Experimental B+-tree for GPU
Krzysztof Kaczmarski · 2011
Abstract. The main intention of this work is to create a dictionary structure which could benefit from massive parallelism of threads when performing computation on all or a selected set of elements, while hav-ing an ability to search for and insert keys very quickly, yet preserving the order of elements. So far, no such structure dedicated for GPU ex-ists. This paper presents results of the first to our knowledge B+-tree implementation in CUDA C language. 1