[PATCH] Fix PR middle-end/28690, indexed load/store performance + reload bug

[Paolo Bonzini]

>> Actually, I meant
>>
>>    if (REG_P (op))
>>      return REG_POINTER (op) ? -1 : -2;
>>    else
>>      return -3;
> 
> Which is how I originally coded it up.

I didn't really understand your example, but you know better than me for 
sure since you tested your patch and I didn't test my thoughts.

>> And then swap_commutative_operands_with_target could die entirely.
> 
> How so, since the change to commutative_operand_precedence doesn't
> account for the preference of wanting set X X Y over set X Y X?

If you can implement your commutative_operand_precedence change and not 
lose performance on i386, the whole 
swap_commutative_operands_with_target should have become pointless.

Paolo