For description of B+ Tree, please contact Amittai Aviram who wrote the original serial code, or see the paper listed below.
B+ Tree application has many internal commands that maintain database and process querries. Only J and K commands had enough parallelism to be ported to parallel languages (OpenMP, CUDA, OpenCL). In these implementations, in case of both J and K, the same algorithms (optimized for exposing fine-grained parallelism) were used for fair comparison purposes. For C/OpenMP execution, it is possible to use the original algorithm.
 J. Fix, A. Wilkes and K. Skadron. "Accelerating Braided B+ Tree Searches on a GPU with CUDA." ACM Transactions on Database Systems. 2009. (pdf)
Input : (tar.gz)
OpenMP Version: (tar.gz)
CUDA Version: (tar.gz)
OpenCL Version: (tar.gz)