From: Peter Zijlstra <peterz@infradead.org>
To: Uros Bizjak <ubizjak@gmail.com>
Cc: Mateusz Guzik <mjguzik@gmail.com>,
torvalds@linux-foundation.org, mingo@redhat.com, x86@kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] x86: prevent gcc from emitting rep movsq/stosq for inlined ops
Date: Fri, 6 Jun 2025 10:19:04 +0200 [thread overview]
Message-ID: <20250606081904.GK30486@noisy.programming.kicks-ass.net> (raw)
In-Reply-To: <CAFULd4Z4SJZErfV1AG_a-+SDd=gnXa88=suRAFN3RHB5M+=bFw@mail.gmail.com>
On Fri, Jun 06, 2025 at 09:27:07AM +0200, Uros Bizjak wrote:
> Now, to use function call instead of "rep movsb", you can
> use -mstringop-strategy=libcall instead of some objtool substitution
> (-O2 -mno-sse -mstringop-strategy=libcall):
>
> foo:
> subq $8, %rsp
> movl $160, %edx
> leaq 16(%rsp), %rsi
> call memcpy
> addq $8, %rsp
> ret
>
> Which IMO is what you are looking for.
Well, no, because I can't replace the 'call memcpy' with "cs;cs;cs;rep
movsb" on good machines.
We want the "rep movsb" / "rep stosb" interface so we can actually use
those instructions when they're good, but then fall back to an
out-of-line function (with the same calling convention as the
instruction) when they're not so very good.
next prev parent reply other threads:[~2025-06-06 8:19 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-06-05 16:47 Mateusz Guzik
2025-06-05 18:32 ` Linus Torvalds
2025-06-05 19:00 ` Peter Zijlstra
2025-06-06 6:13 ` Uros Bizjak
2025-06-06 6:27 ` Uros Bizjak
2025-06-06 7:20 ` Peter Zijlstra
2025-06-09 21:09 ` David Laight
2025-06-15 16:41 ` Uros Bizjak
2025-06-06 7:27 ` Uros Bizjak
2025-06-06 8:19 ` Peter Zijlstra [this message]
2025-06-08 20:51 ` David Laight
2025-06-09 6:04 ` Uros Bizjak
2025-06-09 16:38 ` Linus Torvalds
2025-06-09 19:25 ` Linus Torvalds
2025-06-09 19:04 ` David Laight
2025-06-06 15:36 ` Mateusz Guzik
2025-06-06 17:59 ` Linus Torvalds
2025-06-09 21:19 ` David Laight
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250606081904.GK30486@noisy.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=mjguzik@gmail.com \
--cc=torvalds@linux-foundation.org \
--cc=ubizjak@gmail.com \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®