# US 2019/0244140 A1 · Data path for gpu machine learning training with key value ssd

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

- Number: US 2019/0244140 A1
- Document ID: US20190244140A1
- Status: Published application
- Kind: A1
- Application number: 15/942,218
- Filed: 2018-03-30
- Published: 2019-08-08
- Earliest claimed: 2018-02-02
- Assignees: Samsung Electronics Co., Ltd.
- Applicants: Samsung Electronics Co., Ltd.
- Inventors: Joo Hwan Lee; Yang Seok Ki
- Web page: https://patentgrep.com/patent/US20190244140A1

## CPC

- G06N 20/00: Machine learning
- G06F 3/0611: {in relation to response time}
- G06F 3/064: {Management of blocks}
- G06F 3/0659: {Command handling arrangements, e.g. command buffers, queues, command scheduling}
- G06F 3/0679: {Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]}
- G06F 12/1081: for peripheral access to main memory, e.g. direct memory access [DMA]
- G06F 13/102: {where the program performs an interfacing function, e.g. device driver (G06F13/105 takes precedence; contention policies within device drivers G06F9/4881; scheduling within device drivers G06F9/52)}
- G06F 2212/621: Coherency control relating to peripheral accessing, e.g. from DMA or I/O device
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

A system and method for machine learning. The system includes a GPU with a GPU memory, and a key value storage device connected to the GPU memory. The method includes, writing, by the GPU, a key value request to a key value request queue in a input-output region of the GPU memory, the key value request including a key. The method further includes reading, by the key value storage device, the key value request from the key value request queue, and writing, by the key value storage device, in response to the key value request, a value to the input-output region of the GPU memory, the value corresponding to the key of the key value request.

## Claims

### Claim 1

A method for machine learning, the method comprising:
writing, by a first graphics processing unit, a first key value request to a key value request queue in a first input-output region of a first memory connected to the first graphics processing unit, the first key value request including a key;
reading, by a first key value storage device connected to the first memory, the first key value request from the key value request queue, and
writing, by the first key value storage device, in response to the first key value request, a first value to the first input-output region of the first memory, the first value corresponding to the key of the first key value request.

### Claim 2 (depends on 1)

The method of claim 1, further comprising, performing, by the first key value storage device, a key lookup, in the first key value storage device, to retrieve the first value.

### Claim 3 (depends on 1)

The method of claim 1, wherein the first key value request includes a return-value region, the return-value region being a region allocated for the first value.

### Claim 4 (depends on 3)

The method of claim 3, wherein the writing of the first value to the first input-output region of the first memory comprises writing the first value to the return-value region.

### Claim 5 (depends on 1)

The method of claim 1, wherein the writing of the first value to the first input-output region of the first memory comprises writing the first value to a return-value queue in the first input-output region of the first memory.

### Claim 6 (depends on 1)

The method of claim 1, further comprising configuring, by a host connected to the first key value storage device and to the first graphics processing unit:
the first key value storage device to access the first input-output region of the first memory to receive key value requests and to write values in response to the key value requests; and
the first graphics processing unit to store key value requests in the first input-output region of the first memory and to read values from the first input-output region of the first memory.

### Claim 7 (depends on 6)

The method of claim 6, further comprising, writing, by a second graphics processing unit connected to the host, a second key value request to a key value request queue in an input-output region of a second memory connected to the second graphics processing unit, the second key value request including a key;
reading, by a second key value storage device connected to the host and to the second memory, the second key value request from the key value request queue, and
writing, by the second key value storage device, in response to the second key value request, a second value to the input-output region of the second memory, the second value corresponding to the key of the second key value request.

### Claim 8 (depends on 7)

The method of claim 7, further comprising:
performing, by the first key value storage device, a key lookup, in the first key value storage device, to retrieve the first value, and
performing, concurrently with the performing of the key lookup by the first key value storage device, a key lookup, by the second key value storage device, in the second key value storage device, to retrieve the second value.

### Claim 9 (depends on 1)

The method of claim 1, wherein the reading, by the first key value storage device of the first key value request comprises reading the first key value request via peer-to-peer direct memory access.

### Claim 10 (depends on 1)

The method of claim 1, wherein the writing, by the first key value storage device, of the first value, comprises writing the first value via peer-to-peer direct memory access.

### Claim 11 (depends on 10)

The method of claim 10, wherein the first key value storage device is connected to the first graphics processing unit by a peripheral component interconnect connection.

### Claim 12 (depends on 1)

The method of claim 1, further comprising writing, by the first graphics processing unit, a second key value request to the key value request queue, after the writing, by the first graphics processing unit, of the first key value request and before the writing, by the writing, by the first key value storage device, of the first value.

### Claim 13 (depends on 1)

The method of claim 1, further comprising:
writing, by the first graphics processing unit, a second key value request to a key value request queue in a second input-output region of the first memory, the second key value request including a key;
reading, by a second key value storage device connected to the first memory, the second key value request from the key value request queue of the second input-output region of the first memory, and
writing, by the second key value storage device, in response to the second key value request, a second value to the second input-output region of the first memory, the second value corresponding to the key of the second key value request.

### Claim 14 (depends on 13)

The method of claim 13, further comprising:
performing, by the first key value storage device, a key lookup, in the first key value storage device, to retrieve the first value, and
performing, concurrently with the performing of the key lookup by the first key value storage device, a key lookup, by the second key value storage device, in the second key value storage device, to retrieve the second value.

### Claim 15

A system for machine learning, the system comprising:
a graphics processing unit;
a memory connected to the graphics processing unit; and
a key value storage device;
the key value storage device being connected to the graphics processing unit by a peripheral component interconnect connection;
the graphics processing unit being configured to perform memory-mapped input and output operations in an input-output region of the memory, and to write one or more key value requests to a key value request queue within the input-output region;
the key value storage device being configured to:
perform memory-mapped input and output operations in the input-output region;
read the one or more key value requests from the key value request queue; and
in response to a key value request of the one or more key value requests, write a value in the input-output region of the memory, the value corresponding to a key of the key value request.

### Claim 16 (depends on 15)

The system of claim 15, wherein the key value request includes a return-value region, the return-value region being a region allocated for the value.

### Claim 17 (depends on 16)

The system of claim 16, wherein writing of the value to the input-output region of the memory comprises writing the value to the return-value region.

### Claim 18 (depends on 15)

The system of claim 15, wherein the writing of the value to the input-output region of the memory comprises writing the value to a return-value queue in the input-output region of the memory.

### Claim 19

A system for machine learning, the system comprising:
a graphics processing unit;
a key value storage device; and
shared memory means for communication between the graphics processing unit and the key value storage device;
the graphics processing unit being configured send one or more key value requests to the key value storage device via the shared memory means for communication,
the key value storage device being configured to:
receive the one or more key value requests; and
in response to a key value request of the one or more key value requests, send a value to the graphics processing unit via the shared memory means for communication, the value corresponding to a key of the key value request.

### Claim 20 (depends on 19)

The system of claim 19, wherein the shared memory means for communication comprises a memory connected to the graphics processing unit, and configured to be accessed by the key value storage device via peer-to-peer direct memory access through a peripheral component interconnect connection.
