https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113779
--- Comment #9 from Miro Kropacek <miro.kropacek at gmail dot com> --- A talented Atari developer very nicely described what is the problem with current GCC, completely investigated from scratch: https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-1 https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-2 https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-3 (this is where it gets interesting) https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-4 https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-5 (the actual cause) https://atariscne.org/news/index.php/gcc-for-asm-experts-and-c-c-intermediates-part-6 In the end, his m68k gcc generates super-optimal code now. There are links to his github, so an experienced gcc developer could cherry-pick the changes upstream.
