[Bug regression/55754] New: FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[Bug regression/55754] New: FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[ktkachov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55754

             Bug #: 55754
           Summary: FAIL: gcc.target/arm/unsigned-extend-2.c
                    scan-assembler ands
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: regression
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ktkachov@gcc.gnu.org
                CC: ramana.radhakrishnan@arm.com, richard.earnshaw@arm.com
            Target: arm-none-eabi


FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands
FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler-not uxtb
FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler-not cmp

Bisection shows r194608 introduces the FAILs.

In particular the following snippet causes the test FAIL:

   /* If *op0 is (zero_extend:SI (subreg:QI (reg:SI) 0)) and comparing
      with const0_rtx, change it to (and:SI (reg:SI) (const_int 255)),
      to facilitate possible combining with a cmp into 'ands'.  */
-  if (mode == SImode
+  if (!op0_preserve_value
+      && mode == SImode
       && GET_CODE (*op0) == ZERO_EXTEND
       && GET_CODE (XEXP (*op0, 0)) == SUBREG
       && GET_MODE (XEXP (*op0, 0)) == QImode


This change disables the transformation that the testcase is looking for.

Thanks,
Kyrill

[Bug regression/55754] FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[krebbel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55754

--- Comment #1 from Andreas Krebbel <krebbel at gcc dot gnu.org> 2012-12-20 15:09:21 UTC ---
This hunk needs to be reverted. op0 is modified but it is set to an equivalent
value.

[Bug regression/55754] FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[krebbel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55754

--- Comment #2 from Andreas Krebbel <krebbel at gcc dot gnu.org> 2012-12-20 15:20:17 UTC ---
Author: krebbel
Date: Thu Dec 20 15:20:06 2012
New Revision: 194636

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194636
Log:
2012-12-20  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

    PR target/55754
    * config/arm/arm.c (arm_canonicalize_comparison): Remove
    op0_preserve_value check for zero_extend to and transformation.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/arm/arm.c

[Bug regression/55754] FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55754

--- Comment #3 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2012-12-20 15:44:23 UTC ---
(In reply to comment #1)
> This hunk needs to be reverted. op0 is modified but it is set to an equivalent
> value.

Perhaps you could update the documentation to make that clearer. Eg, by adding
to the example a safe transformation (like this one).

[Bug regression/55754] FAIL: gcc.target/arm/unsigned-extend-2.c scan-assembler ands

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55754

Ramana Radhakrishnan <ramana at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |ramana at gcc dot gnu.org
         Resolution|---                         |FIXED
   Target Milestone|---                         |4.8.0

--- Comment #4 from Ramana Radhakrishnan <ramana at gcc dot gnu.org> ---
fixed for 4.8.0