[Bug rtl-optimization/42694] New: Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[Bug rtl-optimization/42694] New: Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu dot org]

Right now, the compiler has no special code to optimize pow (x, 0.75) into
something like sqrt (x) * sqrt (sqrt (x)) under -ffast-math, nor pow (x, 0.25)
into sqrt (sqrt (x)).  On machines with a builtin sqrt instruction, it is often
times faster to do the calculations using sqrt than calling the pow function. 
In particular, x**0.75 shows up in the bwaves spec 2006 benchmark.

For sqrt (sqrt (x)) vs pow (x, 0.25), I see:
IBM power6: 9.2 times faster
Intel core2 laptop: 5 times faster
AMD K8 system: 3.2 times faster

For sqrt (x) * sqrt (sqrt (x)) vs. pow (x, 0.75), I see:
IBM power6: 6.1 times faster
Intel core2 laptop: 3.4 times faster
AMD K8 system: 2.3 times faster

In addition, the compiler optimizes sqrt (sqrt (x)) into pow (x, 0.25), and
similar optimizations.

This should be fixed by adding a hook to say what the relative speed of sqrt
and cbrt vs. pow, so that the backend can control whether or not this
optimization should be done.  By default, the optimization is probably only
useful if -Os is used on machines that have a hardware sqrt instruction.


-- 
           Summary: Compiler could optimize pow (x, 0.75) into sqrt (x) *
                    sqrt (sqrt (x))
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: meissner at gcc dot gnu dot org
        ReportedBy: meissner at gcc dot gnu dot org
 GCC build triplet: powerpc64-unknown-linux-gnu
  GCC host triplet: powerpc64-unknown-linux-gnu
GCC target triplet: powerpc64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug rtl-optimization/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu dot org]

------- Comment #1 from meissner at gcc dot gnu dot org  2010-01-11 21:42 -------
Created an attachment (id=19542)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19542&action=view)
Program to demonstrate the slowdown

This program calculates the speed of various pow arguments.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2010-01-12 09:08 -------
Confirmed and easy to add in the pow expander (expand_builtin_pow).


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
           Severity|normal                      |enhancement
             Status|UNCONFIRMED                 |NEW
          Component|rtl-optimization            |middle-end
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-01-12 09:08:00
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu dot org]

------- Comment #3 from meissner at gcc dot gnu dot org  2010-01-26 00:40 -------
Created an attachment (id=19706)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19706&action=view)
Proposed patch to fix the problem

Patch to add 3 new switches to tune whether to use a combination of sqrt/cbrt
to replace pow.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2010-01-26 09:57 -------
Ugh, I don't like more switches.  Instead I would say that using two sqrt
calls instead of one pow call is always profitable if not optimizing for size
and the target has an optab for sqrt.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[paolo dot carlini at oracle dot com]

------- Comment #5 from paolo dot carlini at oracle dot com  2010-01-26 10:03 -------
With my "user of the compiler" hat, I must say that I was also put off by the 3
new switches... fwiw


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu dot org]

------- Comment #6 from meissner at gcc dot gnu dot org  2010-04-14 15:02 -------
Subject: Bug 42694

Author: meissner
Date: Wed Apr 14 15:01:40 2010
New Revision: 158346

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158346
Log:
PR 42694: Optimize pow (x, 0.25), pow (x, 0.75), pow (x, 1./6.)

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/builtins.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu dot org]

------- Comment #7 from meissner at gcc dot gnu dot org  2010-04-14 15:03 -------
Fixed with subversion id 158346.


-- 

meissner at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

--- Comment #9 from Michael Meissner <meissner at gcc dot gnu.org> 2010-12-08 16:34:26 UTC ---
Author: meissner
Date: Wed Dec  8 16:34:20 2010
New Revision: 167594

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=167594
Log:
PR 42694: add checks to make sure sqrt is supported

Added:
    trunk/gcc/testsuite/gcc.target/powerpc/ppc-pow.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/builtins.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/42694] Compiler could optimize pow (x, 0.75) into sqrt (x) * sqrt (sqrt (x))

[meissner at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42694

--- Comment #8 from Michael Meissner <meissner at gcc dot gnu.org> 2010-12-07 17:42:04 UTC ---
Created attachment 22675
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22675
Add check for sqrt optab being present before doing some optimizations