[Bug middle-end/42181] New: -fgraphite-identity miscompiles or ICEs on air.f90

[Bug middle-end/42181] New: -fgraphite-identity miscompiles or ICEs on air.f90

[howarth at nitro dot med dot uc dot edu]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 3738 bytes --]

The Polyhedron 2005 benchmark air.f90 is miscompiled at -O2 -fgraphite-identity
such that it produces incorrect results...

         AIRFLOW IN A BOX
            Version 2.0  
   (c) Hanley Innovations  1995 
  1.00000000000000002E-002  0.29999999999999999        10.000000000000000      
0.10000000000000001                0
   272.76999999999998        272.76999999999998        272.76999999999998      
 272.76999999999998        0.0000000000000000     
          11          12
  X-DATA
           1   0.0000000000000000       0.10000000000000001                4
           2  5.00000000000000028E-002  0.22000000000000000                4
           3  0.10000000000000001       0.29999999999999999                4
           4  0.20000000000000001       0.40000000000000002                4
           5  0.29999999999999999       0.50000000000000000                4
           6  0.40000000000000002       0.59999999999999998                4
           7  0.50000000000000000       0.69999999999999996                4
           8  0.59999999999999998       0.80000000000000004                4
           9  0.69999999999999996       0.90000000000000002                4
          10  0.80000000000000004       0.96999999999999997                4
          11  0.90000000000000002        1.0000000000000000                4
  Y-DATA
   0.0000000000000000       0.10000000000000001                4
  4.00000000000000008E-002  0.14999999999999999                4
  8.00000000000000017E-002  0.20000000000000001                4
  0.14999999999999999       0.29999999999999999                4
  0.20000000000000001       0.40000000000000002                4
  0.29999999999999999       0.50000000000000000                4
  0.40000000000000002       0.59999999999999998                4
  0.50000000000000000       0.69999999999999996                4
  0.59999999999999998       0.80000000000000004                4
  0.69999999999999996       0.89000000000000001                4
  0.80000000000000004       0.94999999999999996                4
  0.94999999999999996        1.0000000000000000                4
  ITERATION#  TIME             FINAL          MASS RESIDUAL
        3  100.000005723761   0.01000000         NaN
 deltat, final t, iterations
   100.00000000000000        100.00000572376112                3
           3                       NaN

When compiling air.f90  with -O1 -fgraphite-identity, the gfortran compiler
ICEs...

air.f90: In function ‘state_’:
air.f90:1034:0: internal compiler error: in scan_tree_for_params, at
graphite-sese-to-poly.c:853

Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/sw/lib/gcc4.5/libexec/gcc/x86_64-apple-darwin10.2.0/4.5.0/lto-wrapper
Target: x86_64-apple-darwin10.2.0
Configured with: ../gcc-4.5-20091125/configure --prefix=/sw
--prefix=/sw/lib/gcc4.5 --mandir=/sw/share/man --infodir=/sw/share/info
--enable-languages=c,c++,fortran,objc,obj-c++,java --with-gmp=/sw
--with-libiconv-prefix=/sw --with-ppl=/sw --with-cloog=/sw --with-mpc=/sw
--with-system-zlib --x-includes=/usr/X11R6/include --x-libraries=/usr/X11R6/lib
--disable-libjava-multilib
Thread model: posix
gcc version 4.5.0 20091125 (experimental) (GCC)


-- 
           Summary: -fgraphite-identity miscompiles or ICEs on air.f90
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: howarth at nitro dot med dot uc dot edu
 GCC build triplet: x86_64-apple-darwin10
  GCC host triplet: x86_64-apple-darwin10
GCC target triplet: x86_64-apple-darwin10


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] -fgraphite-identity miscompiles or ICEs on air.f90

[jv244 at cam dot ac dot uk]

------- Comment #1 from jv244 at cam dot ac dot uk  2009-11-26 17:02 -------
*** Bug 41051 has been marked as a duplicate of this bug. ***


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] -fgraphite-identity miscompiles or ICEs on air.f90

[jv244 at cam dot ac dot uk]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1069 bytes --]



------- Comment #2 from jv244 at cam dot ac dot uk  2009-11-26 17:04 -------
the following is a testcase for the ICE (derived from CP2K bug crashes in the
same place):

> cat bug.f90
MODULE powell
  INTEGER, PARAMETER :: dp=8
CONTAINS
  SUBROUTINE newuob (n,npt,x,rhobeg,rhoend,maxfun,xbase,&
       xopt,xnew,xpt,fval,gq,hq,pq,bmat,zmat,ndim,d,vlag,w,opt)
    REAL(dp), DIMENSION(npt, *), &
      INTENT(inout)                          :: xpt
    REAL(dp), DIMENSION(*), INTENT(inout)    :: fval, gq, hq, pq
120 IF (dsq <= 1.0e-3_dp*xoptsq) THEN
       DO k=1,npt
          DO i=1,n
             gq(i)=gq(i)+temp*xpt(k,i)
          END DO
       END DO
    END IF
  END SUBROUTINE newuob
END MODULE powell

 gfortran -fgraphite -O1 bug.f90
bug.f90: In function ‘__powell_MOD_newuob’:
bug.f90:4:0: internal compiler error: in scan_tree_for_params, at
graphite-sese-to-poly.c:853
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[jv244 at cam dot ac dot uk]

------- Comment #3 from jv244 at cam dot ac dot uk  2009-11-26 17:05 -------
the ice is absent in 4.4 but present in 4.5 trunk


-- 

jv244 at cam dot ac dot uk changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-11-26 17:05:43
               date|                            |
            Summary|-fgraphite-identity         |[4.5 Regression] -fgraphite-
                   |miscompiles or ICEs on      |identity miscompiles or ICEs
                   |air.f90                     |on air.f90
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code, wrong-
                   |                            |code
           Priority|P3                          |P1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[spop at gcc dot gnu dot org]

------- Comment #5 from spop at gcc dot gnu dot org  2009-12-14 19:03 -------
Fixed in the Graphite branch.  I will commit it to trunk after test.


-- 

spop at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[spop at gcc dot gnu dot org]

------- Comment #4 from spop at gcc dot gnu dot org  2009-12-14 19:02 -------
Subject: Bug 42181

Author: spop
Date: Mon Dec 14 19:02:37 2009
New Revision: 155233

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155233
Log:
Fix PR42181.

2009-12-14  Sebastian Pop  <sebastian.pop@amd.com>

        PR middle-end/42181
        * graphite-scop-detection.c (graphite_can_represent_scev): Handle more
        carefully PLUS_EXPR, MINUS_EXPR, and MULT_EXPR.

        * testsuite/gfortran.dg/graphite/pr42181.f90: New.

Added:
    branches/graphite/gcc/testsuite/gfortran.dg/graphite/pr42181.f90
Modified:
    branches/graphite/gcc/ChangeLog.graphite
    branches/graphite/gcc/graphite-scop-detection.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[spop at gcc dot gnu dot org]

------- Comment #6 from spop at gcc dot gnu dot org  2009-12-23 07:50 -------
Subject: Bug 42181

Author: spop
Date: Wed Dec 23 07:50:00 2009
New Revision: 155417

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155417
Log:
Fix PR42181.

2009-12-14  Sebastian Pop  <sebastian.pop@amd.com>

        PR middle-end/42181
        * graphite-scop-detection.c (graphite_can_represent_scev): Handle more
        carefully PLUS_EXPR, MINUS_EXPR, and MULT_EXPR.

        * testsuite/gfortran.dg/graphite/pr42181.f90: New.

Added:
    trunk/gcc/testsuite/gfortran.dg/graphite/pr42181.f90
Modified:
    trunk/gcc/ChangeLog.graphite
    trunk/gcc/graphite-scop-detection.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #7 from howarth at nitro dot med dot uc dot edu  2009-12-24 00:14 -------
This still appears to be broken at r155434 on x86_64-apple-darwin10 with...

gfortran -O2 -fgraphite-identity air.f90 -o air
./air
         AIRFLOW IN A BOX
            Version 2.0  
   (c) Hanley Innovations  1995 
  1.00000000000000002E-002  0.29999999999999999        10.000000000000000      
0.10000000000000001                0
   272.76999999999998        272.76999999999998        272.76999999999998      
 272.76999999999998        0.0000000000000000     
          11          12
  X-DATA
           1   0.0000000000000000       0.10000000000000001                4
           2  5.00000000000000028E-002  0.22000000000000000                4
           3  0.10000000000000001       0.29999999999999999                4
           4  0.20000000000000001       0.40000000000000002                4
           5  0.29999999999999999       0.50000000000000000                4
           6  0.40000000000000002       0.59999999999999998                4
           7  0.50000000000000000       0.69999999999999996                4
           8  0.59999999999999998       0.80000000000000004                4
           9  0.69999999999999996       0.90000000000000002                4
          10  0.80000000000000004       0.96999999999999997                4
          11  0.90000000000000002        1.0000000000000000                4
  Y-DATA
   0.0000000000000000       0.10000000000000001                4
  4.00000000000000008E-002  0.14999999999999999                4
  8.00000000000000017E-002  0.20000000000000001                4
  0.14999999999999999       0.29999999999999999                4
  0.20000000000000001       0.40000000000000002                4
  0.29999999999999999       0.50000000000000000                4
  0.40000000000000002       0.59999999999999998                4
  0.50000000000000000       0.69999999999999996                4
  0.59999999999999998       0.80000000000000004                4
  0.69999999999999996       0.89000000000000001                4
  0.80000000000000004       0.94999999999999996                4
  0.94999999999999996        1.0000000000000000                4
  ITERATION#  TIME             FINAL          MASS RESIDUAL
       14  100.000043187820   0.01000000         NaN
 deltat, final t, iterations
   100.00000000000000        100.00004318782042               14
          14                       NaN


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #8 from howarth at nitro dot med dot uc dot edu  2009-12-24 00:15 -------
Still broke on x86_64-apple-darwin10.


-- 

howarth at nitro dot med dot uc dot edu changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #9 from howarth at nitro dot med dot uc dot edu  2009-12-24 00:18 -------
Created an attachment (id=19380)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19380&action=view)
assembly from 'gfortran -O2 -fgraphite-identity air.f90 --save-temps -o air' on
x86_64-apple-darwin10 at r155434


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #10 from howarth at nitro dot med dot uc dot edu  2009-12-24 00:20 -------
Problem doesn't exist for -O1 -fgraphite-identity on x86_64-apple-darwin10.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[dominiq at lps dot ens dot fr]

------- Comment #11 from dominiq at lps dot ens dot fr  2009-12-24 08:12 -------
*** Bug 42480 has been marked as a duplicate of this bug. ***


-- 

dominiq at lps dot ens dot fr changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dominiq at lps dot ens dot
                   |                            |fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[trifunovic at gcc dot gnu dot org]

-- 

trifunovic at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |trifunovic at gcc dot gnu
                   |dot org                     |dot org
             Status|REOPENED                    |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression] -fgraphite-identity miscompiles or ICEs on air.f90

[spop at gcc dot gnu dot org]

------- Comment #12 from spop at gcc dot gnu dot org  2010-01-14 19:24 -------
The kernel from Comment #2 looks identical to the kernel from PR42637 that is
also miscompiled.
These two PRs may be the same.
Konrad, could you look at the testcase from PR42637 that is more precise?

Thanks, 
Sebastian


-- 

spop at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |42637


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles or ICEs on air.f90

[spop at gcc dot gnu dot org]

------- Comment #13 from spop at gcc dot gnu dot org  2010-02-06 17:44 -------
Still fails after PR42637 has been fixed.
NB: that this miscompile is for -fgraphite-identity not with the loop opts.


-- 

spop at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|42637                       |
         AssignedTo|trifunovic at gcc dot gnu   |spop at gcc dot gnu dot org
                   |dot org                     |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #14 from howarth at nitro dot med dot uc dot edu  2010-03-08 23:09 -------
Is this issue to be fixed for gcc 4.5?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #15 from spop at gcc dot gnu dot org  2010-03-08 23:16 -------
Yes.
I think it is important to understand what is miscompiled with the graphite
identity.
I will try to reduce this.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #16 from howarth at nitro dot med dot uc dot edu  2010-03-09 02:26 -------
The miscompiled code in air.f90 is the subroutine SPECTOP. If I pull that
subroutine out into a separate file and compile it at -O3 without
-fgraphite-identity, the remainder of the code can be compiled with
-fgraphite-identity at -O3 and the resulting binary runs fine on
x86_64-apple-darwin10.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #17 from howarth at nitro dot med dot uc dot edu  2010-03-09 02:34 -------
Created an attachment (id=20049)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20049&action=view)
assembly for spectop subroutine compiled at -O2 on x86_64-apple-darwin10


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #18 from howarth at nitro dot med dot uc dot edu  2010-03-09 02:35 -------
Created an attachment (id=20050)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20050&action=view)
assembly for spectop subroutine compiled at -graphite-identity -O2 on
x86_64-apple-darwin10


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #19 from howarth at nitro dot med dot uc dot edu  2010-03-09 02:37 -------
Created an attachment (id=20051)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20051&action=view)
diff between assembly for spectop subroutine at -O2 without and with
-fgraphite-identity


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #20 from howarth at nitro dot med dot uc dot edu  2010-03-09 03:45 -------
The offending optimization for the spectop subroutine at -O2
-fgraphite-identity appears to be -fstrict-overflow. I can compile...

gfortran -fgraphite-identity -O3 -fno-strict-overflow -c spectop.f90
gfortran -O3 -fgraphite-identity air_sans_spectop.f90 spectop.o

and the resulting air benchmark runs fine.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1425 bytes --]



------- Comment #21 from howarth at nitro dot med dot uc dot edu  2010-03-09 03:48 -------
Interestingly, I get...

gfortran -fgraphite-identity -O3 -Wstrict-overflow=5 -c spectop.f90
spectop.f90: In function ‘spectop’:
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when reducing
constant in comparison
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2
spectop.f90:5:0: warning: assuming signed overflow does not occur when changing
X +- C1 cmp C2 to X cmp C1 +- C2

whereas without -fgraphite-identity, I get...

gfortran -O3 -Wstrict-overflow=5 -c spectop.f90
spectop.f90: In function ‘spectop’:
spectop.f90:23:0: warning: assuming signed overflow does not occur when
changing X +- C1 cmp C2 to X cmp C1 +- C2


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[dominiq at lps dot ens dot fr]

------- Comment #22 from dominiq at lps dot ens dot fr  2010-03-09 13:32 -------
The following variant of spectop.f90 is miscompiled with '-fgraphite-identity
-O3':

      SUBROUTINE SPECTOP(Dr,N)
      IMPLICIT REAL*8(A-H,o-Z)
      DIMENSION d1(0:32,0:32) , Dr(0:32,0:32) , x(0:32)
      REAL*8 Dr
!
! PROGRAM TO COMPUTE THE CHEBYSHEV SPECTRAL OPERATOR
!
      ang = DBLE(1)
      s = DBLE(6)
      o = DBLE(1)
      t = DBLE(2)
      pi = t*DASIN(ang)
      DO i = 0 , N
         x(i) = DCOS(pi*DBLE(i)/DBLE(N))
      ENDDO
!
! IF J=K
!
      DO j = 1 , N - 1
         d1(j,j) = -x(j)/(t*(o-x(j)**2))
      ENDDO
      d1(0,0) = (t*DBLE(N)**2+o)/s
      d1(N,N) = -d1(0,0)
!
! IF J.NE.K
!
      fctr = -t
      DO j = 1 , N-1
         d1(0,j) = fctr/(x(0)-x(j))
         fctr = -fctr
      ENDDO
      d1(0,N) = fctr/(t*(x(0)-x(N)))
      fctr1 = -1.0D0
      DO k = 1 , N-1
         d1(k,0) = fctr1/(t*(x(k)-x(0)))
         fctr = -fctr1
         DO j = 1 , N-1
            if (j.ne.k) d1(k,j) = fctr/(x(k)-x(j))
            fctr = -fctr
         ENDDO
         d1(k,N) = fctr/(t*(x(k)-x(N)))
         fctr1 = -fctr1
      ENDDO
      d1(N,0) = fctr1/(x(N)-x(0))
      fctr = -t*fctr1
      DO j = 1 , N-1
         d1(N,j) = fctr/(x(N)-x(j))
         fctr = -fctr
      ENDDO
      DO k = 0 , N
         DO j = 0 , N
            Dr(k,j) = d1(N-k,N-j)
         ENDDO
      ENDDO
      CONTINUE
      END

but is correctly compiled if 

         DO j = 1 , N-1
            if (j.ne.k) d1(k,j) = fctr/(x(k)-x(j))
            fctr = -fctr
         ENDDO

is replaced with

         DO j = 1 , k-1
            d1(k,j) = fctr/(x(k)-x(j))
            fctr = -fctr
         ENDDO
         fctr = -fctr
         DO j = k+1 , N-1
            d1(k,j) = fctr/(x(k)-x(j))
            fctr = -fctr
         ENDDO


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #23 from spop at gcc dot gnu dot org  2010-03-09 22:31 -------
Thanks for reducing this testcase.

On the further reduced kernel:

      SUBROUTINE SPECTOP(Dr,N)
      IMPLICIT REAL*8(A-H,o-Z)
      DIMENSION d1(0:32,0:32) , Dr(0:32,0:32) , x(0:32)
      REAL*8 Dr

      DO k = 1 , N-1
         d1(k,0) = fctr1/(t*(x(k)-x(0)))
         fctr = -fctr1
         DO j = 1 , N-1
            if (j.ne.k) d1(k,j) = fctr/(x(k)-x(j))
            fctr = -fctr
         ENDDO
         d1(k,N) = fctr/(t*(x(k)-x(N)))
         fctr1 = -fctr1
      ENDDO

      Dr(k,j) = d1(N-k,N-j)
      CONTINUE
      END

We have this CLAST generated by CLooG:

if (T_10 >= 3) {
  S4(0) ;
  S5(0,0) ;
  S7(0,0) ;
  for (scat_3=1;scat_3<=T_10-2;scat_3++) {
    S5(0,scat_3) ;
    S6(0,scat_3) ;
    S7(0,scat_3) ;
  }
  S9(0) ;
  S14(0) ;
}
for (scat_1=1;scat_1<=T_10-3;scat_1++) {
  S4(scat_1) ;
  S5(scat_1,scat_1) ;
  S7(scat_1,scat_1) ;
  for (scat_3=scat_1+1;scat_3<=T_10-2;scat_3++) {
    S5(scat_1,scat_3) ;
    S6(scat_1,scat_3) ;
    S7(scat_1,scat_3) ;
  }
  for (scat_3=0;scat_3<=scat_1-1;scat_3++) {
    S5(scat_1,scat_3) ;
    S6(scat_1,scat_3) ;
    S7(scat_1,scat_3) ;
  }
  S9(scat_1) ;
  S14(scat_1) ;
}
if (T_10 >= 3) {
  S4(T_10-2) ;
  S5(T_10-2,T_10-2) ;
  S7(T_10-2,T_10-2) ;
  for (scat_3=0;scat_3<=T_10-3;scat_3++) {
    S5(T_10-2,scat_3) ;
    S6(T_10-2,scat_3) ;
    S7(T_10-2,scat_3) ;
  }
  S9(T_10-2) ;
  S14(T_10-2) ;
}
if (T_10 == 2) {
  S4(0) ;
  S5(0,0) ;
  S7(0,0) ;
  S9(0) ;
  S14(0) ;
}

Where T_10 stands for the parameter N,
S4 stands for
         d1(k,0) = fctr1/(t*(x(k)-x(0)))
         fctr = -fctr1
S5 stands for "if (j.ne.k)"
S6 stands for "d1(k,j) = fctr/(x(k)-x(j))"
S7 stands for "fctr = -fctr".
S14 stands for
         d1(k,N) = fctr/(t*(x(k)-x(N)))
         fctr1 = -fctr1

So the error seems to be that we are splitting the index scat_3 into
two ranges that are not ordered in the same way as in the original
loop nest:

for (scat_1=1;scat_1<=T_10-3;scat_1++) {
  S4(scat_1) ;
  S5(scat_1,scat_1) ;
  S7(scat_1,scat_1) ;
  for (scat_3=scat_1+1;scat_3<=T_10-2;scat_3++) {
    S5(scat_1,scat_3) ;
    S6(scat_1,scat_3) ;
    S7(scat_1,scat_3) ;
  }
  for (scat_3=0;scat_3<=scat_1-1;scat_3++) {
    S5(scat_1,scat_3) ;
    S6(scat_1,scat_3) ;
    S7(scat_1,scat_3) ;
  }
  S9(scat_1) ;
  S14(scat_1) ;
}

We are executing the range [scat_1+1, T_10-2] before executing the
range [0, scat_1-1].


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[dominiq at lps dot ens dot fr]

------- Comment #24 from dominiq at lps dot ens dot fr  2010-03-14 11:40 -------
> We are executing the range [scat_1+1, T_10-2] before executing the
> range [0, scat_1-1].

This would be OK if fctr had been initialized with the right value. Failing to
do so leads to an early NAN.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[dominiq at lps dot ens dot fr]

------- Comment #25 from dominiq at lps dot ens dot fr  2010-03-20 10:45 -------
> We are executing the range [scat_1+1, T_10-2] before executing the
> range [0, scat_1-1].

Do you understand why graphite does not detect that the loop for [scat_1+1,
T_10-2] depends on the one for  [0, scat_1-1]?

Second question why does graphite exchange the order of the split loops?

Note also that in the original loop handling the loop for [0, scat_1-1] is
tricky when scat_1==1 (fctr = fctr1*fctr2 and fctr2 = -o*fctr2 must be handled
correctly).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[sebpop at gmail dot com]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 705 bytes --]



------- Comment #26 from sebpop at gmail dot com  2010-03-21 16:28 -------
Subject: Re:  [4.5 Regression][graphite] 
        -fgraphite-identity miscompiles air.f90

On Sat, Mar 20, 2010 at 05:45, dominiq at lps dot ens dot fr wrote:
> Do you understand why graphite does not detect that the loop for [scat_1+1,
> T_10-2] depends on the one for  [0, scat_1-1]?

Graphite does know this, but it does not ask CLooG to generate
[0, scat_1-1] after [scat_1+1, T_10-2], however CLooG does
generate it, so I am thinking that this is a problem in CLooG.

> Second question why does graphite exchange the order of the split loops?
>

CLooG does that.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

Re: [Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[Sebastian Pop]

On Sat, Mar 20, 2010 at 05:45, dominiq at lps dot ens dot fr wrote:
> Do you understand why graphite does not detect that the loop for [scat_1+1,
> T_10-2] depends on the one for  [0, scat_1-1]?

Graphite does know this, but it does not ask CLooG to generate
[0, scat_1-1] after [scat_1+1, T_10-2], however CLooG does
generate it, so I am thinking that this is a problem in CLooG.

> Second question why does graphite exchange the order of the split loops?
>

CLooG does that.

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #27 from spop at gcc dot gnu dot org  2010-03-23 20:08 -------
I just verified that the problem is well in CLooG-PPL.
With CLooG-ISL we generate a correct code that looks like this:

  for (c2=1;c2<=D.1554_10-3;c2++) {
    S1(c2);
    for (c4=0;c4<=c2-1;c4++) {
      S2(c2,c4);
      S3(c2,c4);
      S4(c2,c4);
    }
    S2(c2,c2);
    S4(c2,c2);
    for (c4=c2+1;c4<=D.1554_10-2;c4++) {
      S2(c2,c4);
      S3(c2,c4);
      S4(c2,c4);
    }
    S5(c2);
    S6(c2);
  }


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #28 from spop at gcc dot gnu dot org  2010-03-23 21:24 -------
I fixed this problem in CLooG-PPL: the code generated with 
the new version looks like the code generated by CLooG-ISL:

  for (c2=1;c2<=D.1554_10-3;c2++) {
    S1(c2);
    for (c4=0;c4<=c2-1;c4++) {
      S2(c2,c4);
      S3(c2,c4);
      S4(c2,c4);
    }
    S2(c2,c2);
    S4(c2,c2);
    for (c4=c2+1;c4<=D.1554_10-2;c4++) {
      S2(c2,c4);
      S3(c2,c4);
      S4(c2,c4);
    }
    S5(c2);
    S6(c2);
  }

I will bootstrap and test the graphite branch with the modified
version of CLooG-PPL and then if everything looks good,
I will prepare a new version CLooG-PPL-0.15.9 and I will 
upload it on ftp://gcc.gnu.org/pub/gcc/infrastructure/

Sebastian


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #29 from spop at gcc dot gnu dot org  2010-03-23 21:27 -------
Also note that with the fix in CLooG-PPL,
gfortran -O2 -fgraphite-identity air.f90 -o air
./air
passes without error.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[spop at gcc dot gnu dot org]

------- Comment #30 from spop at gcc dot gnu dot org  2010-03-24 20:17 -------
Fixed.

Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz


-- 

spop at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[howarth at nitro dot med dot uc dot edu]

------- Comment #31 from howarth at nitro dot med dot uc dot edu  2010-03-24 21:35 -------
(In reply to comment #30)
> Fixed.
> 
> Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz
> 

Shouldn't the required cloog-ppl version in configure be bumped from 0.15.5 to
0.15.9?

  #if CLOOG_VERSION_MAJOR != 0 || CLOOG_VERSION_MINOR != 15 ||
CLOOG_VERSION_REVISION < 5


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

Re: [Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[Sebastian Pop]

On Wed, Mar 24, 2010 at 16:35, howarth at nitro dot med dot uc dot edu
<gcc-bugzilla@gcc.gnu.org> wrote:
>> Fixed.
>>
>> Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz
>>
>
> Shouldn't the required cloog-ppl version in configure be bumped from 0.15.5 to
> 0.15.9?
>

Richi what do you think?

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[sebpop at gmail dot com]

------- Comment #32 from sebpop at gmail dot com  2010-03-25 17:43 -------
Subject: Re:  [4.5 Regression][graphite] 
        -fgraphite-identity miscompiles air.f90

On Wed, Mar 24, 2010 at 16:35, howarth at nitro dot med dot uc dot edu
<gcc-bugzilla@gcc.gnu.org> wrote:
>> Fixed.
>>
>> Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz
>>
>
> Shouldn't the required cloog-ppl version in configure be bumped from 0.15.5 to
> 0.15.9?
>

Richi what do you think?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

[Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[rguenther at suse dot de]

------- Comment #33 from rguenther at suse dot de  2010-03-26 09:58 -------
Subject: Re:  [4.5 Regression][graphite]        -fgraphite-identity
 miscompiles air.f90

On Thu, 25 Mar 2010, Sebastian Pop wrote:

> On Wed, Mar 24, 2010 at 16:35, howarth at nitro dot med dot uc dot edu
> <gcc-bugzilla@gcc.gnu.org> wrote:
> >> Fixed.
> >>
> >> Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz
> >>
> >
> > Shouldn't the required cloog-ppl version in configure be bumped from 0.15.5 to
> > 0.15.9?
> >
> 
> Richi what do you think?

It's a bit late for that change.  You could warn at configure time if
a "broken but acceptable" version is detected, like we do for
gmp or mpfr (I don't remember which one).

Thanks,
Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42181

Re: [Bug middle-end/42181] [4.5 Regression][graphite] -fgraphite-identity miscompiles air.f90

[Richard Guenther]

On Thu, 25 Mar 2010, Sebastian Pop wrote:

> On Wed, Mar 24, 2010 at 16:35, howarth at nitro dot med dot uc dot edu
> <gcc-bugzilla@gcc.gnu.org> wrote:
> >> Fixed.
> >>
> >> Please use ftp://gcc.gnu.org/pub/gcc/infrastructure/cloog-ppl-0.15.9.tar.gz
> >>
> >
> > Shouldn't the required cloog-ppl version in configure be bumped from 0.15.5 to
> > 0.15.9?
> >
> 
> Richi what do you think?

It's a bit late for that change.  You could warn at configure time if
a "broken but acceptable" version is detected, like we do for
gmp or mpfr (I don't remember which one).

Thanks,
Richard.