and computer system

Also Published As

Publication number Publication date
WO2020082813A1 (en) 2020-04-30

Similar Documents

Publication Publication Date Title
CN109298839A (en) Storage controller, storage device, system and method based on PIS
JP4128956B2 (en) Switch / network adapter port for cluster computers using a series of multi-adaptive processors in dual inline memory module format
DE102018004327A1 (en) Systems and methods for accessing mass storage as working memory
US7934025B2 (en) Content terminated DMA
US8990451B2 (en) Controller for direct access to a memory for the direct transfer of data between memories of several peripheral devices, method and computer program enabling the implementation of such a controller
US20230017643A1 (en) Composable infrastructure enabled by heterogeneous architecture, delivered by cxl based cached switch soc
CN101030182A (en) Apparatus and method for performing dma data transfer
JP2017211984A (en) METHOD, SYSTEM AND APPARATUS FOR QoS-AWARE INPUT/OUTPUT MANAGEMENT FOR PCIe STORAGE SYSTEM WITH RECONFIGURABLE MULTI-PORTS
CN102841870A (en) General direct memory access (DMA) structure based on high-speed serial bus and pre-read method
US11082327B2 (en) System and method for computational transport network-on-chip (NoC)
US20220171628A1 (en) Techniques to encode and decode for character class matching
CN117631974A (en) Access request reordering across a multi-channel interface of a memory-based communication queue
CN209044575U (en) Storage controller, storage device and system based on PIS
CN103268299A (en) Universal data compression intellectual property (IP) core for PXI Express bus test system
US9053092B2 (en) System authorizing direct data transfers between memories of several components of that system
CN109739785A (en) The internal connection-wire structure of multiple nucleus system
CN105320630A (en) Heterogeneous multi-core CPU-GPU (Central Processing Unit-Graphics Processing Unit) system architecture based on intelligent flash cache
CN110750210B (en) Storage system
CN110287141A (en) A kind of FPGA reconstructing method and system based on multiple interfaces
CN113688093A (en) Intelligent network card based on Ethernet controller
CN113961489A (en) Data access method, device, equipment and storage medium
CN112232523A (en) Domestic artificial intelligence computing equipment
Zhang et al. A4-Way CC-NUMA System Based on S2500
EP4355047A2 (en) Technology for early abort of compression acceleration
CN110597756B (en) Calculation circuit and data operation method

Legal Events

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