This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Prerelease

>    From: "David O'Brien" <obrien@FreeBSD.org>
>    Date: Fri, 10 May 2002 19:14:41 -0700
> 
>    On Fri, May 10, 2002 at 05:36:19PM -0700, David S. Miller wrote:
>    > I think it's appropriate for 3.1.1
>    
>    Can you raise the seriousness level to what is appropriate so it is
>    on the to-be-fixed list for 3.1.1?
>    
> I would like an x86 expert to analyse it first.

Well, I don't feel to be expert, but it looks like effect of the
FUNCTION_BOUNDARY change from 8 to 16.  According to comments in defaults.h
this is required by C++ ABI as it does use the last bit of pointer for
different purposes, so now GCC limits the alignment to 2, not 1.

Concerning the padding comparison, I guess -falign-loops=0 -falign-jumps=0
is needed to avoid it.

Other bug, that makes me more nervous is target/6104. It is shame that -march=pentium4
appears to be seriously broken for FP.  I was trying to debug this failure and it appears
to be generic reload problem, but I didn't find the exact cause :(

Honza
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]