This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patches] Re: ARM libstdc++ failure fix
> On Sun, Jul 29, 2001 at 06:52:33PM +0200, Jan Hubicka wrote:
> > + /* Given that we've done no other optimizations yet,
> > + the arguments should be immediately available. */
> > + if (GET_CODE (before) == CODE_LABEL && before != boundary)
> > + abort ();
>
> While this comment was applicable to except.c, I don't know that
> it is for gcse.c. There's no particular reason we couldn't have
> hoisted a (common) argument load into a dominator block.
Yes, thats why I've added the before!=boundary check.
except.c calls with boundary = NULL, while gcse with boundary = head
(the code label, if exists).
This way we keep the original behaviour. The comment is out of date,
can you suggest better wording?
Whats about the Graham's idea of checking CALL_INSN too?
>
> While it's probably ok, it'd be better to test on a platform
> that actually uses this code instead of i586.
Choose ppc, mips, sparc (or less prefebraly arm or sh) :)
OK after checking?
Honza
>
>
> r~