This is the mail archive of the gcc-patches@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: [RFA] dwarf2out.c: Don't emit DW_OP_regx for register -1


On Fri, Mar 07, 2003 at 05:56:17PM -0700, Kevin Buettner wrote:
> On Mar 7,  9:27am, cgd at broadcom dot com wrote:
> 
> > this shows up with a set of mips64-linux tools, using the "-O3
> > -mips64" flag.  (-mips64 enables use of the "madd" instruction.)
> > 
> > in te2.cpp, when optimizing (and using madd), i'd expect that 'a9'
> > will live entirely in "LO", from the time that is initialized until
> > its last use.
> 
> It seems clear that the "right thing to do" is to make sure that 'hi'
> and 'lo' have DWARF register numbers associated with them.  (gcc will
> currently output -1 for them.)  But, what should these numbers be?
> I've been searching for much of the day, but have not come up with an
> answer yet.
> 
> The DWARF 3 (draft) specification says:
> 
>     Note that the register number represents a DWARF specific mapping
>     of numbers onto the actual registers of a given architecture.  The
>     mapping should be chosen to gain optimal density and should be
>     shared by all users of a given architecture.  It is recommended
>     that this mapping be defined by the ABI authoring committee for
>     each architecture.
> 
> I have consulted the MIPS supplement for the System V ABI, but have
> not found _any_ information about the mappings between registers and
> the numbers used to represent them in debugging info.  In the past,
> when I've looked at the ABI info for Irix, I don't remember seeing any
> such thing either.
> 
> Does anyone know if this kind of thing is documented anywhere?  (And, if
> so, where?)

I recommend asking the folks at MIPS Technologies; probably Dominic is
a good person for this sort of thing.

> If we're free to choose some numbers for 'hi' and 'lo', what should they
> be?

If they don't know of any existing practice, then numbering the other
interesting registers right after the current ones is probably fine.
 
-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer


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