https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97747
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2022-02-04
Ever confirmed|0 |1
Known to fail| |8.2.0
Summary|missed combine opt with |[9/10/11/12 Regression]
|logical ops after zero |missed combine opt with
|extended load |logical ops after zero
| |extended load
Known to work| |7.5.0
Status|UNCONFIRMED |ASSIGNED
Component|rtl-optimization |tree-optimization
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Mine for GCC 13, bitfield lowering.
Note the store merging pass also messes up and creates:
_5 = _4 & 250;
_6 = _5 | 5;
Which can be optimized to just:
_6 = _4 | 5;
I might look at store merging first.
Note I think this might be a regression in code generation (store merging
caused the regressions).
GCC 7.5.0 produced:
ldr r2, .L3
ldrb r3, [r2] @ zero_extendqisi2
orr r3, r3, #5
strb r3, [r2]
bx lr
While GCC 8.2.0 produced:
ldr r2, .L3
ldrb r3, [r2] @ zero_extendqisi2
bic r3, r3, #5
orr r3, r3, #5
strb r3, [r2]
bx lr
