On Tue, 11 Mar 2014, Richard Biener wrote:
> On Sat, 8 Mar 2014, Gerald Pfeifer wrote:
>
> > Thanks for the time and diligence writing this up, Richi!
> >
> > On Thu, 6 Mar 2014, Richard Biener wrote:
> > > -files; if @option{-flto} is not passed to the linker, no
> > > -interprocedural optimizations are applied.
> > > +files; if @option{-fno-lto} is not passed to the linker, no
> > > +interprocedural optimizations are applied.
> >
> > That looks like one "no" too much?
>
> Fixed.
>
> > > Note that when
> > > +@option{-fno-fat-lto-objects} is enabled the compile-stage is faster
> > > +but you cannot perform a regular, non-LTO link, on them.
> >
> > The comma past "link" appears too much.
>
> Fixed.
>
> > > Additionally, the optimization flags used to compile individual files
> > > are not necessarily related to those used at link time. For instance,
> >
> > That requires -ffat-lto-objects, though? The text above talks more
> > about -fno-fat-lto-objects, not the positive form.
>
> Doesn't require, no. Unfortunately the default depends on some
> configure checks ... so the positive form below is required on
> some systems to make the -fno-lto link work.
>
> > > @smallexample
> > > -gcc -c -O0 -flto foo.c
> > > -gcc -c -O0 -flto bar.c
> > > -gcc -o myprog -flto -O3 foo.o bar.o
> > > +gcc -c -O0 -ffat-lto-objects -flto foo.c
> > > +gcc -c -O0 -ffat-lto-objects -flto bar.c
> > > +gcc -o myprog -O3 foo.o bar.o
> > > @end smallexample
> > >
> > > This produces individual object files with unoptimized assembler
> > > code, but the resulting binary @file{myprog} is optimized at
> > > -@option{-O3}. If, instead, the final binary is generated without
> > > -@option{-flto}, then @file{myprog} is not optimized.
> > > +@option{-O3}. If, instead, the final binary is generated with
> > > +@option{-fno-lto}, then @file{myprog} is not optimized.
> >
> > Would it make sense to use -Os in the example? I assume in the
> > last case myprog would then by optimized with -Os?
>
> You mean -Os instead of -O0?
>
> > I am suggesting this since I believe it's not optimization vs
> > no optimization but "optimization level provided during compilation"?
>
> Yes. But we were motivating the -O0 vs. -On case with fat objects
> because you can get a debug build quickly with -fno-lto and
> an optimized build otherwise (without the need to re-compile).
> Not sure if that matters in practice ... but that's what the example
> tries to tell you how to do that.
>
> [I've merely edited existing parts to reflect reality in 4.9
> due to changed defaults - the whole section should be rewritten
> to be more in a FAQ-like way. That is, "You want to do X? Here is
> now to do it!"]
>
> > > +Currently, the following options and their setting are take from
> > > +the first object file that explicitely specified it:
> > > +@option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
> > > +@option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
> > > +and all the @option{-m} target flags.
> >
> > No -O options in case none are provided during link time?
>
> See below, "If you do not specify an optimization level option ...".
> I've moved this to the very top.
>
> > > +Certain ABI changing flags are required to match in all compilation-units
> > > +and trying to override this at link-time with a conflicting value
> > > +is ignored. This includes options such as @option{-freg-struct-return}
> > > +and @option{-fpcc-struct-return}.
> >
> > If they are required to match, shouldn't a conflicting value during
> > link time trigger a diagnoses -- error or at least warning?
>
> Yes, but unfortunately all diagnoses from link-time are buffered
> by collect2 and thus emitted very late. So we don't emit any
> but fatal diagnostics from lto-wrapper.
>
> > > +Other options such as @option{-ffp-contract},
> > > @option{-fno-strict-overflow},
> > > +@option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
> > > +are passed through to the link stage and merged conservatively for
> > > +conflicting translation units. You can override them at linke-time.
> >
> > What does conservative merging imply? How does that work?
>
> I've added
>
> "Specifically
> @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv}
> take
> precedence and for example @option{-ffp-contract=off} takes precedence
> over @option{-ffp-contract=fast}. You can override them at linke-time."
>
>
> > > +same link with the same options and also specify those options at
> > > +link-time.
> >
> > "link time" (noun)
>
> Fixed.
>
> > > -GCC will not work with an older/newer version of GCC@.
> > > +GCC will not work with an older/newer version of GCC.
> >
> > What is a version here? Release series?
> >
> > Will GCC 4.9.0 and 4.9.1 work, or not?
>
> We make no guarantees ;) Specifically the implemented
> bytecode version check is not strong enough :/
>
> Updated patch below.
>
> Ok?
I have committed this now after no further comments.
Richard.
> Thanks,
> Richard.
>
> 2014-03-11 Richard Biener <rguent...@suse.de>
>
> * doc/invoke.texi (flto): Update for changes in 4.9.
>
> Index: gcc/doc/invoke.texi
> ===================================================================
> --- gcc/doc/invoke.texi (revision 208478)
> +++ gcc/doc/invoke.texi (working copy)
> @@ -8524,8 +8524,9 @@ file. When the object files are linked
> bodies are read from these ELF sections and instantiated as if they
> had been part of the same translation unit.
>
> -To use the link-time optimizer, @option{-flto} needs to be specified at
> -compile time and during the final link. For example:
> +To use the link-time optimizer, @option{-flto} and optimization
> +options should be specified at compile time and during the final link.
> +For example:
>
> @smallexample
> gcc -c -O2 -flto foo.c
> @@ -8555,8 +8556,15 @@ merges them together into a single GIMPL
> them as usual to produce @file{myprog}.
>
> The only important thing to keep in mind is that to enable link-time
> -optimizations the @option{-flto} flag needs to be passed to both the
> -compile and the link commands.
> +optimizations you need to use the GCC driver to perform the link-step.
> +GCC then automatically performs link-time optimization if any of the
> +objects involved were compiled with the @option{-flto}. You generally
> +should specify the optimization options to be used for link-time
> +optimization though GCC will try to be clever at guessing an
> +optimization level to use from the options used at compile-time
> +if you fail to specify one at link-time. You can always override
> +the automatic decision to do link-time optimization at link-time
> +by passing @option{-fno-lto} to the link command.
>
> To make whole program optimization effective, it is necessary to make
> certain whole program assumptions. The compiler needs to know
> @@ -8568,28 +8576,31 @@ the linker plugin is not available, @opt
> used to allow the compiler to make these assumptions, which leads
> to more aggressive optimization decisions.
>
> -Note that when a file is compiled with @option{-flto}, the generated
> -object file is larger than a regular object file because it
> -contains GIMPLE bytecodes and the usual final code. This means that
> +When @option{-fuse-linker-plugin} is not enabled then, when a file is
> +compiled with @option{-flto}, the generated object file is larger than
> +a regular object file because it contains GIMPLE bytecodes and the usual
> +final code (see @option{-ffat-lto-objects}. This means that
> object files with LTO information can be linked as normal object
> -files; if @option{-flto} is not passed to the linker, no
> -interprocedural optimizations are applied.
> +files; if @option{-fno-lto} is passed to the linker, no
> +interprocedural optimizations are applied. Note that when
> +@option{-fno-fat-lto-objects} is enabled the compile-stage is faster
> +but you cannot perform a regular, non-LTO link on them.
>
> Additionally, the optimization flags used to compile individual files
> are not necessarily related to those used at link time. For instance,
>
> @smallexample
> -gcc -c -O0 -flto foo.c
> -gcc -c -O0 -flto bar.c
> -gcc -o myprog -flto -O3 foo.o bar.o
> +gcc -c -O0 -ffat-lto-objects -flto foo.c
> +gcc -c -O0 -ffat-lto-objects -flto bar.c
> +gcc -o myprog -O3 foo.o bar.o
> @end smallexample
>
> This produces individual object files with unoptimized assembler
> code, but the resulting binary @file{myprog} is optimized at
> -@option{-O3}. If, instead, the final binary is generated without
> -@option{-flto}, then @file{myprog} is not optimized.
> +@option{-O3}. If, instead, the final binary is generated with
> +@option{-fno-lto}, then @file{myprog} is not optimized.
>
> -When producing the final binary with @option{-flto}, GCC only
> +When producing the final binary, GCC only
> applies link-time optimizations to those files that contain bytecode.
> Therefore, you can mix and match object files and libraries with
> GIMPLE bytecodes and final object code. GCC automatically selects
> @@ -8598,28 +8609,45 @@ further processing.
>
> There are some code generation flags preserved by GCC when
> generating bytecodes, as they need to be used during the final link
> -stage. Currently, the following options are saved into the GIMPLE
> -bytecode files: @option{-fPIC}, @option{-fcommon} and all the
> -@option{-m} target flags.
> -
> -At link time, these options are read in and reapplied. Note that the
> -current implementation makes no attempt to recognize conflicting
> -values for these options. If different files have conflicting option
> -values (e.g., one file is compiled with @option{-fPIC} and another
> -isn't), the compiler simply uses the last value read from the
> -bytecode files. It is recommended, then, that you compile all the files
> -participating in the same link with the same options.
> +stage. Generally options specified at link-time override those
> +specified at compile-time.
> +
> +If you do not specify an optimization level option @option{-O} at
> +link-time then GCC will compute one based on the optimization levels
> +used when compiling the object files. The highest optimization
> +level will win here.
> +
> +Currently, the following options and their setting are take from
> +the first object file that explicitely specified it:
> +@option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
> +@option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
> +and all the @option{-m} target flags.
> +
> +Certain ABI changing flags are required to match in all compilation-units
> +and trying to override this at link-time with a conflicting value
> +is ignored. This includes options such as @option{-freg-struct-return}
> +and @option{-fpcc-struct-return}.
> +
> +Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
> +@option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
> +are passed through to the link stage and merged conservatively for
> +conflicting translation units. Specifically
> +@option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
> +precedence and for example @option{-ffp-contract=off} takes precedence
> +over @option{-ffp-contract=fast}. You can override them at linke-time.
> +
> +It is recommended that you compile all the files participating in the
> +same link with the same options and also specify those options at
> +link time.
>
> If LTO encounters objects with C linkage declared with incompatible
> types in separate translation units to be linked together (undefined
> behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
> -issued. The behavior is still undefined at run time.
> +issued. The behavior is still undefined at run time. Similar
> +diagnostics may be raised for other languages.
>
> Another feature of LTO is that it is possible to apply interprocedural
> -optimizations on files written in different languages. This requires
> -support in the language front end. Currently, the C, C++ and
> -Fortran front ends are capable of emitting GIMPLE bytecodes, so
> -something like this should work:
> +optimizations on files written in different languages:
>
> @smallexample
> gcc -c -flto foo.c
> @@ -8632,8 +8660,7 @@ Notice that the final link is done with
> runtime libraries and @option{-lgfortran} is added to get the Fortran
> runtime libraries. In general, when mixing languages in LTO mode, you
> should use the same link command options as when mixing languages in a
> -regular (non-LTO) compilation; all you need to add is @option{-flto} to
> -all the compile and link commands.
> +regular (non-LTO) compilation.
>
> If object files containing GIMPLE bytecode are stored in a library archive,
> say
> @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
> @@ -8665,11 +8692,11 @@ The current implementation of LTO makes
> attempt to generate bytecode that is portable between different
> types of hosts. The bytecode files are versioned and there is a
> strict version check, so bytecode files generated in one version of
> -GCC will not work with an older/newer version of GCC@.
> +GCC will not work with an older or newer version of GCC.
>
> Link-time optimization does not work well with generation of debugging
> information. Combining @option{-flto} with
> -@option{-g} is currently experimental and expected to produce wrong
> +@option{-g} is currently experimental and expected to produce unexpected
> results.
>
> If you specify the optional @var{n}, the optimization and code
> @@ -8685,8 +8712,6 @@ You must prepend a @samp{+} to the comma
> for this to work. This option likely only works if @env{MAKE} is
> GNU make.
>
> -This option is disabled by default.
> -
> @item -flto-partition=@var{alg}
> @opindex flto-partition
> Specify the partitioning algorithm used by the link-time optimizer.
>
