Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220511160319.1045812-1-mailhol.vincent@wanadoo.fr>
 <20220511160319.1045812-2-mailhol.vincent@wanadoo.fr> <324f245e-15e2-a1f2-bc8a-46255a1a859f@wanadoo.fr>
In-Reply-To: <324f245e-15e2-a1f2-bc8a-46255a1a859f@wanadoo.fr>
From:   Vincent MAILHOL <mailhol.vincent@wanadoo.fr>
Date:   Thu, 12 May 2022 08:30:45 +0900
Message-ID: <CAMZ6Rq+d_eKXQQZ1p-yUpqC-SOjpa83o_QpZci-GN9qd9UKa5g@mail.gmail.com>
Subject: Re: [PATCH v2 1/2] x86/asm/bitops: ffs: use __builtin_ffs to evaluate
 constant expressions
To:     Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc:     Nick Desaulniers <ndesaulniers@google.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>, x86@kernel.org,
        "H . Peter Anvin" <hpa@zytor.com>,
        Nathan Chancellor <nathan@kernel.org>,
        Tom Rix <trix@redhat.com>, linux-kernel@vger.kernel.org,
        llvm@lists.linux.dev, David Howells <dhowells@redhat.com>,
        Jan Beulich <JBeulich@suse.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Precedence: bulk

On Thu. 12 May 2022 at 05:56, Christophe JAILLET
<christophe.jaillet@wanadoo.fr> wrote:
> Le 11/05/2022 à 18:03, Vincent Mailhol a écrit :
> > For x86_64, the current ffs() implementation does not produce
> > optimized code when called with a constant expression. On the
> > contrary, the __builtin_ffs() function of both GCC and clang is able
> > to simplify the expression into a single instruction.
> >
>
> [...]
>
> >
> > ** Statistics **
> >
> > On a allyesconfig, before applying this patch...:
> >
> > | $ objdump -d vmlinux.o | grep bsf | wc -l
> > | 3607
> >
> > ...and after:
> >
> > | $ objdump -d vmlinux.o | grep bsf | wc -l
> > | 792
> >
> > So, roughly 26.7% of the call to ffs() were using constant expression
> > and were optimized out.
> >
> >
> nitpicking: numbers look odd.

And they are. Thanks for spotting the issue.

>     3607 is the exact same number as in patch 2/2. (ok, could be)
>     26.7% is surprising with these numbers. (I guess it is (total_before
> - remaining) / total_before x 100 = (3607-792)/36.07 = 78.0%)

The 3607 is incorrect (copy/paste issue, sorry). The correct figure is
1081. And:

(1081-792)/1081 = 26.7%

Will amend the comment and send v3 right away.

> (but patch looks great to me :)

Thanks! :)

> CJ