# US 2014/0208072 A1 · User-level manager to handle multi-processing on many-core coprocessor-based systems

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

- Number: US 2014/0208072 A1
- Document ID: US20140208072A1
- Status: Published application
- Kind: A1
- Application number: 13/858,034
- Filed: 2013-04-06
- Published: 2014-07-24
- Earliest claimed: 2013-01-18
- Assignees: NEC Laboratories America, Inc.
- Applicants: NEC Laboratories America, Inc.
- Inventors: Srihari Cadambi; Kunal Rao; Srimat T. Chakradhar; Rajat Phull; Giuseppe Coviello; Murugan Sankaradass; Cheng-Hong Li
- Web page: https://patentgrep.com/patent/US20140208072A1

## CPC

- G06F 9/5044: {considering hardware capabilities}
- G06F 9/5033: {considering data affinity}
- G06F 2209/509: Offload
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method is disclosed to manage a multi-processor system with one or more multiple-core coprocessors by intercepting coprocessor offload infrastructure application program interface (API) calls; scheduling user processes to run on one of the coprocessors; scheduling offloads within user processes to run on one of the coprocessors; and affinitizing offloads to predetermined cores within one of the coprocessors by selecting and allocating cores to an offload, and obtaining a thread-to-core mapping from a user.

## Claims

### Claim 1

A multi-processor system, comprising:
a computer populated with one or more multiple-core coprocessors; and
a management code to control user processes containing offload blocks, including code to:
intercept coprocessor offload infrastructure application program interface (API) calls;
schedule user processes to run on one of the coprocessors;
schedule offloads within user processes to run on one of the coprocessors;
select one or more cores and create a core container consisting of the selected on a coprocessor to be used by an offload; and
obtain an optional thread-to-core mapping from a user and bind threads of an offload to the selected cores.

### Claim 2 (depends on 1)

The system of claim 1, comprising one or more procedures, each of which selecting a process and offloading the process for scheduling.

### Claim 3 (depends on 1)

The system of claim 1, comprising code with thread over-scheduling factors to enhance performance.

### Claim 4 (depends on 1)

The system of claim 1, comprising code with memory over-scheduling factors to enhance performance.

### Claim 5 (depends on 1)

The system of claim 1, comprising code to select cores to be assigned to an offload.

### Claim 6 (depends on 1)

The system of claim 1, comprising code to apply instances of data structures from which a status of cores and information about selected cores are used to enhance core selection.

### Claim 7 (depends on 1)

The system of claim 1, comprising code to communicate using messages between components of the management code.

### Claim 8 (depends on 1)

The system of claim 1, comprising providing management code client module to control each user process and monitor processes from the host-side.

### Claim 9 (depends on 1)

The system of claim 1, wherein each coprocessor comprise a plurality of cores running an operating system, one or more buses connecting one or more host processors and each coprocessor.

### Claim 10 (depends on 1)

The system of claim 1, wherein the management code:
selects and schedules a pending process to a coprocessor with free memory; and
examines offload queues corresponding to each coprocessor, and dispatches an offload to each coprocessor with free threads,
wherein processes and offloads are selected based on an aging-based first-fit heuristic.

### Claim 11

A method to manage a multi-processor system with one or more multiple-core coprocessors, comprising:
intercepting coprocessor offload infrastructure application program interface (API) calls;
scheduling user processes to run on one of the coprocessors;
scheduling offloads within user processes to run on one of the coprocessors; and
affinitizing offloads to predetermined cores within one of the coprocessors by selecting and allocating cores to an offload, and obtaining a thread-to-mapping from a user.

### Claim 12 (depends on 11)

The method of claim 11, comprising applying an aging-based first-fit procedure for process and offload scheduling.

### Claim 13 (depends on 11)

The method of claim 11, comprising applying thread or memory over-scheduling factors to enhance performance.

### Claim 14 (depends on 11)

The method of claim 11, comprising applying greedy core selection such that offloads from the same process get preference to use the same cores.

### Claim 15 (depends on 11)

The method of claim 11, comprising applying bitmaps to enhance core selection.

### Claim 16 (depends on 11)

The method of claim 11, comprising messaging between components of the management code.

### Claim 17 (depends on 11)

The method of claim 11, comprising providing management code client module to control each user process.

### Claim 18 (depends on 11)

The method of claim 11, comprising monitoring processes from the host-side.

### Claim 19 (depends on 11)

The method of claim 11, wherein each coprocessor comprise a plurality of X86 cores running Linux, Peripheral Component Interconnect Express Interface, and memory controllers connected with a bidirectional Interprocessor Network (IPN) ring.

### Claim 20 (depends on 11)

The method of claim 11, further comprising:
selecting and scheduling a pending process to a coprocessor with free memory; and
examining offload queues corresponding to each coprocessor, and dispatch an offload to each coprocessor with free threads,
wherein processes and offloads are selected based on an aging-based first-fit heuristic.
