4 ms·
I almost skipped this thread because of the flamebait title. This is a debate over CPU instruction set performance details, nobody is going to die.
by gary_0 5y ago
I almost skipped this thread because of the flamebait title. This is a debate over CPU instruction set performance details, nobody is going to die.
- yjftsjthsd-h 5y agoIn fairness, this is Hacker News; flame wars^w^w respectful but intense debate over editors, operating systems, and, yes, ISA details, is somewhat expected. (Although, yes, I'm not sure that I would get too worked up about this particular detail; even if the stated claim is 100% true and unmitigated, it means some kinds of code will have potentially bigger binaries. I understand a math library person caring, I don't think I care.)
- rbanffy 5y ago> I understand a math library person caring, I don't think I care. Not wasting much sleep on this one. Not sure there's anything on the spec that stops implementations from recognizing the two instructions and fuse them into a single atomic operation for the backends to deal with. It'll occupy more space in the L1 cache, but that's it.
- dang 5y agoFlamewars are definitely not expected - they're against the rules and something we try to dampen in every way we know. https://hn.algolia.com/?dateRange=all&page=0&prefix=true&query=by%3Adang%20flame&sort=byDate&type=comment https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que... https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html