[Bug tree-optimization/103218] New: (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[Bug tree-optimization/103218] New: (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

            Bug ID: 103218
           Summary: (a < 0) << signbit is not always optimized to a &
                    signbitmask at the gimple level
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, TREE
          Severity: enhancement
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

Take:
int f(signed char a)
{
  signed char t = a < 0;
  return (unsigned char)(t << 7);
}

At the gimple level we get:
int f(signed char a)
{
  signed char t = a < 0;
  return (unsigned char)(t << 7);
}
But combine is able to it:
Trying 9 -> 10:
    9: {r89:QI=r91:SI#0 0>>0x7;clobber flags:CC;}
      REG_DEAD r91:SI
      REG_UNUSED flags:CC
   10: {r90:QI=r89:QI<<0x7;clobber flags:CC;}
      REG_DEAD r89:QI
      REG_UNUSED flags:CC
Successfully matched this instruction:
(parallel [
        (set (reg:QI 90)
            (and:QI (subreg:QI (reg:SI 91) 0)
                (const_int -128 [0xffffffffffffff80])))
        (clobber (reg:CC 17 flags))
    ])
allowing combination of insns 9 and 10
original costs 4 + 4 = 8
replacement cost 4
deferring deletion of insn with uid = 9.
modifying insn i3    10: {r90:QI=r91:SI#0&0xffffffffffffff80;clobber flags:CC;}
      REG_DEAD r91:SI
      REG_UNUSED flags:CC
deferring rescan insn with uid = 10.

If we had wrote the testcase like:
int f(signed char a)
{
  return (a < 0) << 7;
}

GCC does optimize it to:
(int) NON_LVALUE_EXPR <a> & 128

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=103216

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note I noticed this while looking on PR 103216.

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-11-13
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |pinskia at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note another testcase:
signed char  f(signed char a)
{
    if (a < 0)
      return 1u<<7;
    return 0;
}

PHIOPT does convert it to:
  _4 = a_2(D) < 0;
  _5 = (signed char) _4;
  _6 = _5 << 7;

But then nothing converts it into:
  _6 = a_2(D) & 128;


Let me see what I can do.

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
There is code in fold_ternary_loc which does:
      /* A < 0 ? <sign bit of A> : 0 is simply (A & <sign bit of A>).  */

And this works as fold also does (a<0)<<N into (a<0) ? 1<<N : 0 which is not
done in match.pd for reasons.

We should just handle ((cast)(a < 0)) << signbit in match.pd.

Implementing that.

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 51782
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51782&action=edit
patch in testing

Patch in testing.

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://gcc.gnu.org/piperma
                   |                            |il/gcc-patches/2021-Novembe
                   |                            |r/584367.html
           Keywords|                            |patch

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Patch submitted:
https://gcc.gnu.org/pipermail/gcc-patches/2021-November/584367.html

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:11c4a06a6c1a9db0bfdb3ee8509392dd7163709c

commit r12-5305-g11c4a06a6c1a9db0bfdb3ee8509392dd7163709c
Author: Andrew Pinski <apinski@marvell.com>
Date:   Sat Nov 13 04:16:55 2021 +0000

    tree-optimization: [PR103218] Fold ((type)(a<0)) << SIGNBITOFA into
((type)a) & signbit

    This folds Fold ((type)(a<0)) << SIGNBITOFA into ((type)a) & signbit inside
match.pd.
    This was already handled in fold-cost by:
    /* A < 0 ? <sign bit of A> : 0 is simply (A & <sign bit of A>).  */
    I have not removed as we only simplify "a ? POW2 : 0" at the gimple level
to "a << CST1"
    and fold actually does the reverse of folding "(a<0)<<CST" into "(a<0) ?
1<<CST : 0".

    OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

            PR tree-optimization/103218

    gcc/ChangeLog:

            * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".

    gcc/testsuite/ChangeLog:

            * gcc.dg/tree-ssa/pr103218-1.c: New test.

[Bug tree-optimization/103218] (a < 0) << signbit is not always optimized to a & signbitmask at the gimple level

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103218

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |12.0
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed.