Implementace relačního spojení s přihlédnutím k L2 cache

Abstract

This thesis deals with theoretical description of processor’s cache, as well as it’s exploitation in implementing optimalized algorithms. It will focus on algorithms for relational joins. These algorithms are, for example, nested loop and hash-join. The core part of the thesis will explain GRACE hash-join algorithm and radix-cluster hash-join algorithm, which were implemented in C++ with emphasis on processor’s cache exploitation. Furthermore, the tools that can be used during cache aware optimalizations of algorithms will be mentioned. In the conclusion, these algorithms will be compared to each other, as well as with unoptimized hash-join algorithm.