Received-SPF: pass (google.com: domain of linux-kernel+bounces-149926-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Thu, 18 Apr 2024 12:54:00 +0200
From: Borislav Petkov <bp@alien8.de>
To: Uros Bizjak <ubizjak@gmail.com>
CC: linux-kernel@vger.kernel.org, linux-tip-commits@vger.kernel.org,
 Ingo Molnar <mingo@kernel.org>,
 Linus Torvalds <torvalds@linux-foundation.org>,
 Waiman Long <longman@redhat.com>, x86@kernel.org
Subject: =?US-ASCII?Q?Re=3A_=5Btip=3A_locking/core=5D_locking/atomic/x?=
 =?US-ASCII?Q?86=3A_Introduce_arch=5Ftry=5Fcmpxchg64=5Flocal=28=29?=
User-Agent: K-9 Mail for Android
In-Reply-To: <CAFULd4b8kRgy0p3pKwheLRHx7rX+4RuY-45_Zt-CCj5sCpGjCw@mail.gmail.com>
References: <20240414161257.49145-1-ubizjak@gmail.com> <171312759954.10875.1385994404712358986.tip-bot2@tip-bot2> <20240415201119.GBZh2J57f3aouPE_JR@fat_crate.local> <CAFULd4aFSBHNxyVyVt9soPnJXDgiOu6qCCNMLoenFNXtk0W4wA@mail.gmail.com> <20240417184040.GEZiAXqKvw3uRcmTJL@fat_crate.local> <CAFULd4b8kRgy0p3pKwheLRHx7rX+4RuY-45_Zt-CCj5sCpGjCw@mail.gmail.com>
Message-ID: <F0416322-2D03-408E-BC77-E68F024DF3EF@alien8.de>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable

On April 17, 2024 9:21:29 PM GMT+02:00, Uros Bizjak <ubizjak@gmail=2Ecom> w=
rote:
>The above is perhaps a little unfortunate example taken from
>
>if (cmpxchg64(=2E=2E=2E))
>
>where the check is against zero=2E The compiler can optimize the check
>to a TEST insn in this particular case, but otherwise CMP will be
>emitted for different usages=2E Not a big difference, but a register has
>to be kept live across cmpxchg8b=2E
>
=2E=2E=2E
>
>There is one important issue=2E When a register (or two for double-word
>values) has to be kept live for a compare, the register pressure on
>32bit targets around cmpxchg8b goes through the roof, and when using
>the frame pointer (and maybe some fixed register, e=2Eg=2E PIC), the
>register allocator runs out of available registers=2E The number of
>spills around cmpxchg8b signals the troubles register allocator goes
>through to "fix" everything, so from the compiler PoV any relief is
>more than welcome here=2E Even in GCC internal libraries, we have had to
>take a special approach with this insn to avoid internal compiler
>errors=2E The kernel was quite lucky here ;)

That would've been a lot better reason to justify the change=2E I think yo=
u should put those things in the commit messages=2E

Thx=2E


--=20
Sent from a small device: formatting sucks and brevity is inevitable=2E