Transactional memory in the dentry cache--maybe more useful in user space?

Posted Oct 15, 2013 22:57 UTC (Tue) by vomlehn (guest, #45588)
Parent article: Transactional memory in the dentry cache

It's good to see experimentation with the Intel transactional memory, but I think I'm sticking to what I said in lwn.net/Articles/480713/, including the part where I may be missing something. This may actually be useful in user space where, as far as I can tell, people spend less time per line of code on optimization. In that case, you may not get the fine-grained locking that we strive for in the kernel.

Also, there may be rather different ways to do this--Intel doesn't have an exclusive on smart people.