[Bug tree-optimization/19401] New: Trivial loop not unrolled

public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed

* [Bug tree-optimization/19401] New: Trivial loop not unrolled
@ 2005-01-12 16:08 rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:10 ` [Bug tree-optimization/19401] " rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (11 more replies)
  0 siblings, 12 replies; 13+ messages in thread
From: rguenth at tat dot physik dot uni-tuebingen dot de @ 2005-01-12 16:08 UTC (permalink / raw)
  To: gcc-bugs

We do not unroll the loop in

double foo(double __x)
{
        unsigned int __n = 2;
        double __y = __n % 2 ? __x : 1;

        while (__n >>= 1)
          {
            __x = __x * __x;
            if (__n % 2)
              __y = __y * __x;
          }

        return __y;
}

with -O2 which causes us to emit gratiously worse code
for std::pow(x, 2) than for std::pow(x, 2.0).

We should definitely get this right without -funroll-loops
and all its side-effects.

-- 
           Summary: Trivial loop not unrolled
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: tree-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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
@ 2005-01-12 16:10 ` rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:14 ` pinskia at gcc dot gnu dot org
                   ` (10 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: rguenth at tat dot physik dot uni-tuebingen dot de @ 2005-01-12 16:10 UTC (permalink / raw)
  To: gcc-bugs



-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |11706
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:10 ` [Bug tree-optimization/19401] " rguenth at tat dot physik dot uni-tuebingen dot de
@ 2005-01-12 16:14 ` pinskia at gcc dot gnu dot org
  2005-01-12 16:19 ` rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (9 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2005-01-12 16:14 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-12 16:13 -------
Really -funroll-loops should be added to -O3 by default and maybe a tweewed version to -O2/-Os 
(one which only unroll loops which can be unrolled fully and don't produce more code than the loop 
itself).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |missed-optimization
   Last reconfirmed|0000-00-00 00:00:00         |2005-01-12 16:13:54
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:10 ` [Bug tree-optimization/19401] " rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:14 ` pinskia at gcc dot gnu dot org
@ 2005-01-12 16:19 ` rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 16:21 ` pinskia at gcc dot gnu dot org
                   ` (8 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: rguenth at tat dot physik dot uni-tuebingen dot de @ 2005-01-12 16:19 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From rguenth at tat dot physik dot uni-tuebingen dot de  2005-01-12 16:19 -------
In 3.4 one was able to do this by specifying -fpeel-loops and got complete loop
peeling enabled.  In 4.0 this is also the case, but only for the RTL unroller -
the tree unroller is not affected and as such _this_ unrolling does not help
PR11706.  - Just another datapoint.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (2 preceding siblings ...)
  2005-01-12 16:19 ` rguenth at tat dot physik dot uni-tuebingen dot de
@ 2005-01-12 16:21 ` pinskia at gcc dot gnu dot org
  2005-01-12 16:24 ` rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (7 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2005-01-12 16:21 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-12 16:21 -------
Another trivial loop which is not unrolled but should be:

int f(int x)
{
  int i, x1;
  x1 = x;
  for(i = 0;i<2;i++)
   x1 *= x1;
  return x1;
}

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (3 preceding siblings ...)
  2005-01-12 16:21 ` pinskia at gcc dot gnu dot org
@ 2005-01-12 16:24 ` rguenth at tat dot physik dot uni-tuebingen dot de
  2005-01-12 17:05 ` gdr at integrable-solutions dot net
                   ` (6 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: rguenth at tat dot physik dot uni-tuebingen dot de @ 2005-01-12 16:24 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From rguenth at tat dot physik dot uni-tuebingen dot de  2005-01-12 16:24 -------
Or stuff often found in C++ libraries:

template <int Dim>
struct Vector
{
  Vector(float init)
  {
    for (int i=0; i<Dim; ++i)
      v[i] = init;
  }
  float v[Dim];
};

for small Dim (<= 4 should make nearly everyone happy) you want all the
dimension-unaware loops to be unrolled.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (4 preceding siblings ...)
  2005-01-12 16:24 ` rguenth at tat dot physik dot uni-tuebingen dot de
@ 2005-01-12 17:05 ` gdr at integrable-solutions dot net
  2005-01-12 17:06 ` gdr at integrable-solutions dot net
                   ` (5 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: gdr at integrable-solutions dot net @ 2005-01-12 17:05 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From gdr at integrable-solutions dot net  2005-01-12 17:05 -------
Subject: Re:  Trivial loop not unrolled

"pinskia at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| Really -funroll-loops should be added to -O3 by default and maybe a
| tweewed version to -O2/-Os  

I don't think the issue is to have -funroll-loops as default at higher
level. I think it is more about the compiler understanding loops with
constant bounds, just like we handle if(0) or if(1) branches.

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (5 preceding siblings ...)
  2005-01-12 17:05 ` gdr at integrable-solutions dot net
@ 2005-01-12 17:06 ` gdr at integrable-solutions dot net
  2005-01-13 13:48 ` pinskia at gcc dot gnu dot org
                   ` (4 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: gdr at integrable-solutions dot net @ 2005-01-12 17:06 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From gdr at integrable-solutions dot net  2005-01-12 17:06 -------
Subject: Re:  Trivial loop not unrolled

"rguenth at tat dot physik dot uni-tuebingen dot de" <gcc-bugzilla@gcc.gnu.org> writes:

| Or stuff often found in C++ libraries:
| 
| template <int Dim>
| struct Vector
| {
|   Vector(float init)
|   {
|     for (int i=0; i<Dim; ++i)
|       v[i] = init;
|   }
|   float v[Dim];
| };
| 
| for small Dim (<= 4 should make nearly everyone happy) you want all the
| dimension-unaware loops to be unrolled.

I agree.

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (6 preceding siblings ...)
  2005-01-12 17:06 ` gdr at integrable-solutions dot net
@ 2005-01-13 13:48 ` pinskia at gcc dot gnu dot org
  2005-01-14  9:04 ` rakdver at gcc dot gnu dot org
                   ` (3 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2005-01-13 13:48 UTC (permalink / raw)
  To: gcc-bugs



-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
           Severity|normal                      |enhancement


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (7 preceding siblings ...)
  2005-01-13 13:48 ` pinskia at gcc dot gnu dot org
@ 2005-01-14  9:04 ` rakdver at gcc dot gnu dot org
  2005-01-24  9:43 ` rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (2 subsequent siblings)
  11 siblings, 0 replies; 13+ messages in thread
From: rakdver at gcc dot gnu dot org @ 2005-01-14  9:04 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From rakdver at gcc dot gnu dot org  2005-01-14 09:04 -------
Other possible patch:

http://gcc.gnu.org/ml/gcc-patches/2005-01/msg00796.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (8 preceding siblings ...)
  2005-01-14  9:04 ` rakdver at gcc dot gnu dot org
@ 2005-01-24  9:43 ` rguenth at tat dot physik dot uni-tuebingen dot de
  2005-05-06 21:11 ` cvs-commit at gcc dot gnu dot org
  2005-05-07 19:56 ` pinskia at gcc dot gnu dot org
  11 siblings, 0 replies; 13+ messages in thread
From: rguenth at tat dot physik dot uni-tuebingen dot de @ 2005-01-24  9:43 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From rguenth at tat dot physik dot uni-tuebingen dot de  2005-01-24 09:43 -------
Another one - matrix multiplication:

/* A [NxM], B [MxP] */
#define DOLOOP(N, M, P) \
void matmul ## N ## M ## P(double *res, const double *A, const double *B) \
{ \
        int i,j,k; \
        for (k=0; k<P; ++k) \
                for (i=0; i<N; ++i) { \
                        double s = 0.0; \
                        for (j=0; j<M; ++j) \
                                s += A[i*M+j] * B[j*P+k]; \
                        res[i*P+k] = s; \
                } \
}

DOLOOP(1, 1, 1)
DOLOOP(2, 1, 2)
DOLOOP(1, 2, 1)
DOLOOP(2, 2, 2)
DOLOOP(1, 3, 1)
DOLOOP(1, 1024, 1)

all up to 2x2 should be profitable to completely unroll.  Be sure to unroll
one-time rolling loops like for the last case.

Zdeneks patch only does not unroll the DOLOOP(2, 2, 2) case at -O2. Good.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (9 preceding siblings ...)
  2005-01-24  9:43 ` rguenth at tat dot physik dot uni-tuebingen dot de
@ 2005-05-06 21:11 ` cvs-commit at gcc dot gnu dot org
  2005-05-07 19:56 ` pinskia at gcc dot gnu dot org
  11 siblings, 0 replies; 13+ messages in thread
From: cvs-commit at gcc dot gnu dot org @ 2005-05-06 21:11 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-05-06 21:11 -------
Subject: Bug 19401

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	rakdver@gcc.gnu.org	2005-05-06 21:11:29

Modified files:
	gcc            : ChangeLog tree-flow.h tree-ssa-loop-ivcanon.c 
	                 tree-ssa-loop.c 

Log message:
	PR tree-optimization/19401
	* tree-flow.h (tree_unroll_loops_completely): Declaration changed.
	* tree-ssa-loop-ivcanon.c (enum unroll_level): New.
	(estimated_unrolled_size): New function.
	(try_unroll_loop_completely, canonicalize_loop_induction_variables,
	tree_unroll_loops_completely): Always unroll loops if the code size
	does not increase.
	* tree-ssa-loop.c (tree_complete_unroll): Indicate whether all
	loops should be unrolled completely.
	(gate_tree_complete_unroll): Run complete unrolling unconditionally.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.8629&r2=2.8630
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-flow.h.diff?cvsroot=gcc&r1=2.101&r2=2.102
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-ssa-loop-ivcanon.c.diff?cvsroot=gcc&r1=2.11&r2=2.12
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-ssa-loop.c.diff?cvsroot=gcc&r1=2.28&r2=2.29



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

* [Bug tree-optimization/19401] Trivial loop not unrolled
  2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
                   ` (10 preceding siblings ...)
  2005-05-06 21:11 ` cvs-commit at gcc dot gnu dot org
@ 2005-05-07 19:56 ` pinskia at gcc dot gnu dot org
  11 siblings, 0 replies; 13+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2005-05-07 19:56 UTC (permalink / raw)
  To: gcc-bugs


------- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-07 19:56 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19401


^ permalink raw reply	[flat|nested] 13+ messages in thread

end of thread, other threads:[~2005-05-07 19:56 UTC | newest]

Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-01-12 16:08 [Bug tree-optimization/19401] New: Trivial loop not unrolled rguenth at tat dot physik dot uni-tuebingen dot de
2005-01-12 16:10 ` [Bug tree-optimization/19401] " rguenth at tat dot physik dot uni-tuebingen dot de
2005-01-12 16:14 ` pinskia at gcc dot gnu dot org
2005-01-12 16:19 ` rguenth at tat dot physik dot uni-tuebingen dot de
2005-01-12 16:21 ` pinskia at gcc dot gnu dot org
2005-01-12 16:24 ` rguenth at tat dot physik dot uni-tuebingen dot de
2005-01-12 17:05 ` gdr at integrable-solutions dot net
2005-01-12 17:06 ` gdr at integrable-solutions dot net
2005-01-13 13:48 ` pinskia at gcc dot gnu dot org
2005-01-14  9:04 ` rakdver at gcc dot gnu dot org
2005-01-24  9:43 ` rguenth at tat dot physik dot uni-tuebingen dot de
2005-05-06 21:11 ` cvs-commit at gcc dot gnu dot org
2005-05-07 19:56 ` pinskia at gcc dot gnu dot org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).