####Akshay Jajoo [ajajoo@purdue.edu] ####Devesh Kumar Singh [singh349@purdue.edu] ####Sergei Uversky [suversky@purdue.edu]
This is an attempt at a C++ implementation of a non-blocking binary search tree inspired by the algorithm and pseudocode given in the paper by Ellen et al[1]. For simplicity's sake, we are first focusing on tackling the insert() and contains() functions; time permitting, we will then attempt to tackle tree balancing followed by remove().
We have included a reference sequential implementation in SequentialBST.cpp and a transactional-memory based implementation in STMbst.cpp.
The code requires a more modern version of g++ than is available on most of the CS lab machines. The code was tested on the machine mc18.cs.purdue.edu, which has G++ v. 5.3.0. under g++-5.3.0. The Makefile was set up to use this compiler if on mc18 and to otherwise use g++; if testing on a personal machine, please ensure your g++ version is up-to-date to avoid spurious segmentation faults.
To compile the code, run the following command.
$ make
A simple test of the implementation can be reached by typing
$ ./BST test
This test contains examples of inserting into the tree and searching the tree.
Timing of concurrent insertions can be reached by typing
$ ./BST time <numThreads> <iterations>
This will repeat a simple test number of times:
To test out the STM implementation, repeat the same steps as above, but replace BST with STMbst, i.e.
$ ./STMbst test
$ ./STMbst time <numThreads> <iterations>
[1] F. Ellen, P. Fatourou, E. Ruppert., and Van Breugel, F. Non-blocking binary search trees. PODC (2010), 131-140