[Bug target/117192] [15 Regression] wrong code at -O3 with "-fno-unswitch-loops" on x86_64-linux-gnu since r15-4397-g70f59d2a1c51bd

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117192

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Uroš Bizjak from comment #9)
> (In reply to Andrew Pinski from comment #6)
> 
> > The andn is swapped.
> Indeed:
> 
> +(define_expand "andn<mode>3"
> +  [(set (match_operand:MMXMODEI 0 "register_operand")
> +        (and:MMXMODEI
> +          (not:MMXMODEI (match_operand:MMXMODEI 1 "register_operand"))
> +          (match_operand:MMXMODEI 2 "register_operand")))]
> +  "TARGET_SSE2")
> 
> +(define_expand "andn<mode>3"
> +  [(set (match_operand:VI 0 "register_operand")
> +       (and:VI
> +         (not:VI (match_operand:VI 2 "register_operand"))
> +         (match_operand:VI 1 "register_operand")))]
> +  "TARGET_SSE2")
> 
> Which operand order is correct?


The correct one is operand 2 with the not.
>From https://gcc.gnu.org/onlinedocs/gccint/Standard-Names.html :
‘andnm3’
Like andm3, but it uses bitwise-complement of operand 2 rather than operand 2
itself.