[Bug middle-end/53688] New: [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[Bug middle-end/53688] New: [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

             Bug #: 53688
           Summary: [4.8 Regression] 191.fma3d in SPEC CPU 2000
                    miscompiled
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hjl.tools@gmail.com
                CC: rguenth@gcc.gnu.org


On Linux/x86-64, revision 188429:

http://gcc.gnu.org/ml/gcc-cvs/2012-06/msg00339.html

miscompiles 191.fma3d in SPEC CPU 2000:

gfortran     -O2 -ffast-math -fwhole-program -flto=jobserver
-fuse-linker-plugin -DSPEC_CPU2000_LP64 fma3d.o beam_.o include_file_.o
penta_.o segment_set_.o body_force_.o indx_.o periodic_bc_.o
sliding_interface_.o constrained_node_.o layering_.o plate_pair_.o
sliding_node_.o contact_node_.o location_.o platq_.o spot_weld_.o
contact_surface_.o lsold_.o platt_.o spring_.o coord_.o massprop_.o
pressure_bc_.o spring_bc_.o damper_.o material_.o property_.o
state_variables_.o damper_bc_.o mean_stress_.o shared_common_data.o stress_.o
displacement_bc_.o membq_.o qa_record_.o tabulated_function_.o element_set_.o
membt_.o relink_scratch_.o tetra_.o enumerated_sets_.o motion_.o results_.o
tied_bc_.o force_.o nodal_point_mass_.o rigid_body_.o truss_.o force_bc_.o
node_.o rigid_body_mass_.o value_.o gauge1d_.o node_set_.o rigid_wall_bc_.o
velocity_ic_.o gauge2d_.o nonreflecting_bc_.o section_1d_.o gauge3d_.o
nrbc_data_.o section_2d_.o hexah_.o output_.o segment_.o lsold.o damper.o
spring.o material_00.o material_10.o material_11.o material_17.o material_22.o
material_25.o material_32.o material_33.o material_34a.o material_36.o
material_38.o material_dm.o material_sp.o sort.o pdb.o beam.o membq.o membt.o
penta.o tetra.o hexah.o platq.o truss.o platt.o fma1.o getirv.o relink.o
output.o fma2.o partition.o strain.o slide.o     -o fma3d

  Running 191.fma3d ref base lto default
*** Miscompare of fma3d.out, see
/export/gnu/import/git/gcc-test-spec-lto/spec/2000/x86_64/spec/benchspec/CFP2000/191.fma3d
/run/00000003/fma3d.out.mis

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |areg.melikadamyan at gmail
                   |                            |dot com
   Target Milestone|---                         |4.8.0

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2012-06-17
         AssignedTo|unassigned at gcc dot       |rguenth at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-17 11:41:54 UTC ---
Mine.

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-18 13:50:28 UTC ---
NEW_PAGE seems to be miscompiled - we fail to output JOB_ID_RECORD%VERSION
('SPEC CPU'), the comparison difference is lots of

@@ -146,7 +146,7 @@
   43255/3    1.1579E+03   -8.1233E+02    0.0000E+00    2.5290E+02   
0.0000E+00    0.0000E+00         1   41   1.42E+02 43255
   43255/4   -3.5068E+03   -4.3304E+03    0.0000E+00    2.0702E+04   
0.0000E+00    0.0000E+00         1   41   1.42E+02 43255
   43255/5   -3.5204E+03   -3.2181E+03    0.0000E+00    2.4813E+04   
0.0000E+00    0.0000E+00         1   41   1.42E+02 43255
-1 FMA-3D  SPEC CPU       Impulse Loaded Cylindrical Panel (w/ Subcycling) -
100MB Data Set                              Page No:  10
+1 FMA-3D                 Impulse Loaded Cylindrical Panel (w/ Subcycling) -
100MB Data Set                              Page No:  10

     Time T = 4.57200E-07   Time Step =     360
 0   NP    X-DISPLACE   Y-DISPLACE   Z-DISPLACE   X-VELOCITY   Y-VELOCITY  
Z-VELOCITY  X-ACCELERATE Y-ACCELERATE Z-ACCELERATE     NP

by inspecting with gdb you can see that JOB_ID_RECORD%VERSION is properly
populated but HEADLINE is not properly built.

It is sched2 who re-orders

(insn:TI 102 100 250 8 (parallel [
            (set (reg:DI 2 cx [143])
                (const_int 0 [0]))
            (set (reg/f:DI 5 di [141])
                (plus:DI (reg/f:DI 5 di [141])
                    (reg:DI 2 cx [143])))
            (set (mem/c:BLK (reg/f:DI 5 di [141]) [0 MEM[(void *)&headline +
9B]+0 A8])
                (const_int 0 [0]))
            (use (reg:QI 0 ax [142]))
            (use (reg:DI 2 cx [143]))
        ]) t.f90:43848 896 {*rep_stosqi}
     (expr_list:REG_UNUSED (reg/f:DI 5 di [141])
        (nil)))

and

(insn:TI 106 96 107 8 (set (mem/c:DI (const:DI (plus:DI (symbol_ref:DI
("headline.65670") [flags 0x2]  <var_decl 0x7fc398ffbdc0 headline>)
                    (const_int 10 [0xa]))) [0 MEM[(c_char * {ref-all})&headline
+ 10B]+0 S8 A16])
        (reg:DI 1 dx [orig:146 MEM[(c_char * {ref-all})&job_id_record + 12B] ]
[146])) t.f90:43850 62 {*movdi_internal_rex64}
     (expr_list:REG_DEAD (reg:DI 1 dx [orig:146 MEM[(c_char *
{ref-all})&job_id_record + 12B] ] [146])
        (nil)))

quite possibly because the MEM_ATTRs in the setmem pattern do not have
a MEM_EXPR that has a suitable size.

Testcase:

char headline[256];
struct hdr {
  char part1[9];
  char part2[8];
} p;

void __attribute__((noinline,noclone))
init()
{
  __builtin_memcpy (p.part1, "FOOBARFOO", sizeof (p.part1));
  __builtin_memcpy (p.part2, "SPEC CPU", sizeof (p.part2));
}

int main()
{
  char *x;
  int c;
  init();
  __builtin_memcpy (&headline[0], p.part1, 9);
  c = 9;
  x = &headline[0];
  x = x + c;
  __builtin_memset (x, ' ', 245);
  __builtin_memcpy (&headline[10], p.part2, 8);
  c = 18;
  x = &headline[0];
  x = x + c;
  __builtin_memset (x, ' ', 238);
  if (headline[10] != 'S')
    __builtin_abort ();
  return 0;
}

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |georggcc at googlemail dot
                   |                            |com

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-06-19 08:44:30 UTC ---
*** Bug 53686 has been marked as a duplicate of this bug. ***

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[matz at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

--- Comment #4 from Michael Matz <matz at gcc dot gnu.org> 2012-06-21 12:18:31 UTC ---
Author: matz
Date: Thu Jun 21 12:18:23 2012
New Revision: 188852

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188852
Log:
    PR middle-end/53688
    * builtins.c (get_memory_rtx): Always build an all-aliasing MEM_REF
    with correct size.

testsuite/
    * gcc.c-torture/execute/pr53688.c: New test.

Added:
    trunk/gcc/testsuite/gcc.c-torture/execute/pr53688.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/builtins.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/53688] [4.8 Regression] 191.fma3d in SPEC CPU 2000 miscompiled

[matz at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53688

Michael Matz <matz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
                 CC|                            |matz at gcc dot gnu.org
         Resolution|                            |FIXED

--- Comment #5 from Michael Matz <matz at gcc dot gnu.org> 2012-06-21 12:52:18 UTC ---
Fixed.