Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200708054354.25283-1-maochenxi@eswin.com> <mhng-8695bbb1-2ef2-409f-a458-3eb68d87495f@palmerdabbelt-glaptop1>
In-Reply-To: <mhng-8695bbb1-2ef2-409f-a458-3eb68d87495f@palmerdabbelt-glaptop1>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Mon, 20 Jul 2020 09:23:28 +0300
Message-ID: <CAMj1kXEAwuuR_tENNQ-_fTFHkh7g+hiH38SM=yYmoOu6ipU3tw@mail.gmail.com>
Subject: Re: [PATCH 1/1] riscv: Enable compiler optimizations
To:     Palmer Dabbelt <palmer@dabbelt.com>
Cc:     maochenxi@eswin.com, Paul Walmsley <paul.walmsley@sifive.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        linux-riscv <linux-riscv@lists.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        chenxi.mao2013@gmail.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, 19 Jul 2020 at 07:45, Palmer Dabbelt <palmer@dabbelt.com> wrote:
>
> On Tue, 07 Jul 2020 22:43:54 PDT (-0700), maochenxi@eswin.com wrote:
> > Enable ARCH_HAS_FAST_MULTIPLIER and ARCH_SUPPORTS_INT128 for better
> > code generation.
> > These 2 configurations works fine on GCC-9.3 and GCC-10.1
> >
> > Signed-off-by: Chenxi Mao <maochenxi@eswin.com>
> > ---
> >  arch/riscv/Kconfig | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> > index 128192e14ff2..f21b7e5188ca 100644
> > --- a/arch/riscv/Kconfig
> > +++ b/arch/riscv/Kconfig
> > @@ -26,6 +26,8 @@ config RISCV
> >       select ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT if MMU
> >       select ARCH_WANT_FRAME_POINTERS
> >       select ARCH_WANT_HUGE_PMD_SHARE if 64BIT
> > +     select ARCH_HAS_FAST_MULTIPLIER
> > +     select ARCH_SUPPORTS_INT128 if GCC_VERSION >= 50000 || CC_IS_CLANG
>
> IIRC int128 is only supported on 64-bit targets, and we already have this under
> RV64I but with a different set of requirements.  It looks like everyone has a
> different set of requirements, but I feel like CC_HAS_INT128 is the right way
> to go.  I'm not sure why we have the version check there.  Ard: do you know why
> we would need both?
>

That looks like copy/paste from the arm64 tree to me, where the names
of the libgcc routines were changed at some point, and we only support
the newer ones in the kernel.

> The fast multiplier one looks fine to me.  They're supposed to stay
> alphabetical, though.
>
> >       select CLONE_BACKWARDS
> >       select COMMON_CLK
> >       select EDAC_SUPPORT