From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
        id S1754031AbdBHRyF (ORCPT <rfc822;w@1wt.eu>);
        Wed, 8 Feb 2017 12:54:05 -0500
Received: from Galois.linutronix.de ([146.0.238.70]:54458 "EHLO
        Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1752480AbdBHRxq (ORCPT
        <rfc822;linux-kernel@vger.kernel.org>);
        Wed, 8 Feb 2017 12:53:46 -0500
Date: Wed, 8 Feb 2017 18:52:53 +0100 (CET)
From: Thomas Gleixner <tglx@linutronix.de>
To: Andy Lutomirski <luto@amacapital.net>
cc: Vitaly Kuznetsov <vkuznets@redhat.com>, X86 ML <x86@kernel.org>,
        Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
        "K. Y. Srinivasan" <kys@microsoft.com>,
        Haiyang Zhang <haiyangz@microsoft.com>,
        Stephen Hemminger <sthemmin@microsoft.com>,
        Dexuan Cui <decui@microsoft.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        devel@linuxdriverproject.org,
        Linux Virtualization <virtualization@lists.linux-foundation.org>
Subject: Re: [PATCH RFC 2/2] x86/vdso: Add VCLOCK_HVCLOCK vDSO clock read
 method
In-Reply-To: <CALCETrU48en_Zki88eXZEcU7QFf0u9sVAT_qgY0TyzoWkP2-6g@mail.gmail.com>
Message-ID: <alpine.DEB.2.20.1702081849050.3536@nanos>
References: <20170208170744.7632-1-vkuznets@redhat.com> <20170208170744.7632-3-vkuznets@redhat.com> <CALCETrU48en_Zki88eXZEcU7QFf0u9sVAT_qgY0TyzoWkP2-6g@mail.gmail.com>
User-Agent: Alpine 2.20 (DEB 67 2015-01-07)
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, 8 Feb 2017, Andy Lutomirski wrote:
> > +#ifdef CONFIG_HYPERV_CLOCK
> > +/* (a * b) >> 64 implementation */
> > +static inline u64 mul64x64_hi(u64 a, u64 b)
> > +{
> > +       u64 a_lo, a_hi, b_lo, b_hi, p1, p2;
> > +
> > +       a_lo = (u32)a;
> > +       a_hi = a >> 32;
> > +       b_lo = (u32)b;
> > +       b_hi = b >> 32;
> > +       p1 = a_lo * b_hi;
> > +       p2 = a_hi * b_lo;
> > +
> > +       return a_hi * b_hi + (p1 >> 32) + (p2 >> 32) +
> > +               ((((a_lo * b_lo) >> 32) + (u32)p1 + (u32)p2) >> 32);
> > +
> > +}
> 
> Unless GCC is waaay more clever than I think, this is hugely
> suboptimal on 64-bit.  x86 can do this in a single instruction, and
> gcc can express it cleanly using __uint128_t.  I wouldn't be terribly
> surprised if the 32-bit generated code was fine, too.

We already have that: mul_u64_u64_shr() which can be replaced by an arch
specific implementation. Nobody bothered to do that for x86 yet, but we
definitely don't want to open code it another time

Thanks,

	tglx