This is the Bayes benchmark, an adapted version of the Bayes benchmark of the STAMP benchmark suite [Minh2008] ported to Clojure and extended with the use of transactional futures [Swalens2016].
You can run the program using Leiningen.
Run the benchmark as follows (all parameters are optional):
$ lein run -- -n 5 -r 512 -v 48 -x alternatives-parallel
To run the original version of the benchmark, which does not use transactional actors and is similar to the benchmark of the STAMP suite, execute:
$ lein run -- -n 5 -r 512 -v 48
Parameters from original benchmark:
-t: number of threads that process transactions.-v: number of variables.-r: number of records.-e: maximum number of edges learned per variable (-1 for no limit).-i: edge insert penalty.-n: maximum number of parents.-p: percent chance of parent.-q: operation quality factor.-s: seed for random number generator.
Additional parameters:
-m: enable profiling.-x: comma-separated list of "variations" on the original program to enable. (Described below)
(Run lein run -- -h to get this description and more.)
Running the program prints the given options and the total execution time to the screen.
Using command-line options, you can change the behavior of the benchmark. Without any option, it is similar to the original STAMP benchmark.
There are two variations that can be enabled:
alternatives-parallel: in learner/find-best-insert-task, calculate the alternatives in parallel. This uses transactional futures.strict-adtree: in adtree/make-vary, generate the adtree strictly instead of lazily.
In our benchmarks, we compare the alternatives-parallel variation with the original, to measure the benefit of using transactional futures.
Licensed under the MIT license, included in the file LICENSE.
[Swalens2016] J. Swalens, J. De Koster, and W. De Meuter. 2016. "Transactional Tasks: Parallelism in Software Transactions". In Proceedings of the 30th European Conference on Object-Oriented Programming (ECOOP'16).
[Minh2008] C. C. Minh, J. Chung, C. Kozyrakis, and K. Olukotun. 2008. "STAMP: Stanford Transactional Applications for Multi-Processing". In Proceedings of the IEEE International Symposium on Workload Characterization (IISWC'08).