[Bug c/103438] New: Updated documentation for gcc Optimization command line options (sec 3.11)

[Bug c/103438] New: Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

            Bug ID: 103438
           Summary: Updated documentation for gcc Optimization command
                    line options (sec 3.11)
           Product: gcc
           Version: 11.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: nils.smeds at gmail dot com
  Target Milestone: ---

Created attachment 51880
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51880&action=edit
Patch to file gcc/doc/invoke.texi

I was confused about when options --fprefetch-loop-arrays and
--aggressive-loop-optimizations are active and found that these options are
always active (including at level -O0). So I took a pass through the
documentation page and made it reflect this.

I changed the wording for -Os as it to some extent contradicted the options
actually in effect at -Os

I also made minor edits to make chapter 3.11 more consistent in formatting, in
particular making sure that "Enabled/Disabled..." information at the end of
parameter sections is a separate paragraph.

Hope it will be useful

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
-fprefetch-loop-arrays is not enabled by default on all targets.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Thanks for the report, and the patch. After addressing Andrew's comment could
you please send a new patch to the gcc-patches mailing list for review? Patches
attached here tend to get missed or forgotten about.

https://gcc.gnu.org/contribute.html#patches

See also https://gcc.gnu.org/dco.html

Thanks again!

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Oh the "enabled" thing when doing --help is confusing in the case of
-fprefetch-loop-arrays:
fprefetch-loop-arrays
Common Var(flag_prefetch_loop_arrays) Init(-1) Optimization
Generate prefetch instructions, if available, for arrays in loops.


-1 in this case is auto.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #4 from Nils Smeds <nils.smeds at gmail dot com> ---
(In reply to Andrew Pinski from comment #1)
> -fprefetch-loop-arrays is not enabled by default on all targets.

$ gcc --version
gcc (GCC) 11.2.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

$ for lvl in O0 O1 O2 O3 Os; do gcc -Q --help=optimizer -$lvl hello.c >
level-$lvl-default-opt.txt  2>&1 ; done

$ grep fprefetch-loop-arrays level-O*-opt.txt
level-O0-default-opt.txt:  -fprefetch-loop-arrays               [enabled]
level-O1-default-opt.txt:  -fprefetch-loop-arrays               [enabled]
level-O2-default-opt.txt:  -fprefetch-loop-arrays               [enabled]
level-O3-default-opt.txt:  -fprefetch-loop-arrays               [enabled]
level-Os-default-opt.txt:  -fprefetch-loop-arrays               [enabled]

$ grep aggressive-loop level-O*-opt.txt
level-O0-default-opt.txt:  -faggressive-loop-optimizations      [enabled]
level-O1-default-opt.txt:  -faggressive-loop-optimizations      [enabled]
level-O2-default-opt.txt:  -faggressive-loop-optimizations      [enabled]
level-O3-default-opt.txt:  -faggressive-loop-optimizations      [enabled]
level-Os-default-opt.txt:  -faggressive-loop-optimizations      [enabled]

$ grep inline-functions   *txt
level-O0-default-opt.txt:  -finline-functions                   [disabled]
level-O0-default-opt.txt:  -finline-functions-called-once       [disabled]
level-O1-default-opt.txt:  -finline-functions                   [disabled]
level-O1-default-opt.txt:  -finline-functions-called-once       [enabled]
level-O2-default-opt.txt:  -finline-functions                   [enabled]
level-O2-default-opt.txt:  -finline-functions-called-once       [enabled]
level-O3-default-opt.txt:  -finline-functions                   [enabled]
level-O3-default-opt.txt:  -finline-functions-called-once       [enabled]
level-Os-default-opt.txt:  -finline-functions                   [enabled]
level-Os-default-opt.txt:  -finline-functions-called-once       [enabled]

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
I think --help is confused here, the flag is mostly set under target control
only and the -1 init should be removed in favor of opts_set.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-11-26
           Assignee|unassigned at gcc dot gnu.org      |marxin at gcc dot gnu.org

--- Comment #6 from Martin Liška <marxin at gcc dot gnu.org> ---
Lemme take a look.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #7 from Nils Smeds <nils.smeds at gmail dot com> ---
(In reply to Martin Liška from comment #6)
> Lemme take a look.

Great, thanks. 

What happens in real life inside the code is beyond me, I am mostly after that
the documentation should describe what is really happening. If --help is not
reflecting what is happening I think that is a bug that needs to be filed by
someone smarter than I  ;-)

I still am happy to provide patches to the documentation once --help is known
to reflect reality.

I will stay tuned to see what enables the options -fprefetch-loop-arrays and
-faggressive-loop-optimizations but it appears to be the target used and not
the level of Optimization applied? Where should I look in the code to find
this?

If so, I will modify the docs accordingly: "Enabled by default on targets that
support xxxx" (whatever xxxx may be).

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #8 from Nils Smeds <nils.smeds at gmail dot com> ---
(In reply to Nils Smeds from comment #7)
> (In reply to Martin Liška from comment #6)
> > Lemme take a look.
> 
> Great, thanks. 

prefetch-loop-arrays appears to be activated on arm,s390 and i386 at level -O3
and higher on platforms that support software prefetch (and in the case of i386
also have TARGET_SOFTWARE_PREFETCHING_BENEFICIAL defined). For aarch64 it
appears to get activated for optimization level
aarch64_tune_params.prefetch->default_opt_level and higher when
aarch64_tune_params.prefetch->default_opt_level is defined

See:
gcc/common.opt
gcc/config/arm/arm.c
gcc/config/i386/i386-options.c
gcc/config/aarch64/aarch64.c
gcc/config/s390/s390.c

So it does appear to be a bug in the code path for -Q --help

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #9 from Martin Liška <marxin at gcc dot gnu.org> ---
I've just sent a patch:
https://gcc.gnu.org/pipermail/gcc-patches/2021-November/585650.html

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #10 from Nils Smeds <nils.smeds at gmail dot com> ---
Is there a benefit to augment the -Q output as follows?

[tooladmin@login2302 build]$ diff -r -p a b
diff -r -p a/gcc/opts.c b/gcc/opts.c
*** a/gcc/opts.c        Wed Jul 28 08:55:07 2021
--- b/gcc/opts.c        Thu Dec  2 13:50:34 2021
*************** print_filtered_help (unsigned int includ
*** 1531,1536 ****
--- 1531,1538 ----
                    strcat (new_help, _("[enabled]"));
                  else if (ena == 0)
                    strcat (new_help, _("[disabled]"));
+           else
+                   strcat (new_help, _("[deferred]"))
                }
            }

------
This should lead to the following diff:

-  -fdelete-null-pointer-checks                 [enabled]
+  -fdelete-null-pointer-checks                 [deferred]
@@ -332 +332 @@
-  -fleading-underscore                 [enabled]
+  -fleading-underscore                 [deferred]
@@ -393 +393 @@
-  -fprefetch-loop-arrays               [enabled]
+  -fprefetch-loop-arrays               [deferred]
@@ -502 +502 @@
-  -fstrict-volatile-bitfields          [enabled]
+  -fstrict-volatile-bitfields          [deferred]
@@ -533 +533 @@
-  -ftree-loop-if-convert               [enabled]
+  -ftree-loop-if-convert               [deferred]

Indicating to the user that he/she should dig further if they are interested to
find out how/if the option is in effect.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #11 from Nils Smeds <nils.smeds at gmail dot com> ---
I also noticed the following with this patch (if I applied it correctly)

[tooladmin@login2302 gcc-11.2.0]$ gcc -O0 -Q --help=optimize t.c | grep agg
  -faggressive-loop-optimizations       [enabled]

[tooladmin@login2302 gcc-11.2.0]$ type gcc
gcc is hashed (/hpc/base/ctt/packages/compiler/gnu/11.2.0/bin/gcc)

I notice that you don't see -faggressive-loop-optimizations in your diff output
of the patch.

/Nils

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #12 from Martin Liška <marxin at gcc dot gnu.org> ---
> I notice that you don't see -faggressive-loop-optimizations in your diff
> output of the patch.

Correct. The option should be enabled at all levels:

faggressive-loop-optimizations
Common Var(flag_aggressive_loop_optimizations) Optimization Init(1)
Aggressively optimize loops using language constraints.

There's Init(1).

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #13 from Nils Smeds <nils.smeds at gmail dot com> ---
OK, then it makes sense

On Thu, Dec 2, 2021 at 3:02 PM marxin at gcc dot gnu.org <
gcc-bugzilla@gcc.gnu.org> wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438
>
> --- Comment #12 from Martin Liška <marxin at gcc dot gnu.org> ---
> > I notice that you don't see -faggressive-loop-optimizations in your diff
> > output of the patch.
>
> Correct. The option should be enabled at all levels:
>
> faggressive-loop-optimizations
> Common Var(flag_aggressive_loop_optimizations) Optimization Init(1)
> Aggressively optimize loops using language constraints.
>
> There's Init(1).
>
> --
> You are receiving this mail because:
> You reported the bug.
> You are on the CC list for the bug.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[nils.smeds at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #14 from Nils Smeds <nils.smeds at gmail dot com> ---
OK, the wording "aggressive" makes it look counter-intuitive that it is set
at -O0, but apparently it is correct.

/Nils

On Thu, Dec 2, 2021 at 3:02 PM marxin at gcc dot gnu.org <
gcc-bugzilla@gcc.gnu.org> wrote:
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438
>
> --- Comment #12 from Martin Liška <marxin at gcc dot gnu.org> ---
> > I notice that you don't see -faggressive-loop-optimizations in your diff
> > output of the patch.
>
> Correct. The option should be enabled at all levels:
>
> faggressive-loop-optimizations
> Common Var(flag_aggressive_loop_optimizations) Optimization Init(1)
> Aggressively optimize loops using language constraints.
>
> There's Init(1).
>
> --
> You are receiving this mail because:
> You reported the bug.
> You are on the CC list for the bug.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

--- Comment #15 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Martin Liska <marxin@gcc.gnu.org>:

https://gcc.gnu.org/g:8e836af61b7027c0819da62c12a8d18b7c46f3fc

commit r12-5824-g8e836af61b7027c0819da62c12a8d18b7c46f3fc
Author: Martin Liska <mliska@suse.cz>
Date:   Mon Nov 29 14:46:47 2021 +0100

    Fix --help -Q output

            PR middle-end/103438

    gcc/ChangeLog:

            * config/s390/s390.c (s390_valid_target_attribute_inner_p):
            Use new enum CLVC_INTEGER.
            * opt-functions.awk: Use new CLVC_INTEGER.
            * opts-common.c (set_option): Likewise.
            (option_enabled): Return -1,0,1 for CLVC_INTEGER.
            (get_option_state): Use new CLVC_INTEGER.
            (control_warning_option): Likewise.
            * opts.h (enum cl_var_type): Likewise.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

Martin Liška <marxin at gcc dot gnu.org> changed:

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

--- Comment #16 from Martin Liška <marxin at gcc dot gnu.org> ---
Fixed on master, not planning any backport.

[Bug middle-end/103438] Updated documentation for gcc Optimization command line options (sec 3.11)

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103438

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mcn at onet dot pl

--- Comment #17 from Martin Liška <marxin at gcc dot gnu.org> ---
*** Bug 106173 has been marked as a duplicate of this bug. ***