# US 2021/0240473 A1 · Processor device

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0240473 A1
- Document ID: US20210240473A1
- Status: Published application
- Kind: A1
- Application number: 17/049,197
- Filed: 2019-04-23
- Published: 2021-08-05
- Earliest claimed: 2018-04-24
- Assignees: ArchiTek Corporation
- Applicants: ArchiTek Corporation
- Inventors: Shuichi TAKADA
- Web page: https://patentgrep.com/patent/US20210240473A1

## CPC

- G06F 9/30043: {LOAD or STORE instructions; Clear instruction}
- G06F 5/16: Multiplexed systems, i.e. using two or more similar devices which are alternately accessed for enqueue and dequeue operations, e.g. ping-pong buffers
- G06F 9/3012: {Organisation of register space, e.g. banked or distributed register file}
- G06F 9/3013: {according to data content, e.g. floating-point registers, address registers}
- G06F 9/3824: {Operand accessing}
- G06F 9/3887: {controlled by a single instruction for multiple data lanes [SIMD]}
- G06F 12/0875: with dedicated cache, e.g. instruction or stack
- G06F 15/167: using a common memory, e.g. mailbox
- G06F 15/80: comprising an array of processing units with common control, e.g. single instruction multiple data processors (G06F15/82 takes precedence {; for correlation function computation G06F17/15})
- G06N 20/00: Machine learning
- G06F 2212/452: Instruction code
- G06N 3/063: using electronic means
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

In a processor device according to the present invention, a memory access unit reads data to be processed from an external memory and writes the data to a first register group that a plurality of processors does not access among a plurality of register groups. A control unit sequentially makes each of the plurality of processors implement a same instruction, in parallel with changing an address of a register group that stores the data to be processed. A scheduler, based on specified scenario information, specifies an instruction to be implemented and a register group to be accessed for the plurality of processors, and specifies a register group to be written to among the plurality of register groups and data to be processed that is to be written for the memory access unit.

## Claims

### Claim 1

A processor device, comprising:
an external memory;
a plurality of processors;
a plurality of register groups, each of the plurality of register groups including a plurality of registers to store data to be processed;
a memory access unit configured to read data to be processed from the external memory and write the data to a first register group that the plurality of processors does not access among the plurality of register groups;
a control unit configured to sequentially make each of the plurality of processors implement a same instruction in parallel with changing an address of one register group that stores the data to be processed; and
a scheduler configured to, based on specified scenario information, specify an instruction to be implemented and a register group to be accessed for the plurality of processors, and specify a register group to be written to among the plurality of register groups and data to be processed that is to be written for the memory access unit.

### Claim 2 (depends on 1)

The processor device according to claim 1, wherein a bank that is defined for each of the plurality of processors is further provided, and each of the plurality of processors is configured to read data from a register group that is defined as the bank and implement a given instruction.

### Claim 3 (depends on 1)

The processor device according to claim 1, wherein the plurality of processors is configured to write data that is obtained as a result of implementing the instruction to the first register group from which the data to be processed that is an original data of the data is read, and the memory access unit is configured to read the data to be written to the first register group by the plurality of processors and write the data to the external memory.

### Claim 4 (depends on 1)

The processor device according to claim 1, wherein, when the plurality of processors implements the same instruction by using the data stored in the first register group, the memory access unit is configured to read the data to be processed from the external memory according to an instruction to be implemented next by the plurality of processors and write the data to be processed to a second register group, the second register group being included in the plurality of register groups and including a plurality of registers different from the first register group.

### Claim 5 (depends on 1)

The processor device according to claim 1, wherein the scheduler is configured to increase operation speed of the plurality of processors according to an instruction to be implemented by the plurality of processors.

### Claim 6 (depends on 1)

The processor device according to claim 1, wherein the scheduler is configured to be inputted a leaning model for inference processing and learning processing in deep learning, the learning model comprised of a node and a link, and store input data to each node of the learning model in each register of the plurality of register groups, and the plurality of processors is configured to perform an arithmetic operation to be performed in each node.

### Claim 7 (depends on 2)

The processor device according to claim 2, wherein the plurality of processors is configured to write data that is obtained as a result of implementing the instruction to the first register group from which the data to be processed that is an original data of the data is read, and the memory access unit is configured to read the data to be written to the first register group by the plurality of processors and write the data to the external memory.

### Claim 8 (depends on 2)

The processor device according to claim 2, wherein, when the plurality of processors implements the same instruction by using the data stored in the first register group, the memory access unit is configured to read the data to be processed from the external memory according to an instruction to be implemented next by the plurality of processors and write the data to be processed to a second register group, the second register group being included in the plurality of register groups and including a plurality of registers different from the first register group.

### Claim 9 (depends on 3)

The processor device according to claim 3, wherein, when the plurality of processors implements the same instruction by using the data stored in the first register group, the memory access unit is configured to read the data to be processed from the external memory according to an instruction to be implemented next by the plurality of processors and write the data to be processed to a second register group, the second register group being included in the plurality of register groups and including a plurality of registers different from the first register group.

### Claim 10 (depends on 7)

The processor device according to claim 7, wherein, when the plurality of processors implements the same instruction by using the data stored in the first register group, the memory access unit is configured to read the data to be processed from the external memory according to an instruction to be implemented next by the plurality of processors and write the data to be processed to a second register group, the second register group being included in the plurality of register groups and including a plurality of registers different from the first register group.
