17942424. SINGLE INSTRUCTION MULTIPLE DATA SIMD INSTRUCTION GENERATION AND PROCESSING METHOD AND RELATED DEVICE simplified abstract (Huawei Technologies Co., Ltd.)

From WikiPatents
Jump to navigation Jump to search

SINGLE INSTRUCTION MULTIPLE DATA SIMD INSTRUCTION GENERATION AND PROCESSING METHOD AND RELATED DEVICE

Organization Name

Huawei Technologies Co., Ltd.

Inventor(s)

Chen Wu of Shanghai (CN)

Yifan Lin of Shanghai (CN)

Xiaoqiang Dan of Shenzhen (CN)

SINGLE INSTRUCTION MULTIPLE DATA SIMD INSTRUCTION GENERATION AND PROCESSING METHOD AND RELATED DEVICE - A simplified explanation of the abstract

This abstract first appeared for US patent application 17942424 titled 'SINGLE INSTRUCTION MULTIPLE DATA SIMD INSTRUCTION GENERATION AND PROCESSING METHOD AND RELATED DEVICE

Simplified Explanation

The abstract describes a method and device for generating and processing SIMD (Single Instruction, Multiple Data) instructions. The method involves obtaining the length of each loop dimension of a tensor formula, selecting information about a second SIMD instruction model that matches the tensor formula, and generating a first SIMD instruction based on the length of at least one loop dimension and the selected SIMD instruction model.

  • The method involves obtaining the length of each loop dimension of a tensor formula.
  • Information about a second SIMD instruction model is selected based on the length of each loop dimension of the tensor formula.
  • A first SIMD instruction is generated based on the length of at least one loop dimension and the selected SIMD instruction model.

Potential applications of this technology:

  • High-performance computing: The method can be used to optimize SIMD instruction generation and processing in applications that require high computational power, such as scientific simulations, data analytics, and machine learning.
  • Image and video processing: SIMD instructions are commonly used in image and video processing algorithms. This method can improve the efficiency and speed of such algorithms.
  • Signal processing: SIMD instructions are also widely used in signal processing applications, such as audio and speech processing. This technology can enhance the performance of these applications.

Problems solved by this technology:

  • Efficient SIMD instruction generation: The method provides a systematic approach to selecting the appropriate SIMD instruction model based on the characteristics of the tensor formula, resulting in optimized instruction generation.
  • Improved performance: By generating SIMD instructions tailored to the specific tensor formula, the method can enhance the performance of computations, leading to faster execution times and improved efficiency.
  • Simplified development process: The method automates the selection and generation of SIMD instructions, reducing the complexity and effort required in manually optimizing code for SIMD architectures.

Benefits of this technology:

  • Enhanced computational efficiency: The optimized SIMD instructions generated by this method can significantly improve the computational efficiency of applications, allowing for faster and more efficient processing of large datasets.
  • Improved performance scalability: By tailoring SIMD instructions to the specific tensor formula, the method enables better performance scalability across different hardware architectures, maximizing the utilization of SIMD capabilities.
  • Simplified code optimization: The method automates the process of selecting and generating SIMD instructions, reducing the need for manual code optimization and making it easier for developers to leverage SIMD capabilities in their applications.


Original Abstract Submitted

An SIMD instruction generation and processing method and a related device are provided. The method may include: obtaining a length of each loop dimension of a first tensor formula; selecting, from a plurality of groups of information about a first SIMD instruction model based on the length of each loop dimension of a first tensor formula, information about a second SIMD instruction model matching the first tensor formula; generating, based on a length of at least one loop dimension of the first tensor formula and the second SIMD instruction model, a first SIMD instruction obtained after the first tensor formula is converted. The information about a second SIMD instruction model is selected from the plurality of groups of information about a first SIMD instruction model based on the length of each loop dimension of the tensor formula.