No Sane Compiler Would Optimize Atomics https://github.com/jfbastien/papers/blob/master/source/N4455.rst [cue evil laugh]
miniblog.
Related Posts
"the LLVM optimizers could optimize debug code just as well as non-debug code" https://llvm.org/docs/SourceLevelDebugging.html#debugging-optimized-code (impressive!)
Arch Linux feels a lot more snappy after running pacman-optimize and rankmirrors. Should have done it earlier.
I've never liked arguments of the form OPTIMIZE=y or FAST=y. Why would I want slow? Clearer would be DEBUG_LEVEL i.e. show the tradeoff.