[Bug c/61588] New: Optimization defaults are not what documentation say they should be

[Bug c/61588] New: Optimization defaults are not what documentation say they should be

[af300wsm at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61588

            Bug ID: 61588
           Summary: Optimization defaults are not what documentation say
                    they should be
           Product: gcc
           Version: 4.4.7
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: af300wsm at gmail dot com

Created attachment 32988
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=32988&action=edit
Os settings for gcc 4.4.7

The default settings for optimization levels O2 and Os are not what the
documentation say they should be.  This is true for gcc 4.4.7 on CentOS 6.5 (it
is also true for gcc 4.4.4 on CentOS 6.0).

The documentation found here:
https://gcc.gnu.org/onlinedocs/gcc-4.4.7/gcc/Optimize-Options.html#Optimize-Options
shows what is turned on/off between O2 and Os optimization options. 
Specifically, the documentation implies that Os is O2 but favoring size of the
binary over speed.  The following are the noted exceptions between Os (off) and
O2 (on):

  - falign-functions
  - falign-jumps
  - falign-loops
  - falign-labels
  - freorder-blocks
  - freorder-blocks-and-partition
  - fprefetch-loop-arrays
  - ftree-vect-loop-version

There are attachments to this bug for the O2 and Os settings for 4.4.4 (CentOS
6.0) and 4.4.7 (CentOS 6.5).  As you'll notice with diffing between OS
versions, the list is identical (except that in CentOS 6.5, gcc 4.4.7,
fstrict-enums exists), and diffing between gcc versions shows only 1
difference: finline-functions.

Contrary to the gcc documentation for 4.4.7, all values noted above are enabled
except falign-loops, freorder-blocks-and-partition and fprefetch-loop-arrays. 
Most concerning, however, is that all listed optimization settings between O2
and Os are identical with one exception.  The one difference between these two
is finline-functions.  Oddly, this is *disabled* in O2 and *enabled* in Os. 
This seems contrary to the stated goal of Os.

The attached files were generated with this command:

$ gcc -Q -Ox --help=optimizers > <output_file>  # where x is either 2 or s


This link,
https://gcc.gnu.org/onlinedocs/gcc-4.4.7/gcc/Overall-Options.html#Overall-Options
, shows the use of the -c option as well.  However, I noted this made no
difference to the output.

[Bug c/61588] Optimization defaults are not what documentation say they should be

[af300wsm at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61588

--- Comment #2 from af300wsm at gmail dot com ---
Created attachment 32990
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=32990&action=edit
O2 settings for 4.4.4

[Bug c/61588] Optimization defaults are not what documentation say they should be

[af300wsm at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61588

--- Comment #1 from af300wsm at gmail dot com ---
Created attachment 32989
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=32989&action=edit
O2 settings for 4.4.7

[Bug c/61588] Optimization defaults are not what documentation say they should be

[af300wsm at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61588

--- Comment #3 from af300wsm at gmail dot com ---
Created attachment 32991
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=32991&action=edit
Os settings for gcc 4.4.4

[Bug c/61588] Optimization defaults are not what documentation say they should be

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61588

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mpolacek at gcc dot gnu.org
           Severity|critical                    |normal