Device and method for executing vector inner product operation

Abstract

The invention provides a device and method for executing vector inner product operation, and aims at cooperating with a corresponding instruction set to execute vector inner product operation. The device comprises an instruction storage unit, a register unit and a vector inner product unit, wherein the storage unit is internally stored with vectors; the register unit is internally stored with addresses for vector storage; and the vector inner product unit is used for obtaining a vector address in the register unit according to an assorted instruction, obtaining a corresponding vector in the storage unit according to the vector address and carrying out vector inner product operation according to the obtained vector so as to obtain an operation result. According to the device and method, vector data participating in the calculation is temporarily stored on a scratch pad memory, so that the data with different widths can be supported more flexibly and effectively in the vector inner product operation process and then the execution performance of applications comprising a large number of vector inner products is enhanced.

Description

technical field [0001] The invention relates to a vector inner product operation device and method, which are used for efficiently and flexibly executing vector inner product according to a vector inner product instruction, and can well solve the problem that more and more algorithms in the current computer field include a large number of vector inner products. Background technique [0002] A vector inner product (dot product; scalar product, also known as a dot product) is a binary operation that takes two vectors over a real number R and returns a real-valued scalar. It is the standard inner product of Euclidean space. [0003] In the field of mathematical operations, the inner product of two vectors a = [a1, a2, ..., an] and b = [b1, b2, ..., bn] is defined as: [0004] a·b=a1b1+a2b2+...+anbn [0005] Using matrix multiplication and treating the (column) vectors as n×1 matrices, the dot product can also be written as: [0006] a·b=a^T*b [0007] Among them, a^T repres...

AI Technical Summary

Problems solved by technology

[0011] The object of the present invention is to provide a vector inner product device and method to solve the problems existing in the prior art such as limited inter-chip communication, insufficient on-chip cache, and insufficient flexibility of supported vector lengths.

Images