This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Excessive stack usage on m68k - BUG?
- To: Kamil Iskra <iskra@icsr.agh.edu.pl>
- Subject: Re: Excessive stack usage on m68k - BUG?
- From: Andreas Schwab <schwab@suse.de>
- Date: 26 Aug 1999 16:03:27 +0200
- Cc: gcc@gcc.gnu.org, gcc-patches@gcc.gnu.org
- References: <Pine.A32.4.05.9908261418430.17351-100000@val.icsr.agh.edu.pl>
Kamil Iskra <iskra@icsr.agh.edu.pl> writes:
|> Now, with gcc-2.95:
|>
|> #NO_APP
|> gcc2_compiled.:
|> ___gnu_compiled_c:
|> .text
|> .even
|> .globl _main
|> _main:
|> link a5,#-16
|> jbsr ___main
|> pea a5@(-16)
|> jbsr _f
|> moveq #0,d0
|> unlk a5
|> rts
|>
|> Why 16 bytes?
This is a generic bug in the backend:
Thu Aug 26 16:00:17 1999 Andreas Schwab <schwab@suse.de>
* function.c (assign_stack_temp_for_type): Fix change of Mar 5 for
the fact that ALIGN is measured in bits, not bytes.
Index: function.c
===================================================================
RCS file: /egcs/carton/cvsfiles/egcs/gcc/function.c,v
retrieving revision 1.98
diff -u -a -u -r1.98 function.c
--- function.c 1999/08/24 12:04:54 1.98
+++ function.c 1999/08/26 14:00:02
@@ -674,11 +674,12 @@
So for requests which depended on the rounding of SIZE, we go ahead
and round it now. We also make sure ALIGNMENT is at least
BIGGEST_ALIGNMENT. */
- if (mode == BLKmode && align < (BIGGEST_ALIGNMENT / BITS_PER_UNIT))
+ if (mode == BLKmode && align < BIGGEST_ALIGNMENT)
abort();
p->slot = assign_stack_local (mode,
- mode == BLKmode
- ? CEIL_ROUND (size, align) : size,
+ (mode == BLKmode
+ ? CEIL_ROUND (size, align / BITS_PER_UNIT)
+ : size),
align);
p->align = align;
--
Andreas Schwab "And now for something
schwab@suse.de completely different."
SuSE GmbH, Schanzäckerstr. 10, D-90443 Nürnberg