[Bug middle-end/23391] New: [4.1 regression] Tree checking failure due to scev

[Bug middle-end/23391] New: [4.1 regression] Tree checking failure due to scev

[steven at gcc dot gnu dot org]

Test case: 
========================================== 
void 
foo (int N) 
{ 
  int C; 
  double R; 
 
  R = 0.0; 
  do 
    { 
      R += 0.001; 
      C = (int) (R * N); 
      if (-R * N <= R * N) 
        { 
          C++; 
        } 
    } 
  while (C < 0); 
 
  return; 
} 
========================================== 
 
$ ./cc1 -fno-unit-at-a-time -O1 -ffast-math t.c 
 foo 
t.c: In function 'foo': 
t.c:3: internal compiler error: tree check: expected real_cst, have 
integer_cst in const_binop, at fold-const.c:1512 
 
The test case comes from SPEC twolf.  The problem appeared after this patch 
was committed: http://gcc.gnu.org/ml/gcc-patches/2005-08/msg00801.html

-- 
           Summary: [4.1 regression] Tree checking failure due to scev
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P2
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: steven at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org,spop at gcc dot gnu dot
                    org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-15 00:01 -------
Confirmed, only -O1 is required to reproduce this.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|middle-end                  |tree-optimization
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-15 00:01:26
               date|                            |
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-08-15 00:05 -------
Note that part of the problem is that build_int_cst_type, which scev uses 
here, should check that the type given to it is an integral type.  That would 
have caught the checking failure much earlier.  Right now scev tries to use 
build_int_cst_type with type==double. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-15 00:09 -------
This is related to PR 19899 which was fixed.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[sebastian dot pop at cri dot ensmp dot fr]

------- Additional Comments From sebastian dot pop at cri dot ensmp dot fr  2005-08-15 00:34 -------
Subject: Re:  [4.1 regression] Tree checking failure due to scev

This patch should fix the problem.  There are some more cases that use
build_int_cst instead of build_real.  I'll propose a more complete
patch later.

*** tree-scalar-evolution.c.~2.35.~	2005-08-13 19:06:26.000000000 +0200
--- tree-scalar-evolution.c	2005-08-15 02:36:50.000000000 +0200
***************
*** 1680,1686 ****
        opnd10 = TREE_OPERAND (opnd1, 0);
        chrec10 = analyze_scalar_evolution (loop, opnd10);
        chrec10 = chrec_convert (type, chrec10, at_stmt);
!       res = chrec_fold_minus (type, build_int_cst (type, 0), chrec10);
        break;
  
      case MULT_EXPR:
--- 1680,1688 ----
        opnd10 = TREE_OPERAND (opnd1, 0);
        chrec10 = analyze_scalar_evolution (loop, opnd10);
        chrec10 = chrec_convert (type, chrec10, at_stmt);
!       res = chrec_fold_multiply (type, chrec10, SCALAR_FLOAT_TYPE_P (type)
! 				  ? build_real (type, dconstm1)
! 				  : build_int_cst_type (type, -1));
        break;
  
      case MULT_EXPR:
*** tree-chrec.c.~2.23.~	2005-08-13 19:06:26.000000000 +0200
--- tree-chrec.c	2005-08-15 02:20:51.000000000 +0200
***************
*** 284,291 ****
  	    return build_polynomial_chrec 
  	      (CHREC_VARIABLE (op1), 
  	       chrec_fold_minus (type, op0, CHREC_LEFT (op1)),
! 	       chrec_fold_multiply (type, CHREC_RIGHT (op1),
! 				    build_int_cst_type (type, -1)));
  
  	default:
  	  {
--- 284,292 ----
  	    return build_polynomial_chrec 
  	      (CHREC_VARIABLE (op1), 
  	       chrec_fold_minus (type, op0, CHREC_LEFT (op1)),
! 	       chrec_fold_multiply (type, CHREC_RIGHT (op1), SCALAR_FLOAT_TYPE_P (type)
! 				    ? build_real (type, dconstm1)
! 				    : build_int_cst_type (type, -1)));
  
  	default:
  	  {


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-08-15 12:26 -------
Subject: Bug 23391

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	spop@gcc.gnu.org	2005-08-15 12:26:13

Modified files:
	gcc            : ChangeLog Makefile.in tree-chrec.c 
	                 tree-scalar-evolution.c 
Added files:
	gcc/testsuite/gcc.dg/tree-ssa: pr23391.c 

Log message:
	PR 23391
	* Makefile.in (tree-chrec.o): Depends on real.h.
	* tree-chrec.c: Include real.h.
	(chrec_fold_plus_poly_poly, chrec_fold_multiply_poly_poly,
	chrec_fold_plus_1): Use build_real for SCALAR_FLOAT_TYPE_P.
	* tree-scalar-evolution.c (add_to_evolution_1,
	interpret_rhs_modify_expr): Ditto.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.9732&r2=2.9733
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/Makefile.in.diff?cvsroot=gcc&r1=1.1535&r2=1.1536
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-chrec.c.diff?cvsroot=gcc&r1=2.23&r2=2.24
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-scalar-evolution.c.diff?cvsroot=gcc&r1=2.35&r2=2.36
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/tree-ssa/pr23391.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[sebastian dot pop at cri dot ensmp dot fr]

------- Additional Comments From sebastian dot pop at cri dot ensmp dot fr  2005-08-15 12:33 -------
Subject: Re:  [4.1 regression] Tree checking failure due to scev

pinskia at gcc dot gnu dot org wrote:
> This is related to PR 19899 which was fixed.
> 

Yes, PR is related to PR19899, but same pattern occured in several
places and the fix to PR19899 was not complete.  Fixed now.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[sebastian dot pop at cri dot ensmp dot fr]

------- Additional Comments From sebastian dot pop at cri dot ensmp dot fr  2005-08-15 12:35 -------
Subject: Re:  [4.1 regression] Tree checking failure due to scev

Sebastian Pop wrote:
> 
> Yes, PR is related to PR19899, but same pattern occured in several
> places and the fix to PR19899 was not complete.  Fixed now.
> 

We'll probably need this fix for 4.0 branch as well...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-15 14:53 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391

[Bug tree-optimization/23391] [4.1 regression] Tree checking failure due to scev

[reichelt at gcc dot gnu dot org]

------- Additional Comments From reichelt at gcc dot gnu dot org  2005-08-17 11:23 -------
*** Bug 23433 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |micis at gmx dot de


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23391