This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: reload_indi with scratch == target?
- From: Richard Henderson <rth at redhat dot com>
- To: Ulrich Weigand <Ulrich dot Weigand at de dot ibm dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Mon, 11 Mar 2002 14:55:32 -0800
- Subject: Re: reload_indi with scratch == target?
- References: <OF7748471A.58A6ADC1-ONC1256B79.0071A345@de.ibm.com>
On Mon, Mar 11, 2002 at 09:57:26PM +0100, Ulrich Weigand wrote:
> Did I misread the manual or is this a bug in reload?
Call it a design flaw.
There are two kinds of scratch registers allocated by reload that
serve distinct purposes. Ports used them interchangably, leading
to bug fixes for things that weren't bugs, leading to a mess that
will be hard to untangle.
That's probably an oversimplification; I don't recall all of the
details at the moment. It probably wasn't realized from the start
that the two classes of scratches were in fact distinct.
> Should I implement a similar workaround to sparc64 for our platform?
Yes. We do this for Alpha as well.
r~