This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug c/78380] New: GCC crash with internal compiler error: in gen_reg_rtx, at emit-rtl.c:1025


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78380

            Bug ID: 78380
           Summary: GCC crash with internal compiler error: in
                    gen_reg_rtx, at emit-rtl.c:1025
           Product: gcc
           Version: 6.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: elmar at yasara dot org
  Target Milestone: ---

Created attachment 40057
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40057&action=edit
Preprocessed source file

Dear GCC team,

I encountered this crash when compiling for MacOSX, either directly in MacOSX
(precompiled GCC 6.2.0 downloaded from hpc.sourceforge.net) or when
cross-compiling from Linux (GCC 6.2.0 compiled using
https://github.com/tpoechtrager/osxcross scripts)

The bug database found three possibly related bugs, but in emit-rtl.c line
1027, not line 1025 like mine:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70420
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69401
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68538

The preprocessed source is attached.

This is the complete command line:

/usr/local/bin/gcc -v -O0 -ffast-math -m32 -mfpmath=sse -ffast-math
-march=core-avx2 -mavx2 -Wall -Werror -x c -c crash.c -o crash.o

Using built-in specs.
COLLECT_GCC=/usr/local/bin/gcc
Target: x86_64-apple-darwin15.6.0
Configured with: ../gcc-6.2.0/configure --enable-languages=c++,fortran
--with-gmp=/usr/local
Thread model: posix
gcc version 6.2.0 (GCC) 
COLLECT_GCC_OPTIONS='-O0' '-m32' '-mfpmath=sse' '-ffast-math'
'-march=core-avx2' '-mavx2' '-Wall' '-Werror' '-c' '-o' 'crash.o' '-v'
'-mmacosx-version-min=10.11.3'
 /usr/local/libexec/gcc/x86_64-apple-darwin15.6.0/6.2.0/cc1 -quiet -v
-imultilib i386 -D__DYNAMIC__ crash.c -fPIC -quiet -dumpbase crash.c -m32
-mfpmath=sse -march=core-avx2 -mavx2 -mmacosx-version-min=10.11.3
-auxbase-strip crash.o -O0 -Wall -Werror -version -ffast-math -o
/var/folders/vx/qnr8bkmj1_l5cq70_9h578p80000gn/T//ccU4jeMd.s
GNU C11 (GCC) version 6.2.0 (x86_64-apple-darwin15.6.0)
        compiled by GNU C version 6.2.0, GMP version 6.0.0, MPFR version
3.1.2-p10, MPC version 1.0.2, isl version none
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
ignoring nonexistent directory
"/usr/local/lib/gcc/x86_64-apple-darwin15.6.0/6.2.0/../../../../x86_64-apple-darwin15.6.0/include"
ignoring nonexistent directory "/usr/include"
#include "..." search starts here:
#include <...> search starts here:
 /usr/local/lib/gcc/x86_64-apple-darwin15.6.0/6.2.0/include
 /usr/local/include
 /usr/local/lib/gcc/x86_64-apple-darwin15.6.0/6.2.0/include-fixed
 /System/Library/Frameworks
 /Library/Frameworks
End of search list.
GNU C11 (GCC) version 6.2.0 (x86_64-apple-darwin15.6.0)
        compiled by GNU C version 6.2.0, GMP version 6.0.0, MPFR version
3.1.2-p10, MPC version 1.0.2, isl version none
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
Compiler executable checksum: 226fa07ddbd61344d84e06f1b58d4ece
/home/twinset/yasara/inout.c: In function ‘cl_measurebandwidth’:
/home/twinset/yasara/inout.c:2340:2: internal compiler error: in gen_reg_rtx,
at emit-rtl.c:1025

/home/twinset/yasara/inout.c:2340:2: internal compiler error: Abort trap: 6
gcc: internal compiler error: Abort trap: 6 (program cc1)
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

------------------------------------------

While fiddling with crash.c, I found that removing the 'exectime' from line
149493, i.e.

           mbps=testsize/(1024.*1024.*1e-6);

yields a different error:
x86_64-apple-darwin9-as -arch i386 -force_cpusubtype_ALL -o crash.o
/tmp/ccQKjbFH.s

/tmp/ccQKjbFH.s:954:no such instruction: `shlx %eax, %edx,%eax'
/tmp/ccQKjbFH.s:1040:no such instruction: `vxorpd %xmm0, %xmm0,%xmm0'
/tmp/ccQKjbFH.s:1041:no such instruction: `vcvtsi2sd -36(%ebp), %xmm0,%xmm0'
/tmp/ccQKjbFH.s:1043:no such instruction: `vmovsd (%eax), %xmm1'
/tmp/ccQKjbFH.s:1044:no such instruction: `vdivsd %xmm1, %xmm0,%xmm0'
/tmp/ccQKjbFH.s:1045:no such instruction: `vcvttsd2si %xmm0, %eax'

which is strange, since these instructions exist in 32bit mode.

And BTW, I wonder why the preprocessor inserts myriads of useless line number
directives now, this did not happen in older GCC versions:

# 2268 "/home/twinset/yasara/inout.c" 3 4

# 2268 "/home/twinset/yasara/inout.c"

# 2268 "/home/twinset/yasara/inout.c" 3 4

# 2268 "/home/twinset/yasara/inout.c"
       else testbuff=cl_CreateBuffer(context->id,0x40,testsize,                
                                                                      ((void
*)0)                                                                           
               ,                                                               
                            ((void *)0)                                        
                                                       ); }

# 2269 "/home/twinset/yasara/inout.c" 3 4

# 2269 "/home/twinset/yasara/inout.c"

# 2269 "/home/twinset/yasara/inout.c" 3 4

# 2269 "/home/twinset/yasara/inout.c"
     else testbuff=cl_CreateBuffer(context->id,0,testsize,                     
                                   ((void *)0)                                 
                           ,                                                   
          ((void *)0)                                                          
       );

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]