This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PR2076
- To: dje at watson dot ibm dot com
- Subject: Re: PR2076
- From: Geoff Keating <geoffk at geoffk dot org>
- Date: Thu, 26 Apr 2001 20:14:28 -0700
- CC: dalej at apple dot com, gcc at gcc dot gnu dot org, shebs at apple dot com
- References: <200104270019.UAA29006@makai.watson.ibm.com>
- Reply-to: Geoff Keating <geoffk at redhat dot com>
> cc: Geoff Keating <geoffk@redhat.com>, gcc@gcc.gnu.org,
> Stan Shebs <shebs@apple.com>
> Date: Thu, 26 Apr 2001 20:19:13 -0400
> From: David Edelsohn <dje@watson.ibm.com>
>
> The only options is to conform to the ABI. Period. GCC must
> interact with other compilers. If the ABI says that the values will be in
> FPRs, another compiler may expect to find them there.
>
> I will check into this further next week, but I do not believe
> that the compiler has any leeway in the implementation. Improving code
> quality is an optimization. One should not hard-code optimizations like
> this; the compiler must deduce the optimization when the information
> allows it to do so.
This seems OK---in any case, if we added such an optimisation, it
should be a separate patch, not as part of a bug fix. So could you
look at Dale's second patch? It should generate equivalent code in
all situations but for fixing the aliasing bug.
--
- Geoffrey Keating <geoffk@geoffk.org>