https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116662
--- Comment #10 from Levi Zim <rsworktech at outlook dot com> --- (In reply to Jeffrey A. Law from comment #9) > So the question in my mind, how important is this? On modern kernels & > toolchains it's possible to query the cboz extension & its block size which > effectively gives you the size of a cache line. But not every design has > the cboz extension, so we'd have to have a fallback if we went the route of > dynamic discovery. Usually std::hardware_destructive_interference_size is used at compile time in alignas. If this doesn't match L1 cacheline size, probably it will only make performance given it is used correctly. > Or we could just update the constant. 64b cache lines are likely going to > be the sweet spot for a while. I wouldn't lose sleep just setting it to > 64bit and getting on with our lives. Yes. I think that's the solution.
