EON regression due to pass ordering problem (PR tree-optimize/24653)

[Paolo Bonzini]

>>and so on.  Until dom1 pass we still maintain pointer "this" so the
>>startPoint must live in memory.  DCE properly transforms it into:
>>
>>this = &startPoint;
>>startPoint.e[0] = code;
> 
> teaches forwprop to do this transformation (I guess there really are multiple
> uses of 'this').  I wonder why DCE does such transformation, and why
> DCE does not help the tramp3d cases I invented the forwprop change for.

I think Honza meant that DOM does this transformation:

>> and because we never run dce after first DOM and before next SRA, we end
>> up with this still being considered as with address taken by the alias
>> analysis pass we do before SRA.

Paolo