[Bug rtl-optimization/46888] New: missed optimization of zero_extract with constant inputs

[Bug rtl-optimization/46888] New: missed optimization of zero_extract with constant inputs

[ams at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

           Summary: missed optimization of zero_extract with constant
                    inputs
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ams@gcc.gnu.org


The compiler fails to do constant folding for bit fields, at least on ARM
targets, and instead builds the constant at run time.

Test case:

  struct bits
  {
     unsigned a:5;
     unsigned b:5;
     unsigned c:5;
     unsigned d:5;
  };

  struct bits
  f (unsigned int a)
  {
     struct bits bits = {0,0,0,0};
     bits.a = 1;
     bits.b = 2;
     bits.c = 3;
     bits.d = a;
     return bits;
  }

Output, compiled for ARM with "-O2 -mcpu=cortex-a8 -mthumb":
        movs    r2, #1
        movs    r3, #0
        bfi     r3, r2, #0, #5
        movs    r2, #2
        bfi     r3, r2, #5, #5
        movs    r2, #3
        bfi     r3, r2, #10, #5
        bfi     r3, r0, #15, #5
        mov     r0, r3
        bx      lr

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[ams at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

--- Comment #1 from Andrew Stubbs <ams at gcc dot gnu.org> 2010-12-10 17:02:05 UTC ---
Two different patches have been posted to fix this:

 http://gcc.gnu.org/ml/gcc-patches/2010-12/msg00778.html

 http://gcc.gnu.org/ml/gcc-patches/2010-12/msg00784.html

One, or both should be applied to GCC 4.7, when the time comes.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.12.10 19:42:11
            Version|unknown                     |4.6.0
     Ever Confirmed|0                           |1
           Severity|normal                      |enhancement

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> 2010-12-10 19:42:11 UTC ---
Here is a simpler example. Take:
struct b
{
  int t;
  int tt;
};
union a
{
  long long t;
  struct b b;
};
long long f(int i, int t)
{
  long long ii = ((long long)i) << 32;
  union a a;
  a.t = 0;
  a.b.t = i;
  a.b.tt = t;
  return a.t;
}

--- CUT ---
For MIPS64r2 (both n64 and n32):
    move    $2,$0
    dins    $2,$4,32,32
    j    $31
    dins    $2,$5,0,32
--- CUT ---
is produced but the first two instructions can really be done as one shift:
        dsll    $2,$4,32
    j    $31
    dins    $2,$5,0,32

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2010-12-13 21:37:37 UTC ---
(In reply to comment #1)
> Two different patches have been posted to fix this:
> 
>  http://gcc.gnu.org/ml/gcc-patches/2010-12/msg00778.html
> 
>  http://gcc.gnu.org/ml/gcc-patches/2010-12/msg00784.html

The testcase in comment #2 is not optimized with the CSE patch but is with the
combine patch.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |pinskia at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-10-12 08:07:49 UTC ---
Created attachment 28428
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28428
A third patch

Here is a third patch which improves the code a different way, one which seems
better as combine.c already had the code which is supposed to do the combining
of:
(set (x) (const))
(set (zero_extract (x a b) (const))
But it forgot that x can be a non subreg if we are using a zero extract.

Also it is easy add the case where the second set does not have a const there
and optimize it to a shift followed by an and which gets my testcase too.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-10-12 08:23:35 UTC ---
It actually does not work for the arm case but it can be improved to work for
it which I am doing now.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #28428|0                           |1
        is obsolete|                            |

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-10-12 08:49:35 UTC ---
Created attachment 28431
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28431
New patch

Still does not fix the ARM case but similar thing can be done for the three
combine case.  It does fix the mips case which I was looking into; though
really there should be only one insert and no shift.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-10-12 08:50:55 UTC ---
Oh with my current patch, we are able to remove at least one bfi, the first
one.

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-11-13 01:18:52 UTC ---
http://gcc.gnu.org/ml/gcc-patches/2012-11/msg00952.html
Will improve the testcase at the tree level for little-endian:
  _4 = (unsigned char) a_3(D);
  _5 = (<unnamed-unsigned:5>) _4;
  BIT_FIELD_REF <D.1727, 10, 0> = 65;
  D.1727.c = 3;
  D.1727.d = _5;
  return D.1727;

And will fix it on big-endian:
  D.1354_2 = (unsigned char) a_1(D);
  D.1355_3 = (<unnamed-unsigned:5>) D.1354_2;
  BIT_FIELD_REF <D.1356, 15, 0> = 1091;
  D.1356.d = D.1355_3;

[Bug rtl-optimization/46888] missed optimization of zero_extract with constant inputs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46888

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
   Target Milestone|---                         |10.0
         Resolution|---                         |FIXED

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed fully in GCC 10+:
        movw    r3, #3137
        bfi     r3, r0, #15, #5
        mov     r0, r3
        bx      lr

Had also improved in GCC 8:
        movw    r2, #3137
        movs    r3, #0
        bfi     r3, r2, #0, #16
        bfi     r3, r0, #15, #5
        mov     r0, r3
        bx      lr