https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122996
Bug ID: 122996
Summary: Missed optimisation for branch folding with unsigned
arithmetic wraparound
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: mital at mitalashok dot co.uk
Target Milestone: ---
This function <https://godbolt.org/z/7GeTcPqj8>:
unsigned inc(unsigned x) {
return x == -1u ? 0u : x + 1u;
}
Currently produces this assembly:
inc:
xor edx, edx
lea eax, [rdi+1]
cmp edi, -1
cmove eax, edx
ret
When it could produce:
inc:
lea eax, [rdi+1]
ret
Because the function is equivalent to `return x + 1u;`. This optimisation is
done for `return x == 0u ? 1u : x + 1u;`.
This comes up when dealing with std::string{,_view}::npos in C++
