[Buildroot] [git commit] Config.in: add -Ofast option

[Buildroot] [git commit] Config.in: add -Ofast option

[Thomas Petazzoni]

commit: https://git.buildroot.net/buildroot/commit/?id=ed6a7e18af8f4dfddee6e6b144f20dfaf6926315
branch: https://git.buildroot.net/buildroot/commit/?id=refs/heads/master

-Ofast (introduced in GCC 4.6) It combines the existing optimization level -O3
with options that can affect standards compliance but result in better optimized
code. For example, -Ofast enables -ffast-math.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
---
 Config.in           | 11 +++++++++++
 package/Makefile.in |  3 +++
 2 files changed, 14 insertions(+)

diff --git a/Config.in b/Config.in
index 0002df5176..037ea2265b 100644
--- a/Config.in
+++ b/Config.in
@@ -527,6 +527,17 @@ config BR2_OPTIMIZE_S
 	  -ftree-vect-loop-version
 	  This is the default.
 
+config BR2_OPTIMIZE_FAST
+	bool "optimize for fast"
+	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_6
+	help
+	  Optimize for fast. Disregard strict standards
+	  compliance. -Ofast enables all -O3 optimizations. It also
+	  enables optimizations that are not valid for all
+	  standard-compliant programs. It turns on -ffast-math and the
+	  Fortran-specific -fstack-arrays, unless -fmax-stack-var-size
+	  is specified, and -fno-protect-parens.
+
 endchoice
 
 config BR2_GOOGLE_BREAKPAD_ENABLE
diff --git a/package/Makefile.in b/package/Makefile.in
index e387ce67fe..828e12e7d7 100644
--- a/package/Makefile.in
+++ b/package/Makefile.in
@@ -128,6 +128,9 @@ endif
 ifeq ($(BR2_OPTIMIZE_S),y)
 TARGET_OPTIMIZATION = -Os
 endif
+ifeq ($(BR2_OPTIMIZE_FAST),y)
+TARGET_OPTIMIZATION = -Ofast
+endif
 ifeq ($(BR2_DEBUG_1),y)
 TARGET_DEBUGGING = -g1
 endif

[Buildroot] [git commit] Config.in: add -Ofast option

[Joshua Henderson]

Thomas, All, 

On 03/26/2018 01:19 PM, Thomas Petazzoni wrote:
> commit: https://git.buildroot.net/buildroot/commit/?id=ed6a7e18af8f4dfddee6e6b144f20dfaf6926315
> branch: https://git.buildroot.net/buildroot/commit/?id=refs/heads/master
> 
> -Ofast (introduced in GCC 4.6) It combines the existing optimization level -O3
> with options that can affect standards compliance but result in better optimized
> code. For example, -Ofast enables -ffast-math.
> 
> Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> ---
>  Config.in           | 11 +++++++++++
>  package/Makefile.in |  3 +++
>  2 files changed, 14 insertions(+)
> 
> diff --git a/Config.in b/Config.in
> index 0002df5176..037ea2265b 100644
> --- a/Config.in
> +++ b/Config.in
> @@ -527,6 +527,17 @@ config BR2_OPTIMIZE_S
>  	  -ftree-vect-loop-version
>  	  This is the default.
>  
> +config BR2_OPTIMIZE_FAST
> +	bool "optimize for fast"
> +	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_6
> +	help
> +	  Optimize for fast. Disregard strict standards
> +	  compliance. -Ofast enables all -O3 optimizations. It also
> +	  enables optimizations that are not valid for all
> +	  standard-compliant programs. It turns on -ffast-math and the
> +	  Fortran-specific -fstack-arrays, unless -fmax-stack-var-size
> +	  is specified, and -fno-protect-parens.
> +
>  endchoice
>  

...

For reference, I just found an unfriendly package with this option.

sqlite3.c: In function ?sqlite3IsNaN?:
sqlite3.c:28554:3: error: #error SQLite will not work correctly with the -ffast-math option of GCC.
 # error SQLite will not work correctly with the -ffast-math option of GCC.
   ^
Makefile:541: recipe for target 'sqlite3.lo' failed

sqlite3 refuses to be built with -ffast-math (a side effect of -Ofast) when it
falls back to implementing its own isnan() function.  From what I can see,
SQLITE_HAVE_ISNAN can be defined [1] to make it use the system library isnan().
But, how to tell if the system library supports a proper isnan() seems to open
up a whole new can of worms.  Thoughts?

[1]: https://www.sqlite.org/compile.html

Josh

[Buildroot] [git commit] Config.in: add -Ofast option

[Yann E. MORIN]

Joshua, All,

On 2018-04-05 14:46 -0700, Joshua Henderson spake thusly:
> On 03/26/2018 01:19 PM, Thomas Petazzoni wrote:
> > commit: https://git.buildroot.net/buildroot/commit/?id=ed6a7e18af8f4dfddee6e6b144f20dfaf6926315
> > branch: https://git.buildroot.net/buildroot/commit/?id=refs/heads/master
> > 
> > -Ofast (introduced in GCC 4.6) It combines the existing optimization level -O3
> > with options that can affect standards compliance but result in better optimized
> > code. For example, -Ofast enables -ffast-math.
> > 
> > Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
> > Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
> > ---
> >  Config.in           | 11 +++++++++++
> >  package/Makefile.in |  3 +++
> >  2 files changed, 14 insertions(+)
> > 
> > diff --git a/Config.in b/Config.in
> > index 0002df5176..037ea2265b 100644
> > --- a/Config.in
> > +++ b/Config.in
> > @@ -527,6 +527,17 @@ config BR2_OPTIMIZE_S
> >  	  -ftree-vect-loop-version
> >  	  This is the default.
> >  
> > +config BR2_OPTIMIZE_FAST
> > +	bool "optimize for fast"
> > +	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_6
> > +	help
> > +	  Optimize for fast. Disregard strict standards
> > +	  compliance. -Ofast enables all -O3 optimizations. It also
> > +	  enables optimizations that are not valid for all
> > +	  standard-compliant programs. It turns on -ffast-math and the
> > +	  Fortran-specific -fstack-arrays, unless -fmax-stack-var-size
> > +	  is specified, and -fno-protect-parens.
> > +
> >  endchoice
> For reference, I just found an unfriendly package with this option.
> sqlite3.c: In function ?sqlite3IsNaN?:
> sqlite3.c:28554:3: error: #error SQLite will not work correctly with the -ffast-math option of GCC.
>  # error SQLite will not work correctly with the -ffast-math option of GCC.
>    ^
> Makefile:541: recipe for target 'sqlite3.lo' failed
> 
> sqlite3 refuses to be built with -ffast-math (a side effect of -Ofast) when it
> falls back to implementing its own isnan() function.  From what I can see,
> SQLITE_HAVE_ISNAN can be defined [1] to make it use the system library isnan().
> But, how to tell if the system library supports a proper isnan() seems to open
> up a whole new can of worms.  Thoughts?

Assuming those packages are not too many, I'd say we go to overriding
the CFLAGS for those, with code something like:

    SQLITE_CFLAGS = $(subst -Ofast,-O3,$(TARGET_CFLAGS))

If it turns out that the number of broken packages is too high, then we
can come up with an alternate solution...

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] [git commit] Config.in: add -Ofast option

[Joshua Henderson]

Yann, All,

On 04/07/2018 07:37 AM, Yann E. MORIN wrote:
> Joshua, All,
> 
> On 2018-04-05 14:46 -0700, Joshua Henderson spake thusly:
>> On 03/26/2018 01:19 PM, Thomas Petazzoni wrote:
>>> commit: https://git.buildroot.net/buildroot/commit/?id=ed6a7e18af8f4dfddee6e6b144f20dfaf6926315
>>> branch: https://git.buildroot.net/buildroot/commit/?id=refs/heads/master
>>>
>>> -Ofast (introduced in GCC 4.6) It combines the existing optimization level -O3
>>> with options that can affect standards compliance but result in better optimized
>>> code. For example, -Ofast enables -ffast-math.
>>>
>>> Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
>>> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
>>> ---
>>>  Config.in           | 11 +++++++++++
>>>  package/Makefile.in |  3 +++
>>>  2 files changed, 14 insertions(+)
>>>
>>> diff --git a/Config.in b/Config.in
>>> index 0002df5176..037ea2265b 100644
>>> --- a/Config.in
>>> +++ b/Config.in
>>> @@ -527,6 +527,17 @@ config BR2_OPTIMIZE_S
>>>  	  -ftree-vect-loop-version
>>>  	  This is the default.
>>>  
>>> +config BR2_OPTIMIZE_FAST
>>> +	bool "optimize for fast"
>>> +	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_6
>>> +	help
>>> +	  Optimize for fast. Disregard strict standards
>>> +	  compliance. -Ofast enables all -O3 optimizations. It also
>>> +	  enables optimizations that are not valid for all
>>> +	  standard-compliant programs. It turns on -ffast-math and the
>>> +	  Fortran-specific -fstack-arrays, unless -fmax-stack-var-size
>>> +	  is specified, and -fno-protect-parens.
>>> +
>>>  endchoice
>> For reference, I just found an unfriendly package with this option.
>> sqlite3.c: In function ?sqlite3IsNaN?:
>> sqlite3.c:28554:3: error: #error SQLite will not work correctly with the -ffast-math option of GCC.
>>  # error SQLite will not work correctly with the -ffast-math option of GCC.
>>    ^
>> Makefile:541: recipe for target 'sqlite3.lo' failed
>>
>> sqlite3 refuses to be built with -ffast-math (a side effect of -Ofast) when it
>> falls back to implementing its own isnan() function.  From what I can see,
>> SQLITE_HAVE_ISNAN can be defined [1] to make it use the system library isnan().
>> But, how to tell if the system library supports a proper isnan() seems to open
>> up a whole new can of worms.  Thoughts?
> 
> Assuming those packages are not too many, I'd say we go to overriding
> the CFLAGS for those, with code something like:
> 
>     SQLITE_CFLAGS = $(subst -Ofast,-O3,$(TARGET_CFLAGS))

Something like this gets the job done.  The -Ofast still hangs around in other
*_ENV flags because it's still in TARGET_CFLAGS, but not where I can find it
matters.  Patch to follow.

> 
> If it turns out that the number of broken packages is too high, then we
> can come up with an alternate solution...

This is the only one I have found so far.  sqlite3 was diligent enough to make
this a build time issue.  Other packages may not.

Josh