[Bug tree-optimization/55022] New: [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[Bug tree-optimization/55022] New: [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

             Bug #: 55022
           Summary: [4.8 Regression] air.f90 is miscompliled with -m64 -O2
                    -fgraphite-identity after revision 190619
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dominiq@lps.ens.fr
                CC: grosser@gcc.gnu.org, rguenth@gcc.gnu.org


The test air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after
revision 190619 (r188694 is OK):

...
  ITERATION#  TIME             FINAL          MASS RESIDUAL
        9    0.000045757110   0.01000000    0.597638
       17    0.000091473627   0.01000000    0.448609
       25    0.000137176442   0.01000000    0.608353
       33    0.000182824593   0.01000000    0.619686
       41    0.000228306169   0.01000000    0.628800
       49    0.000273480245   0.01000000    0.813862
       58    0.000323343055   0.01000000    1.026950
       67    0.000371631805   0.01000000    1.238993
       77    0.000422628642   0.01000000    1.474307
       88    0.000474256730   0.01000000    1.679987
      100    0.000524089193   0.01000000    1.789565
      114    0.000575138355   0.01000000    1.834548
      129    0.000625547555   0.01000000    1.945421
      144    0.000674419139   0.01000000    2.198194
      166    0.000728576510   0.01000000         NaN
      168  100.000735868323   0.01000000         NaN
 deltat, final t, iterations
   100.00000000000000        100.00073586832266              168
           9  0.59763770273003503     
          17  0.44860912267491976     
          25  0.60835306073381645     
          33  0.61968645389575716     
          41  0.62879989012582926     
          49  0.81386186515901682     
          58   1.0269501832675307     
          67   1.2389929088800784     
          77   1.4743070523641029     
          88   1.6799869908106582     
         100   1.7895645561305713     
         114   1.8345478482485826     
         129   1.9454206794378737     
         144   2.1981944269573130     
         166                       NaN
         168                       NaN

The miscompiled subroutines are botwall and topwall.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #1 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-10-22 14:24:51 UTC ---
Created attachment 28507
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28507
compressed archive for the test

The attachment contains the original source air_main.f90 without the
subroutines botwall and topwall which are in the files botwall.f90 and
topwall.f90, and the files datax.inp and datay.inp needed to run the test. It
can be used as

gfc -c -O2 -fgraphite-identity air_main.f90
gfortran -c -O2 -fgraphite-identity botwall.f90 topwall.f90
gfc air_main.o botwall.o topwall.o

With gfc any post r190619 gfortran and gfortran any version pre r188694, this
will give a working executable. If botwall.f90 and/or topwall.f90 is compiled
with a post r190619 gfortran, the executable will give NANs.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-10-23 13:35:44 UTC ---
It likely has started after revision 189156.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.0

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |x86_64-*-*

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2013-01-04 03:10:36 UTC ---
Does this work now?

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-01-04 07:29:09 UTC ---
> Does this work now?

Nope! -(r194872).

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[aldyh at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aldyh at gcc dot gnu.org
         AssignedTo|unassigned at gcc dot       |aldyh at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #5 from Aldy Hernandez <aldyh at gcc dot gnu.org> 2013-02-05 15:57:55 UTC ---
I'll take a peek.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #6 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-02-05 16:00:21 UTC ---
> I'll take a peek.

Great!-)

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[grosser at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #7 from Tobias Grosser <grosser at gcc dot gnu.org> 2013-02-05 16:10:21 UTC ---
(In reply to comment #5)
> I'll take a peek.

Wonderful, feel free to CC me if you have any questions!

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[aldyh at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-02-05
     Ever Confirmed|0                           |1

--- Comment #8 from Aldy Hernandez <aldyh at gcc dot gnu.org> 2013-02-05 23:52:30 UTC ---
Confirmed with cloog 0.18 and isl 0.11, both installed in /build/install, and
configuring GCC with:

/source/gcc/configure --disable-bootstrap --disable-multilib
--enable-languages=c,c++,fortran --with-cloog=/build/install
--with-isl=/build/install

houston:/build/55022-f90/gcc/pr55022$ ls
air_main.f90  botwall.f90  datax.inp  datay.inp  topwall.f90
houston:/build/55022-f90/gcc/pr55022$ ../xgcc -B../ -O2 -fgraphite-identity
air_main.f90 botwall.f90  topwall.f90 -L
../../x86_64-unknown-linux-gnu/libgfortran/.libs/ -lgfortran -lm
houston:/build/55022-f90/gcc/pr55022$ ./a.out
...
...
      129    0.000625547555   0.01000000    1.945421
      144    0.000674419139   0.01000000    2.198194
      166    0.000728576510   0.01000000         NaN
      168  100.000735868323   0.01000000         NaN

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[aldyh at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING

--- Comment #9 from Aldy Hernandez <aldyh at gcc dot gnu.org> 2013-02-06 18:18:54 UTC ---
I'm sorry, but the reporter is going to have to reduce this significantly
before we can debug this.  This is a huge mess of spaghetti code with tons of
global and common variables scattered throughout, as well as state being passed
around in files.

Dominique, please reduce this to the smallest amount of reproducible code, so
we can attack this.  Avoid reading and writing to files (as I see, these have
only a minimal amount of data anyhow), and global variables if at all possible.

Please only keep the code that is relevant to reproduce the problem.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[aldyh at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |matz at gcc dot gnu.org,
                   |                            |sebpop at gmail dot com

--- Comment #10 from Aldy Hernandez <aldyh at gcc dot gnu.org> 2013-02-06 21:08:02 UTC ---
This regression started when we moved graphite from using PPL over to ISL.  The
patch submission started here:

    http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01773.html

    2012-07-02  Richard Guenther  <rguenther@suse.de>
        Michael Matz  <matz@suse.de>
        Tobias Grosser <tobias@grosser.es>
        Sebastian Pop <sebpop@gmail.com>

and the revision is trunk@189156.

Provided this test gets further reduced as I have mentioned in comment 9, can
any of you gentlemen look at this?

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW

--- Comment #11 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-02-07 14:57:40 UTC ---
I have reduced botwall.f90 to

!*==BOTWALL.spg  processed by SPAG 6.55Dc at 09:26 on 23 Sep 2005
!
      SUBROUTINE BOTWALL(Spx,Epx,Apy,Mxpx,U1,U2,U3,U4,Iv)
      IMPLICIT REAL*8(A-H,O-Z)
      PARAMETER (NX=150,NY=150)
      DIMENSION U(NX,NY) , V(NX,NY) , RHO(NX,NY) , T(NX,NY) , E(NX,NY)
      DIMENSION AS1(NX) , AS2(NX) , AS3(NY) , AS4(NY) , NX1(NX) ,       &
     &          NY1(NX)
      DIMENSION NX2(NX) , NY2(NX) , SIG(NX) , P(NX,NY)
      DIMENSION U1(NX,NY) , U2(NX,NY) , U3(NX,NY) , U4(NX,NY)
      INTEGER Spx , Epx , Apy
      REAL*8 NX1 , NX2 , NY1 , NY2 , MINlet(NY)
      COMMON /BNDRY / U , V , P , RHO , T , E , AS1 , AS2 , AS3 , AS4 , &
     &                NX1 , NY1 , NX2 , NY2 , SIG , GMA , S0 , T0 , P0 ,&
     &                PE , HOO , RR , MINlet
!
      TWX_0 = 272.77
      E_0 = RR*TWX_0/(GMA-1.)
      R_0 = 1.0/(TWX_0*RR)
      A_0 = DSQRT(GMA*TWX_0*RR)
      DO i = Spx , Epx
!
! BOTTOM WALL CONDITIONS FOR SUBSONIC FLOW
!
         vn1 = U(i,Apy)*NX1(i) + V(i,Apy)*NY1(i)
         P(i,Apy) = P(i,Apy)/DEXP(GMA*vn1/AS1(i))
         U(i,Apy) = 0.0
         V(i,Apy) = 0.0
         T(i,Apy) = TWX_0
         RHO(i,Apy) = P(i,Apy)*R_0
         AS1(i) = A_0
         E(i,Apy) = E_0
         U1(i,Apy) = RHO(i,Apy)
         U2(i,Apy) = 0.0
         U3(i,Apy) = 0.0
         U4(i,Apy) = RHO(i,Apy)*E_0
      ENDDO
!      print *, U1(Spx:Epx,Apy)
!      print *, U2(Spx:Epx,Apy)
!      print *, U3(Spx:Epx,Apy)
!      print *, U4(Spx:Epx,Apy)
      print *, U1(Spx:Spx+4,Apy)
      print *, U2(Spx:Spx+4,Apy)
      print *, U3(Spx:Spx+4,Apy)
      print *, U4(Spx:Spx+4,Apy)
      stop
      CONTINUE
      END

With the following compilation (gfc: gcc 4.8 r195848, gfortran: 4.7.2)

gfc -c -O2 -fgraphite-identity air_main.f90
gfortran -c -O2 -fgraphite-identity topwall.f90
gfortran -c -O2 -fgraphite-identity botwall_db.f90
gfc air_main.o botwall_db.o topwall.o

I get

   1.2901579736800930        1.2901579736800932        1.2901579736800932      
 1.2901579736800932        1.2901579736800932     
   0.0000000000000000        0.0000000000000000        0.0000000000000000      
 0.0000000000000000        0.0000000000000000     
   0.0000000000000000        0.0000000000000000        0.0000000000000000      
 0.0000000000000000        0.0000000000000000     
   252499.99999999997        252500.00000000003        252500.00000000003      
 252500.00000000003        252500.00000000003     

If I use

gfc -c -O2 -fgraphite-identity botwall_db.f90
gfc air_main.o botwall_db.o topwall.o

I get

   1.2901579217165384        1.2000000476837158        1.2000000476837158      
 1.2000000476837158        1.2000000476837158     
   0.0000000000000000        6.5847860013093756E-016  -2.6339144005237502E-015 
 0.0000000000000000        7.7354314600658297E-015
   0.0000000000000000        2.3705229604713751E-014   2.3705229604713751E-014 
 2.3705229604713751E-014   2.6300466964223820E-014
   252500.00000000000        252500.00000000009        252500.00000000009      
 252500.00000000009        252500.00000000009     

(In reply to comment #9)
> I'm sorry, but the reporter is going to have to reduce this significantly
> before we can debug this.  This is a huge mess of spaghetti code with tons of
> global and common variables scattered throughout, as well as state being passed
> around in files.
>
> Dominique, please reduce this to the smallest amount of reproducible code, so
> we can attack this.  Avoid reading and writing to files (as I see, these have
> only a minimal amount of data anyhow), and global variables if at all possible.

Well, I am only the reporter. The code is not mine and this makes the task to
reduce it while keeping the wrong code quite difficult (and cumbersome). 

> Please only keep the code that is relevant to reproduce the problem.

I'll try to provide a self contained test, but not before the week-end.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #12 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2013-02-07 15:11:23 UTC ---
If I replace the loop
      DO i = Spx , Epx
         vn1 = U(i,Apy)*NX1(i) + V(i,Apy)*NY1(i)
         P(i,Apy) = P(i,Apy)/DEXP(GMA*vn1/AS1(i))
         U(i,Apy) = 0.0
         V(i,Apy) = 0.0
         T(i,Apy) = TWX_0
         RHO(i,Apy) = P(i,Apy)*R_0
         AS1(i) = A_0
         E(i,Apy) = E_0
         U1(i,Apy) = RHO(i,Apy)
         U2(i,Apy) = 0.0
         U3(i,Apy) = 0.0
         U4(i,Apy) = RHO(i,Apy)*E_0
      ENDDO

with

      DO i = Spx , Epx
         vn1 = GMA*(U(i,Apy)*NX1(i) + V(i,Apy)*NY1(i))/AS1(i)
         P(i,Apy) = P(i,Apy)/DEXP(vn1)
         RHO(i,Apy) = P(i,Apy)*R_0
         AS1(i) = A_0
         E(i,Apy) = E_0
         U1(i,Apy) = RHO(i,Apy)
         U4(i,Apy) = RHO(i,Apy)*E_0
      ENDDO
      U(Spx:Epx,Apy) = 0.0
      V(Spx:Epx,Apy) = 0.0
      T(Spx:Epx,Apy) = TWX_0
      U2(Spx:Epx,Apy) = 0.0
      U3(Spx:Epx,Apy) = 0.0

Then the test does not abort with NaNs, but the result seems wrong:

        1729  0.40924060378447885     
        1737  0.41934356330600248     

compared to

        1729  0.40411964185599653     
        1737  0.41528583240265665

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> 2013-02-20 13:58:41 UTC ---
Graphite -> P2.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.0                       |4.8.1

--- Comment #14 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-03-22 14:41:58 UTC ---
GCC 4.8.0 is being released, adjusting target milestone.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.1                       |4.8.2

--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.1 has been released.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.8.2                       |4.8.3

--- Comment #16 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 4.8.2 has been released.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P2                          |P4

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[mircea.namolaru at inria dot fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Mircea Namolaru <mircea.namolaru at inria dot fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mircea.namolaru at inria dot fr

--- Comment #17 from Mircea Namolaru <mircea.namolaru at inria dot fr> ---
Vladimir Kargov, Tobias Grosser and me found that the problem is caused by
incorrect folding of the floord operator. As a result Cloog translates the
expression
-4294967296*floord(_19-i_17,4294967296)
to the tree-SSA expression
4294967296*floord(_19-i_17,-4294967296)

This is wrong, in the first case floord is 0 and in the second is -1.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

--- Comment #18 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Mircea Namolaru from comment #17)
> Vladimir Kargov, Tobias Grosser and me found that the problem is caused by
> incorrect folding of the floord operator. As a result Cloog translates the
> expression
> -4294967296*floord(_19-i_17,4294967296)
> to the tree-SSA expression
> 4294967296*floord(_19-i_17,-4294967296)
> 
> This is wrong, in the first case floord is 0 and in the second is -1.

Not sure where that "folding" happens or what "floord" is, but if you
are refering to negate_expr () negating a /[fl] CST as a /[fl] -CST
then yes, that looks suspicious.  OTOH the division variant tree codes
don't appear in the IL very often so it's hard to notice these mistakes.
Testing

Index: gcc/fold-const.c
===================================================================
*** gcc/fold-const.c    (revision 209292)
--- gcc/fold-const.c    (working copy)
*************** negate_expr_p (tree t)
*** 484,491 ****

      case TRUNC_DIV_EXPR:
      case ROUND_DIV_EXPR:
-     case FLOOR_DIV_EXPR:
-     case CEIL_DIV_EXPR:
      case EXACT_DIV_EXPR:
        /* In general we can't negate A / B, because if A is INT_MIN and
         B is 1, we may turn this into INT_MIN / -1 which is undefined
--- 484,489 ----
*************** fold_negate_expr (location_t loc, tree t
*** 682,689 ****

      case TRUNC_DIV_EXPR:
      case ROUND_DIV_EXPR:
-     case FLOOR_DIV_EXPR:
-     case CEIL_DIV_EXPR:
      case EXACT_DIV_EXPR:
        /* In general we can't negate A / B, because if A is INT_MIN and
         B is 1, we may turn this into INT_MIN / -1 which is undefined
--- 680,685 ----

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[kargov at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #19 from Vladimir Kargov <kargov at gmail dot com> ---
> if you are refering to negate_expr () negating a /[fl] CST as a /[fl] -CST
then yes, that looks suspicious
Yes, the one.

The function converts expressions "x * -C" into "-x * C". In the case where x
is an expression of form "a /[fl] b" or "a /[cl] b" this optimization is
generally invalid. For example:
(1 /[fl] 2) * -1 = 0 * -1 = 0 will turn into (1 /[fl] -2) * 1 = -1 * 1 = -1
(1 /[cl] 2) * -1 = 1 * -1 = -1 will turn into (1 /[cl] -2) * 1 = 0 * 1 = 0
In general, it appears that the optimization will be incorrect for any floor
and ceiling divisions with a non-zero remainder since after negation they will
be rounded to a different integer.

The problem can be seen in the following function when compiled with "-O3
-fgraphite-identity":

void f(int *limit, int minLen, int maxLen)
{
  int i;

  for (i = minLen; i <= maxLen; i++) {
    limit[i] = i;
  }
}

In this case constant folding will be transforming expression "(((signed long)
maxLen_5(D) - (signed long) minLen_3(D)) /[fl] 4294967296) * -1" (i.e. 0) into
"((signed long) maxLen_5(D) - (signed long) minLen_3(D)) /[fl] -4294967296"
(i.e. -4294967296).

Since Graphite abundantly uses floor division, this breaks a lot of code,
including bzip2 build
(https://groups.google.com/forum/#!topic/gcc-graphite/WaCUu0xg4Kc) and this
bug.

Removing floor and ceiling division as in comment 18 will fix it.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenther at suse dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #20 from rguenther at suse dot de <rguenther at suse dot de> ---
On Fri, 11 Apr 2014, kargov at gmail dot com wrote:

> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022
> 
> --- Comment #19 from Vladimir Kargov <kargov at gmail dot com> ---
> > if you are refering to negate_expr () negating a /[fl] CST as a /[fl] -CST
> then yes, that looks suspicious
> Yes, the one.
> 
> The function converts expressions "x * -C" into "-x * C". In the case where x
> is an expression of form "a /[fl] b" or "a /[cl] b" this optimization is
> generally invalid. For example:
> (1 /[fl] 2) * -1 = 0 * -1 = 0 will turn into (1 /[fl] -2) * 1 = -1 * 1 = -1
> (1 /[cl] 2) * -1 = 1 * -1 = -1 will turn into (1 /[cl] -2) * 1 = 0 * 1 = 0
> In general, it appears that the optimization will be incorrect for any floor
> and ceiling divisions with a non-zero remainder since after negation they will
> be rounded to a different integer.
> 
> The problem can be seen in the following function when compiled with "-O3
> -fgraphite-identity":
> 
> void f(int *limit, int minLen, int maxLen)
> {
>   int i;
> 
>   for (i = minLen; i <= maxLen; i++) {
>     limit[i] = i;
>   }
> }
> 
> In this case constant folding will be transforming expression "(((signed long)
> maxLen_5(D) - (signed long) minLen_3(D)) /[fl] 4294967296) * -1" (i.e. 0) into
> "((signed long) maxLen_5(D) - (signed long) minLen_3(D)) /[fl] -4294967296"
> (i.e. -4294967296).
> 
> Since Graphite abundantly uses floor division, this breaks a lot of code,
> including bzip2 build
> (https://groups.google.com/forum/#!topic/gcc-graphite/WaCUu0xg4Kc) and this
> bug.
> 
> Removing floor and ceiling division as in comment 18 will fix it.

I can't seem to observe the above with the testcase or sth like
the following suitable for the testsuite:

/* { dg-do run } */
/* { dg-options "-O3 -fgraphite-identity" } */

extern void abort (void);

void __attribute__((noinline,noclone))
f(int *limit, int minLen, int maxLen)
{
  int i;

  for (i = minLen; i <= maxLen; i++) {
      limit[i] = i;
  }
}

int main()
{
  int limit[256], i;
  f (limit, 0, 255);
  for (i = 0; i < 256; ++i)
    {
      if (limit[i] != i)
        abort ();
      __asm__ volatile ("" : : : "memory");
    }
  return 0;
}

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[kargov at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #21 from Vladimir Kargov <kargov at gmail dot com> ---
(In reply to rguenther@suse.de from comment #20)

> I can't seem to observe the above with the testcase or sth like
> the following suitable for the testsuite

Can you please check the following:

#include <stdio.h>

void __attribute__ ((noinline))
BZ2_hbCreateDecodeTables(int *limit, int minLen, int maxLen)
{
  int i;

  for (i = minLen; i <= maxLen; i++) {
    limit[i] = i;
  }
}

void main()
{
  int limit[2] = {0, 0};
  int pos;

  BZ2_hbCreateDecodeTables (limit, 0, 1);

  printf("%d\n", limit[1]);
}

When compiled with "-fgraphite-identity -O3" on trunk it will give "0" which is
incorrect.

The patched version should give the correct "1".

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #22 from Richard Biener <rguenth at gcc dot gnu.org> ---
Ah, bogus fix for another bug in my tree.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[kargov at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #23 from Vladimir Kargov <kargov at gmail dot com> ---
(In reply to rguenther@suse.de from comment #20)
Your test also exploits this bug for me on trunk rev. 208043.

[Bug tree-optimization/55022] [4.8/4.9 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[kargov at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #24 from Vladimir Kargov <kargov at gmail dot com> ---
(In reply to Vladimir Kargov from comment #19)
For the sake of correctness, I had an obvious typo in my comment:
> "((signed long) maxLen_5(D) - (signed long) minLen_3(D)) /[fl] -4294967296" (i.e. -4294967296).
((signed long) maxLen_5(D) - (signed long) minLen_3(D)) /[fl] -4294967296 = -1,
not -4294967296.

[Bug tree-optimization/55022] [4.8/4.9/4.10 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #25 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Mon Apr 14 08:11:08 2014
New Revision: 209356

URL: http://gcc.gnu.org/viewcvs?rev=209356&root=gcc&view=rev
Log:
2014-04-14  Richard Biener  <rguenther@suse.de>

    PR middle-end/55022
    * fold-const.c (negate_expr_p): Don't negate directional rounding
    division.
    (fold_negate_expr): Likewise.

    * gcc.dg/graphite/pr55022.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/graphite/pr55022.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fold-const.c
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #26 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Tue Apr 22 13:33:37 2014
New Revision: 209633

URL: http://gcc.gnu.org/viewcvs?rev=209633&root=gcc&view=rev
Log:
2014-04-22  Richard Biener  <rguenther@suse.de>

    Backport from mainline
    2014-04-14  Richard Biener  <rguenther@suse.de>

    PR middle-end/55022
    * fold-const.c (negate_expr_p): Don't negate directional rounding
    division.
    (fold_negate_expr): Likewise.

    * gcc.dg/graphite/pr55022.c: New testcase.

Added:
    branches/gcc-4_9-branch/gcc/testsuite/gcc.dg/graphite/pr55022.c
Modified:
    branches/gcc-4_9-branch/gcc/ChangeLog
    branches/gcc-4_9-branch/gcc/fold-const.c
    branches/gcc-4_9-branch/gcc/testsuite/ChangeLog

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[mircea.namolaru at inria dot fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #27 from Mircea Namolaru <mircea.namolaru at inria dot fr> ---
Hi,

Many thanks. 

I've passed over the meta-bug opened by you for Graphite,
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59859 and seems to me
that many of the problem have been already solved (some by you) or at least 
as in case of some compile-time/memory usage/missed optimization issues
we know how to solve it - some form of delinearization of array accesses 
that for sure was not for GCC 4.9. 

Btw, there is also another Graphite bug in this list
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59586 that was solved but didn't 
make it for GCC 4.9.0.

Mircea


----- Original Message -----
> From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
> To: "mircea namolaru" <mircea.namolaru@inria.fr>
> Sent: Tuesday, April 22, 2014 3:34:09 PM
> Subject: [Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity
> after revision 190619
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022
> 
> --- Comment #26 from Richard Biener <rguenth at gcc dot gnu.org> ---
> Author: rguenth
> Date: Tue Apr 22 13:33:37 2014
> New Revision: 209633
> 
> URL: http://gcc.gnu.org/viewcvs?rev=209633&root=gcc&view=rev
> Log:
> 2014-04-22  Richard Biener  <rguenther@suse.de>
> 
>     Backport from mainline
>     2014-04-14  Richard Biener  <rguenther@suse.de>
> 
>     PR middle-end/55022
>     * fold-const.c (negate_expr_p): Don't negate directional rounding
>     division.
>     (fold_negate_expr): Likewise.
> 
>     * gcc.dg/graphite/pr55022.c: New testcase.
> 
> Added:
>     branches/gcc-4_9-branch/gcc/testsuite/gcc.dg/graphite/pr55022.c
> Modified:
>     branches/gcc-4_9-branch/gcc/ChangeLog
>     branches/gcc-4_9-branch/gcc/fold-const.c
>     branches/gcc-4_9-branch/gcc/testsuite/ChangeLog
> 
> --
> You are receiving this mail because:
> You are on the CC list for the bug.
>

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
      Known to work|                            |4.8.3
         Resolution|---                         |FIXED
      Known to fail|                            |4.8.2

--- Comment #28 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.

[Bug tree-optimization/55022] [4.8 Regression] air.f90 is miscompliled with -m64 -O2 -fgraphite-identity after revision 190619

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55022

--- Comment #29 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Tue May  6 12:01:21 2014
New Revision: 210104

URL: http://gcc.gnu.org/viewcvs?rev=210104&root=gcc&view=rev
Log:
2014-05-06  Richard Biener  <rguenther@suse.de>

    Backport from mainline
    2014-04-14  Richard Biener  <rguenther@suse.de>

    PR middle-end/55022
    * fold-const.c (negate_expr_p): Don't negate directional rounding
    division.
    (fold_negate_expr): Likewise.

    * gcc.dg/graphite/pr55022.c: New testcase.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gcc.dg/graphite/pr55022.c
Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/fold-const.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog