Matrix-based XML Stream Processing Using a GPU

Abstract

With the advent of GPGPU computing, a major paradigm shift in parallel computation in recent years, massive parallel processing, has become available at a relatively low cost. However, conventional XML stream processing algorithms do not utilize the advantages of GPUs since the algorithms were intended for single-thread execution. In this research, we propose a GPU-accelerated, matrix-based XML stream processing methodology in which a large collection of XPath queries are transformed into matrixes of binary values and the XML stream is also converted to two matrix indexes. Then, the processing of a number of queries is transformed to simple bit-AND Boolean operations. With the XMark benchmark data set, we show that the proposed algorithm outperforms the conventional algorithms by about eight times.