This is the mail archive of the 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: Fix dealing with large (>2GB) static data initalizers

> On Sun, 11 May 2003, Jan Hubicka wrote:
> > on all 64bit targets.  This is because all sizes in ASM_OUTPUT_* machinery are
> > ints.  This patch makes them HOST_WIDE_INTs that brings large amount of
> > anoyance in updating the targets.  I updated x86_64, alpha and partly pa/s390
> > targets to deal with the sizes properly and added simple cast in the rest (to
> > reduce amount of mistakes in the patch).  Given the amount of changes needed I
> > would like to update the rest of 64bit targets incrementally.
> Since you're changing all these macro definitions, and the interface for
> targets to these macros, perhaps you could move all these macros into the
> target structure (thereby giving them properly defined function types)?  
> (There is a case for doing the two steps separately, in which case this
> patch to fix the interface may be a reasonable start.)  Some ASM_OUTPUT
> macros have already moved there, but there are many more to do.

I originally  started to do that but the patch run out of controll.
There are too many macros affected and it is dificult to track down the
dependencies on each other.  I would preffer to do this separately from
the bugfix one by one.
> As a general comment on this patch, the move to ISO C90 means that you can 
> now use string concatenation with HOST_WIDE_INT_PRINT_DEC: you don't need 
> to split up the printfs.
This is cool, I forgot about that.  I can send updated patch once we
settle down on the other issue.

> -- 
> Joseph S. Myers

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