[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAK7LNARfoNg1wFi8oGar3bhzUXtNRMyuz0ksSh2hhzG_PdAMVg@mail.gmail.com>
Date: Fri, 11 Sep 2020 03:07:49 +0900
From: Masahiro Yamada <masahiroy@...nel.org>
To: Nick Desaulniers <ndesaulniers@...gle.com>
Cc: ian.bearman@...rosoft.com,
clang-built-linux <clang-built-linux@...glegroups.com>,
LKML <linux-kernel@...r.kernel.org>,
Sami Tolvanen <samitolvanen@...gle.com>
Subject: Re: [PATCH v2 00/28] Add support for Clang LTO
On Fri, Sep 11, 2020 at 2:46 AM 'Nick Desaulniers' via Clang Built
Linux <clang-built-linux@...glegroups.com> wrote:
>
> Hey Ian,
> It was nice to meet you at Linux plumbers. You might want to take a
> look at this series. It implements builds of the Linux kernel with
> LTO. It would be good to get eyes on it and help review it from folks
> working on this from the GCC angle. The series has some configs that
> split where Clang specific changes need to be made; it might be of
> interest to think about what would the similar changes be needed for
> GCC. Also, congrats on your LWN article!
> https://lwn.net/Articles/830300/
FYI.
Some people may remember GCC LTO was posted in 2014
(and rejected by Linus).
https://lore.kernel.org/linux-kbuild/20140407201919.GA15838@sepie.suse.cz/
In my understanding, GCC LTO does not work for relocatable objects
because GNU linker is agnostic about LTO.
GCC LTO was implemented in the final link stage (i.e. vmlinux)
by using $(CC) as a linker driver.
On the other hand, Clang LTO, in this series implementation,
works in the pre-mospost stage (i.e. vmlinux.o).
--
Best Regards
Masahiro Yamada
Powered by blists - more mailing lists