[FFmpeg-devel] [PATCH] The definition of av_always_inline with --enable-small or --disable-optimizations
Martin Storsjö
martin
Tue Dec 14 23:56:29 CET 2010
On Tue, 14 Dec 2010, Michael Niedermayer wrote:
> On Tue, Dec 14, 2010 at 10:48:49PM +0200, Martin Storsj? wrote:
> > On Tue, 14 Dec 2010, Michael Niedermayer wrote:
> >
> > > On Tue, Dec 14, 2010 at 11:28:36AM +0200, Martin Storsj? wrote:
> > > > Hi,
> > > >
> > > > At the end of the configure script, we currently do this:
> > > >
> > > > if enabled small || disabled optimizations; then
> > > > echo "#undef av_always_inline" >> $TMPH
> > > > echo "#define av_always_inline" >> $TMPH
> > > > fi
> > > >
> > > > This causes a lot of warnings when building with either of these
> > > > configurations, warnings like this one:
> > > >
> > > > ./libavutil/x86/bswap.h:32: warning: ?av_bswap16? defined but not used
> > > >
> > > > These warnings can be silenced either by adding -Wno-unused-function to
> > > > the cflags, or by changing the definition of av_always_inline for this
> > > > case. Possible options are:
> > > >
> > > > - empty, as currently, spewing lots of warnings
> > > > - inline (as opposed to the __attribute__((always_inline) used otherwise),
> > > > but that won't avoid all inlineing even if the user requested
> > > > --disable-optimizations
> > >
> > > > - __attribute__((unused)), which seems to be supported since gcc 2.95
> > >
> > > This would suppress unused warnings incorrectly
> >
> > Only for builds with --disable-optimizations or for --enable-small.
> > Currently, in those build configurations, it's really hard to see the
> > relevant warnings since the console is totally spammed with the warnings
> > for all the defined but unused static av_always_inline functions.
> >
> > For all occurrances of av_always_inline functions in headers, I see no
> > issue with __attribute__((unused)).
> >
> > Would you prefer the version where av_always_inline is redefined as
> > "inline" for --enable-small builds (but as __attribute__((unused)) for
> > --disable-optimizations)?
>
> I would slightly prefer this unless it makes small bigger by a non trivial
> amount
This version is attached now.
It doesn't seem to make it bigger at all in a normal build with
--enable-small, probably since gcc doesn't choose to inline any of those
functions when -Os is specified.
// Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-configure-Avoid-warnings-if-av_always_inline-is-disa.patch
Type: text/x-diff
Size: 1708 bytes
Desc:
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20101215/59eb1f9f/attachment.patch>
More information about the ffmpeg-devel
mailing list