This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: builtin_setjmp_receiver expander for SH
- To: gcc-patches at gcc dot gnu dot org
- Subject: Re: builtin_setjmp_receiver expander for SH
- From: kaz Kojima <kkojima at rr dot iij4u dot or dot jp>
- Date: Mon, 09 Apr 2001 11:24:14 +0900
- References: <org0fi6e67.fsf@guarana.lsd.ic.unicamp.br>
Alexandre Oliva <aoliva@redhat.com> wrote:
> On Apr 8, 2001, Richard Henderson <rth@redhat.com> wrote:
>
>> On Sun, Apr 08, 2001 at 09:35:09AM +0900, kaz Kojima wrote:
>>> + emit_insn (gen_GOTaddr2picreg ());
>>> + emit_insn (gen_rtx_USE (VOIDmode, pic_offset_table_rtx));
>
>> Why do you emit the use? The PIC register should be fixed
>> (or you have other more serious problems) at which point I
>> don't think we remove references to it.
>
> Presumably because flow might delete restores after which the register
> is not explicitly used. I don't think that's necessary, though. All
> cases that need the PIC register to be properly set are marked as
> such, so, if flow finds a restore is not necessary, it's very likely
> that it really isn't.
Agreed. The reason why I emit the use rtl is just as you say. I was
affected with the comments at the end of load_pic_register in sparc.c
and arm_finalize_pic in arm.c. I attach the revised patch.
Thanks!
kaz
--
* sh.md (builtin_setjmp_receiver): New expander.
Index: sh.md
===================================================================
RCS file: /cvs/gcc/egcs/gcc/config/sh/sh.md,v
retrieving revision 1.82
diff -u -r1.82 sh.md
--- sh.md 2001/04/03 15:06:11 1.82
+++ sh.md 2001/04/09 02:07:54
@@ -3752,6 +3752,15 @@
operands[1] = gen_rtx_SYMBOL_REF (VOIDmode, GOT_SYMBOL_NAME);
}
")
+
+(define_expand "builtin_setjmp_receiver"
+ [(label_ref (match_operand 0 "" ""))]
+ "flag_pic"
+ "
+{
+ emit_insn (gen_GOTaddr2picreg ());
+ DONE;
+}")
(define_expand "call_site"
[(unspec [(match_dup 0)] UNSPEC_CALLER)]