Tauchain Development Update: TML Performance
The current development emphasis for TML is on performance improvements (hardware accelerated BDDs?)
- The current attention of TML development is on improving performance.
- The strategy being taken thus far is parallel computing optimizations.
- The method of implementation is based on the work of Tom Van Dijk and his multi-core implementation of binary decision diagrams.
Binary decision diagrams are a data structure which can allow for efficient implementation of model checking, automated reasoning, etc. We know these are features critical to Tauchain so we know it is also important that these features be implemented in the way which is highly optimized for performance. Multi-core BDD algorithms are a specific form of binary decision diagram algorithms constructed in such a way so as to benefit from parallelism. For those who may be more familiar with EOS you may also recall that Dan Larimer took a similar approach of developing for performance by putting an emphasis on parallel processing.
The parallelization approach is in essence useful because it takes advantage of the trends in modern hardware. Most computers today are multi-core and because of this there is great benefit to adapting TML to leverage the multi-core environment. The specifics of this approach are highlighted in Tom's thesis and in my opinion this appears to be a viable way to improve performance. The reasoning for the layman reader is that by better taking advantage of available hardware (by making use of multiple cores) there can be a time surplus (time saved) as a result of the increase in speed.
Anything which can save either time or space is measurably of higher efficiency. In this case from what I can understand this would save primarily time because the processes can be more intelligently distributed to the available hardware. I'm not entirely sure how space can be saved specifically but will look deeper into it. More technical minds on Steem are welcome to study the thesis in the references to understand in more detail how Slyvan works and to get a general idea on how it might benefit TML.
Conclusion
TML is looking pretty good. It's functioning and bugs are being fixed. Performance improvements are being continuously attempted. The strategic approach (in my humble opinion) is correct and likely to work. The exact implementation will have to undergo testing of course but I'm hopeful that it will produce vastly improved performance.
Join the Tauchain Community Mailing List for the latest updates:
Tauchain Community Mailing List