[Bug c++/11680] New: member access of C++ class with all methods inline should be optimized better

[Bug c++/11680] New: member access of C++ class with all methods inline should be optimized better

[dbaron at dbaron 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=11680

           Summary: member access of C++ class with all methods inline
                    should be optimized better
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dbaron at dbaron dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu

This is a reduced testcase of the problem described in
http://bugzilla.mozilla.org/show_bug.cgi?id=206682#c25 , being filed as
mentioned in bug 11376 comment 6 and bug 11376 comment 37.  The bug is that gcc
didn't optimize the code in question as well as I think in should.  (Optimizing
code written in a "generic programming" style is needed to encourage people to
write such code...)  In particular, I was working on a class that did UTF-16 to
UTF-8 conversion (and could be used as one of many character sinks with
Mozilla's string classes).

In particular, the code involved a large inline method that involved repeated
use of a member variable (being used as a cursor into the output buffer being
written).  (By default, gcc didn't inline the method.  This was hand-optimized
by using |__attribute__((always_inline))|.  Should I file a separate bug on
this, or is it unlikely to be fixed?)  Once the method was inlined, gcc did not
optimize the use of the member variable as well as it could have.  This was
hand-optimized by assigning from the member variable at the beginning of the
|write| function and assigning back to it at the end.  Since all the methods of
the class are inlined and the compiler can tell that a pointer to the member
variable |mBuffer| is never taken, gcc should have been able to produce code
that was just as fast with or without this second hand-optimization.

This can be tested by comparing the performance of the attached testcase with
the arguments "slow" (forced inline, but without the local variable
hand-optimization) and "fast" (with the local variable hand-optimization).  I
claim that gcc should be producing essentially identical code for these two
cases, so they should perform equivalently.  This is not the case:

> /usr/local/gcc-trunk/bin/g++ -O2 perf.cpp

> time ./a.out fast
1.270u 0.010s 0:01.24 103.2%    0+0k 0+0io 224pf+0w
> time ./a.out fast
1.270u 0.000s 0:01.25 101.6%    0+0k 0+0io 224pf+0w
> time ./a.out fast
1.300u 0.010s 0:01.27 103.1%    0+0k 0+0io 224pf+0w

> time ./a.out slow
4.030u 0.010s 0:04.03 100.2%    0+0k 0+0io 224pf+0w
> time ./a.out slow
4.030u 0.000s 0:04.11 98.0%     0+0k 0+0io 224pf+0w
> time ./a.out slow
4.060u 0.000s 0:04.10 99.0%     0+0k 0+0io 224pf+0w

These measurements were made with the gcc trunk as of today.  The numbers for
the gcc 3.3 release are similar (although gcc 3.3 was slightly slower for the
"fast" version and slightly faster for the "slow" version).

[Bug c++/11680] member access of C++ class with all methods inline should be optimized better

[dbaron at dbaron 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=11680



------- Additional Comments From dbaron at dbaron dot org  2003-07-27 05:16 -------
Created an attachment (id=4487)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=4487&action=view)
test program

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[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=11680


pinskia at physics dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
             Status|UNCONFIRMED                 |NEW
          Component|c++                         |optimization
     Ever Confirmed|                            |1
  GCC build triplet|i686-pc-linux-gnu           |
   GCC host triplet|i686-pc-linux-gnu           |
 GCC target triplet|i686-pc-linux-gnu           |*-*-*
           Keywords|                            |pessimizes-code
           Priority|P2                          |P3
   Last reconfirmed|0000-00-00 00:00:00         |2003-07-27 05:31:00
               date|                            |


------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-27 05:31 -------
I can confirm this on the mainline (20030726) on powerpc-apple-darwin6.6 on a powerpc 750.  
The problem is that are large number of stores in the loop which is redudent which might be able 
to fixed by store motion but are not.

On the mainline, GCC inlines the method so there is no need for always_inline any more.

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[dberlin 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=11680



------- Additional Comments From dberlin at gcc dot gnu dot org  2003-07-27 05:33 -------
Try the tree-ssa branch, which should and does help quite a lot for these types of problems

My numbers:

mainline: 
[dberlin@dberlin dberlin]$ time ./a.out  slow

real    0m3.777s
user    0m3.770s
sys     0m0.010s
[dberlin@dberlin dberlin]$ time ./a.out  fast

real    0m1.065s
user    0m1.070s
sys     0m0.000s

tree-ssa:
[dberlin@dberlin dberlin]$ time ./a.out  slow

real    0m1.356s
user    0m1.360s
sys     0m0.000s
[dberlin@dberlin dberlin]$ time ./a.out  fast

real    0m1.120s
user    0m1.110s
sys     0m0.010s

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[gdr at integrable-solutions dot net]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680



------- Additional Comments From gdr at integrable-solutions dot net  2003-07-27 09:34 -------
Subject: Re:  New: member access of C++ class with all methods inline should be optimized better

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

| In particular, the code involved a large inline method that involved repeated
| use of a member variable (being used as a cursor into the output buffer being
| written).  (By default, gcc didn't inline the method.  This was hand-optimized
| by using |__attribute__((always_inline))|.  Should I file a separate bug on
| this, or is it unlikely to be fixed?)

Please fill a bug report. GCC should implement the obvious and original
meaning of "inline".  That it failed is a bug.  
That one should ressort to contorsions and unspeakable syntax like
__attribute__((always_inline)) while we have the obivous syntax is
also a bug.

-- Gaby

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[dhazeghi at yahoo dot com]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680


dhazeghi at yahoo dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|*-*-*                       |
   Target Milestone|3.4                         |---

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[dbaron at dbaron 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=11680



------- Additional Comments From dbaron at dbaron dot org  2003-09-08 23:01 -------
Re comment 3: That's good to see.  However, I would think that there should be
an optimization (really application of existing optimizations to this case) that
would cause the generated code to be exactly the same.  That was my hope when
filing this bug.  (After all, making clearer code run slower discourages clear
code.)

Re comment 4: Filed as bug 12219.

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-11-27 07:32 -------
Suspending this one as it is fixed on the tree-ssa (might even be better than quoted now).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |SUSPENDED
   Target Milestone|---                         |tree-ssa


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680

[Bug optimization/11680] member access of C++ class with all methods inline should be optimized better

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-03-03 07:10 -------
GCC mainline:
zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out slow
4.510u 0.000s 0:04.59 98.2%     0+0k 0+0io 0pf+0w
[zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out fast
3.560u 0.010s 0:03.91 91.3%     0+0k 0+0io 0pf+0w
[zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out original
4.010u 0.010s 0:04.42 90.9%     0+0k 0+0io 0pf+0w

TREE-SSA:
[zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out original
3.350u 0.010s 0:03.50 96.0%     0+0k 0+1io 0pf+0w
[zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out fast
3.620u 0.020s 0:04.62 78.7%     0+0k 0+0io 0pf+0w
[zhivago2:gcc/objdir_c/gcc] pinskia% time ./a.out slow
3.610u 0.000s 0:03.73 96.7%     0+0k 0+0io 0pf+0w

Either it is already fixed on the mainline or the tree-ssa needs improvements.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|SUSPENDED                   |NEW
   Last reconfirmed|2003-07-27 05:31:00         |2004-03-03 07:10:01
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680

[Bug rtl-optimization/11680] member access of C++ class with all methods inline should be optimized better

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|3.5.0                       |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680

[Bug rtl-optimization/11680] member access of C++ class with all methods inline should be optimized better

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-03 04:42 -------
This was fixed in the merging of the tree-ssa, all the asm of the different ways are exactly the same 
now (and better than the 3.4.0 also).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11680