# US 2021/0173701 A1 · Systems and methods for an operating system module for a data processing accelerator

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

- Number: US 2021/0173701 A1
- Document ID: US20210173701A1
- Status: Published application
- Kind: A1
- Application number: 16/709,468
- Filed: 2019-12-10
- Published: 2021-06-10
- Earliest claimed: 2019-12-10
- Assignees: Baidu USA LLC
- Applicants: Baidu USA LLC
- Inventors: Yueqiang Cheng; Yong Liu
- Web page: https://patentgrep.com/patent/US20210173701A1

## CPC

- G06F 9/4881: {Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues}
- G06F 9/5016: {the resource being the memory}
- G06F 9/5027: {the resource being a machine, e.g. CPUs, Servers, Terminals}
- G06N 20/00: Machine learning
- G06F 2209/509: Offload
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Embodiments disclose systems and methods of an operating system module for a data processing accelerator (DPA). The operating system module can schedule jobs and manage resources (e.g., computing units, memory, IO bandwidth) of the DPA. For one embodiment, a system receives a request, by a DPA, the request sent by an application to execute a data processing task. The system schedules, by an operating system module of the DPA, one or more commands based on one or more kernels to complete the data processing task. The system generates, by the operating system module of the DPA, a completion signal upon completion of the one or more scheduled commands. The system sends a result back to the application acknowledging completion of the one or more commands for the data processing task based on the completion signal.

## Claims

### Claim 1

A computer-implemented method for processing data by a data processing accelerator, the method comprising:
receiving a request by a data processing accelerator, the request sent by an application to execute a data processing task;
scheduling, by an operating system module of the data processing accelerator, one or more commands based on one or more kernels to complete the data processing task;
generating, by the operating system module of the data processing accelerator, a completion signal upon completion of the one or more scheduled commands; and
sending a result back to the application acknowledging completion of the one or more scheduled commands for the data processing task based on the completion signal.

### Claim 2 (depends on 1)

The method of claim 1, further comprising creating an application session context for the data processing task, wherein the application session context is isolated from contexts of application sessions of other data processing tasks of the data processing accelerator.

### Claim 3 (depends on 2)

The method of claim 2, wherein the isolation isolates resources for the data processing task from resources for the other data processing tasks of the data processing accelerator.

### Claim 4 (depends on 1)

The method of claim 1, further comprising allocating, by the operating system module of the data processing accelerator, resources for the data processing task, wherein the resources include computing resources and memory resources.

### Claim 5 (depends on 1)

The method of claim 1, wherein the request is sent by an application through a host, wherein the host assigns one or more data processing tasks to one or more data processing accelerators.

### Claim 6 (depends on 5)

The method of claim 5, wherein the host is relieved of the scheduling and managing of the one or more commands required to complete a data processing task.

### Claim 7 (depends on 1)

The method of claim 1, further comprising receiving one or more kernels from the host, the one or more kernels comprising instructions to execute the one or more commands to complete a data processing task.

### Claim 8

A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform one or more operations, the operations comprising:
receiving a request by a data processing accelerator, the request sent by an application to execute a data processing task;
scheduling, by an operating system module of the data processing accelerator, one or more commands based on one or more kernels to complete the data processing task;
generating, by the operating system module of the data processing accelerator, a completion signal upon completion of the one or more scheduled commands; and
sending a result back to the application acknowledging completion of the one or more scheduled commands for the data processing task based on the completion signal.

### Claim 9 (depends on 8)

The non-transitory machine-readable medium of claim 8, wherein the operations further comprise creating an application session context for the data processing task, wherein the application session context is isolated from contexts of application sessions of other data processing tasks of the data processing accelerator.

### Claim 10 (depends on 9)

The non-transitory machine-readable medium of claim 9, wherein the isolation isolates resources for the data processing task from resources for the other data processing tasks of the data processing accelerator.

### Claim 11 (depends on 8)

The non-transitory machine-readable medium of claim 8, wherein the operations further comprise allocating, by the operating system module of the data processing accelerator, resources for the data processing task, wherein the resources include computing resources and memory resources.

### Claim 12 (depends on 8)

The non-transitory machine-readable medium of claim 8, wherein the request is sent by an application through a host, wherein the host assigns one or more data processing tasks to one or more data processing accelerators.

### Claim 13 (depends on 12)

The non-transitory machine-readable medium of claim 12, wherein the host is relieved of the scheduling and managing of the one or more commands required to complete a data processing task.

### Claim 14 (depends on 8)

The non-transitory machine-readable medium of claim 8, wherein the operations further comprise receiving one or more kernels from the host, the one or more kernels comprising instructions to execute the one or more commands to complete a data processing task.

### Claim 15

A data processing system, comprising:
a processor; and
a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations, the operations including receiving a request by a data processing accelerator, the request sent by an application to execute a data processing task;
scheduling, by an operating system module of the data processing accelerator, one or more commands based on one or more kernels to complete the data processing task;
generating, by the operating system module of the data processing accelerator, a completion signal upon completion of the one or more scheduled commands; and
sending a result back to the application acknowledging completion of the one or more scheduled commands for the data processing task based on the completion signal.

### Claim 16 (depends on 15)

The system of claim 15, wherein the operations further comprise creating an application session context for the data processing task, wherein the application session context is isolated from contexts of application sessions of other data processing tasks of the data processing accelerator.

### Claim 17 (depends on 16)

The system of claim 16, wherein the isolation isolates resources for the data processing task from resources for the other data processing tasks of the data processing accelerator.

### Claim 18 (depends on 15)

The system of claim 15, wherein the operations further comprise allocating, by the operating system module of the data processing accelerator, resources for the data processing task, wherein the resources include computing resources and memory resources.

### Claim 19 (depends on 15)

The system of claim 15, wherein the request is sent by an application through a host, wherein the host assigns one or more data processing tasks to one or more data processing accelerators.

### Claim 20 (depends on 19)

The system of claim 19, wherein the host is relieved of the scheduling and managing of the one or more commands required to complete a data processing task.

### Claim 21 (depends on 15)

The system of claim 15, wherein the operations further comprise receiving one or more kernels from the host, the one or more kernels comprising instructions to execute the one or more commands to complete a data processing task.
