[prev in list] [next in list] [prev in thread] [next in thread] 

List:       pcc-list
Subject:    Re: invalid OREGs
From:       Anders Magnusson <ragge () ludd ! ltu ! se>
Date:       2008-03-18 21:52:43
Message-ID: 47E039AB.40307 () ludd ! ltu ! se
[Download RAW message or body]

Gregory McGarry skrev:
>
>> Gregory McGarry skrev:
>>> Here's a patch to avoid generating OREGs where the target doesn't 
>>> support them:
>> Looks reasonable, for targets that accept OREGs they will be 
>> converted back later
>> in canon().  But actually the problem is bigger;  maybe there should 
>> be a routine that
>> checks with notoff() and unwindles OREGs in pass2?  The f77 frontend 
>> makes heavy
>> use of OREGs that must be removed when entering pass2.  Or maybe this 
>> is not a
>> big problem, most targets accept gladly OREGs.
>
> Actually, most architectures (should) have logic in notoff().
>
> Additionally, OREG won't be converted to R2PACK-style index 
> addressing, if required.
>
> I can see that fcom is building OREGs.  I don't think that decomposing 
> an OREG in pass2 is the best approach.
You're right here, better to do that in fcom when the references are 
created.

-- Ragge
[prev in list] [next in list] [prev in thread] [next in thread]