[Bug optimization/11710] New: loop optimizer cannot figure out # iterations for shift and / * "increments"

[Bug optimization/11710] New: loop optimizer cannot figure out # iterations for shift and / * "increments"

[rguenth at tat dot physik dot uni-tuebingen dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710

           Summary: loop optimizer cannot figure out # iterations for shift
                    and / * "increments"
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P2
         Component: optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at tat dot physik dot uni-tuebingen dot de
                CC: gcc-bugs at gcc dot gnu dot org

All loops in the following testcases run a small constant amout of iterations.
If compiled with -O2 -funroll-loops they dont get unrolled, though, probably due
to the lack of the loop optimizer finding out the constant number of iterations.

This hurts optimization of libstdc++ pow(T, int) implementation for constant
second argument.

int shift()
{
        int n = 5, i = 0;
        while (n <<= 2)
                ++i;
        return i;
}

int divide()
{
        int n = 5, i = 0;
        while (n /= 2)
                ++i;
        return i;
}

int multiply()
{
        int n = 5, i = 0;
        while (n *= 2 < 20)
                ++i;
        return i;
}

[Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[pinskia at physics dot uc dot edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710


pinskia at physics dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |pessimizes-code
   Last reconfirmed|0000-00-00 00:00:00         |2003-07-29 18:22:53
               date|                            |


------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-29 18:22 -------
I can confirm this on the mainline (20030729) on powerpc-apple-darwin6.6 but it is not fixed on 
the rtlopt branch sorry.

Re: [Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[Zdenek Dvorak]

Hello,

> I can confirm this on the mainline (20030729) on powerpc-apple-darwin6.6 but it is not fixed on 
> the rtlopt branch sorry.

yes, this feature is not supported neither by old nor by new loop
optimizer just now.  It might be with the new iv analysis that is being
written for tree-ssa (although this is only my guess).  Handling
non-linear induction variables brings very little gain in practice,
as they are rare, and it has very small priority for me now.

Zdenek

[Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[rakdver at atrey dot karlin dot mff dot cuni dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710



------- Additional Comments From rakdver at atrey dot karlin dot mff dot cuni dot cz  2003-07-29 19:27 -------
Subject: Re:  loop optimizer cannot figure out # iterations for shift and / * "increments"

Hello,

> I can confirm this on the mainline (20030729) on powerpc-apple-darwin6.6 but it is not fixed on 
> the rtlopt branch sorry.

yes, this feature is not supported neither by old nor by new loop
optimizer just now.  It might be with the new iv analysis that is being
written for tree-ssa (although this is only my guess).  Handling
non-linear induction variables brings very little gain in practice,
as they are rare, and it has very small priority for me now.

Zdenek

[Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[neroden at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710


neroden at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|3.4                         |---


------- Additional Comments From neroden at gcc dot gnu dot org  2003-08-03 19:34 -------
Not gonna happen for 3.4

[Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[rakdver at gcc dot gnu dot org]

------- Additional Comments From rakdver at gcc dot gnu dot org  2004-03-12 01:45 -------
Fixed on lno-branch by

http://gcc.gnu.org/ml/gcc-patches/2004-03/msg00995.html

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710

[Bug optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-03-12 04:27 -------
Fixed on the lno so suspending as fixed on the lno branch.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |SUSPENDED
   Target Milestone|---                         |lno


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710

[Bug rtl-optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-20 01:02 -------
This is fixed on the mainline with the option -ftree-loop-ivcanon which should be enabled by default 
but is not yet.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rakdver at gcc dot gnu dot
                   |                            |org
             Status|SUSPENDED                   |NEW
      Known to fail|2.97 3.0 3.0.1 3.0.2 3.0.3  |2.97 3.0 3.0.1 3.0.2 3.0.3
                   |3.0.4 3.1 3.1.1 3.1.2 3.2   |3.0.4 3.1 3.1.1 3.1.2 3.2
                   |3.2.1 3.2.2 3.2.3 3.3 3.3.1 |3.2.1 3.2.2 3.2.3 3.3 3.3.1
                   |3.3.2 3.3.3 3.4.0 4.0       |3.3.2 3.3.3 3.4.0 4.0.0
      Known to work|                            |lno
   Last reconfirmed|2004-06-11 00:51:41         |2004-09-20 01:02:53
               date|                            |
   Target Milestone|lno                         |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710

[Bug rtl-optimization/11710] loop optimizer cannot figure out # iterations for shift and / * "increments"

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-02 04:04 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
      Known to fail|2.97 3.0 3.0.1 3.0.2 3.0.3  |2.97 3.0 3.0.1 3.0.2 3.0.3
                   |3.0.4 3.1 3.1.1 3.1.2 3.2   |3.0.4 3.1 3.1.1 3.1.2 3.2
                   |3.2.1 3.2.2 3.2.3 3.3 3.3.1 |3.2.1 3.2.2 3.2.3 3.3 3.3.1
                   |3.3.2 3.3.3 3.4.0 4.0.0     |3.3.2 3.3.3 3.4.0
         Resolution|                            |FIXED
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11710