https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90693
--- Comment #11 from Piotr Siupa <piotrsiupa at gmail dot com> ---
Thanks! Now the generated assembly is one instruction shorter.
It works for:
bool foo(unsigned x)
{
[[assume(x != 0)]];
return std::has_single_bit(x);
}
and for:
bool foo(unsigned x)
{
if (x == 0)
std::unreachable();
else
return std::has_single_bit(x);
}
However, I've noticed that:
bool foo(unsigned x)
{
if (x == 0)
return true;
else
return std::has_single_bit(x);
}
still uses (X ^ (X - 1)) > (X - 1).
