3 ms·
IBM's Hardware Transactional Memory
- aidenn0 15y agoThis should be interesting. My maxim is "Shared-state multithreading using locks for mutual exclusion is the most performant way to write a buggy program" I typically use message-passing to avoid this (it also scales a lot better) but it will be interesting to see how transactional memory works for smaller scale parallelism (or the lowest level of a multi-tiered parallel architecture).