Re: optimization/7005: Sibling and tail recursive calls optimisation with -fomit-frame-pointer: strange code generated.

[Andrew Pinski <pinskia@physics.uc.edu>]

The following reply was made to PR optimization/7005; it has been noted by GNATS.

From: Andrew Pinski <pinskia@physics.uc.edu>
To: fshvaige@cisco.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: optimization/7005: Sibling and tail recursive calls optimisation with -fomit-frame-pointer: strange code generated.
Date: Wed, 12 Jun 2002 09:36:09 -0400 (EDT)

 This has been fixed in gcc version 3.2 20020607 (experimental) at least for the i686 version.
 I did not check the ppc-eabi version but I do not think that the PowerPC version can be changed because the abi specifies this.
 
 Thanks,
 Andrew Pinski
 
 
 > >Description:
 > File test2.c:
 > 
 > int f1 (void);
 > int f2 (void) {
 >   return f1 ();
 > }
 > 
 > i686 -Os (best code):
 > 
 > f2:
 >   jmp     f1
 > 
 > i686 -O2/-O3 (Ooops):
 > 
 > f2:
 >   subl  $12, %esp
 >   addl  $12, %esp
 >   jmp f1
 > 
 > PowerPC -Os/-O2/-O3 (Ooops):
 > 
 > f2:
 >   stwu 1,-8(1)
 >   mflr 0
 >   stw 0,12(1)
 >   bl f1 
 >   lwz 0,12(1)
 >   addi 1,1,8 
 >   mtlr 0
 >   blr   
 > >How-To-Repeat:
 > gcc -S -Os test2.c -Wall -fomit-frame-pointer
 > gcc -S -O2 test2.c -Wall -fomit-frame-pointer
 > gcc -S -O3 test2.c -Wall -fomit-frame-pointer
 > ppc-eabi-gcc -S -Os test2.c -Wall -fomit-frame-pointer
 > ppc-eabi-gcc -S -O2 test2.c -Wall -fomit-frame-pointer
 > ppc-eabi-gcc -S -O3 test2.c -Wall -fomit-frame-pointer