https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89670
--- Comment #6 from Jörn Engel <joern at purestorage dot com> ---
True for one, but not the other.
return mask ? __builtin_ctz(mask) : 32;
1099: 83 f6 ff xor $0xffffffff,%esi
109c: 74 47 je 10e5 <main+0x85>
109e: f3 0f bc f6 tzcnt %esi,%esi
I used:
gcc-8 -std=gnu11 -Wall -Wextra -g -march=core-avx2 -mbmi -fPIC -O3 %
_tzcnt_u32() works as you said it should. Nicer than inline asm and allows
type checking. Thank you for that hint!
