https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51428
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |missed-optimization
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |pinskia at gcc dot
gnu.org
Component|middle-end |tree-optimization
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Mine for GCC 13 for bitfield lowering.
With the current code I have (which is applied to GCC 10 base compiler), we
get:
_7 = MEM <unsigned char> [(struct f *)a_5(D)];
_8 = _7 & 3;
_3 = _8 != 0;
_10 = BIT_INSERT_EXPR <_7, _3, 0 (1 bits)>;
MEM <unsigned char> [(struct f *)a_5(D)] = _10;
And:
_8 = _7 & 3;
_6 = _8 == 3;
_10 = BIT_INSERT_EXPR <_7, _6, 0 (1 bits)>;
both are not good really as we don't need to extract bit 0 really.
