RE: Error on invalid asm registers

[Hans-Peter Nilsson <hp at bitrange dot com>]

On Fri, 9 Apr 1999, Simon Marlow wrote:
> Hans-Peter,
> 
> As I understand it, your patch disallows the use of the frame pointer (%ebp
> on x86) as a global register variable.  This is a plea for re-instatement:
> we've been using %ebp as a global register variable for several years now,
> and absolutely *depend* on it for decent performance from our Haskell
> compiler on x86 platforms (see
> http://research.microsoft.com/users/t-simonm/ghc/).  I fear that we won't be
> able to make use of future versions of egcs unless a solution to the problem
> can be found.
>
> Cheers,
>         Simon

Thanks for your comments, but please direct comments about the patch to
egcs-patches@egcs.cygnus.com.  It is not for me to decide if it will be
installed.

The current situation makes for hard-to-trace bugs at runtime and is
inconsistent; it allows the frame-pointer register to be specified in some
contexts and disallows it in others.
 Disallowing the frame-pointer register as an asm-reg totally seems more
right; if not, then checks need to be added to emit errors in cases where
the frame pointer can not be eliminated (and the frame pointer register is
needed for internal use by gcc).

brgds, H-P