[Bug c++/28871] New: massive C++ compile time slowdown

[Bug c++/28871] New: massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

Sometime between Aug 1, 2006 and Aug 22, 2006 the time needed to run the g++
and libstdc++ testsuite ballooned. 

See this message and followups:
http://gcc.gnu.org/ml/gcc/2006-08/msg00398.html

Here are the specifics that I have:

20060724, make check-target-libstdc++
1145.687u 137.804s 23:13.98 92.0%       0+0k 0+0io 23pf+0w

20060822, make check-target-libstdc++
2624.876u 247.539s 55:42.82 85.9%       0+0k 0+0io 45pf+0w

Ouch. In that time, there have been some additions in the libstdc++ testsuite,
however it did not double in size. I've been trying to compile mainline
libstdc++ with 4.1.1 g++ to see if I can pin down the issue to a particular
component but there are a lot of incompatible changes, so comparing directly is
difficult to impossible.


-- 
           Summary: massive C++ compile time slowdown
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bkoz 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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-08-28 13:23 -------
It might not have doubled but one testcase that was added might have been
always slow.  We really need one testcase that has slowed down or one testcase
that has always been slow that was added.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|major                       |normal
             Status|UNCONFIRMED                 |WAITING
           Keywords|                            |compile-time-hog


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

------- Comment #2 from bkoz at gcc dot gnu dot org  2006-08-29 10:36 -------

More info.

I was able to take mainline libstdc++ sources, modify limits,
--disable-visibility and then run check-compile on the same machine/os/sources
with both 4_1-branch and mainline (20060827).

The time just to compile the libstdc++ testsuite is:

gcc-4_1-branch --disable-visibility
1626.889u 142.268s 34:30.18 85.4%       0+0k 0+0io 0pf+0w

gcc --disable-visibility
1713.939u 152.437s 32:24.81 95.9%       0+0k 0+0io 4pf+0w

gcc
1885.657u 126.063s 34:21.40 97.5%       0+0k 0+0io 0pf+0w

This is compiling with a PCH (which appears to be used from -Winvalid-pch and
-H), but no linking or executing. Still seems slower than I remember it being a
month ago, but I'll try to isolate more, and then come up with something simple
that can be used to fix this.

It does not look like the visibilty support impacts compile time. That's good.

These times are close enough to be a wash. So, I think we can remove the C++FE
from the list of slowdown candidates. (Which includes libstdc++, g++,
optmizers.)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

------- Comment #3 from bkoz at gcc dot gnu dot org  2006-08-29 10:44 -------

Actually, this is -S, so not just compile only. Hmm.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

------- Comment #4 from bkoz at gcc dot gnu dot org  2006-09-04 15:25 -------

...this is a PCH vs. anonymous namespace issue. Crack smoking! If PCH is
incommpatible with anonymous namespaces, then -Winvalid-pch needs to warn about
it at the time of pch creation. 

I'm going to check in a fix to this, in that I'm going to revert to using
stdc++.h and remove the one use of anonymous namespaces.

I'll also make an enhancement request about -Winvalid-pch vs. anonymous
namespaces.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

------- Comment #5 from bkoz at gcc dot gnu dot org  2006-09-04 15:26 -------

Fixed.


-- 

bkoz at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871

[Bug c++/28871] massive C++ compile time slowdown

[bkoz at gcc dot gnu dot org]

------- Comment #6 from bkoz at gcc dot gnu dot org  2006-09-04 15:41 -------
Subject: Bug 28871

Author: bkoz
Date: Mon Sep  4 15:41:18 2006
New Revision: 116680

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=116680
Log:
2006-09-04  Benjamin Kosnik  <bkoz@redhat.com>

        PR c++/28871
        * include/ext/bitmap_allocator.h: Add comment for end of anonymous
        namespace.
        * include/ext/rope: Same.
        * include/bits/cpp_type_traits.h: Same.
        * include/tr1/tuple: Same.
        * include/tr1/functional_iterate.h: Same.

        * include/bits/cpp_type_traits.h: Revert anonymous namespace
        change, use nested detail instead.

        * testsuite/lib/libstdc++.exp (libstdc++_init): PCHFLAGS revert to
        stdc++.h.


Modified:
    trunk/libstdc++-v3/ChangeLog
    trunk/libstdc++-v3/include/bits/cpp_type_traits.h
    trunk/libstdc++-v3/include/ext/bitmap_allocator.h
    trunk/libstdc++-v3/include/ext/rope
    trunk/libstdc++-v3/include/tr1/functional_iterate.h
    trunk/libstdc++-v3/include/tr1/tuple
    trunk/libstdc++-v3/testsuite/lib/libstdc++.exp


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28871