# US 2024/0005136 A1 · Dynamic precision for neural network compute operations

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

- Number: US 2024/0005136 A1
- Document ID: US20240005136A1
- Status: Published application
- Kind: A1
- Application number: 18/351,124
- Filed: 2023-07-12
- Published: 2024-01-04
- Earliest claimed: 2017-04-24
- Assignees: Intel Corporation
- Applicants: Intel Corporation
- Inventors: Kamal Sinha; Balaji Vembu; Eriko Nurvitadhi; Nicolas C. Galoppo Von Borries; Rajkishore Barik; Tsung-Han Lin; Joydeep Ray; Ping T. Tang; Michael S. Strickland; Xiaoming Chen; Anbang Yao; Tatiana Shpeisman; Abhishek R. Appu; Altug Koker; Farshad Akhbari; Narayan Srinivasa; Feng Chen; Dukhwan Kim; Nadathur Rajagopalan Satish; John C. Weast; Mike B. MacPherson; Linda L. Hurd; Vasanth Ranganathan; Sanjeev Jahagirdar
- Web page: https://patentgrep.com/patent/US20240005136A1

## CPC

- G06N 3/063: using electronic means
- G06F 1/3287: by switching off individual functional units in the computer system
- G06F 1/3293: by switching to a less power-consuming processor, e.g. sub-CPU
- G06F 9/30014: {with variable precision}
- G06F 9/30036: {Instructions to perform operations on packed data, e.g. vector, tile or matrix operations}
- G06F 9/30038: {using a mask}
- G06F 15/76: Architectures of general purpose stored program computers (with program plugboard G06F15/08; multicomputers G06F15/16)
- G06F 15/78: comprising a single central processing unit
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/098: Distributed learning, e.g. federated learning
- G06T 1/20: Processor architectures; Processor configuration, e.g. pipelining
- G06T 15/005: {General purpose rendering architectures}
- G06T 1/60: Memory management
- 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)
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- 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

In an example, an apparatus comprises a compute engine comprising a high precision component and a low precision component; and logic, at least partially including hardware logic, to receive instructions in the compute engine; select at least one of the high precision component or the low precision component to execute the instructions; and apply a gate to at least one of the high precision component or the low precision component to execute the instructions. Other embodiments are also disclosed and claimed.

## Claims

### Claim at position 1

-11. (canceled)

### Claim 12

An apparatus comprising:
processor circuitry coupled to a memory, the processor circuitry to:
track precision level data relating to neural network operations; and
expose the precision level data in one or more model specific registers.

### Claim 13 (depends on 12)

The apparatus of claim 12, wherein tracking further includes tracking weights associated with the neural network operations, wherein tracking is performed while a neural network application is being executed on the processor circuitry.

### Claim 14 (depends on 12)

The apparatus of claim 12, wherein the one or more model specific registers are used to tune one or more neural network applications executing on the processor circuitry.

### Claim 15 (depends on 12)

The apparatus of claim 12, wherein the processor circuitry is further to automatically schedule, based on the precision level data, runtimes for workloads associated with the processor circuitry.

### Claim 16 (depends on 12)

The apparatus of claim 12, wherein the processor circuitry comprises graphics processor circuitry.

### Claim 17

A method comprising:
tracking, by a processor of a computing device, precision level data relating to neural network operations; and
exposing, by the processor, the precision level data in one or more model specific registers.

### Claim 18 (depends on 17)

The method of claim 17, wherein tracking further includes tracking weights associated with the neural network operations, wherein tracking is performed while a neural network application is being executed on the processor.

### Claim 19 (depends on 17)

The method of claim 17, wherein the one or more model specific registers are used to tune one or more neural network applications executing on the processor.

### Claim 20 (depends on 17)

The method of claim 17, further comprising automatically scheduling, based on the precision level data, runtimes for workloads associated with the processor.

### Claim 21 (depends on 17)

The method of claim 17, wherein the processor comprises a graphics processor, wherein the graphics processor is coupled a memory and an application processor.

### Claim 22

A computer-readable medium having stored thereon instructions which, when executed, cause a computing device to perform operations comprising:
tracking precision level data relating to neural network operations; and
exposing the precision level data in one or more model specific registers.

### Claim 23 (depends on 22)

The computer-readable medium of claim 22, wherein tracking further includes tracking weights associated with the neural network operations, wherein tracking is performed while a neural network application is being executed on one or more processors of the computing device.

### Claim 24 (depends on 22)

The computer-readable medium of claim 22, wherein the one or more model specific registers are used to tune one or more neural network applications executing on the one or more processors.

### Claim 25 (depends on 22)

The computer-readable medium of claim 22, wherein the operations further comprise automatically scheduling, based on the precision level data, runtimes for workloads associated with one or more processors.

### Claim 26 (depends on 22)

The computer-readable medium of claim 22, wherein the one or more processors comprise one or more graphics processors coupled to a memory and one or more application processors.
