This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: rfc: unwinder, ppc-spe dual sized registers, and a patch...
- From: Aldy Hernandez <aldyh at redhat dot com>
- To: Zack Weinberg <zack at codesourcery dot com>
- Cc: gcc at gcc dot gnu dot org, geoffk at geoffk dot org, dje at watson dot ibm dot com, rth at redhat dot com, jason at redhat dot com
- Date: Sun, 9 Mar 2003 20:36:18 -0800
- Subject: Re: rfc: unwinder, ppc-spe dual sized registers, and a patch...
- References: <20030309190409.A5975@redhat.com> <87heab3nvm.fsf@egil.codesourcery.com>
- Reply-to: aldyh at redhat dot com
On Sun, Mar 09, 2003 at 07:24:13PM -0800, Zack Weinberg wrote:
> Aldy Hernandez <aldyh at redhat dot com> writes:
>
> > There is one heads up-- the ABI says the register pairs should be N and
> > N+1200 but this creates a huge hole in the unwinder tables, thus defeating
> > the whole point, which was (well, to fix exception handling and...)
> > to save frame size. I've set the pairs to N and N+113, unless someone
> > has a huge objection. This creates a bit of a confusion with the
> > debugging info which is ABI compliant and has DW_OP_pieces of N and
> > N+1200. I'm willing to live with that. Suggestions welcome.
>
> This concerns me, because I (that is, CodeSourcery) have clients who
> expect to be able to interoperate C++ code generated by GCC and by
> their proprietary compiler, on the e500. Can't you use a sparse array
> or something?
Meanwhile, back at the ranch, there's a discussion on irc about this...
Making the unwinder use sparse data structures may be overkill, and
it's time consuming ;-). Daniel Jacobowitz suggested having a target macro
mapping plain registers to dwarf registers. Easy, simple. I'll be doing
that tomorrow.
Richard, I need confirmation thet upping DWARF_FRAME_REGISTERS won't break
ABI compatability... I'm 98.7% sure.
Aldy