[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[froydnj at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

Nathan Froyd <froydnj at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |froydnj at gcc dot gnu.org
         Resolution|                            |WONTFIX

--- Comment #10 from Nathan Froyd <froydnj at gcc dot gnu.org> 2011-03-22 00:29:53 UTC ---
This bug has been open for almost five years.  When I tried Andrew's posted
patch, I think bootstrap slowed down slightly--possibly because of the extra
parsing and logic in TREE_CODE_CLASS?

Anyway, LTO should take care of this.  I'm going to close this; if somebody
wants to keep the bug around for posterity, feel free to reopen.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

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

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reopening due to GET_MODE_SIZE getting the same attention.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
since we require C++11 now, shouldn't they be constexpr now?

I will try doing that in the next couple of weeks.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #8839|0                           |1
        is obsolete|                            |

--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 53921
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53921&action=edit
C++11 constexprify these variables

This "constexpr"ify these variables. I am going to do a bootstrap/test on it
(building cc1 cc1plus works so ...).

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

--- Comment #14 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Compile time using the same base compiler (without the patch and with
--enable-checking=yes)

Without the patch:
apinski@xeond:~/src/upstream-gcc/gcc/objdir/gcc$ time make -j16 generic-match.o
CXXFLAGS="-O2 -g"
/home/apinski/src/upstream-gcc/gcc/objdir/./prev-gcc/xg++
-B/home/apinski/src/upstream-gcc/gcc/objdir/./prev-gcc/
-B/home/apinski/upstream-gcc/x86_64-pc-linux-gnu/bin/ -nostdinc++
-B/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-B/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs

-I/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/include/x86_64-pc-linux-gnu

-I/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/include
 -I/home/apinski/src/upstream-gcc/gcc/libstdc++-v3/libsupc++
-L/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-L/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs
 -fno-PIE -c   -O2 -g -DIN_GCC     -fno-exceptions -fno-rtti
-fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic
-Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common
-Wno-unused -DHAVE_CONFIG_H -I. -I. -I/home/apinski/src/upstream-gcc/gcc/gcc
-I/home/apinski/src/upstream-gcc/gcc/gcc/.
-I/home/apinski/src/upstream-gcc/gcc/gcc/../include
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libcpp/include
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libcody
-I/home/apinski/src/upstream-gcc/gcc/objdir/./gmp
-I/home/apinski/src/upstream-gcc/gcc/gmp
-I/home/apinski/src/upstream-gcc/gcc/objdir/./mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpc/src 
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libdecnumber
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libdecnumber/bid -I../libdecnumber
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libbacktrace
-I/home/apinski/src/upstream-gcc/gcc/objdir/./isl/include
-I/home/apinski/src/upstream-gcc/gcc/isl/include  -o generic-match.o -MT
generic-match.o -MMD -MP -MF ./.deps/generic-match.TPo generic-match.cc

real    1m48.703s
user    1m47.137s
sys     0m1.458s


with the patch:
apinski@xeond:~/src/upstream-gcc/gcc/objdir/gcc$ time make -j16 generic-match.o
CXXFLAGS="-O2 -g"
/home/apinski/src/upstream-gcc/gcc/objdir/./prev-gcc/xg++
-B/home/apinski/src/upstream-gcc/gcc/objdir/./prev-gcc/
-B/home/apinski/upstream-gcc/x86_64-pc-linux-gnu/bin/ -nostdinc++
-B/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-B/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs

-I/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/include/x86_64-pc-linux-gnu

-I/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/include
 -I/home/apinski/src/upstream-gcc/gcc/libstdc++-v3/libsupc++
-L/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-L/home/apinski/src/upstream-gcc/gcc/objdir/prev-x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs
 -fno-PIE -c   -O2 -g -DIN_GCC     -fno-exceptions -fno-rtti
-fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic
-Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -Werror -fno-common
-Wno-unused -DHAVE_CONFIG_H -I. -I. -I/home/apinski/src/upstream-gcc/gcc/gcc
-I/home/apinski/src/upstream-gcc/gcc/gcc/.
-I/home/apinski/src/upstream-gcc/gcc/gcc/../include
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libcpp/include
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libcody
-I/home/apinski/src/upstream-gcc/gcc/objdir/./gmp
-I/home/apinski/src/upstream-gcc/gcc/gmp
-I/home/apinski/src/upstream-gcc/gcc/objdir/./mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpfr/src
-I/home/apinski/src/upstream-gcc/gcc/mpc/src 
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libdecnumber
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libdecnumber/bid -I../libdecnumber
-I/home/apinski/src/upstream-gcc/gcc/gcc/../libbacktrace
-I/home/apinski/src/upstream-gcc/gcc/objdir/./isl/include
-I/home/apinski/src/upstream-gcc/gcc/isl/include  -o generic-match.o -MT
generic-match.o -MMD -MP -MF ./.deps/generic-match.TPo generic-match.cc

real    1m41.673s
user    1m39.165s
sys     0m1.462s

So slightly faster (~7%). I wonder if it is because there is less debug info
produced. 

generic-match.cc is the best case of what this can do really.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

--- Comment #15 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Looking at the -fdump-tree-optimized for generic-match.cc (-O2 -g) in terms of
lines, we get:
  832068 before
  718872 after

That is 17% less lines. That is nice improvments. Majority is debugging info
for arguments which are no longer used any more.
A good example is:
   # DEBUG BEGIN_STMT
-  # DEBUG __t => _39
-  # DEBUG __f => "generic-match.cc"
-  # DEBUG __l => 8705
-  # DEBUG __g => "generic_simplify_127"
-  # DEBUG INLINE_ENTRY non_type_check
-  # DEBUG BEGIN_STMT
-  _273 = tree_code_type[26];
-  if (_273 == 2)
-    goto <bb 38>; [0.00%]
-  else
-    goto <bb 39>; [100.00%]
-
-  <bb 38> [count: 0]:
-  # DEBUG BEGIN_STMT
-  tree_not_class_check_failed (_39, 2, "generic-match.cc", 8705,
"generic_simplify_127");
-
-  <bb 39> [local count: 104528766]:
-  # DEBUG BEGIN_STMT
-  # DEBUG __t => NULL
-  # DEBUG __f => NULL
-  # DEBUG __l => NULL
-  # DEBUG __g => NULL
   _43 = BIT_FIELD_REF <_39->base, 8, 16>;


and you can see constant folding is there and removes the comparison.
I think 26 is INTEGER_CST even.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

--- Comment #16 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Andrew Pinski <pinskia@gcc.gnu.org>:

https://gcc.gnu.org/g:5c021f17e7d09a0eae2d6fb875c9a5484bd4e043

commit r13-4170-g5c021f17e7d09a0eae2d6fb875c9a5484bd4e043
Author: Andrew Pinski <apinski@marvell.com>
Date:   Fri Nov 18 05:05:03 2022 +0000

    constexprify some tree variables

    Since we use C++11 by default now, we can
    use constexpr for some const decls in tree-core.h.

    This patch does that and it allows for better optimizations
    of GCC code with checking enabled and without LTO.

    For an example generic-match.cc compiling is speed up due
    to the less number of basic blocks and less debugging info
    produced. I did not check the speed of compiling the same source
    but rather the speed of compiling the old vs new sources here
    (but with the same compiler base).

    The small slow down in the parsing of the arrays in each TU
    is migrated by a speed up in how much code/debugging info
    is produced in the end.

    Note I looked at generic-match.cc since it is one of the
    compiling sources which causes parallel building to stall and
    I wanted to speed it up.

    OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

    gcc/ChangeLog:

            PR middle-end/14840
            * tree-core.h (tree_code_type): Constexprify
            by including all-tree.def.
            (tree_code_length): Likewise.
            * tree.cc (tree_code_type): Remove.
            (tree_code_length): Remove.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |13.0
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #17 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed finally.

[Bug middle-end/14840] fold tree_code_type[CST] and tree_code_length[CST] in GCC itself

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2006-04-08 01:46 -------
This is not a missed optimization, just a compile time fix really.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|missed-optimization         |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14840