[Bug target/54619] New: GCC aborts during compilation with '-O2 -mips16'

[Bug target/54619] New: GCC aborts during compilation with '-O2 -mips16'

[sje at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54619

             Bug #: 54619
           Summary: GCC aborts during compilation with '-O2 -mips16'
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: sje@gcc.gnu.org
            Target: mips*-*-*


I am getting a Floating point exception from the compiler while compiling the
following small test program with -O2 -mips16.  I can reproduce it while using
a native or cross compiler targeting mips and while generating code for big or
little endian.  I could also reproduce it on targets for linux and for elf.

% cat q.c
swapfunc(char *a , char *b , int n , int swaptype)
{
    long i = (n) / sizeof (char);
    char *pi = (char *) (a);
    char *pj = (char *) (b);
    do {
      *pi++ = *pj;
    } while (--i > 0);
}
qsort(void *a)
{
 char *pb;
 int r;
 int swaptype;
 swapfunc(a, pb - r, r, swaptype);
}


% obj*/gcc/initial/gcc/cc1 -quiet -mel -O2 -mips16 q.c
q.c: In function ‘qsort’:
q.c:10:1: internal compiler error: Floating point exception
 qsort(void *a)
 ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.


% obj*/gcc/gcc/cc1 -version
GNU C (GCC) version 4.8.0 20120913 (experimental) (mipsel-unknown-linux-gnu)
    compiled by GNU C version 4.8.0 20120913 (experimental), GMP version 4.3.2,
MPFR version 3.0.0-p3, MPC version 0.8.2
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096

[Bug target/54619] GCC aborts during compilation with '-O2 -mips16'

[sje at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54619

--- Comment #1 from Steve Ellcey <sje at gcc dot gnu.org> 2012-09-18 18:34:21 UTC ---
It looks like this is coming from mips16_unextended_reference_p,
if mode is VOIDmode then GET_MODE_SIZE (mode) is zero and we wind
up doing 'offset % 0'.   I think we should be checking for
'GET_MODE_SIZE == 0' instead of mode != 'BLKmode'. before doing
the modulo instruction.

[Bug tree-optimization/54619] GCC aborts during compilation with '-O2 -mips16'

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54619

vries at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-11-14
                 CC|                            |vries at gcc dot gnu.org
          Component|target                      |tree-optimization
     Ever Confirmed|0                           |1

--- Comment #2 from vries at gcc dot gnu.org 2012-11-14 11:20:31 UTC ---
Confirmed.

Discussed here: http://gcc.gnu.org/ml/gcc-patches/2012-09/msg01378.html.

Changing component to tree-optimization since problem is in ivopts passing
VOIDmode to TARGET_ADDRESS_COST.

[Bug tree-optimization/54619] GCC aborts during compilation with '-O2 -mips16'

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54619

--- Comment #3 from vries at gcc dot gnu.org 2012-11-15 16:50:04 UTC ---
Author: vries
Date: Thu Nov 15 16:49:45 2012
New Revision: 193537

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=193537
Log:
2012-11-15  Tom de Vries  <tom@codesourcery.com>

    PR tree-optimization/54619
    * tree-ssa-loop-ivopts.c (get_use_type): New function.
    (get_computation_at): Use get_use_type.
    (get_computation_cost_at): Declare and set mem_mode.  Use mem_mode.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-ssa-loop-ivopts.c

[Bug tree-optimization/54619] GCC aborts during compilation with '-O2 -mips16'

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54619

vries at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
         AssignedTo|unassigned at gcc dot       |vries at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #4 from vries at gcc dot gnu.org 2012-11-15 16:55:13 UTC ---
Patch checked in, resolving as fixed.