This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: other/7114: ICE building strcoll.op from glibc-2.2.5
- From: Alan Modra <amodra at bigpond dot net dot au>
- To: David Edelsohn <dje at watson dot ibm dot com>, geoffk at redhat dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 15 Jul 2002 21:50:18 +0930
- Subject: Re: other/7114: ICE building strcoll.op from glibc-2.2.5
- References: <20020625081846.10430.qmail@sources.redhat.com> <20020715092603.GS30362@bubble.sa.bigpond.net.au>
On Mon, Jul 15, 2002 at 06:56:03PM +0930, Alan Modra wrote:
> This patch cures the testcase. The !using_store_multiple code tests
> whether regs are live before saving. We need to do something similar
> for using_store_multiple, in case all regs need not be saved.
>
> * config/rs6000/rs6000.c (rs6000_emit_prologue): Trim saved regs
> for -mmultiple case like -mno-multiple case.
> (rs6000_emit_epilogue): Likewise.
While this does cure the ICE, on further investigation, I'm not happy.
I think trimming off the saved regs in the -mno-multiple case is
wrong, and the above patch just copies wrong code. Why isn't
first_reg_to_save giving us the right number?
The answer to this is that r30 isn't being marked as used in the
current_function_needs_context case. PR 5967 is one result.
What's the right way to fix this? Using get_hard_reg_initial_val
for the static chain reg?
--
Alan Modra
IBM OzLabs - Linux Technology Centre