CUDA（26）之Parallel Reduction Intro

Abstract
This blog will talk about reduction.

1. Reduction
Reduction implements an efficient version of non-strict evaluation. For example, an example of evaluating an arithmetic expression follows: “2+2+2+2+3+3”. This mathematical expression can be represented as a tree, where the root node represents the final result of the expression. .

2. Importance
The concept of reduction is a fundamental technique in the field of parallel computing, in which a big problem can be divided into several small problems and solved in parallel. The complexity of the problem-solving algorithm will reduce from O(n*n) to O(log(n)) in this case.