This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

[patch] i960 nonlocal goto vs fp



That change broke gcc.c-torture/execute/920428-2.c -O3
-fomit-frame-pointer.  This reverts just this bit of that change.

2001-02-21  DJ Delorie  <dj@redhat.com>

	* config/i960/i960.h (FRAME_POINTER_REQUIRED): Revert removal of
	check for current_function_has_nonlocal_goto from 1999-11-12.

Index: config/i960/i960.h
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/i960/i960.h,v
retrieving revision 1.38
diff -p -3 -r1.38 i960.h
*** i960.h	2000/11/02 23:29:10	1.38
--- i960.h	2001/02/22 04:16:10
*************** extern int target_flags;
*** 617,623 ****
  /* ??? It isn't clear to me why this is here.  Perhaps because of a bug (since
     fixed) in the definition of INITIAL_FRAME_POINTER_OFFSET which would have
     caused this to fail.  */
! #define FRAME_POINTER_REQUIRED (! leaf_function_p ())
  
  /* Definitions for register eliminations.
  
--- 617,628 ----
  /* ??? It isn't clear to me why this is here.  Perhaps because of a bug (since
     fixed) in the definition of INITIAL_FRAME_POINTER_OFFSET which would have
     caused this to fail.  */
! /* ??? Must check current_function_has_nonlocal_goto, otherwise frame pointer
!   elimination messes up nonlocal goto sequences.  I think this works for other
!   targets because they use indirect jumps for the return which disables fp
!   elimination.  */
! #define FRAME_POINTER_REQUIRED \
!   (! leaf_function_p () || current_function_has_nonlocal_goto)
  
  /* Definitions for register eliminations.
  


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]