WEAVER: An Energy Efficient, General-Purpose Acceleration Architecture for String Operations in Big Data Applications

The explosion of data volume in big data era drives architecture designers to find more efficient processor architectures. Different from the conventional scientific computing, big data applications are usually simple computing patterns with massive input data, which makes current general-purpose processors suffer poor performance and huge energy consumption. Among various computing patterns in big data applications, string operations are common but very important parts of data processing. Dealing with simple and fixed computing patterns with high-performance processors is wasteful. In this paper, we propose WEAVER, a high-efficiency, general-purpose acceleration architecture for string operations, which can be integrated into current processors and work beside on-chip memory controller. The WEAVER can both reduce the latency of data transfer between memory and processing cores and accelerate the string processing by using customized architecture. The experimental results show WEAVER gains an average speedup of 3.2x over a general-purpose Intel processor, and reduces the energy consumption by 3.79x on average, with tiny hardware overhead.