This issue would also come up if the atomic blocks were locked using a normal lock via .lock() and .unlock(), although it might be a bit clearer to a reader where the error is in the program if locks were used.

That's a great point! ^ I was wondering in class if transactional memory is always better than locks. Are there any characteristics of TM that would make them worse than standard locks and mutexes?

@pranil I believe there's a graph comparing their runtimes on slide 27 and a few slides ago it talks about when you'd want to choose one over the other. It is interesting that CS110 spends so long introducing locks but doesn't mention transitional memory-- maybe it's because robust, working locks came around first and we've been using them since.

@jiaju I think it's also because locks can be used for more control of threads beyond atomic sections and provide more granularity like was mention elsewhere in these slides.