[PATCH] Use the LTO linker plugin by default

[Jan Hubicka]

> On Thu, 6 Mar 2014, Jan Hubicka wrote:
> 
> > > On Tue, 4 Mar 2014, Jan Hubicka wrote:
> > > 
> > > > > 
> > > > > The following patch addresses the common (?) issue of people
> > > > > omitting -flto from the linker command-line which gets more
> > > > > severe with GCC 4.9 where slim LTO objects are emitted by
> > > > > default.  The patch simply _always_ arranges for the linker
> > > > > plugin to be used, so if there are any (slim) LTO objects
> > > > > on the link LTO will be done on them.  Similarly the
> > > > > non-linker-plugin path in collect2 is adjusted.
> > > > > 
> > > > > You can still disable this by specifying -fno-lto on the 
> > > > > linker command-line.
> > > > > 
> > > > > One side-effect of enabling the linker-plugin by default
> > > > > (for HAVE_LTO_PLUGIN == 2) is that collect2 will then
> > > > > use the configured plugin ld rather than the default ld.
> > > > > Not sure if that is desired.
> > > > > 
> > > > > Comments?
> > > > 
> > > > I like it; it was on my TODO list, but I was only worried about
> > > > --with-plugin-ld and did not find time, yet, to look into the consequences.
> > > > These days, I do not think we need to worry much aboud --with-plugin-ld.
> > > 
> > > Yeah, I think we should eventually remove that capability.
> > > 
> > > Now as of the two patches (compute a default link-time optimization
> > > level and this patch, make considering LTO at link-time the default),
> > > they only make sense together (at least the default opt level at
> > > link-time doesn't improve real-world situations without also considering
> > > all links to be possibly -flto).
> > > 
> > > So the question remains if we want to have both patches at this
> > > stage or if we want to wait with them for 4.10 (we do have the
> > > user-visible change of slim-lto objects by default with 4.9
> > > already).
> > 
> > I think it makes things easier, so I would like to see this in 4.9
> 
> Ok.  I'll push this to trunk now and then prepare a documentation
> update for LTO opts handling (as promised some time ago ...).
> 
> If any issues show up with these two patches then we'll revert
> and revisit this for 4.10.

Thanks,
I definitely saw issues without these patches even on projects that are kind of
LTO enabled, like libreoffice (whose --enable-lto does linking of couple 3rd
party libraries without -flto) and firefox (that links everything with -flto
but sometimes forget optimization flags).  So for a while I am using wrapper
script that adds -O3 -flto to gcc command line just to be sure setup is not
completely insane.  Having this done automatically by a wrapper is nice.

Honza
> 
> Thanks,
> Richard.