CN216388083U - DPU device based on FPGA - Google Patents

DPU device based on FPGA Download PDF

Info

Publication number
CN216388083U
CN216388083U CN202122964619.2U CN202122964619U CN216388083U CN 216388083 U CN216388083 U CN 216388083U CN 202122964619 U CN202122964619 U CN 202122964619U CN 216388083 U CN216388083 U CN 216388083U
Authority
CN
China
Prior art keywords
module
fpga
capacitor
memory
interface
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.)
Expired - Fee Related
Application number
CN202122964619.2U
Other languages
Chinese (zh)
Inventor
严和平
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.)
Shanghai Xinpu Information Technology Co ltd
Original Assignee
Shanghai Xinpu Information Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Xinpu Information Technology Co ltd filed Critical Shanghai Xinpu Information Technology Co ltd
Priority to CN202122964619.2U priority Critical patent/CN216388083U/en
Application granted granted Critical
Publication of CN216388083U publication Critical patent/CN216388083U/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Logic Circuits (AREA)

Abstract

本新型公开了一种基于FPGA的DPU装置,包括:FPGA模块:接收接口模块的外部数据进行运算并将运算结果通过接口模块发送至外部设备;RISC模块:接收FPGA模块运算时的内存管理信号发送至内存管理模块;内存管理模块:管理内存模块与FPGA模块间的数据传输通道;内存模块:包括多个内存组,用于存储FPGA模块进行运算时的临时数据;接口模块:用于提供装置与外部设备的数据互交接口:时序模块:用于产生FPGA模块工作所需的时钟信号;电源模块:为整个DPU装置提供所需工作电压,有效控制内存组与FPGA模块进行数据互交,提高了FPGA模块的数据处理效率,在FPGA模块计算负载过大时保证了溢出数据的安全稳定,并能为FPGA的提供多路计算通道保证的多项目计算正常运行。

Figure 202122964619

The new model discloses an FPGA-based DPU device, comprising: an FPGA module: receiving external data of an interface module for operation and sending the operation result to an external device through the interface module; RISC module: receiving a memory management signal when the FPGA module operates To the memory management module; the memory management module: manages the data transmission channel between the memory module and the FPGA module; the memory module: includes a plurality of memory groups, used to store the temporary data when the FPGA module performs operations; the interface module: used to provide the device with Data interaction interface of external devices: timing module: used to generate the clock signal required for the FPGA module to work; power module: to provide the required operating voltage for the entire DPU device, effectively control the memory bank and the FPGA module for data exchange, and improve the The data processing efficiency of the FPGA module ensures the safety and stability of the overflow data when the calculation load of the FPGA module is too large, and can provide the multi-project calculation of the FPGA with the guarantee of multiple calculation channels.

Figure 202122964619

Description

DPU device based on FPGA
Technical Field
The application relates to the field of embedded technology, in particular to a DPU device based on an FPGA.
Background
The DPU is an artificial intelligence big data analysis system project. It consists of a series of pipeline algorithms, from sorting, indexing, grouping and optimization to processing and analyzing matrix and tensor large data. The sort and index pipeline arranges the data in order, providing a descriptive basis for better understanding of the order of things that occur in all dimensions. Packet pipelines place data in proportion or range, generating a diagnostic basis for better signal detection, finding useful structures, defining logical or salient features. The scale is inherent to the data structure, and the range is a choice of human preference. Optimizing the pipeline ultimately makes the data valuable, extracting intelligent information from the data and building predictive models from which we can infer what will happen in the future. In the process, a large amount of data calculation processing is required, and the requirements on processing speed and stability of data in the memory are high. The traditional memory architecture of the DPU device generally needs a plurality of clocks to finish quick access to uncleaned data and derivative arrays, and the data processing speed of the DPU is seriously influenced.
Accordingly, a solution is desired to solve or at least mitigate the above-mentioned deficiencies of the prior art.
SUMMERY OF THE UTILITY MODEL
The present invention is directed to a DPU device based on an FPGA to solve at least one of the above problems.
An FPGA-based DPU device, comprising: an FPGA module: receiving external data of the interface module, performing operation and sending an operation result to external equipment through the interface module; RISC module: receiving a memory management signal when the FPGA module operates, and sending the memory management signal to a memory management module; a memory management module: the system comprises at least two triggers, wherein the triggers respectively receive memory control signals of a RISC module and an FPGA module and manage a data transmission channel between the memory module and the FPGA module; a memory module: the FPGA module comprises a plurality of memory groups and a plurality of storage modules, wherein the memory groups are used for storing temporary data when the FPGA module performs operation; an interface module: the data interaction interface for providing the device and the external equipment comprises: a time sequence module: the FPGA module comprises at least two counters and a clock signal generator, wherein the counters are used for generating clock signals required by the operation of the FPGA module; a power supply module: providing the required operating voltage for the entire DPU device.
Further, the memory management module specifically includes: the circuit comprises a first trigger U3 and a second trigger U4, wherein the enable end of the first trigger U3 is connected with the memory control end of the RISC module, the input pins of the first trigger U3 are respectively connected with the pins of the memory selection signal end of the RISC module, and the output pins of the first trigger U are grounded through a forward diode and a resistor; the enabling end of the second trigger U4 is connected with the memory control end of the FPGA module, the input end of the second trigger U4 is connected with the reference voltage, and the output end of the second trigger U4 is connected with the control end of each memory group through a forward diode; the anodes of the diodes of the first flip-flop U3 are connected to the anodes of the corresponding diodes of the second flip-flop U4.
Further, the timing module specifically includes: the circuit comprises a crystal oscillator Y1, wherein two ends of the crystal oscillator Y1 are grounded through a capacitor C1 and a capacitor C2 respectively, and a resistor R1 is connected between two ends of the crystal oscillator Y1; one end of the crystal oscillator Y1 is connected with the RST pin of the first counter U1, and the other end of the crystal oscillator Y1 is connected with the PI pin of U1; the output pin Q8 of the first counter U1 is connected to the clkb pin of the second counter U2, and the qa pin of the second technician outputs the clock signal.
Further, the power module comprises a transformer and a rectifier which are connected in sequence, and external alternating current is rectified and transformed and then is transmitted to the first power module, the second power module and the third power module.
Further, the first power module includes a chip U5, an input terminal of which is grounded through a forward polarity capacitor C3, an output terminal of which is grounded through a capacitor C7, a capacitor C8 and a forward polarity capacitor C5, respectively, an output terminal of which outputs a first operating voltage through an inductor L1, and an output terminal of which is grounded through a capacitor C4 and a forward polarity capacitor C6, respectively.
Further, the second power module includes a chip U6, an input terminal of which is grounded through a forward polarity capacitor C9, a ground terminal of which is grounded through a resistor R11, a resistor R10 is connected between an output terminal of which and the ground terminal, an output terminal of which outputs a second operating voltage, and output terminals of which are grounded through a capacitor C11, a capacitor C12 and a forward polarity capacitor C10, respectively; the negative terminal of the polar capacitor C10 is connected to the digital block ground AGND via the inductor L2.
Further, the third power module includes a chip U7, an output terminal of which is grounded through a forward polarity capacitor C13, a ground terminal of which is grounded through a resistor R14, a resistor R12 is connected between the output terminal and the ground terminal, the output terminal of which outputs a third operating voltage, and the output terminals of which are grounded through a resistor R13, a capacitor C15, a capacitor C16, and a forward polarity capacitor C14, respectively.
Further, the interface module includes: PCIe interface, QSFP interface, SSD interface, USB interface, SSD interface, GigE interface and GPIO interface.
Furthermore, the FPGA module and the RISC module are also connected with a safety module.
Furthermore, the FPGA module is also connected with a JTAG module.
Advantageous effects
When the application is used for actually carrying out data operation, the internal memory trigger signal and the control signal which are provided by the FPGA module during operation are processed by the internal instruction set of the RISC module to generate the internal memory selection signal, and then the internal memory group in the internal memory module is controlled by the internal memory management module to carry out data interaction with the FPGA module, so that the data processing efficiency of the FPGA module is effectively improved, the safety and the stability of overflowing data are ensured when the calculation load of the FPGA module is overlarge, and the normal operation of multi-project calculation which is ensured by a multi-path calculation channel can be provided for the FPGA.
The application also provides a plurality of external data interfaces to meet the requirements of various data working environments, and normal use of each internal module is ensured through multi-stage power output.
Drawings
Fig. 1 is a schematic diagram of the present application.
Fig. 2 is a schematic circuit diagram of a memory management module according to an embodiment of the present application.
Fig. 3 is a schematic circuit diagram of a sequential module according to an embodiment of the present application.
Fig. 4 is a circuit schematic diagram of a first power module according to an embodiment of the present application.
Fig. 5 is a circuit schematic diagram of a second power module according to an embodiment of the present application.
Fig. 6 is a circuit schematic diagram of a third power module according to an embodiment of the present application.
Detailed Description
In order to make the implementation objects, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be described in more detail below with reference to the drawings in the embodiments of the present application. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are a subset of the embodiments in the present application and not all embodiments in the present application. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application. Embodiments of the present application will be described in detail below with reference to the accompanying drawings.
As shown in fig. 1, an FPGA-based DPU apparatus includes: an FPGA module: receiving external data of the interface module, performing operation and sending an operation result to external equipment through the interface module; RISC module: receiving a memory management signal when the FPGA module operates, and sending the memory management signal to a memory management module; a memory management module: the system comprises at least two triggers, wherein the triggers respectively receive memory control signals of a RISC module and an FPGA module and manage a data transmission channel between the memory module and the FPGA module; a memory module: the FPGA module comprises a plurality of memory groups and a plurality of storage modules, wherein the memory groups are used for storing temporary data when the FPGA module performs operation; an interface module: the data interaction interface for providing the device and the external equipment comprises: a time sequence module: the FPGA module comprises at least two counters and a clock signal generator, wherein the counters are used for generating clock signals required by the operation of the FPGA module; a power supply module: the power supply device provides required working voltage for the whole DPU device, comprises a transformer and a rectifier which are connected in sequence, and transmits external alternating current to a first power module, a second power module and a third power module after rectifying and transforming the external alternating current. The FPGA module and the RISC module are also connected with a safety module. The FPGA module is also connected with a JTAG module. And the safety module is used for detecting the working states of the FPGA module and the RISC module and timely informing the device of emergency treatment when the working states are abnormal. The JTAG module is used for presetting working states for the FPGA module so as to meet different calculation requirements.
As shown in fig. 2, the memory management module includes a first flip-flop U3 and a second flip-flop U4, an enable terminal of the first flip-flop U3 is connected to a memory control terminal of the RISC module, input pins of the first flip-flop U3 are respectively connected to pins of a memory selection signal terminal of the RISC module, and pins of an output terminal of the first flip-flop U are grounded through a forward diode and a resistor; the enabling end of the second trigger U4 is connected with the memory control end of the FPGA module, the input end of the second trigger U4 is connected with the reference voltage, and the output end of the second trigger U4 is connected with the control end of each memory group through a forward diode; the anodes of the diodes of the first flip-flop U3 are connected to the anodes of the corresponding diodes of the second flip-flop U4. The first flip-flop U3 and the second flip-flop U4 both use eight flip-flops, that is, the number of memory banks in the corresponding memory module is eight. The memory group adopts DDR4/QDDR + specification.
As shown in fig. 3, the timing module includes a crystal oscillator Y1, two ends of which are grounded through a capacitor C1 and a capacitor C2, respectively, and a resistor R1 is connected between the two ends of the crystal oscillator Y1; one end of the crystal oscillator Y1 is connected with the RST pin of the first counter U1, and the other end of the crystal oscillator Y1 is connected with the PI pin of U1; the output pin Q8 of the first counter U1 is connected to the clkb pin of the second counter U2, and the qa pin of the second technician outputs the clock signal. The two counters form a frequency division circuit to process the original pulse signal of the crystal oscillator Y1 so as to meet the working requirement of the FPGA module. When necessary, a plurality of groups of frequency division circuits can be arranged to generate a plurality of groups of time sequence signals, and the application environment of the FPGA module is further improved by controlling the required time sequence signals through the multiplexer.
As shown in fig. 4, the first power module includes a chip U5, an input terminal of which is grounded through a forward polarity capacitor C3, an output terminal of which is grounded through a capacitor C7, a capacitor C8 and a forward polarity capacitor C5, respectively, an output terminal of which outputs a first operating voltage through an inductor L1, and an output terminal of which is grounded through a capacitor C4 and a forward polarity capacitor C6, respectively. The polar capacitors C3, C5 and C6 are all 10 muF, the capacitor C7 is 0.01 muF, the capacitor C8 and the capacitor C4 are 0.1 muF, and the output first working voltage is 1.2V.
As shown in fig. 5, the second power module includes a chip U6, an input terminal of which is grounded through a forward polarity capacitor C9, a ground terminal of which is grounded through a resistor R11, a resistor R10 connected between an output terminal of which and the ground terminal, an output terminal of which outputs the second operating voltage, and output terminals of which are grounded through a capacitor C11, a capacitor C12 and a forward polarity capacitor C10, respectively; the negative terminal of the polar capacitor C10 is connected to the digital block ground AGND via the inductor L2. The polar capacitors C9 and C10 are 10 muF, the capacitor C11 is 0.01 muF, the capacitor C12 is 0.1 muF, the resistor R11 and the resistor R10 are 100 omega, and the output second working voltage is 2.5V.
As shown in fig. 6, the third power module includes a chip U7, an output terminal of which is grounded through a forward polarity capacitor C13, a ground terminal of which is grounded through a resistor R14, a resistor R12 connected between the output terminal and the ground terminal of which outputs a third operating voltage, and output terminals of which are grounded through a resistor R13, a capacitor C15, a capacitor C16, and a forward polarity capacitor C14, respectively. The polar capacitors C13 and C14 are 10 muF, the capacitor C15 is 0.01 muF, the capacitor C16 is 0.1 muF, the resistor R14 is 162 omega, the resistor R12 is 100 omega, and the output third working voltage is 3.3V. In the embodiment, the chips U5, U6 and U7 all adopt LT 1084.

Claims (10)

1. An FPGA-based DPU apparatus, comprising:
an FPGA module: receiving external data of the interface module, performing operation and sending an operation result to external equipment through the interface module;
RISC module: receiving a memory management signal when the FPGA module operates, and sending the memory management signal to a memory management module;
a memory management module: the system comprises at least two triggers, wherein the triggers respectively receive memory control signals of a RISC module and an FPGA module and manage a data transmission channel between the memory module and the FPGA module;
a memory module: the FPGA module comprises a plurality of memory groups and a plurality of storage modules, wherein the memory groups are used for storing temporary data when the FPGA module performs operation;
an interface module: the data interaction interface for providing the device and the external equipment comprises:
a time sequence module: the FPGA module comprises at least two counters and a clock signal generator, wherein the counters are used for generating clock signals required by the operation of the FPGA module;
a power supply module: providing the required operating voltage for the entire DPU device.
2. The FPGA-based DPU device of claim 1, wherein the memory management module is specifically: the circuit comprises a first trigger U3 and a second trigger U4, wherein the enable end of the first trigger U3 is connected with the memory control end of the RISC module, the input pins of the first trigger U3 are respectively connected with the pins of the memory selection signal end of the RISC module, and the output pins of the first trigger U are grounded through a forward diode and a resistor; the enabling end of the second trigger U4 is connected with the memory control end of the FPGA module, the input end of the second trigger U4 is connected with the reference voltage, and the output end of the second trigger U4 is connected with the control end of each memory group through a forward diode; the anodes of the diodes of the first flip-flop U3 are connected to the anodes of the corresponding diodes of the second flip-flop U4.
3. The FPGA-based DPU device of claim 1, wherein the timing module is specifically to: the circuit comprises a crystal oscillator Y1, wherein two ends of the crystal oscillator Y1 are grounded through a capacitor C1 and a capacitor C2 respectively, and a resistor R1 is connected between two ends of the crystal oscillator Y1; one end of the crystal oscillator Y1 is connected with the RST pin of the first counter U1, and the other end of the crystal oscillator Y1 is connected with the PI pin of U1; the output pin Q8 of the first counter U1 is connected to the clkb pin of the second counter U2, and the qa pin of the second technician outputs the clock signal.
4. The FPGA-based DPU device of claim 1, wherein the power module comprises a transformer and a rectifier connected in sequence, and the transformer and the rectifier rectify and transform external alternating current and transmit the rectified alternating current to the first power module, the second power module and the third power module.
5. The FPGA-based DPU device of claim 4, wherein the first power module comprises a chip U5 having an input terminal coupled to ground through a forward polarity capacitor C3, an output terminal coupled to ground through a capacitor C7, a capacitor C8 and a forward polarity capacitor C5, respectively, an output terminal coupled to the first operating voltage through an inductor L1, and an output terminal coupled to ground through a capacitor C4 and a forward polarity capacitor C6, respectively.
6. The FPGA-based DPU device of claim 4, wherein the second power module comprises a chip U6, an input terminal of which is grounded through a forward polarity capacitor C9, a ground terminal of which is grounded through a resistor R11, a resistor R10 is connected between an output terminal of which and the ground terminal, an output terminal of which outputs the second operating voltage, and output terminals of which are grounded through a capacitor C11, a capacitor C12 and a forward polarity capacitor C10, respectively; the negative terminal of the polar capacitor C10 is connected to the digital block ground AGND via the inductor L2.
7. The FPGA-based DPU device of claim 4, wherein the third power module comprises a chip U7, an output terminal of which is grounded through a forward polarity capacitor C13, a ground terminal of which is grounded through a resistor R14, a resistor R12 is connected between the output terminal and the ground terminal, an output terminal of which outputs the third operating voltage, and output terminals of which are grounded through a resistor R13, a capacitor C15, a capacitor C16 and a forward polarity capacitor C14, respectively.
8. The FPGA-based DPU device of claim 1, wherein the interface module comprises: PCIe interface, QSFP interface, SSD interface, USB interface, SSD interface, GigE interface and GPIO interface.
9. The FPGA-based DPU device of claim 1, wherein the security module is further coupled to the FPGA module and the RISC module.
10. The FPGA-based DPU device of claim 1, wherein the FPGA module is further coupled to a JTAG module.
CN202122964619.2U 2021-11-29 2021-11-29 DPU device based on FPGA Expired - Fee Related CN216388083U (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202122964619.2U CN216388083U (en) 2021-11-29 2021-11-29 DPU device based on FPGA

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202122964619.2U CN216388083U (en) 2021-11-29 2021-11-29 DPU device based on FPGA

Publications (1)

Publication Number Publication Date
CN216388083U true CN216388083U (en) 2022-04-26

Family

ID=81219381

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202122964619.2U Expired - Fee Related CN216388083U (en) 2021-11-29 2021-11-29 DPU device based on FPGA

Country Status (1)

Country Link
CN (1) CN216388083U (en)

Similar Documents

Publication Publication Date Title
CN106875012B (en) A Pipelined Acceleration System of FPGA-Based Deep Convolutional Neural Network
CN102262604B (en) Concurrent access method, system and interface device
CN106844256B (en) A Design Method of Internal Interface of Active Distribution Network Real-time Simulator Based on Multiple FPGAs
CN106527575B (en) A kind of trigger grouping Clock gating method based on activity similarity
CN107169244A (en) A kind of electromechanical electromagnetic transient hybrid simulation interface system and method
CN119621313A (en) A dynamic allocation management system for real-time monitoring of computing resources
CN216388083U (en) DPU device based on FPGA
CN104850516B (en) A kind of DDR Frequency Conversion Designs method and apparatus
CN109596262A (en) Full-automatic pressure sensor verification system
CN104537498A (en) Electric power information device management system based on two-dimensional codes
WO2019047398A1 (en) Data concentration system for inner detector of oil-gas pipeline, and timing control method
CN109919826B (en) A graph data compression method for a graph computing accelerator and a graph computing accelerator
CN104932864A (en) Merging-sorting method based on assembly line process and valve control device using merging-sorting method
CN109472734A (en) A kind of target detection network and its implementation based on FPGA
CN109271566A (en) A kind of computer data processing system
CN119090225A (en) Power network energy storage data matching management system
CN102646262A (en) Reconfigurable visual preprocessor and visual processing system
CN111445019A (en) A device and method for realizing channel shuffling operation in packet convolution
CN212782003U (en) Intelligent USB concentrator with controllable function
CN114270956B (en) Synchronous circuit and synchronous chip
CN203574628U (en) Button expansion circuit
CN108197365A (en) High-performance data processing unit circuit based on Power PC Processor
CN111260070B (en) Operation method, device and related product
CN208623363U (en) Pluggable intelligent battery group screen display tool
CN108983714B (en) Method and device for monitoring state of unit of million-kilowatt nuclear power station and terminal equipment

Legal Events

Date Code Title Description
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20220426

CF01 Termination of patent right due to non-payment of annual fee