https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118755
--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So fwprop can't do this
```
propagating insn 11 into insn 12, replacing:
(set (reg:QI 107 [ _3 ])
(subreg:QI (reg:SI 108) 0))
failed to match this instruction:
(set (reg:QI 107 [ _3 ])
(subreg:QI (eq:SI (reg:CC 66 cc)
(const_int 0 [0])) 0))
```
But combine does:
```
Trying 11 -> 12:
11: r108:SI=cc:CC==0
REG_DEAD cc:CC
12: r107:QI=r108:SI#0
REG_DEAD r108:SI
Successfully matched this instruction:
(set (reg:QI 107 [ _3 ])
(eq:QI (reg:CC 66 cc)
(const_int 0 [0])))
allowing combination of insns 11 and 12
```
This is due to this code in gen_lowpart_for_combine:
```
/* If X is a comparison operator, rewrite it in a new mode. This
probably won't match, but may allow further simplifications. */
else if (COMPARISON_P (x)
&& SCALAR_INT_MODE_P (imode)
&& SCALAR_INT_MODE_P (omode))
return gen_rtx_fmt_ee (GET_CODE (x), omode, XEXP (x, 0), XEXP (x, 1));
```
The question should we try to do another replacement in fwprop/late_combine for
stuff like this to simplify?
We have the same issue with (zero_extend (comparison)) too.
Let me see if adding them can help here too.
