https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127693
--- Comment #27 from Richard Biener <rguenth at gcc dot gnu.org> ---
OK, so here we end up with
<bb 2> [local count: 14598063]:
D.165232.m_size = 1;
D.165232.m_data = "\x12\"\x00";
sought = D.165232;
.ASSUME (_ZNK11QStringView4sizeEv._assume.0, &sought);
goto <bb 33>; [100.00%]
... (with calls to qt_assert)
<bb 31> [local count: 1073741824]:
# i_12 = PHI <i_15(30), 0(56)>
if (_8 > i_12)
goto <bb 5>; [89.00%]
else
goto <bb 32>; [11.00%]
<bb 32> [local count: 118111600]:
tokensMatched_17 = tokensMatched_16 + 1;
<bb 33> [local count: 132709663]:
# idx_19 = PHI <-1(2), _98(30)>
# tokensMatched_16 = PHI <0(2), tokensMatched_17(30)>
_20 = idx_19 + 1;
s = sought;
_67 = s.m_size;
_68 = __builtin_constant_p (_67);
if (_68 != 0)
goto <bb 34>; [50.00%]
else
goto <bb 54>; [50.00%]
<bb 34> [local count: 4880644631]:
_69 = (long unsigned int) _67;
if (_69 <= 4611686018427387902)
goto <bb 36>; [100.00%]
else
goto <bb 35>; [0.00%]
<bb 35> [count: 0]:
__builtin_unreachable ();
and the .ASSUME makes 'sought' address-taken and thus clobbered
by calls. I have a patch to fix that, with that it's optimized.
