Deep CSE bug!

[Dave Korn]

> -----Original Message-----
> From: gcc-owner On Behalf Of Dave Korn
> Sent: 17 June 2004 16:39
> To: gcc
> Subject: RE: Deep CSE bug!

> > > is it valid that src_eqv is
> > > 
> > > (const_double -1 [0xffffffff] 0 [0x0] 0 [0x0] 0 [0x0] 0 
> > [0x0] 0 [0x0])
> > 
> > ...or I suppose I should wonder if the REG_EQUAL note is a 
> > valid thing to
> > attach to the SImode subreg rather than the DImode reg?
> 
> 
>   Hmm.  I see that Jie Zhang had pretty much this exact same 
> problem over a
> year ago, but I can't find any followups to this post:
> 
> http://gcc.gnu.org/ml/gcc/2003-04/msg01397.html
 

And I notice that Kazu Hirata stumbled across this issue back in December
2002 and also didn't get any answer:

http://gcc.gnu.org/ml/gcc/2002-12/msg01508.html

Hmm.  This has been extant for some time, then.  I'm starting to think that
the reg_equal note when applied to a SUBREG should only contain the relevant
subword of the full constant.  But I'm just not sure yet.


    cheers, 
      DaveK
-- 
Can't think of a witty .sigline today....