RFA: Use _Unwind_Ptr (Was: Re: RFC / RFA: dwarf2 unwinding for targets with call-part-clobbered registers)

[Richard Henderson]

On Thu, Jul 10, 2003 at 06:53:44PM +0100, Joern Rennecke wrote:
> > Hum.  I'm not thrilled.  And I think it may miss out on external
> > users of _Unwind_GetGR etc.
> 
> I can't find any evidence of relevant external users.

Perhaps not _Unwind_GetGR in gcc directly, but it is an ABI
specified interface.  We do use _Unwind_SetGR in gcc source.

> I think you need a cast here to avoid a warning.

Not for extension of an integer, surely.

> That would appear to give the right result - and even with more
> functionality consistently available.
> However, won't that kill unwinding performance?

I wouldn't expect so.  That was one of the reasons that I
checked only specific sizeof's, and didn't use read_[248],
which would fall back to unaligned loads (which fall back
to byte-by-byte loads).

I would expect the difference on x86 to be

	cmpl	$4, mem
	jne	to_abort

which should be non-measurable.


r~