Ubuntu
gcc-4.6 package

cselim tree optimizer generates incorrect code

Bug #1029454 reported by Stéphane Glondu on 2012-07-26

This bug affects 1 person

	Status	Importance	Assigned to	Milestone
Linaro GCC	Fix Released	High	Matthew Gretton-Dann	Linaro GCC 4.6-2012.10
gcc-4.6 (Ubuntu)	Fix Released	High	Unassigned
Precise	Won't Fix	High	Unassigned	Ubuntu precise-updates

Bug Description

I think gcc (4.6.3-1ubuntu5, amd64) generates incorrect code with the attached test case.

The expected output is "m = 8241071186". However, when compiled with "-O3", it outputs "m = 1". Looking further, it seems that the following lines:

*m = 1UL;
ularith_div_2ul_ul_ul_r (m, 0UL, 1UL, s);

are being swapped in the assembly output.

The output is correct when compiled with "-O2" or with "-O3 -fno-tree-cselim".

On Debian (version 4.6.3-1), the output is always correct. I have reports that the output is correct with upstream gcc compiled on Ubuntu, so this behaviour looks Ubuntu-specific.

Tags:

Related branches

lp://staging/~matthew-gretton-dann/gcc-linaro/4.6-cselim-fix

Merged into lp://staging/gcc-linaro/4.6 at revision 106911

Christophe Lyon: Approve on 2012-10-03

Michael Hope: Pending requested 2012-09-20

lp://staging/debian/gcc-4.6

lp://staging/debian/gnat-4.6

Revision history for this message

Stéphane Glondu (glondu) wrote on 2012-07-26:

Test case Edit (698 bytes, text/plain)

Revision history for this message

Stéphane Glondu (glondu) wrote on 2012-07-30:

The version in quantal (4.6.3-8ubuntu1) is not affected.

Revision history for this message

Stéphane Glondu (glondu) wrote on 2012-07-30:

Sorry, I meant the default gcc (which is gcc-4.7). With gcc-4.6, I observe the same behaviour.

Revision history for this message

Stéphane Glondu (glondu) wrote on 2012-07-30:

Same issue in oneiric (4.6.1-9ubuntu3).

Revision history for this message

Stéphane Glondu (glondu) wrote on 2012-07-30:

I've tried various historic versions of the gcc-4.6 package in a clean precise chroot.

The first faulty version appears to be 4.6.0-7ubuntu1. The previous version, 4.6.0-6ubuntu3, generates correct code.

Moreover, the last version (4.6.3-1ubuntu5) with all Linaro patches out generates correct code.

Revision history for this message

Michael Hope (michaelh1) wrote on 2012-08-01:

Thank you for the bug report. I've confirmed this with gcc-linaro-4.6+bzr106903 on x86_64:

michaelh@crucis:~/linaro/bugs$ ~/linaro/gcc/build/native-4.6/install/bin/gcc -O3 testcase3.c
michaelh@crucis:~/linaro/bugs$ ./a.out
m = 1

The fault is probably introduced by r106733 which added condtional store sinking support.

I've set it to high priority as bad code is generated. Likely trunk fixes include r113242, r111593, r110,443, and r114986 of lp:gcc-linaro/4.7.

Changed in gcc-linaro:
status:	New → Triaged
importance:	Undecided → High

Matthias Klose (doko) on 2012-09-06

Changed in gcc-4.6 (Ubuntu):
importance:	Undecided → High
milestone:	none → precise-updates
status:	New → Triaged

Revision history for this message

Matthew Gretton-Dann (matthew-gretton-dann) wrote on 2012-09-10:

Confirmed reverting lp:gcc-linaro/4.6 r106733 fixes the issue.

Michael Hope (michaelh1) on 2012-09-11

Changed in gcc-linaro:
assignee:	nobody → Matthew Gretton-Dann (matthew-gretton-dann)

Kate Stewart (kate.stewart) on 2012-09-12

tags:

added: precise quantal rls-q-incoming

Matthias Klose (doko) on 2012-09-13

Changed in gcc-4.6 (Ubuntu Precise):
importance:	Undecided → High
milestone:	none → precise-updates
status:	New → Triaged
Changed in gcc-4.6 (Ubuntu):
milestone:	precise-updates → none

Revision history for this message

Matthew Gretton-Dann (matthew-gretton-dann) wrote on 2012-09-18:

The revisions Michael suggested do not fix the issue - either singly or when applied on top of each other.

Revision history for this message

Matthew Gretton-Dann (matthew-gretton-dann) wrote on 2012-09-18:

Investigations lead me to believe this is upstream bug PR51987 (http://gcc.gnu.org/PR51987) made visible because of the conditional store sinking support being added.

I believe upstream fix svn r183524 fixes the issue. This is currently under test

Matthew Gretton-Dann (matthew-gretton-dann) on 2012-09-18

Changed in gcc-linaro:
milestone:	none → 4.6-2012.10

Revision history for this message

Matthias Klose (doko) wrote on 2012-09-18:

#10

the backport fixes the issue, no regressions running the testsuite. however the testcase fails, because 4.6 doesn't know about ia32, so s/ia32/ilp32/

Revision history for this message

Launchpad Janitor (janitor) wrote on 2012-09-18:

#11

This bug was fixed in the package gcc-4.6 - 4.6.3-10ubuntu1

---------------
gcc-4.6 (4.6.3-10ubuntu1) quantal; urgency=low

* Merge with Debian.

gcc-4.6 (4.6.3-10) unstable; urgency=low

  * Update to SVN 20120918 (r191439) from the gcc-4_6-branch.
    - Fix PR c/54552 (ice on valid), PR c/54103 (ice on valid),
      PR target/54536 (AVR), PR middle-end/54515 (ice on valid),
      PR target/45070 (ARM, wrong code), PR target/54220 (AVR),
      PR driver/54335, PR rtl-optimization/54369 (mips, wrong code),
      PR c++/54511 (ice on valid), PR fortran/54225 (ice on invalid),
      PR fortran/53306 (ice on invalid), PR fortran/54556 (wrong code),
      PR fortran/54208 (rejects valid).

[ Nobuhiro Iwamatsu ]
* Remove sh4-enable-ieee.diff, -mieee enabled by default. Closes: #685974.

  [ Matthias Klose ]
  * Fix PR tree-optimization/51987, backport from the trunk, Linaro only
   (Matthew Gretton-Dann). LP: #1029454.
-- Matthias Klose <email address hidden> Tue, 18 Sep 2012 22:44:00 +0200