Bug 58431 - [4.9 Regression] wrong code at -O3 on x86_64-linux-gnu (in 32-bit mode)
Summary: [4.9 Regression] wrong code at -O3 on x86_64-linux-gnu (in 32-bit mode)
Status: RESOLVED DUPLICATE of bug 58418
Alias: None
Product: gcc
Classification: Unclassified
Component: tree-optimization (show other bugs)
Version: 4.9.0
: P3 normal
Target Milestone: 4.9.0
Assignee: Jeffrey A. Law
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-09-16 06:56 UTC by Zhendong Su
Modified: 2013-09-16 19:29 UTC (History)
3 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2013-09-16 00:00:00


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Zhendong Su 2013-09-16 06:56:41 UTC
The current gcc trunk mis-compiles the following code on x86_64-linux at -O3 in 32-bit mode. 

This is a regression from 4.8.x. 


$ gcc-trunk -v
Using built-in specs.
COLLECT_GCC=gcc-trunk
COLLECT_LTO_WRAPPER=/usr/local/gcc-trunk/libexec/gcc/x86_64-unknown-linux-gnu/4.9.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-trunk/configure --enable-languages=c,c++,objc,obj-c++,fortran,lto --with-gmp=/usr/local/gcc-trunk --with-mpfr=/usr/local/gcc-trunk --with-mpc=/usr/local/gcc-trunk --with-cloog=/usr/local/gcc-trunk --prefix=/usr/local/gcc-trunk
Thread model: posix
gcc version 4.9.0 20130915 (experimental) [trunk revision 202599] (GCC) 
$ gcc-trunk -m32 -O2 small.c
$ a.out
$ gcc-4.8 -m32 -O3 small.c
$ a.out
$ gcc-trunk -m32 -O3 small.c
$ a.out
Aborted (core dumped)
$ 


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


char a, h;
int b, d, e, g, j, k;
volatile int c;
short i;

int
main ()
{
  int m = i ^= 1;
  for (b = 0; b < 1; b++)
    {
      char o = m;
      g = k;
      j = j || c;
      if (a != o)
	for (; d < 1; d++)
	  ;
      else
	{
	  char *p = &h;
	  *p = 1;
	  for (; e; e++)
	    ;
	}
    }

  if (h != 0)
    __builtin_abort(); 

  return 0;
}
Comment 1 Jakub Jelinek 2013-09-16 07:34:37 UTC
Started with r202489.
Comment 2 Jeffrey A. Law 2013-09-16 18:37:53 UTC
Jakub,

That doesn't make *any* sense.  r202489 simply *avoids* doing any jump threading in certain cases.  If that change is indeed the trigger, then the root cause is going to be a latent bug elsewhere.
Comment 3 Marc Glisse 2013-09-16 19:28:35 UTC
-fdisable-tree-phicprop2 lets it pass (the dumps appear identical before and after this pass, but with verbose dumps we see some memory PHIs disappear)
-fdisable-tree-cddce2 also lets it pass, with larger differences in the dumps.
Comment 4 Jeffrey A. Law 2013-09-16 19:29:23 UTC
Verified as a duplicate.

*** This bug has been marked as a duplicate of bug 58418 ***