# US 2026/0037478 A1 · Systems and methods for performing direct memory access data transfers

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

- Number: US 2026/0037478 A1
- Document ID: US20260037478A1
- Status: Published application
- Kind: A1
- Application number: 18/790,015
- Filed: 2024-07-31
- Published: 2026-02-05
- Earliest claimed: 2024-07-31
- Assignees: NVIDIA Corporation
- Applicants: NVIDIA Corporation
- Inventors: Andrew Peter TAUSSIG; Ravi Pratap SINGH; Ching-Yu HUNG; Sreenivas KRISHNAN; Jagadeesh SANKARAN; Yen-Te SHIH
- Web page: https://patentgrep.com/patent/US20260037478A1

## CPC

- G06F 15/8023: {Two dimensional arrays, e.g. mesh, torus}
- G06F 9/3001: {Arithmetic instructions}
- G06F 9/30181: {Instruction operation extension or modification}
- G06F 9/3888: {controlled by a single instruction for multiple threads [SIMT] in parallel}
- G06F 9/3897: {with adaptable data path}
- G06F 9/3887: {controlled by a single instruction for multiple data lanes [SIMD]}
- 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})
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

In various examples, systems and methods are disclosed that relate to programming multi-dimensional single instruction, multiple data (SIMD) processors (also referred to as an accelerator). In one example, a processor can obtain instructions to be performed by the accelerator. The processor can determine one or more operations to be performed by the accelerator based at least on the instructions and generate a set of accelerator instructions. In examples, the processor can then provide data associated with the accelerator instructions to cause the accelerator to perform at least a portion of the one or more operations.

## Claims

### Claim 1

One or more processors comprising:
one or more circuits to:
obtain an instruction to be performed by an accelerator;
determine that the instruction corresponds to one or more operations compatible with operations the accelerator is configured to execute;
based at least on the determined compatibility, determine the one or more operations to be performed by the accelerator based at least on the instruction and a datapath associated with the accelerator, the datapath representing one or more transfers of data within the accelerator via one or more components of the accelerator, and the data having a datatype that is compatible with the accelerator;
generate a set of accelerator instructions in an instruction set architecture (ISA) of the accelerator based at least on the one or more operations to be performed by the accelerator; and
provide the data associated with the set of accelerator instructions to a system to cause the system to coordinate operation of the accelerator in accordance with the set of accelerator instructions.

### Claim 2 (depends on 1)

The one or more processors of claim 1, wherein the accelerator is a pixel processing engine (PPE) comprising a plurality of processing elements,
wherein the plurality of processing elements are logically arranged in a two-dimensional array, and
wherein the one or more circuits determine the one or more operations to be performed by the PPE based at least on the datapath associated with the PPE.

### Claim 3 (depends on 2)

The one or more processors of claim 2, wherein the plurality of processing elements of the PPE are configured to communicate with one or more other processing elements of the PPE in accordance with one or more connection sets; and
wherein the one or more circuits generate the set of accelerator instructions based at least on the one or more operations to be performed and the one or more connection sets.

### Claim 4 (depends on 1)

The one or more processors of claim 1, wherein the one or more circuits receive the data associated with the instruction to be performed by the accelerator, the instruction to be performed represented in a first programming language,
wherein the one or more circuits are to determine a correspondence between the instruction to be performed and the one or more operations to be performed; and
wherein the one or more circuits determine the one or more operations to be performed by the accelerator based at least on correspondence between the instruction to be performed and the one or more operations to be performed.

### Claim 5 (depends on 1)

The one or more processors of claim 1, wherein the one or more circuits generate the set of accelerator instructions based at least on the one or more operations to be performed by the accelerator, the one or more operations involving shifting data from a first register of the accelerator to a second register of the accelerator.

### Claim 6 (depends on 5)

The one or more processors of claim 5, wherein the accelerator is a pixel processing engine (PPE) comprising a plurality of processing elements, and
wherein the one or more operations involving shifting the data from the first register of the PPE to the second register of the PPE involve shifting data from a register of a first processing element (PE) of the PPE to a register of a second PE of the PPE.

### Claim 7 (depends on 1)

The one or more processors of claim 1, wherein the one or more circuits generate the set of accelerator instructions based at least on the one or more operations to be performed by the accelerator, the one or more operations involving:
shifting data from a first register of the accelerator to a second register of the accelerator, and
performing at least one of: an addition operation, a subtraction operation, a multiplication operation, or a division operation.

### Claim 8 (depends on 1)

The one or more processors of claim 1, wherein the one or more circuits are to cause register data to be loaded into at least one register of a decoupled load/store unit (DLSU); and
provide the data associated with the set of accelerator instructions to the system to cause the system to coordinate operation of the accelerator in accordance with the set of accelerator instructions based at least on the register data located into the at least one register of the DLSU.

### Claim 9 (depends on 1)

The one or more processors of claim 1, wherein or more processors are comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system implemented using a robot;
an aerial system;
a medical system;
a boating system;
a smart area monitoring system;
a system for performing deep learning operations;
a system for performing simulation operations;
a system for generating or presenting virtual reality (VR) content, augmented reality (AR) content, or mixed reality (MR) content;
a system for performing digital twin operations;
a system implemented using an edge device;
a system incorporating one or more virtual machines (VMs);
a system for generating synthetic data;
a system implemented at least partially in a data center;
a system for performing conversational artificial intelligence (AI) operations;
a system for performing generative AI operations;
a system implementing language models;
a system for performing generative AI operations;
a system for implementing vision language models (VLMs);
a system implementing large language models (LLMs);
a system for implementing multi-modal language models;
a system for hosting one or more real-time streaming applications;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets; or
a system implemented at least partially using cloud computing resources.

### Claim 10

A system comprising:
an accelerator, and
at least one processor comprising one or more circuits to:
obtain an instruction to be performed by the accelerator;
determine that the instruction corresponds to one or more operations compatible with operations the accelerator is configured to execute;
based at least on the determined compatibility, determine the one or more operations to be performed by the accelerator based at least on the instruction and a datapath associated with the accelerator, the datapath representing one or more transfers of data within the accelerator via one or more components of the accelerator, the data having a datatype that is compatible with the accelerator;
generate a set of accelerator instructions in an instruction set architecture (ISA) of the accelerator based at least on the one or more operations to be performed by the accelerator; and
provide the data associated with the set of accelerator instructions to cause the accelerator to operate in accordance with the set of accelerator instructions.

### Claim 11 (depends on 10)

The system of claim 10, wherein the accelerator is a pixel processing engine (PPE) comprising a plurality of processing elements,
wherein the plurality of processing elements are logically arranged in a two-dimensional array, and
wherein the at least one processor determines the one or more operations to be performed by the PPE based at least on the datapath associated with the PPE.

### Claim 12 (depends on 11)

The system of claim 11, wherein the plurality of processing elements of the PPE are configured to communicate with one or more other processing elements of the PPE in accordance with one or more connection sets; and
wherein the at least one processor generates the set of accelerator instructions based at least on the one or more operations to be performed and the one or more connection sets.

### Claim 13 (depends on 10)

The system of claim 10, wherein the at least one processor receives the data associated with the instruction to be performed by the accelerator, the instruction to be performed represented in a first programming language,
wherein the at least one processor determines a correspondence between the instruction to be performed and the one or more operations to be performed; and
wherein the at least one processor determines the one or more operations to be performed by the accelerator based at least on correspondence between the instruction to be performed and the one or more operations to be performed.

### Claim 14 (depends on 10)

The system of claim 10, wherein the at least one processor generates the set of accelerator instructions based at least on the one or more operations to be performed by the accelerator, the one or more operations involving shifting data from a first register of the accelerator to a second register of the accelerator.

### Claim 15 (depends on 14)

The system of claim 14, wherein the accelerator is a pixel processing engine (PPE) comprising a plurality of processing elements, and
wherein the one or more operations involving shifting the data from the first register of the PPE to the second register of the PPE involve shifting data from a register of a first processing element (PE) of the PPE to a register of a second PE of the PPE.

### Claim 16 (depends on 10)

The system of claim 10, wherein the at least one processor generates the set of accelerator instructions based at least on the one or more operations to be performed by the accelerator, the one or more operations involving:
shifting data from a first register of the accelerator to a second register of the accelerator, and
performing at least one of: an addition operation, a subtraction operation, a multiplication operation, or a division operation.

### Claim 17 (depends on 10)

The system of claim 10, wherein the at least one processor causes register data to be loaded into at least one register of a decoupled load/store unit (DLSU); and
provide the data associated with the set of accelerator instructions to the system to cause the system to coordinate operation of the accelerator in accordance with the set of accelerator instructions based at least on the register data located into the at least one register of the DLSU.

### Claim 18 (depends on 10)

The system of claim 10, wherein the at least one accelerator is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system implemented using a robot;
an aerial system;
a medical system;
a boating system;
a smart area monitoring system;
a system for performing deep learning operations;
a system for performing simulation operations;
a system for generating or presenting virtual reality (VR) content, augmented reality (AR) content, or mixed reality (MR) content;
a system for performing digital twin operations;
a system implemented using an edge device;
a system incorporating one or more virtual machines (VMs);
a system for generating synthetic data;
a system implemented at least partially in a data center;
a system for performing conversational artificial intelligence (AI) operations;
a system for performing generative AI operations;
a system implementing language models;
a system for performing generative AI operations;
a system for implementing vision language models (VLMs);
a system for implementing large language models (LLMs);
a system for implementing multi-modal language models;
a system for hosting one or more real-time streaming applications;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets; or
a system implemented at least partially using cloud computing resources.

### Claim 19

A method comprising:
obtaining an instruction to be performed by an accelerator;
determining that the instruction corresponds to one or more operations compatible with operations the accelerator is configured to execute;
based at least on the determined compatibility, determining the one or more operations to be performed by the accelerator based at least on the instruction and a datapath associated with the accelerator, the datapath representing one or more transfers of data within the accelerator via one or more components of the accelerator, and the data having a datatype that is compatible with the accelerator;
generating a set of accelerator instructions in an instruction set architecture (ISA) of the accelerator based at least on the one or more operations to be performed by the accelerator; and
providing data associated with the set of accelerator instructions to a system to cause the system to coordinate operation of the accelerator in accordance with the set of accelerator instructions.

### Claim 20 (depends on 19)

The method of claim 19, wherein the accelerator is a pixel processing engine (PPE) comprising a plurality of processing elements, and
wherein the plurality of processing elements are logically arranged in a two-dimensional array, the method comprising determining the one or more operations to be performed by the PPE based at least on the datapath associated with the PPE.
