CN111258635B

CN111258635B - Data processing method, processor, data processing device and storage medium

Info

Publication number: CN111258635B
Application number: CN201811454786.9A
Authority: CN
Inventors: 不公告发明人
Original assignee: Shanghai Cambricon Information Technology Co Ltd
Current assignee: Shanghai Cambricon Information Technology Co Ltd
Priority date: 2018-11-30
Filing date: 2018-11-30
Publication date: 2022-12-09
Anticipated expiration: 2038-11-30
Also published as: CN111258635A

Abstract

The application relates to a data processing method, a processor, a data processing device and a storage medium. The data processing method comprises the following steps: reading the subdata segments from the first storage device according to the operation instruction, storing the currently read subdata segments into the second storage device, comparing N subdata in the subdata segments one by one according to the operation instruction to obtain a minimum value in the N subdata, and storing the minimum value as a first comparison result in the first storage device; and reading the first comparison result and the other sub-data segment to perform minimum comparison to obtain a second comparison result, and storing the second comparison result in the first storage device. The big data is divided into the small data to be subjected to circular operation processing, and the operand is circularly read according to the data reading capacity, so that the size of the data which can be accommodated during the operation processing meets the requirement, and the operation speed is accelerated.

Description

Data processing method, processor, data processing device and storage medium

Technical Field

The present application relates to the field of data processing technologies, and in particular, to a data processing method, a processor, a data processing apparatus, and a storage medium.

Background

Atomic operation refers to an operation that is not interrupted by a thread scheduling mechanism, and once the operation is started, the operation is run to the end, and no thread switching is performed in the middle (for example, the shared variable i performs accumulation, and the result of i + + is generated by multiple cores at the same time in the case of non-atomic operation, and the error occurs). In a multi-core processor system, a plurality of processor cores share the same memory space, and a common data transmission technology may not guarantee atomicity, that is, a plurality of processor cores may access the same address at the same time.

In an actual program, an operation result is stored in a memory space, the memory space has a certain address range, and since the memory space may be accessed by other processor cores before the operation is completed, the conventional method is to read data in the memory space to a memory unit, then store the operation result in the memory unit, and write the result of the memory unit back to the memory space after the instruction operation is completed. However, if another processor core accesses the memory space during the operation, an erroneous result is obtained, destroying the atomicity of the accumulation.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a data processing method, a processor, a data processing apparatus, and a storage medium, which are capable of achieving independent access to an off-chip storage space during an atomic operation.

A method of data processing, the method comprising:

obtaining an operation instruction, wherein the operation instruction is used for implementing comparison operation of a source operand, the source operand comprises at least one sub-data segment, and each sub-data segment comprises N sub-data;

reading the sub data segments from a first storage device according to data reading capacity and the operation instruction and according to a preset data reading mode, and storing the currently read sub data segments into a second storage device, wherein the first storage device is an off-chip storage device, and the second storage device is an on-chip storage device;

according to the operation instruction, comparing N pieces of sub data in the sub data segments one by one to obtain a minimum value in the N pieces of sub data, and storing the minimum value serving as a first comparison result into the second storage device and the first storage device;

reading the first comparison result and the other sub data segment from the first storage device, and storing the currently read first comparison result and the other sub data segment to a second storage device;

and comparing the first comparison result with the N pieces of sub data in the other sub data segment one by one according to the operation instruction, and storing a minimum value obtained by comparison into the second storage device and the first storage device as a second comparison result.

In one embodiment, the instruction format of the operation instruction comprises instruction type, source operand, target operand and operation code;

the instruction type is used for determining whether the operation instruction is an atomic operation instruction;

the instruction type is used for determining the operation type of the operation instruction;

the operation code is used for configuring the number of source operands;

the target operand is used to represent a current comparison result.

In one embodiment, the method further comprises:

and controlling the counter to accumulate once or decrement once, and then returning to the step of reading the first subdata from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode until the counter accumulates to a target cycle number from an initial value or the counter decrements to the initial value from the target cycle number, so as to finish the operation corresponding to the operation instruction.

A processor for use in a data processing method, the processor comprising an arithmetic circuitry, a read-write circuitry and a second storage means disposed adjacent the arithmetic circuitry, the second storage means being connectable by the read-write circuitry to a first storage means external to the processor;

the arithmetic circuit is used for acquiring an arithmetic instruction and sending a read-write request to the first storage device according to the arithmetic instruction;

the operation instruction is used for realizing comparison operation of a source operand, wherein the source operand comprises at least one sub-data segment, and each sub-data segment comprises N sub-data;

the read-write circuit is used for reading the subdata segments from the first storage device according to the read-write request and storing the subdata segments to the second storage device;

the arithmetic circuit is used for reading the sub data segments from the first storage device according to data reading capacity and the arithmetic instruction and according to a preset data reading mode, comparing N sub data in the sub data segments one by one to obtain a minimum value in the N sub data, and storing the minimum value serving as a first comparison result into the second storage device; storing the first comparison result of the second storage device into the first storage device through the read-write circuit; reading the first comparison result and the other sub-data segment from the first storage device, comparing the first comparison result with the N pieces of sub-data in the other sub-data segment one by one, and storing a minimum value obtained by comparison into the second storage device as a second comparison result; storing, by the read-write circuit, the second comparison result of the second storage device in the first storage device.

In one embodiment, the arithmetic circuit comprises a master processing circuit and more than one slave processing circuits, and the more than one slave processing circuits are all connected to the master processing circuit;

the monocular minimum value operation module is arranged in the main processing circuit.

A data processing apparatus, the apparatus comprising:

the system comprises an acquisition module, a comparison module and a comparison module, wherein the acquisition module is used for acquiring an operation instruction, the operation instruction is used for realizing comparison operation of a source operand, the source operand comprises at least one sub-data segment, and each sub-data segment comprises N pieces of sub-data;

a reading module, configured to read the sub data segment from a first storage device according to a data reading capacity and the operation instruction and according to a preset data reading manner, and store the currently read sub data segment in a second storage device, where the first storage device is an off-chip storage device, and the second storage device is an on-chip storage device;

the operation module is used for comparing N pieces of sub data in the sub data segments one by one according to the operation instruction to obtain a minimum value in the N pieces of sub data, and storing the minimum value serving as a first comparison result into the second storage device and the first storage device; reading the first comparison result and the other sub data segment from the first storage device, and storing the currently read first comparison result and the other sub data segment to a second storage device; and comparing the first comparison result with the N pieces of sub data in the other sub data segment one by one according to the operation instruction, and storing a minimum value obtained by comparison into the second storage device and the first storage device as a second comparison result.

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

According to the data processing method, the processor, the data processing device and the storage medium, according to an operation instruction, according to a preset data reading mode, first sub data is read from a first storage device, the currently read first sub data is stored in a second storage device, then according to the operation instruction, N sub data in a sub data segment are compared one by one to obtain a minimum value of the N sub data, the minimum value is used as a first comparison result and stored in the second storage device and the first storage device, then the first comparison result and another sub data segment are read from the first storage device, the first comparison result and the N sub data in another sub data segment are compared one by one, and the obtained minimum value is used as a second comparison result and stored in the first storage device until relevant operation corresponding to the operation instruction is completed. The data are read circularly for comparison, the comparison result of each circulation is continuously stored in the first storage device, exclusive access to the first storage device is realized, other processor cores are prevented from accessing the first storage device, and atomicity of atomic operation is guaranteed. The arithmetic operation function of the processor is further expanded, and the operation efficiency during the atomic operation is improved by realizing exclusive access to the first storage device.

Drawings

FIG. 1 is a block diagram of a processor in one embodiment;

FIG. 2 is a block diagram of an embodiment of an operational module;

FIG. 3 is a schematic diagram of a processor according to another embodiment;

FIG. 4 is a schematic diagram of a processor according to another embodiment;

FIG. 5 is a schematic diagram of a processor according to another embodiment;

FIG. 6 is a flow diagram illustrating a data processing method according to one embodiment;

FIG. 7 is a flow chart illustrating a data processing method according to another embodiment;

FIG. 8 is a flowchart illustrating a method for instruction disassembly in accordance with another embodiment;

FIG. 9 is a flowchart illustrating step S800 according to an embodiment;

FIG. 10 is a flowchart illustrating step S800 in another embodiment;

FIG. 11 is a schematic flow chart of the Atomic MIN _ SCALAR method in one embodiment;

FIG. 12 is a schematic flow chart of the Atomic MIN _ SCALAR method in another embodiment;

FIG. 13 is a block diagram showing the structure of a data processing apparatus according to an embodiment;

FIG. 14 is a block diagram of an instruction disassembling apparatus according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The terms "first," "second," and "third," etc. in the description and claims of this application and in the accompanying drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus. Off-chip refers to the outside of the processor, i.e., off-chip memory means a memory device disposed outside the processor; on-chip refers to the inside of the processor, i.e. on-chip memory means refers to memory means arranged inside the processor.

The data processing method provided by the present application can be applied to the processor 1000 shown in fig. 1. The processor 1000 includes an arithmetic circuit 12, a read/write circuit 203, and a second storage device 201. The second storage 201 may be a buffer and/or a register arranged inside the processor 1000. The second storage device 201 may be connected to the first storage device 13 provided outside the processor 1000 through the read-write circuit 203. The first storage device 13 and the second storage device 201 may be a non-volatile memory or a volatile memory, and are not limited herein. The read and write circuits 203 may be I/O circuits.

The operation circuit 12 and the read/write circuit 203 can be connected to the second storage device 201, respectively, and the read/write circuit 203 can be connected to the first storage device 13. The second memory device 201 can be connected to the first memory device 13 outside the processor 1000 via the read/write circuit 203. The second storage device 201 can read the first source operand from the first storage device 13 through the read-write circuit 203 and transmit the first source operand to the operation circuit 12 for operation. The operation circuit 12 may store the operation result and the intermediate operation result obtained by the operation circuit in the second storage device 201, and the second storage device 201 may also write the operation result back to the first storage device 13 through the read/write circuit 203. In the embodiment of the present application, the intermediate operation result is continuously written back from the second storage device 201 to the first storage device 13 outside the processor 1000, so that the operation circuit 12 can exclusively use the first storage device 13, and the atomicity of the operation and the accuracy of the operation result can be ensured.

The arithmetic circuit 12 is configured to receive an arithmetic instruction, analyze the arithmetic instruction, and implement a corresponding arithmetic operation according to the arithmetic instruction. Optionally, the operation instruction may have a specific instruction format, and the operation circuit may analyze the instruction format according to the operation instruction to obtain instruction information such as an instruction type, a source operand, and an operation code of the operation instruction, so as to implement a corresponding operation according to the operation instruction.

Alternatively, the operation instruction in the embodiment of the present application may be an atomic operation instruction, and as shown in table 1 below, the instruction format of the operation instruction may include an instruction type Name, an instruction type Op, a first source operand, a second source operand, a destination operand Dst, an opcode Src Op, and the like.

The instruction class Name is used for determining the class of the instruction (the class of the instruction includes an atomic operation instruction and other common operation instructions), that is, the instruction class is used for determining whether the operation instruction is an atomic operation instruction. The instruction type Op is used to determine the operation type of the operation instruction, and the operation type is used to indicate what kind of operation the operation instruction implements, so as to distinguish the specific function of the operation, for example, the operation type may be an accumulation operation, a decrement operation, a maximum value operation, a minimum value operation, a logical and operation, a logical or operation, a logical xor operation, an alternative operation, a swap operation, and the like. The operation code Src Op is used to configure the number of source operands involved in the operation instruction. The target operand dstaddr is used to indicate a current operation result obtained after at least one source operand is operated, specifically, the target operand dstaddr may refer to a storage address of the current comparison result, and an operation result corresponding to the operation instruction may be stored in a storage space indicated by the storage address corresponding to the dstaddr. The first and second source operands may represent data participating in an operation, the first source operand may be data stored on the off-chip first storage means 13, i.e. the first source operand may represent data stored in the address Src0 addr. The second source operand may represent data stored in an immediate or address in an instruction.

Further, the instruction format of the operation instruction may further include an identification bit Src1 vec for identifying whether the source operand a is an immediate or an address, and an identification bit Src2 vec for identifying whether the source operand B is an immediate or an address.

Specifically, when Src1 vec is 0, it indicates that the source operand a is an immediate number, and when Src1 vec is 1, it indicates that the source operand a is data stored in an address; when Src2 vec is 0, it indicates that the source operand B is an immediate number, and when Src2 vec is 1, it indicates that the source operand B is data stored in an address.

Furthermore, the instruction format of the operation instruction further includes a Data size for indicating the first source operand and a Data stream IO config for requesting splitting, indicating that the target loop number is calculated.

The instruction format of the operation instruction may be as follows, as shown in table 1:

instruction field	Bit width	Means of
			Name	8	Instruction class, atomic class 15
Op	8	Instruction type, distinguishing particular functions
			Src0 addr	49	Source operand 0 Address, off-chip only, aligned by byte
Dst addr	32	Destination address, on-chip only
			Src1	32	Source operand A, immediate/address (determined by Src1 vec)
Src2	32	Source operand B, immediate/address (determined by Src2 vec)
			IO config	9	Atomic operations read-write data stream ID for requesting splitting
Data size	32	Atomic operations read and write data size, aligned by byte
			Scr Op	3	Operation code, configuration source operand number
Data type	3	Data type
			Src1 vec	1	Source operand A type (immediate/address)
Src2 vec	1	Source operand B type (immediate/address)

Where Src0addr represents the address of the first source operand and dstaddr represents the storage address of the target operand.

In one embodiment, the source operand A or the source operand B is used as the second source operand according to the operation instruction.

Optionally, the second source operand includes source operand A (Src 1), source operand B (Src 2), source operand A selected identification bits, source operand B selected identification bits. Specifically, when the selected identification bit of the source operand A is valid, the source operand A is taken as a second source operand; and when the selected identification bit of the source operand B is valid, taking the source operand B as a second source operand. When the selected identification bit of the source operand A is valid and the selected identification bit of the source operand B is valid, the source operand A and the source operand B can be simultaneously used as second source operands, and the number of the second source operands is two. Further, the bit width of the operation code Scr Op may include 3 bits, where 2 bits are used to distinguish the number of source operands participating in the operation, and 1 bit is used to select the source operand a (Src 1) and/or the source operand B (Src 2) as the second source operand to participate in the operation. Reference may be made to table 2:

when the operation code Scr Op is "000", it indicates that the source operand of the operation instruction is 1, which is the first source operand Src0. When the opcode Scr Op is "010", it indicates that the source operands of the operation instruction are 2, including the first source operand Src0 and the second source operand, and the selected identification bit of the source operand a is valid, and the second source operand is the source operand a (Src 1). When the opcode Scr Op is "011", it indicates that the source operands of the operation instruction are 2, including the first source operand Src0 and the second source operand Src2, and the selected identification bit of the source operand B is valid. When the operation code Scr Op is "100", it indicates that the source operands of the operation instruction are 3, including the first source operand Src0, the source operand a (Src 1), and the source operand B (Src 2).

In the embodiment of the present application, it may be default that the first source operand Src0 is always valid.

Optionally, the Data Type represents a Data Type, and the instruction supports, but is not limited to, the following Data types:

data type	Data Type
		Int16	000
Uint16	001
		Int32	010
Uint32	011

Optionally, the operation instruction may include an arithmetic operation instruction, and may also include a logic operation instruction: the arithmetic operation instruction may include: the system comprises a monocular maximum value operation instruction Atomic MAX _ SCALAR, a monocular minimum value operation instruction Atomic MIN _ SCALAR, a binocular maximum value operation instruction Atomic MAX _ VEC, a binocular minimum value operation instruction Atomic MIN _ VEC, a replacement operation instruction Atomic CAS, an exchange operation instruction Atomic EXCH, an addition operation instruction Atomic ADD, an accumulation operation instruction Atomic INC and a subtraction operation instruction Atomic DEC. The logical operation instruction may include: the logical AND operation instruction Atomic AND, the logical OR operation instruction Atomic OR, the logical XOR operation instruction Atomic XOR, AND the logical NOT operation instruction Atomic NOT.

The monocular maximum value operation instruction Atomic MAX _ SCALAR is used for solving the maximum value of the plurality of first subdata in the first source operand.

And the monocular minimum value operation instruction Atomic _ SCALAR is used for solving the minimum value of a plurality of first subdata in the first source operand.

The binary maximum value operation instruction Atomic MAX _ VEC is used for calculating the maximum value of the first source operand and the second source operand.

The binary minimum value operation instruction Atomic MIN _ VEC is used for calculating the minimum value of the first source operand and the second source operand.

The ADD operation instruction Atomic ADD is used for adding a first source operand and a second source operand.

And the accumulation operation instruction Atomic INC is used for performing accumulation operation between the first source operand and the second source operand.

A subtraction instruction Atomic DEC to perform a subtraction operation between a first source operand and a second source operand.

AND the logical AND operation instruction Atomic AND is used for performing an AND logical operation between the first source operand AND the second source operand.

The logical OR instruction Atomic OR is used for carrying out logical OR operation between the first source operand and the second source operand.

And the logic exclusive-OR operation instruction Atomic XOR is used for performing logic exclusive-OR operation between the first source operand and the second source operand.

And the logic NOT instruction is used for carrying out NOT operation between the first source operand and the second source operand.

The replacement operation instruction Atomic CAS is used for replacing among the first source operand, the second source operand and the third source operand.

The system comprises an exchange operation instruction Atomic EXCH and an operation instruction used for exchanging between a first source operand and a second source operand.

In this embodiment, to ensure atomicity of operation, the operation instruction may be implemented by dividing the same operation into multiple sub-operation operations, and the first storage device is monopolized by continuously writing back the intermediate calculation result to the first storage device.

Specifically, the first source operand includes at least one first subdata, the arithmetic circuit 12 receives an arithmetic instruction, sends a read-write request to the first storage device 13 according to the arithmetic instruction, the read-write circuit 203 reads the first subdata from the first storage device 13 according to the read-write request and stores the first subdata into the second storage device 201 according to a data reading mode, the arithmetic circuit 12 obtains the second source operand according to the arithmetic instruction, executes an arithmetic operation to obtain a current arithmetic result, stores the obtained current arithmetic result into the second storage device 201, and stores the current arithmetic result of the second storage device 201 into the first storage device 13 through the read-write circuit 203. After that, the arithmetic circuit 12 may send a read/write request to the first storage device 13 again to read the first sub-data from the first storage device 13 again, and execute the arithmetic operation for multiple times in a loop until the arithmetic operation corresponding to the arithmetic instruction is completed.

Optionally, the processor may further comprise a counter, which may be connected to the arithmetic circuitry 12, for recording a target number of cycles of the arithmetic instruction. Specifically, each time the read/write circuit 203 stores the current operation result of the second storage device 201 in the first storage device 13, the operation circuit 12 may control the counter to increment once, and send the read/write request to the first storage device 13 again until the counter is incremented from the initial value to the target cycle number. In this embodiment, the initial value of the counter may be 0, that is, when the counter is incremented from 0 to the target cycle number, the corresponding operation of the operation instruction is completed. Alternatively, the arithmetic circuit 12 may control the counter to decrement once and send the read/write request to the first storage device 13 again until the counter is decremented from the target number of cycles to the initial value. In this embodiment, the initial value of the counter may be 0, that is, when the counter is decremented from the target number of cycles to 0, the corresponding operation of the operation instruction is completed.

Further, the arithmetic circuit 12 may be provided with an arithmetic module corresponding to each arithmetic instruction. Specifically, referring to fig. 2, the arithmetic circuit 12 may include a binoculus maximum value operation module 121, a binoculus minimum value operation module 122, a logical and operation module 123, a logical or operation module 124, a logical exclusive-or operation module 125, an exchange operation module 126, a replacement operation module 127, a monocular maximum value operation module 128, a monocular minimum value operation module 129, an addition operation module 130, an accumulation operation module 131, a subtraction operation module 132, a logical not operation module 133, and the like.

The two-purpose maximum value operation module 121 is configured to implement the operation of the two-purpose maximum value operation instruction Atomic MAX _ VEC, that is, to implement the maximum value operation of the first source operand and the second source operand.

The binary minimum operation module 122 is configured to implement the operation of the binary minimum operation instruction Atomic MIN _ VEC, that is, to implement the minimum operation of the first source operand and the second source operand.

The AND logic module 123 is configured to implement the above-mentioned AND operation of the AND logic instruction Atomic AND, that is, to implement the AND logic operation between the first source operand AND the second source operand.

The OR logic module 124 is configured to implement the operation of the above-mentioned OR logic instruction Atomic OR, that is, to implement a logical OR operation between the first source operand and the second source operand.

The exclusive or operation module 125 is configured to implement the operation of the above-mentioned exclusive or operation instruction Atomic XOR, that is, to implement the exclusive or operation between the first source operand and the second source operand.

The swap operation module 126 is used to implement the operation of the swap operation instruction Atomic EXCH, i.e. the swap operation between the first source operand and the second source operand.

The replacement operation module 127 is configured to implement the operation of the replacement operation instruction Atomic CAS, that is, to implement a replacement operation between the first source operand, the second source operand, and the third source operand.

The monocular maximum value calculating module 128 is configured to implement the operation of the above-mentioned monocular maximum value calculating instruction Atomic MAX _ SCALAR, that is, to implement the maximum value calculation of the plurality of first sub-data in the first source operand.

The monocular minimum value operation module 129 is configured to implement the operation of the above-mentioned monocular minimum value operation instruction Atomic MIN _ SCALAR, that is, to implement the minimum value operation of the plurality of first sub-data in the first source operand.

The addition operation module 130 is configured to implement the operation of the above-mentioned addition operation instruction Atomic ADD, that is, to implement the operation of adding the first source operand and the second source operand.

The accumulation operation module 131 is configured to implement the operation of the accumulation operation instruction Atomic INC, that is, to implement the operation of accumulating the first source operand and the second source operand.

The subtraction module 132 is configured to implement the operation of the subtraction instruction Atomic DEC, that is, to implement the subtraction operation between the first source operand and the second source operand.

The logical NOT operation module 133 is configured to implement the operation of the above-mentioned logical NOT operation instruction Atomic NOT, that is, to implement the logical NOT operation between the first source operand and the second source operand.

Alternatively, each operation module may include an operation unit and a result output unit connected to the operation unit. The operation unit is used for executing specific operation steps, and the result output unit is used for taking the result obtained in the operation steps as the current operation result.

Further, as shown in fig. 1 and 2, the processor may further include a data selector 14, and the data selector 14 is connected between the arithmetic circuit 12 and the read-write circuit 203. The data selector 14 is configured to gate the connection paths between each operation block in the operation circuit 12 and the read/write circuit 203. For example, if the operation command is an Atomic MAX _ VEC, the data selector 14 is used to gate the connection path between the binary maximum operation module 121 and the read/write circuit 203. At this time, the maximum two-entry operation module 121 is configured to obtain the second subdata, determine whether the currently read first subdata is greater than or equal to the second subdata according to the operation instruction, store the obtained current comparison result in the second storage device 201, and store the current comparison result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation command is the binary minimum operation command Atomic MIN _ VEC, the data selector 14 is configured to gate the connection path between the binary minimum operation module 122 and the read/write circuit 203. At this time, the two-entry minimum value operation module 122 is configured to obtain the second sub data, determine whether the currently read first sub data is smaller than the second sub data according to the operation instruction, store the obtained current comparison result in the second storage device 201, and store the current comparison result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is an AND operation instruction Atomic AND, the data selector 14 is configured to gate a connection path between the AND operation block 123 AND the read/write circuit 203. At this time, the and logic module 123 is configured to obtain the second sub-data, perform and logic operation on the currently read first sub-data and second sub-data according to the operation instruction, store the obtained current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is an and operation instruction Atomic OR, the data selector 14 is configured to gate the connection path between the OR operation module 124 and the read/write circuit 203. At this time, the or logic module 124 is configured to obtain the second sub data, perform a or logic operation on the currently read first sub data and second sub data according to the operation instruction, store the obtained current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is the and operation instruction Atomic XOR, the data selector 14 is used to gate the connection between the XOR operation module 125 and the read/write circuit 203. At this time, the xor operation module 125 is configured to obtain the second sub data, perform xor operation on the currently read first sub data and second sub data according to the operation instruction, store the obtained current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is the replacement operation instruction Atomic CAS, the data selector 14 is used for gating the connection path between the replacement operation module 127 and the read/write circuit 203. At this time, the replacement operation module 127 is configured to obtain the second sub-data and the third sub-data, perform replacement operation on the currently read first sub-data, second sub-data, and third sub-data according to the operation instruction, store the obtained current operation result into the second storage device 201, and store the current operation result of the second storage device 201 into the first storage device 13 through the read-write circuit 203 and the data selector 14.

If the operation instruction is an exchange operation instruction Atomic EXCH, the data selector 14 is used for gating the connection path between the exchange operation module 126 and the read/write circuit 203. At this time, the exchange operation module 126 is configured to obtain the second sub data, perform an exchange operation on the currently read first sub data and second sub data according to the operation instruction, store the obtained current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation command is the monocular maximum value operation command Atomic MAX _ SCALAR, the data selector 14 is configured to gate the connection path between the monocular maximum value operation module 128 and the read/write circuit 203. At this time, the monocular maximum value operating module 128 is configured to compare N pieces of sub data in the sub data segments of the source operand one by one to obtain a maximum value of the N pieces of sub data, store the maximum value as a current comparison result into the second storage device 201, and store the current comparison result of the second storage device 201 into the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation command is the monocular minimum value operation command Atomic MIN _ SCALAR, the data selector 14 is configured to gate the connection path between the monocular minimum value operation module 129 and the read/write circuit 203. At this time, the monocular minimum value operating module 129 is configured to compare N pieces of sub data in the sub data segments of the source operand one by one to obtain a minimum value of the N pieces of sub data, store the minimum value as a current comparison result in the second storage device 201, and store the current comparison result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is an ADD operation instruction Atomic ADD, the data selector 14 is configured to gate a connection path between the ADD operation module 130 and the read/write circuit 203. At this time, the addition operation module 130 is configured to obtain the second sub data, add the currently read first sub data and the second sub data according to the operation instruction to obtain a current operation result, store the current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

If the operation instruction is the accumulation operation instruction Atomic INC, the data selector 14 is used for gating the connection path between the accumulation operation module 131 and the read/write circuit 203. At this time, the accumulation operation module 131 is configured to obtain the second sub-data, determine whether the currently read first sub-data is greater than or equal to the second sub-data according to the operation instruction, reset the first sub-data when the first sub-data is greater than or equal to the second sub-data, store the reset first sub-data as a current comparison result in the second storage device 201, and store the current comparison result of the second storage device 201 in the first storage device 13 through the read-write circuit 203 and the data selector 14.

If the operation instruction is the subtraction instruction Atomic DEC, the data selector 14 is configured to gate a connection path between the subtraction module 132 and the read/write circuit 203. At this time, the subtraction module 132 is configured to obtain second sub-data, determine whether the currently read first sub-data is larger than the second sub-data according to the operation instruction, and store the second sub-data as a current comparison result in the second storage device 201 when the first sub-data is larger than the second sub-data; when the first sub-data is smaller than or equal to the second sub-data, the first sub-data is subtracted from the first preset value, the subtracted first sub-data is stored in the second storage device 201 as a current comparison result, and the current comparison result of the second storage device 201 is stored in the first storage device 13 through the read-write circuit 203 and the data selector 14.

If the operation instruction is the logic NOT operation instruction Atomic NOT, the data selector 14 is configured to gate a connection path between the logic NOT operation module 133 and the read/write circuit 203. At this time, the logical negation operation module 133 is configured to obtain the second sub data, perform a logical negation operation on the currently read first sub data and second sub data according to the operation instruction, store the obtained current operation result in the second storage device 201, and store the current operation result of the second storage device 201 in the first storage device 13 through the read/write circuit 203 and the data selector 14.

In one embodiment, with continued reference to fig. 3-5, the second memory device 201 and the read/write circuit 203 may be packaged as a memory circuit 10. The arithmetic circuit 12 includes a master processing circuit 101 and at least one slave processing circuit 102, the at least one slave processing circuit 102 each being connected to the master processing circuit 101, the master processing circuit 101 being connected to a branch processing circuit(s) 103, the branch processing circuit 103 being connected to the one or more slave processing circuits 102; the branch processing circuit 103 is configured to execute forwarding data or instructions between the master processing circuit 101 and the slave processing circuit 102. The main processing circuit 101 is used for performing preamble processing on a source operand and transmitting data and an operation instruction with a plurality of slave processing circuits; the plurality of slave processing circuits 102 are configured to perform an intermediate operation in parallel according to the data and the operation instruction transmitted from the master processing circuit to obtain a plurality of intermediate results, and transmit the plurality of intermediate results to the master processing circuit; the main processing circuit 101 is configured to perform subsequent processing on the plurality of intermediate results to obtain a calculation result of the calculation instruction.

The main processing circuit 101 may include the aforementioned two-purpose maximum operation module 121, two-purpose minimum operation module 122, logical and operation module 123, logical or operation module 124, logical exclusive or operation module 125, replacement operation module 126, swap operation module 127, one-purpose maximum operation module 128, one-purpose minimum operation module 129, addition operation module 130, accumulation operation module 131, subtraction operation module 132, and logical not operation module 133. The data selector 14 described above may be connected between the main processing circuit 101 and the read/write circuit 203.

In one embodiment, the processor may further include a controller circuit 11, the controller circuit 11 including: instruction cache circuit 110, instruction processing circuit 111, and store queue circuit 113.

The instruction cache circuit 110 is configured to store a calculation instruction associated with an artificial neural network operation.

The instruction processing circuit 111 is configured to analyze the calculation instruction to obtain a plurality of operation instructions.

A store queue circuit 113 for storing an instruction queue, the instruction queue comprising: and a plurality of operation instructions or calculation instructions to be executed according to the front and back sequence of the queue.

Further, the controller circuit 11 may include a split granularity circuit 114, a cycle number processing circuit 115, and a data read capacity calculation circuit 116.

The split granularity circuit 114 is connected to the cycle number processing circuit 115, the cycle number processing circuit 115 is connected to the instruction processing circuit 111 and the data reading capacity calculation circuit 116, the data reading capacity calculation circuit 116 is connected to the operation circuit 12, and the second storage device 201 can be connected to the first storage device 13 outside the processor through the read/write circuit 203.

The instruction processing circuit 111 is configured to obtain an operation instruction, parse the data size of the first source operand according to the operation instruction, and transmit the data size of the first source operand to the loop number processing circuit 115.

The split granularity circuit 114 is used to store a preset split granularity. In this embodiment, the split-granularity circuit 114 may be a buffer or a segment of a storage space in the second storage device, for example, the split-granularity circuit 114 may be a storage space corresponding to a specified address interval in the second storage device.

The cycle number processing circuit 115 is configured to obtain a target cycle number according to the size of the first operand and a preset splitting granularity, and transmit the target cycle number to the operation circuit 12. In this embodiment, the loop number processing circuit 115 may be a counter.

The data reading capacity calculation circuit 116 is configured to obtain a data reading capacity according to the size of the first operand and a preset splitting granularity, and transmit the data reading capacity to the operation circuit 12. The arithmetic circuit 12 is configured to send a read/write request to the first storage device 13 according to the arithmetic instruction, so as to read first sub-data from the first storage device 13, where the size of the first sub-data is equal to the data reading capacity. After that, the operation circuit 12 may perform an operation according to the read first sub data and second sub data, and after the current operation is completed, the cycle number processing circuit 115 increments the target cycle number once from the initial value, and sends the read/write request to the first storage device 13 again until the counter increments from the initial value to the target cycle number. In this embodiment, the initial value may be 0, that is, until the current cycle count is accumulated from 0 to the target cycle count, the corresponding operation of the operation instruction is completed. Alternatively, the loop count processing circuit 115 decrements the target loop count once, and transmits the read/write request to the first storage device 13 again until the target loop count is decremented to 0. Namely, when the current cycle number is decreased to 0, the corresponding operation of the operation instruction is completed.

The embodiment splits data by adding the split granularity circuit 114, the cycle number processing circuit 115 and the data reading capacity calculation circuit 116, so that the processed data size is larger than the memory access bandwidth which can be accommodated by a single clock cycle.

Referring to fig. 6 or fig. 7, when receiving the operation instruction, the processor may perform the following steps:

s100, an operation instruction is obtained.

The operation instruction is used for realizing operation among source operands, and the first source operand comprises at least one first subdata.

And S200, reading the first subdata from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode, and storing the currently read first subdata into the second storage device.

The data reading capacity represents the number of data read at a time, and can be calculated. The first memory device 13 is an off-chip memory device and the second memory device 201 is an on-chip memory device. Specifically, after the arithmetic circuit 12 obtains the arithmetic instruction, it sends a read-write request to the first storage device 13 according to the arithmetic instruction, and then the read-write circuit 203 reads the first sub-data from the first storage device 13 according to the read-write request in a preset data reading manner, and stores the currently read first sub-data in the second storage device 201.

S300, executing operation according to the operation instruction, and storing the obtained current operation result into the second storage device and the first storage device.

Specifically, the arithmetic circuit 12 performs a corresponding arithmetic operation according to the obtained arithmetic instruction, so as to obtain a current arithmetic result, and then stores the obtained current arithmetic result in the second storage device 201, and then stores the current arithmetic result of the second storage device 201 in the first storage device 13 through the read/write circuit 203.

And S400, returning to the step of reading the first subdata from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode until the operation corresponding to the operation instruction is completed.

Specifically, step S400 may include: and controlling the counter to accumulate once or decrement once, then returning to the step S200, and reading the first subdata from the first storage device according to the operation instruction and the data reading capacity and a preset data reading mode until the counter is accumulated to the target cycle number from the initial value or the counter is decremented to the initial value from the target cycle number. In the embodiment of the present application, the initial value of the counter may be 0.

Further, the target number of cycles is calculated according to the data size of the first source operand. After the current operation result of the second storage device 201 is stored in the first storage device 13 through the read-write circuit 203, the counter is controlled to accumulate once, and then the first subdata is continuously read from the first storage device 13 according to the operation instruction and the data reading capacity until the counter is accumulated from 0 to the target cycle number. Or controls the counter to decrement once until the counter decrements the target number of cycles to 0, and stops reading the first sub data from the first storage device 13.

In another embodiment, referring to fig. 8, the data processing method may further include the following steps:

s500, an operation instruction is obtained, and the data size of the first source operand is analyzed according to the operation instruction.

Specifically, the instruction processing circuit 111 obtains the operation instruction, parses the data size of the first source operand according to the operation instruction, and sends the data size of the first source operand to the loop number processing circuit 115.

And S600, obtaining the cycle number and the data reading capacity according to the data size of the first source operand and the preset splitting granularity.

Specifically, the preset split granularity is stored in the split granularity circuit 114, which may be a certain storage space in the static memory on the chip. The cycle number processing circuit 115 receives the data size of the first source operand, and calculates the target cycle number according to the data size of the first source operand and the preset splitting granularity. The data read capacity calculation circuit 116 calculates the data read capacity according to the cycle count sent by the cycle count processing circuit 115, the data size of the first source operand sent by the instruction processing circuit 111, and the preset split granularity, and sends the data read capacity and the target cycle count to the operation circuit 12.

Alternatively, the loop number processing circuit 115 may calculate the target loop number according to the following formula:

wherein, count is expressed as a target cycle number, data size is expressed as a data size of the first source operand, and the splitting granularity is a preset splitting granularity. In the embodiment of the application, the quotient obtained by dividing the data size by the preset splitting granularity is rounded up to obtain the target cycle number Count.

The data read capacity calculation circuit 116 may calculate the data read capacity according to the following formula:

data real size = min { unprocessed data size, split granularity }

The data real size represents data reading capacity, the splitting granularity is preset splitting granularity, and the unprocessed data size represents the data size of the first source operand minus the data reading capacity.

For example, the controller circuit 11 analyzes that the data size of the first source operand is 1000 bytes, the preset splitting granularity is 512 bytes, the number of cycles is 2, and the data reading capacities of the two times are 512 bytes and 488 bytes, respectively, according to the operation instruction.

Further, the data processing method may further include:

according to the operation instruction, the number of source operands included in the operation instruction is determined. The number of source operands may include one, two, or three, among others.

Specifically, the operation circuit 12 may perform the determination according to the operation code Scr Op in the operation instruction, and when the operation code Scr Op is "000", it indicates that the source operand of the operation instruction is 1, and is the first source operand Src0. When the opcode Scr Op is "010", it indicates that the source operands of the operation instruction are 2, including the first source operand Src0 and the second source operand, and the second source operand is the source operand a (Src 1). When the operation code Scr Op is "011", it indicates that the source operands of the operation instruction are 2, including the first source operand Src0 and the second source operand, and the second source operand is the source operand B (Src 2). When the operation code Scr Op is "100", it indicates that the source operands of the operation instruction are 3, including the first source operand Src0, the source operand a (Src 1), and the source operand B (Src 2).

Optionally, when there is one source operand, the step S300 may further include the following steps:

and executing the operation according to the operation instruction and the sub-data segment, and storing the obtained current operation result/comparison result into the second storage device and the first storage device.

Specifically, the arithmetic circuit 12 executes corresponding operations according to the received instructions and the sub-data segments to obtain current operation results/comparison results, stores the obtained current operation results/comparison results in the second storage device 201, and stores the current operation results/comparison results in the second storage device 201 in the first storage device 13 through the read/write circuit 203.

Specifically, when the number of the source operands is one, the operation instruction may be a monocular maximum value operation instruction Atomic MAX _ SCALAR, a monocular minimum value operation instruction Atomic MIN _ SCALAR, a logical NOT operation instruction Atomic NOT, and the like, and the specific execution process of each operation instruction may specifically refer to the above description.

Optionally, when the number of the source operands is two, referring to fig. 9 together, the two source operands are a first source operand and a second source operand respectively, and the step S300 may further include the following steps:

s310, acquiring second subdata according to the operation instruction.

Specifically, two source operands participating in the operation are determined according to the operation instruction, and the operation circuit 12 obtains the second sub-data according to the operation instruction.

And S320, executing operation according to the operation instruction, the first subdata and the second subdata, and storing the obtained current operation result into a second storage device and a first storage device.

Specifically, the arithmetic circuit 12 executes corresponding operations according to the received instruction, the first sub-data and the second sub-data, obtains a current arithmetic result, stores the obtained current arithmetic result in the second storage device 201, and stores the current arithmetic result in the second storage device 201 in the first storage device 13 through the read/write circuit 203.

Specifically, when the number of the source operands is two, the operation instruction may be an Atomic maximum operation instruction Atomic MAX _ VEC, an Atomic minimum operation instruction Atomic MIN _ VEC, an Atomic AND operation instruction Atomic AND, an Atomic OR operation instruction Atomic OR, an exclusive OR operation instruction Atomic XOR, an exchanging operation instruction Atomic EXCH, an adding operation instruction Atomic ADD, an accumulating operation instruction Atomic INC, a subtracting operation instruction Atomic DEC, AND the like, AND the specific execution process of each operation instruction is described above.

Optionally, when the number of the source operands is three, referring to fig. 10, the three source operands are a first source operand, a second source operand, and a third source operand, respectively, the step S300 may further include the following steps:

s330, acquiring the second sub data and the third sub data according to the operation instruction.

Wherein the third source operand includes at least one third child data.

Specifically, three source operands participating in the operation are determined according to the operation instruction, and the operation circuit 12 obtains the second sub-data and the third sub-data according to the operation instruction.

And S340, executing operation according to the operation instruction, the first sub-data, the second sub-data and the third sub-data, and storing the obtained current operation result into the second storage device and the first storage device.

Specifically, the arithmetic circuit 12 executes corresponding operations according to the received instruction, the first sub-data, the second sub-data, and the third sub-data, obtains a current operation result, stores the obtained current operation result in the second storage device 201, and stores the current operation result in the second storage device 201 in the first storage device 13 through the read/write circuit 203.

In one embodiment, the source operand A or the source operand B is used as the second source operand according to the instruction format of the operation instruction.

Specifically, referring to the format of Src Op in table 2, when Src Op in the received operation instruction is 010, which indicates that source operand a is valid, source operand a is used as the second source operand; when Src Op in the received operation instruction corresponds to 011, indicating that source operand B is valid, and using source operand B as a second source operand; when Src Op in the received operation instruction corresponds to 100, which indicates that source operand a is valid and source operand B is valid, source operand a is used as the second source operand and source operand B is used as the third source operand.

In this embodiment, the source operand a or the source operand B is selected as the second source operand for selection according to the format of the operation code Src Op in the instruction format.

Specifically, when the above-mentioned operation instruction is Atomic MIN _ SCALAR, the data processing method shown in fig. 12 may include the following steps:

s962, an operation instruction is acquired.

The operation instruction is used for realizing comparison operation of a source operand, the source operand comprises sub data segments, and each sub data segment comprises N pieces of sub data.

Specifically, the arithmetic circuit 12 obtains an arithmetic instruction for performing a comparison operation of source operands.

And S964, reading the subdata segments from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode, and storing the currently read subdata segments into the second storage device.

The first storage device 13 is an off-chip storage device, and the second storage device 201 is an on-chip storage device.

Specifically, after the arithmetic circuit 12 obtains the arithmetic instruction, it sends a read-write request to the first storage device 13 according to the arithmetic instruction and the data reading capacity, and then the read-write circuit 203 reads the sub-data segment from the first storage device 13 according to the read-write request in a preset data reading manner, and stores the currently read sub-data segment in the second storage device 203.

S966, comparing the N sub data in the sub data segments one by one according to the operation instruction to obtain the minimum value of the N sub data, and storing the minimum value as a first comparison result into the second storage device and the first storage device.

Wherein the first comparison result refers to a minimum value in the first segment of sub data segments.

Specifically, the arithmetic circuit 12 compares the N pieces of sub data in the sub data segment one by one according to the obtained arithmetic instruction to obtain a minimum value of the N pieces of sub data, and uses the minimum value as a first comparison result, and then stores the obtained first comparison result in the second storage device 201, and then stores the first comparison result in the first storage device 13 through the read/write circuit 203.

S968, reading the first comparison result and the another sub data segment from the first storage device, and storing the currently read first comparison result and the another sub data segment in the second storage device.

Specifically, after the arithmetic circuit 12 obtains the arithmetic instruction, it sends a read-write request to the first storage device 13 according to the arithmetic instruction and the data reading capacity, and then the read-write circuit 203 reads the first comparison result and the other sub-data segment from the first storage device 13 according to the preset data reading mode and according to the read-write request, and stores the currently read first comparison result and the other sub-data segment to the second storage device 203.

Optionally, another sub-data segment may be sequentially read according to the order of storing addresses in the second storage device 203, or the sub-data segment may not be read according to the order of storing addresses, which is not specifically limited herein.

And S970, comparing the first comparison result with the N pieces of sub data in the other sub data segment one by one according to the operation instruction, and storing the minimum value obtained by comparison into the second storage device and the first storage device as a second comparison result.

And the second comparison result refers to the minimum value of the first segment of sub-data segment and the other segment of sub-data segment.

Specifically, the arithmetic circuit 12 compares the first comparison result with the N pieces of sub data in the other sub data segment one by one according to the obtained arithmetic instruction, that is, compares the minimum value in the first sub data segment with the N pieces of sub data in the other sub data segment one by one, and uses the minimum value obtained by the comparison as the second comparison result, and then stores the obtained second comparison result in the second storage device 201, and stores the second comparison result in the first storage device 13 through the read/write circuit 203.

And performing circular comparison operation in the same way until the comparison operation of all the subdata segments is completed to obtain the minimum value of all the subdata in the source operand.

Specifically, the loop comparison operation may be implemented by controlling the counter to increment once or decrement once until the counter is incremented from the initial value to the target number of loops, or the counter is decremented from the target number of loops to the initial value. In the embodiment of the present application, the initial value of the counter may be 0.

Further, the target cycle number is calculated according to the data size of the source operand. After the first comparison result of the second storage device 201 is stored in the first storage device 13 through the read-write circuit 203, the counter is controlled to accumulate once, and then the first comparison result and another sub data segment are continuously read from the first storage device 13 according to the operation instruction and the data reading capacity until the counter is accumulated from 0 to the target cycle number. Or controls the counter to decrement once until the counter decrements the target number of cycles to 0, stopping the read operation from the first memory means 13.

According to the data processing method, the data are read circularly and compared, the comparison result of each circulation is stored into the first storage device continuously, exclusive access is realized on the first storage device, other processor cores are prevented from accessing the first storage device, and atomicity of atomic operation is guaranteed.

In one embodiment, the step S966 may further include:

and comparing the first subdata with the second subdata in the subdata segment, and when the first subdata is smaller than or equal to the second subdata, continuously comparing the first subdata with the third subdata until the comparison with the Nth subdata is completed to obtain the minimum value in the N subdata.

Specifically, after the sub data segment is obtained, the arithmetic circuit 12 compares a first sub data with a second sub data in the currently read sub data segment, and when the first sub data is smaller than or equal to the second sub data, the first sub data is continuously compared with a third sub data until the comparison with the nth sub data is completed, and a minimum value of the N sub data is used as a first comparison result.

Further, after the arithmetic circuit 12 obtains the first comparison result and the other sub-data segment, the currently read first comparison result and each sub-data in the other sub-data segment are sequentially compared until the comparison with the nth sub-data in the other sub-data segment is completed, and the minimum value obtained by the comparison is used as the first comparison result.

For example, the currently read first sub-data segment is a plurality of data, and the first sub-data segment has a ₁ = {1,2,3,4,5,6,7,8}, will a ₁ A plurality of subdata in the data processing unit are compared one by one to obtain a ₁ Minimum value 1 of (1), and storing the minimum value 1 in the first storage device 13 and the second storage device 201 as a first comparison result; reading the first comparison result and the second sub data segment a again ₂ = {0,3,4,5,2, 10,7,2}, and the first comparison result 1 is compared with a ₂ Each sub-data in the first storage device 13 is compared in sequence to obtain a minimum value 0, and the minimum value 0 is stored into the second storage device 201 as a second comparison result; and so on.

In this embodiment, the N pieces of sub data in the sub data segments are compared one by one, and the minimum value obtained by the comparison is stored in the first storage device 13, thereby completing the search for the minimum value.

In one embodiment, the data processing method shown in fig. 11 may include the following steps:

and S1000, acquiring the data length of the source operand according to the operation instruction.

Specifically, in the process of analyzing the operation instruction, the controller circuit 11 may obtain the Data length of the source operand according to the Data size field of the operation instruction.

S2000, increasing the data length of the source operand by an operation bit to obtain the data size of the source operand; wherein the incremented operation bit is used to store the first comparison result.

Specifically, in the decoding stage, the controller circuit 11 needs to add an operation bit after the actual length of the source operand for storing the first comparison result, and take the data length after the addition of the operation bit as the data size of the source operand. Optionally, one or more operation bits may be added, which is not specifically limited herein, and is preferably added.

For example: the controller circuit 11 analyzes that the data size of the source operand is 1000 bytes according to the operation instruction, adds operation bits (assuming that the operation bits are 1 byte) to the source operand, and obtains the data size of the source operand of 1001 bytes.

In one embodiment, with continued reference to fig. 8, an instruction unpacking method is provided and may include:

Specifically, the instruction processing circuit 111 obtains the operation instruction, analyzes the data size of the first source operand according to the operation instruction, and sends the data size of the first source operand to the loop number processing circuit 115.

S600, obtaining target cycle times and data reading capacity according to the data size of the first source operand and preset splitting granularity.

The preset splitting granularity is stored in the splitting granularity circuit 114, and may be a certain storage space in the on-chip static memory.

Specifically, the cycle count processing circuit 115 receives the data size of the first source operand, and calculates the target cycle count according to the data size of the first source operand and the preset splitting granularity. The data reading capacity calculation circuit 116 calculates the data reading capacity according to the cycle count sent by the cycle count processing circuit 115, the data size of the first source operand sent by the instruction processing circuit 111, and the preset splitting granularity. And sends the data reading capacity and the number of cycles to the arithmetic circuit 12.

Alternatively, the formula for calculating the target number of cycles may be:

the Count is expressed as a target cycle number, the data size is expressed as a data size of the first source operand, and the splitting granularity is a preset splitting granularity.

The formula for calculating the data read capacity may be:

data real size = min { unprocessed data size, split granularity }

And S700, reading the first subdata from the first storage device according to the operation instruction and the data reading capacity and a preset data reading mode, and storing the currently read first subdata into the second storage device.

Specifically, the arithmetic circuit 12 receives the arithmetic instruction, initiates a read-write request to the first storage device 13 according to the arithmetic instruction and the data reading capacity, reads the first subdata from the first storage device 13 through the read-write circuit 203 according to a preset data reading mode, and stores the currently read first subdata into the second storage device 201.

And S800, executing operation according to the operation instruction, and storing the obtained current operation result into the second storage device and the first storage device.

Specifically, the arithmetic circuit 12 executes a corresponding arithmetic operation according to the received arithmetic instruction, obtains a current arithmetic result, stores the current arithmetic result in the second storage device 201, and stores the current arithmetic result in the second storage device 201 in the first storage device 13.

And S900, returning to the step of obtaining the target cycle number and the data reading capacity according to the data size of the first source operand and the preset splitting granularity until the current cycle number is equal to the initial value or the current cycle number is equal to the target cycle number, and finishing the operation corresponding to the operation instruction.

Specifically, the steps may include: and decreasing the target cycle number once or accumulating the target cycle number once from the initial value, returning to the step S600, and determining the data reading capacity according to the data size of the first operand and the preset splitting granularity until the target cycle number is decreased to the initial value or accumulated to the target cycle number from the initial value. In the embodiment of the present application, the initial value of the counter may be 0.

Further, after the arithmetic circuit 12 finishes executing the arithmetic operation and stores the current arithmetic result in the first storage device 13, the loop time processing circuit 115 decrements the target loop time by one time, and then continues to execute step S600 until the target loop time is decremented to 0. Or once from 0, and then continues to step S600 until the target cycle number is reached.

Further, the arithmetic circuit 12 may perform judgment according to the operation code Scr Op in the arithmetic instruction, determine the number of the source operands, and when the number of the source operands is two, obtain the second sub-data according to the instruction, and then perform the arithmetic operation. And when the number of the source operands is three, acquiring second subdata and third subdata according to the instruction, and then executing operation. For reading the second sub data and/or the third sub data, the above description may be specifically referred to.

In the instruction disassembling method in this embodiment, the data size of the first source operand is obtained in the decoding stage, the first source operand is disassembled and cyclically read, the large data is disassembled into the small data for operation, and the operand is cyclically read according to the data reading capacity, so that the size of the data that can be accommodated during operation meets the requirement, and the operation speed is increased.

When it can be determined that three source operands participate in the operation according to the operation instruction, according to the instruction format of the operation instruction, the source operand A is used as the second source operand, and the source operand B is used as the third source operand.

It should be understood that although the various steps in the flow charts of fig. 6-12 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Also, at least some of the steps in fig. 6-12 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternatingly with other steps or at least some of the sub-steps or stages of other steps.

In one embodiment, as shown in fig. 13, there is provided a data processing apparatus including: an obtaining module 100, a reading module 200, an operation module 300 and a counting module 400, wherein:

the obtaining module 100 is configured to obtain an operation instruction.

The reading module 200 is configured to read first subdata from a first storage device according to a data reading capacity and an operation instruction in a preset data reading manner, and store the currently read first subdata in a second storage device.

The operation module 300 is configured to execute an operation according to the operation instruction, obtain a current operation result, store the current operation result in the second storage device and the first storage device, and then circularly call the reading module 200 and the operation module 300 until the operation corresponding to the operation instruction is completed.

Further, the data processing apparatus may include a counting module 400 for controlling the counter to increment or decrement once after the current operation result in the second storage device is stored in the first storage device, and then, circularly calling the reading module 200, the operation module 300, and the counting module 400 until the counter is incremented from the initial value to the target number of cycles or the counter is decremented from the target number of cycles to the initial value. In the embodiment of the present application, the initial value may be 0.

For specific limitations of the data processing apparatus, reference may be made to the above limitations of the data processing method, which are not described herein again. The various modules in the data processing apparatus described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

When the above operation instruction is Atomic MIN _ SCALAR, the operation module 300 may include a monocular minimum value operation module, wherein the obtaining module 100 is configured to obtain the operation instruction, wherein the operation instruction is used to implement a comparison operation of a source operand, the source operand includes sub-data segments, and each sub-data segment includes N sub-data; a reading module 200, configured to read a sub data segment from a first storage device according to a data reading capacity and an operation instruction and according to a preset data reading manner, and store the currently read sub data segment in a second storage device, where the first storage device is an off-chip storage device, and the second storage device is an on-chip storage device; the monocular minimum value operation module 313 is used for comparing the N sub-data in the sub-data segments one by one according to the operation instruction to obtain the minimum value in the N sub-data, and storing the minimum value as a first comparison result into the second storage device and the first storage device; reading the first comparison result and the other subdata segment from the first storage device, and storing the currently read first comparison result and the other subdata segment into a second storage device; and comparing the first comparison result with the N pieces of sub data in the other sub data segment one by one according to the operation instruction, and storing the minimum value obtained by comparison into a second storage device and a first storage device as a second comparison result. Then, the reading module 200 and the monocular minimum value operation module 313 are called in a loop until the operation corresponding to the operation instruction is completed. Further, the data processing apparatus may include a counting module 400 for controlling the counter to increment or decrement once after the current operation result in the second storage device is stored in the first storage device, and then, the reading module 200, the monocular minimum value operation module 313 and the counting module 400 are called in a loop until the counter is incremented from the initial value to the target number of loops or the counter is decremented from the target number of loops to the initial value. In the embodiment of the present application, the initial value may be 0.

In the embodiment of the present invention, the specific structure of the operation module is similar to that of the operation circuit in the embodiment, and refer to fig. 2 and the description above.

In one embodiment, as shown in fig. 14, there is provided an instruction dismantling device including: an obtaining module 500, a cycle number processing module 600, a data reading capacity calculating module 700, a reading module 800, and an operation module 900, wherein:

the obtaining module 500 is configured to obtain an operation instruction, and analyze the data size of the first source operand according to the operation instruction.

The cycle number processing module 600 is configured to obtain a cycle number according to the data size of the first source operand and a preset splitting granularity. The data reading capacity calculation module 700 is configured to obtain a data reading capacity according to the data size of the first source operand and a preset splitting granularity. The reading module 800 is configured to read the first subdata from the first storage device according to the operation instruction and the data reading capacity in a preset data reading manner, and store the currently read first subdata into the second storage device. The operation module 900 is configured to execute an operation according to the operation instruction, and store the obtained current operation result in the second storage device and the first storage device. The cycle number processing module 600 is configured to decrement the cycle number once or accumulate the cycle number once from 0, and then call the data reading capacity calculating module 700, the reading module 800, and the operation module 900 until the current cycle number is equal to the initial value or the current cycle number is equal to the target cycle number, so as to complete the operation corresponding to the operation instruction.

In the embodiment of the present application, the operation principle of the operation module 900 is the same as the operation principle of the operation module 300 and the operation circuit, and the description thereof can be referred to specifically. For specific limitations of the instruction disassembling device, reference may be made to the above limitations on the instruction disassembling method, and details are not described herein again. The various modules in the data processing apparatus described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

and acquiring an operation instruction, wherein the operation instruction is used for realizing operation among source operands, and the first source operand comprises at least one first subdata.

Reading first subdata from a first storage device according to the data reading capacity and the operation instruction and a preset data reading mode, and storing the currently read first subdata into a second storage device; the first storage device 13 is an off-chip storage device, and the second storage device 201 is an on-chip storage device.

And executing the operation according to the operation instruction, and storing the obtained current operation result into the second storage device and the first storage device.

And then, returning to the step of reading the first subdata from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode until the operation corresponding to the operation instruction is completed.

It should be clear that, the steps implemented when the computer program in the embodiment of the present application is executed by the processor are consistent with the execution process of each step of the method in the above embodiments, and specific reference may be made to the above description, and no further description is given here.

acquiring an operation instruction, and analyzing the data size of a first source operand according to the operation instruction, wherein the first source operand comprises at least one piece of first subdata;

obtaining cycle times and data reading capacity according to the data size of the first source operand and a preset splitting granularity;

reading first subdata from a first storage device according to an operation instruction and a data reading capacity and a preset data reading mode, and storing the currently read first subdata into a second storage device;

executing operation according to the operation instruction, storing the obtained current operation result into the second storage device, and storing the current operation result in the second storage device into the first storage device;

and then, returning to the step of determining the data reading capacity according to the data size of the first source operand and the preset splitting granularity until the operation corresponding to the operation instruction is completed.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent application shall be subject to the appended claims.

Claims

1. A method of data processing, the method comprising:

obtaining an operation instruction, wherein the operation instruction is used for realizing comparison operation of a source operand, the source operand comprises sub data segments, and each sub data segment comprises N pieces of sub data;

reading the sub data segment from a first storage device according to data reading capacity and the operation instruction and according to a preset data reading mode, and storing the currently read sub data segment to a second storage device, wherein the first storage device is an off-chip storage device, and the second storage device is an on-chip storage device;

according to the operation instruction, comparing the first comparison result with N pieces of sub data in another sub data segment one by one, and storing a minimum value obtained by comparison into the second storage device and the first storage device as a second comparison result;

and comparing the cyclically read data, storing the minimum value obtained by each comparison into the second storage device as a current comparison result, and storing the current comparison result of the second storage device into the first storage device through the read-write circuit.

2. The data processing method according to claim 1, wherein the step of comparing the N sub-data in the sub-data segment one by one to obtain a minimum value of the N sub-data, and storing the minimum value as a first comparison result in the second storage device and the first storage device comprises:

and comparing the first subdata with the second subdata in the subdata segment, and when the first subdata is larger than or equal to the second subdata, continuously comparing the first subdata with the third subdata until the comparison with the Nth subdata is completed to obtain the minimum value in the N subdata.

3. The data processing method of claim 1, wherein the method further comprises:

and after the first comparison result is stored in the first storage device, taking the address next to the end address of the sub-data segment read last time as the starting address of the currently read sub-data segment.

4. The data processing method of claim 1, wherein the method further comprises:

and when the first comparison result is stored in the first storage device, the storage address of the first comparison result is consistent with the storage address of the currently read subdata segment.

5. The data processing method of claim 1, wherein the method further comprises:

obtaining the data size of the source operand according to the operation instruction;

and obtaining the target cycle number according to the data size of the source operand and the preset splitting granularity.

6. The data processing method of claim 5, wherein the step of obtaining the data size of the source operand according to the operation instruction comprises:

obtaining the data length of the source operand according to the operation instruction;

increasing the data length of the source operand by an operation bit to obtain the data size of the source operand;

wherein the increased operation bits are used to store the first comparison result.

7. The data processing method of claim 1,

the instruction format of the operation instruction comprises an instruction type, a source operand, a target operand and an operation code;

the operation code is used for configuring the number of source operands;

the target operand is used to represent a current comparison result.

8. The data processing method of claim 1, wherein the method further comprises:

and controlling the counter to accumulate once or decrement once, and then returning to the step of reading the first subdata from the first storage device according to the data reading capacity and the operation instruction and a preset data reading mode until the counter accumulates to the target cycle number from an initial value or the counter decrements to the initial value from the target cycle number, so as to finish the operation corresponding to the operation instruction.

9. A processor for use in a data processing method, the processor comprising an arithmetic circuitry, a read-write circuitry, and a second storage means disposed adjacent to the arithmetic circuitry, the second storage means being connectable by the read-write circuitry to a first storage means external to the processor;

the operation instruction is used for realizing comparison operation of a source operand, wherein the source operand comprises sub-data segments, and each sub-data segment comprises N pieces of sub-data;

the arithmetic circuit is used for reading the sub data segments from the first storage device according to data reading capacity and the arithmetic instruction and according to a preset data reading mode, comparing N sub data in the sub data segments one by one to obtain a minimum value in the N sub data, and storing the minimum value as a first comparison result into the second storage device; storing the first comparison result of the second storage device into the first storage device through the read-write circuit; reading the first comparison result and the other sub data segment from the first storage device, comparing the first comparison result with the N sub data in the other sub data segment one by one, and storing a minimum value obtained by comparison into the second storage device as a second comparison result; storing, by the read-write circuitry, the second comparison result of the second storage device in the first storage device; and comparing the cyclically read data, storing the minimum value obtained by each comparison into the second storage device as a current comparison result, and storing the current comparison result of the second storage device into the first storage device through the read-write circuit.

10. The processor of claim 9, further comprising a data selector, wherein the arithmetic circuit comprises a minimum value operation module, the data selector is connected between the arithmetic circuit and the read/write circuit, and the data selector is configured to gate a connection path between the minimum value operation module and the read/write circuit;

the minimum value operation module is used for comparing N pieces of sub data in the sub data segments one by one according to the operation instruction to obtain a minimum value in the N pieces of sub data, and storing the minimum value serving as a first comparison result into the second storage device; storing the first comparison result of the second storage device into the first storage device through the read-write circuit and the data selector; reading the first comparison result and the other sub data segment from the first storage device, and storing the currently read first comparison result and the other sub data segment to a second storage device; storing the second comparison result of the second storage device in the first storage device through the read-write circuit and the data selector.

11. The processor according to claim 10, wherein the minimum operation module comprises a comparison unit and a result output unit connected to the comparison unit;

the comparison unit is used for comparing a first subdata with a second subdata in the subdata segment, and when the first subdata is larger than or equal to the second subdata, continuously comparing the first subdata with a third subdata until the comparison with the Nth subdata is completed;

the comparison unit is further configured to compare the first comparison result with a first sub-data in another sub-data segment, and when the first comparison result is greater than or equal to the first sub-data in the another sub-data segment, continue to compare the first comparison result with a second sub-data in the another sub-data segment until the comparison with the nth sub-data in the another sub-data segment is completed;

the result output unit is used for taking the minimum value of N pieces of sub data in the sub data segment as the first comparison result;

the result output unit is further configured to take the first comparison result and a minimum value of the N sub data in the other sub data segment as the second comparison result.

12. The processor according to any of claims 9-11, wherein the arithmetic circuitry comprises a master processing circuit and more than one slave processing circuits, each of the more than one slave processing circuits being connected to the master processing circuit;

the minimum value operation module is arranged in the main processing circuit.

13. A data processing apparatus, characterized in that the apparatus comprises:

the system comprises an acquisition module, a comparison module and a comparison module, wherein the acquisition module is used for acquiring an operation instruction, the operation instruction is used for realizing comparison operation of a source operand, the source operand comprises sub-data segments, and each sub-data segment comprises N pieces of sub-data;

the operation module is used for comparing N pieces of sub data in the sub data segments one by one according to the operation instruction to obtain a minimum value in the N pieces of sub data, and storing the minimum value serving as a first comparison result into the second storage device and the first storage device; reading the first comparison result and the other sub data segment from the first storage device, and storing the currently read first comparison result and the other sub data segment to a second storage device; according to the operation instruction, comparing the first comparison result with N pieces of sub data in another sub data segment one by one, and storing a minimum value obtained by comparison to the second storage device and the first storage device as a second comparison result; and comparing the cyclically read data, storing the minimum value obtained by each comparison into the second storage device as a current comparison result, and storing the current comparison result of the second storage device into the first storage device through the read-write circuit.

14. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 8.