The gcc-in-cxx branch now completes bootstrap

The gcc-in-cxx branch now completes bootstrap

[Ian Lance Taylor]

I'm happy to report that the gcc-in-cxx branch can now bootstrap.  That
is, the code in gcc proper can now be compiled with a C++ compiler.

My plan going forward is as follows (when we are back in stage 1):

* For each difference between trunk and gcc-in-cxx:
  + Try to implement a -Wc++-compat warning which detects the change.
    - If it is possible, implement the warning, and make the changes to
      let gcc bootstrap with the warning.
    - If a warning is not possible for some reason, I will simply
      propose the change by itself.  I expect this will be a small
      subset of the changes, mostly related to the build system and to
      low-level configuration like ansidecl.h.
  + This process will eventually eliminate all differences between trunk
    and gcc-in-cxx, at which point gcc-in-cxx can be retired.

* Implement a configure option, --enable-c++-build or something like
  that, which builds gcc with a C++ compiler.
  + Start running regular builds with that option, to avoid any
    regressions in C++ buildability for cases for which there is no
    -Wc++-compat warning.

* Begin the lobbying process for changing the default value of the
  configure option.

I certainly welcome help on any of these steps.

Ian

Re: The gcc-in-cxx branch now completes bootstrap

[Kaveh R. GHAZI]

On Fri, 6 Mar 2009, Ian Lance Taylor wrote:

> I'm happy to report that the gcc-in-cxx branch can now bootstrap.  That
> is, the code in gcc proper can now be compiled with a C++ compiler.

Great work, thanks!

I'm curious whether there are any detectable differences in the resulting
compiler when built with g++ rather than gcc.  E.g. testsuite regressions,
changes in the speed or size of cc1, etc.  Also, is cc1 linked with
libstdc++.so ?  Stuff like that.

Would you please consider checking this?

		Thanks,
		--Kaveh
--
Kaveh R. Ghazi			ghazi@caip.rutgers.edu

Re: The gcc-in-cxx branch now completes bootstrap

[Laurent GUERBY]

On Fri, 2009-03-06 at 18:44 -0800, Ian Lance Taylor wrote:
> I'm happy to report that the gcc-in-cxx branch can now bootstrap.  That
> is, the code in gcc proper can now be compiled with a C++ compiler.

Hi, did you test with Ada enabled? There are some C files in the
Ada compiler and RTS.

Laurent

Re: The gcc-in-cxx branch now completes bootstrap

[Gabriel Dos Reis]

On Fri, Mar 6, 2009 at 8:44 PM, Ian Lance Taylor <iant@google.com> wrote:
> I'm happy to report that the gcc-in-cxx branch can now bootstrap.  That
> is, the code in gcc proper can now be compiled with a C++ compiler.
>
> My plan going forward is as follows (when we are back in stage 1):
>
> * For each difference between trunk and gcc-in-cxx:
>  + Try to implement a -Wc++-compat warning which detects the change.
>    - If it is possible, implement the warning, and make the changes to
>      let gcc bootstrap with the warning.
>    - If a warning is not possible for some reason, I will simply
>      propose the change by itself.  I expect this will be a small
>      subset of the changes, mostly related to the build system and to
>      low-level configuration like ansidecl.h.
>  + This process will eventually eliminate all differences between trunk
>    and gcc-in-cxx, at which point gcc-in-cxx can be retired.
>
> * Implement a configure option, --enable-c++-build or something like
>  that, which builds gcc with a C++ compiler.
>  + Start running regular builds with that option, to avoid any
>    regressions in C++ buildability for cases for which there is no
>    -Wc++-compat warning.
>
> * Begin the lobbying process for changing the default value of the
>  configure option.
>
> I certainly welcome help on any of these steps.
>
> Ian
>

Hi Ian,

  This is a really fantastic job you've done.
Please send copies of your messages to my address at TAMU for
patches to implement the need -Wc++-compat so that I can review
as quickly as possible.

-- Gaby

Re: The gcc-in-cxx branch now completes bootstrap

[Ian Lance Taylor]

"Kaveh R. GHAZI" <ghazi@caip.rutgers.edu> writes:

> I'm curious whether there are any detectable differences in the resulting
> compiler when built with g++ rather than gcc.  E.g. testsuite regressions,
> changes in the speed or size of cc1, etc.  Also, is cc1 linked with
> libstdc++.so ?  Stuff like that.
>
> Would you please consider checking this?

I haven't tested any of that stuff yet.  I will at some point, or of
course anybody else can checkout gcc-in-cxx branch.

On gcc-in-cxx branch cc1 is linked with libstdc++.so.

Ian

Re: The gcc-in-cxx branch now completes bootstrap

[Ian Lance Taylor]

Laurent GUERBY <laurent@guerby.net> writes:

> On Fri, 2009-03-06 at 18:44 -0800, Ian Lance Taylor wrote:
>> I'm happy to report that the gcc-in-cxx branch can now bootstrap.  That
>> is, the code in gcc proper can now be compiled with a C++ compiler.
>
> Hi, did you test with Ada enabled? There are some C files in the
> Ada compiler and RTS.

I did not test with Ada enabled.  That would have to be addressed during
the merge.

Ian

Re: The gcc-in-cxx branch now completes bootstrap

[Ben Elliston]

> I'm curious whether there are any detectable differences in the resulting
> compiler when built with g++ rather than gcc.  E.g. testsuite regressions,
> changes in the speed or size of cc1, etc.  Also, is cc1 linked with
> libstdc++.so ?  Stuff like that.

Also, is there any significant difference in bootstrap times?

Ben

Re: The gcc-in-cxx branch now completes bootstrap

[Ian Lance Taylor]

Ben Elliston <bje@au1.ibm.com> writes:

>> I'm curious whether there are any detectable differences in the resulting
>> compiler when built with g++ rather than gcc.  E.g. testsuite regressions,
>> changes in the speed or size of cc1, etc.  Also, is cc1 linked with
>> libstdc++.so ?  Stuff like that.
>
> Also, is there any significant difference in bootstrap times?

I haven't actually measured, but subjectively bootstrap does seem to
take longer.

I encourage anybody interested in these sorts of questions to check out
the branch and find out.  gcc-in-cxx is something I do entirely in my
spare time, and it's not like I have more spare time than anybody else
does.

Ian

Re: The gcc-in-cxx branch now completes bootstrap

[Ben Elliston]

On Mon, 2009-03-09 at 21:25 -0700, Ian Lance Taylor wrote:

> > Also, is there any significant difference in bootstrap times?
> 
> I haven't actually measured, but subjectively bootstrap does seem to
> take longer.

A subjective assessment was all I was interested in.

Thanks,
Ben

Re: The gcc-in-cxx branch now completes bootstrap

[Thomas Neumann]

>> Also, is there any significant difference in bootstrap times?
> 
> I haven't actually measured, but subjectively bootstrap does seem to
> take longer.
I tried this out of curiosity. The numbers below are the bootstrap times on 
a 64bit 2.6.28 Linux system (Core 2 E8400), building single threaded with   
--enable-languages=c,c++ --disable-multilib

Regular gcc:

real    59m6.914s
user    53m53.702s
sys     3m24.073s

gcc-in-cxx:

real    68m15.366s
user    61m32.255s
sys     4m24.481s

Of course the bootstrap times are not that useful themselves, as they 
compare two quite different compilation tasks (one C, one C++). To get a 
better idea about the different compiler speeds, I compiled some random 
(reasonably complex) C++ code I had at hand, and compared the compile times.

Regular gcc:

real    0m30.478s
user    0m27.842s
sys     0m1.888s

gcc-in-cxx:

real    0m35.926s
user    0m34.386s
sys     0m1.208s

Again the comparison is not 100% fair ("regular gcc" is current mainline, 
while the gcc-in-cxx branch is older), but apparently the C++ version is 
quite a bit slower.

Admittedly gcc-in-cxx just recently managed to bootstrap at all, so perhaps 
performance comparisons are a bit unfair. But I do not mean this as critique 
of gcc-in-cxx, I want to help improve it and to bring it to the same speed 
as regular mainline.
Is there any reasonably simple way to find out why the C++ version is 
slower? I can use something like oprofile, of course, but I thought gcc can 
somehow give statistics about its internal times, which might be more useful 
for a first approximation.

Thomas

Re: The gcc-in-cxx branch now completes bootstrap

[Segher Boessenkool]

> Is there any reasonably simple way to find out why the C++ version is
> slower? I can use something like oprofile, of course, but I thought  
> gcc can
> somehow give statistics about its internal times, which might be  
> more useful
> for a first approximation.

I think you're thinking about the -Q option?


Segher

Re: The gcc-in-cxx branch now completes bootstrap

[Ben Elliston]

On Sat, 2009-04-11 at 22:37 +0200, Thomas Neumann wrote:

> Is there any reasonably simple way to find out why the C++ version is 
> slower? I can use something like oprofile, of course, but I thought
> gcc can somehow give statistics about its internal times, which might
> be more useful for a first approximation.

Try using -ftime-report.

Ben

Re: The gcc-in-cxx branch now completes bootstrap

[Thomas Neumann]

Ben Elliston wrote:
> Try using -ftime-report.
thanks, that was what I had in mind.

The largest difference seems to be in "tree STMT verifier" (36% runtime 
increase), a few others increased slightly, most seem to be nearly 
identical. (This distribution could be an artifact of my example code, of 
course).

I will have to take a closer look to find the exact cause of the slow-down.

Thomas

Re: The gcc-in-cxx branch now completes bootstrap

[Thomas Neumann]

Curious. I ran both g++ variants in oprofile, and then compared the 
generated assembler code for the most critical functions.

The top 1 function in both cases is pointer_set_insert, and there the 
assembler code is 100% identical (module one choice between r14 and r15).

The second most critical function in the gcc-in-cxx build is walk_tree_1, 
which is only place 4 in mainline gcc.
There the code seems to be identical, too, except for code layout: The 
compiler arranges the code in a different order, and apparently has 
different a different branch prediction. The non-branching code is nearly 
identical, too.
The "hottest" assembler instructions in walk_tree_1 are memory accesses, 
apparently the mainline version causes slightly less cache misses or better 
prediction? (my interpretation, not measured yet)

I am a bit unsure how to proceed. The gcc-in-cxx assembler code looks ok, as 
it is nearly identical to the mainline code. The main differences are in the 
code/branch layout, and I wouldn't know how to debug this.

Thomas

Re: The gcc-in-cxx branch now completes bootstrap

[Richard Guenther]

On Sun, Apr 12, 2009 at 9:56 AM, Thomas Neumann
<tneumann@users.sourceforge.net> wrote:
> Curious. I ran both g++ variants in oprofile, and then compared the
> generated assembler code for the most critical functions.
>
> The top 1 function in both cases is pointer_set_insert, and there the
> assembler code is 100% identical (module one choice between r14 and r15).
>
> The second most critical function in the gcc-in-cxx build is walk_tree_1,
> which is only place 4 in mainline gcc.
> There the code seems to be identical, too, except for code layout: The
> compiler arranges the code in a different order, and apparently has
> different a different branch prediction. The non-branching code is nearly
> identical, too.
> The "hottest" assembler instructions in walk_tree_1 are memory accesses,
> apparently the mainline version causes slightly less cache misses or better
> prediction? (my interpretation, not measured yet)
>
> I am a bit unsure how to proceed. The gcc-in-cxx assembler code looks ok, as
> it is nearly identical to the mainline code. The main differences are in the
> code/branch layout, and I wouldn't know how to debug this.

I suggest to compare against trunk at the latest merge point for
gcc-in-cxx (which was likely before the alias-improvements branch merge
which brought quite some changes in the tree verifiers and improved
compile-time overall).

Or just ask Ian to merge again from trunk ...

Richard.

> Thomas
>
>
>
>

Re: The gcc-in-cxx branch now completes bootstrap

[Mark Mitchell]

Ian Lance Taylor wrote:

> My plan going forward is as follows (when we are back in stage 1):

FWIW, I think this is a great plan.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713