[Bug middle-end/37730] New: [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[Bug middle-end/37730] New: [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[pinskia at gcc dot gnu dot org]

Since on x86-darwin, SSE2 is enabled by default, I see a failure with this
testcase:
output
is:/Users/apinski/src/local/gcc/gcc/testsuite/gcc.c-torture/compile/pr37713.c:
In function 'add_opush':^M
/Users/apinski/src/local/gcc/gcc/testsuite/gcc.c-torture/compile/pr37713.c:8:
internal compiler error: in set_mem_alias_set, at emit-rtl.c:1784^M Please
submit a full bug report,^M
with preprocessed source if appropriate.^M
See <http://gcc.gnu.org/bugs.html> for instructions.^M

--- CUT ---
void add_opush(void)
{
    unsigned char formats[] = { 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0xff };
    void *dtds[sizeof(formats)];
    unsigned int i;
    unsigned char dtd = 0x08;
    for (i = 0; i < sizeof(formats); i++)
        dtds[i] = &dtd;
    sdp_seq_alloc(dtds);
}
--- CUT ---


-- 
           Summary: [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at
                    -O3 -msse2
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
GCC target triplet: i686-*-*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2008-10-03 20:29 -------
Confirmed.  At first sight this looks like an expansion problem of the
initializer.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-10-03 20:29:59
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[jakub at gcc dot gnu dot org]

------- Comment #2 from jakub at gcc dot gnu dot org  2008-10-06 06:18 -------
This looks like a vectorizer bug to me.  Vectorizer creates:
  vector void * * vect_pdtds.39;
  vector void * * vect_pdtds.34;
  vector unsigned char * vect_cst_.33;
...
  vect_cst_.33_33 = {&dtd, &dtd, &dtd, &dtd};
  vect_pdtds.39_34 = (vector void * *) &dtds;
  vect_pdtds.34_35 = vect_pdtds.39_34;
...
  # vect_pdtds.34_36 = PHI <vect_pdtds.34_37(9), vect_pdtds.34_35(3)>
  # ivtmp.40_38 = PHI <ivtmp.40_39(9), 0(3)>
  *vect_pdtds.34_36 = vect_cst_.33_33;

Shouldn't that use VCE instead?  void * and unsigned char * certainly have
different alias sets, so storing vect_cst.33 with element type's alias set 3
into dtds variable, which has element alias set 2, doesn't work very well.


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |irar at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2008-10-07 21:05 -------
Also ICEs the same way on spu-elf.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|i686-*-*                    |i686-*-*, spu-elf


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2008-10-08 08:01 -------
Well, the RHS scalar alias set is IMHO irrelevant.  What is relevant is the
alias set of the store.  Which is why I think the assert is somewhat bogus.
Note that the gimple type-system allows assignment of any pointer to a void *
pointer without a conversion, which is what happens here.

It looks like on RTL we treat the RHS as memory load (because of the
constructor)
and for this reason verify that the alias sets are compatible.  But I think
there is no reason to assert that?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[mmitchel at gcc dot gnu dot org]

-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[jakub at gcc dot gnu dot org]

------- Comment #5 from jakub at gcc dot gnu dot org  2008-10-30 12:49 -------
Subject: Bug 37730

Author: jakub
Date: Thu Oct 30 12:49:31 2008
New Revision: 141461

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=141461
Log:
        PR middle-end/37730
        * expr.c (store_constructor): For vectors, if target is a MEM, use
        target's MEM_ALIAS_SET instead of elttype alias set.

        * gcc.dg/vect/pr37730.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/vect/pr37730.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/expr.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730

[Bug middle-end/37730] [4.4 Regression] gcc.c-torture/compile/pr37713.c ICEs at -O3 -msse2

[jakub at gcc dot gnu dot org]

------- Comment #6 from jakub at gcc dot gnu dot org  2008-10-30 12:51 -------
Fixed.


-- 

jakub at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37730