18093929. MATRIX OPERATION METHOD AND ACCELERATOR simplified abstract (Huawei Technologies Co., Ltd.)
Contents
MATRIX OPERATION METHOD AND ACCELERATOR
Organization Name
Inventor(s)
MATRIX OPERATION METHOD AND ACCELERATOR - A simplified explanation of the abstract
This abstract first appeared for US patent application 18093929 titled 'MATRIX OPERATION METHOD AND ACCELERATOR
Simplified Explanation
The abstract describes a method for performing matrix operations using a dedicated matrix operation accelerator. The accelerator stores subsets of two matrices in memory and performs matrix operations based on instructions to obtain results. This offloads the matrix operation burden from a processor and allows for faster completion of large-scale matrix operations.
- Matrix operation method applied to a dedicated accelerator
- Subsets of first and second matrices stored in memory
- Subsets of first matrix multiplied by subsets of second matrix and stored in memory
- Matrix operations performed on subsets of matrices based on instructions
- Matrix operation results obtained using the dedicated accelerator
Potential Applications
- Scientific research and simulations requiring large-scale matrix operations
- Data analysis and machine learning algorithms that involve matrix calculations
- Image and signal processing applications that utilize matrix operations
Problems Solved
- Offloads matrix operation burden from the processor, improving overall system performance
- Enables faster completion of large-scale matrix operations
- Reduces the time required for complex calculations involving matrices
Benefits
- Allows for efficient execution of matrix operations
- Increases the speed and performance of systems performing matrix calculations
- Enables the handling of larger and more complex matrices
- Reduces the computational load on the processor, freeing it up for other tasks
Original Abstract Submitted
A matrix operation method is provided, applied to an accelerator configured to perform a matrix operation. A matrix operation accelerator () respectively stores, in response to a received matrix operation instruction, subsets of a first matrix and subsets of a second matrix in a first storage space and a second storage space of a memory (); stores, in a third storage space of the memory (), subsets obtained after the subsets of the first matrix are multiplied by the subsets of the second matrix; and performs matrix operations on the subsets of the first matrix and the subsets of the second matrix based on the matrix operation instruction, to obtain matrix operation results. The dedicated matrix operation accelerator () is used to perform a matrix operation, so that a large-scale matrix operation can be completed in relatively short time, thereby offloading a matrix operation burden of a processor.
