Title :
A Scaled-Up Dot Product Processor for Searching
Author_Institution :
Cisco Syst. Inc., San Jose, CA, USA
Abstract :
Dot product of large vectors is speed critical operation in a search engine. We present a dot product processor which can handle large vectors having 10000 components. Our architecture can finish computation in less than 80 clock cycles on average. This was achieved by a simpler circuitry instead of traditional Bloom Filters.
Keywords :
data structures; search engines; Bloom filters; scaled-up dot product processor; search engine; Clocks; Complexity theory; Computer architecture; Hardware; Search engines; Semantics; Vectors; Dot product hardware accelerator; comparison of leaf-labelled trees; fine grained parallelization; information search and retrieval; large vectors;
Conference_Titel :
Computer, Consumer and Control (IS3C), 2014 International Symposium on
Conference_Location :
Taichung
DOI :
10.1109/IS3C.2014.272
