[Bug rtl-optimization/99863] [10/11 Regression] wrong code with -O -fno-tree-forwprop -mno-sse2 since r10-7268-g529ea7d9596b26ba

[rguenther at suse dot de via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99863

--- Comment #11 from rguenther at suse dot de <rguenther at suse dot de> ---
On Thu, 1 Apr 2021, jakub at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99863
> 
> Jakub Jelinek <jakub at gcc dot gnu.org> changed:
> 
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                  CC|                            |jakub at gcc dot gnu.org
> 
> --- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
> I think the REG_DEAD note isn't the problem here.
> At least, when I dump bb 2 in rest_of_handle_dse right before df_analyze call,
> there is
> (insn 6 3 7 2 (set (reg:CCZ 17 flags)
>         (compare:CCZ (reg/v:DI 203 [ y ])
>             (const_int 0 [0]))) "pr99863.c":14:12 8 {*cmpdi_ccno_1}
>      (nil))
> (insn 7 6 8 2 (set (reg:QI 205)
>         (eq:QI (reg:CCZ 17 flags)
>             (const_int 0 [0]))) "pr99863.c":14:12 802 {*setcc_qi}
>      (expr_list:REG_DEAD (reg:CCZ 17 flags)
>         (nil)))
> ...
> (insn 40 38 41 2 (set (reg:QI 220)
>         (ne:QI (reg:CCZ 17 flags)
>             (const_int 0 [0]))) "pr99863.c":15:11 802 {*setcc_qi}
>      (expr_list:REG_DEAD (reg:CCZ 17 flags)
>         (nil)))
> but if I dump it right after df_analyze call, I see:
> (insn 6 3 7 2 (set (reg:CCZ 17 flags)
>         (compare:CCZ (reg/v:DI 203 [ y ])
>             (const_int 0 [0]))) "pr99863.c":14:12 8 {*cmpdi_ccno_1}
>      (expr_list:REG_DEAD (reg/v:DI 203 [ y ])
>         (nil)))
> (insn 7 6 8 2 (set (reg:QI 205)
>         (eq:QI (reg:CCZ 17 flags)
>             (const_int 0 [0]))) "pr99863.c":14:12 802 {*setcc_qi}
>      (nil))
> ...
> (insn 40 38 41 2 (set (reg:QI 220)
>         (ne:QI (reg:CCZ 17 flags)
>             (const_int 0 [0]))) "pr99863.c":15:11 802 {*setcc_qi}
>      (expr_list:REG_DEAD (reg:CCZ 17 flags)
>         (nil)))
> 
> i.e. the problematic REG_DEAD note is gone and I think at this point DF knows
> that CC is live there.
> But replace_read has code to check for live hard regs, see
> note_stores (this_insn, look_for_hardregs, regs_set);
> in there etc.

But find_shift_sequence doesn't seem to check whether the emitted
insns contain any extra hardreg clobbers that make the replacement
invalid.