From: Jason Gunthorpe <jgg@nvidia.com>
To: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>,
Niklas Schnelle <schnelle@linux.ibm.com>,
Leon Romanovsky <leon@kernel.org>, Arnd Bergmann <arnd@arndb.de>,
linux-arch@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
linux-rdma@vger.kernel.org, llvm@lists.linux.dev,
Michael Guralnik <michaelgur@mellanox.com>,
Nathan Chancellor <nathan@kernel.org>,
Nick Desaulniers <ndesaulniers@google.com>,
Will Deacon <will@kernel.org>, Marc Zyngier <maz@kernel.org>
Subject: Re: [PATCH rdma-next 1/2] arm64/io: add memcpy_toio_64
Date: Thu, 25 Jan 2024 13:43:33 -0400 [thread overview]
Message-ID: <20240125174333.GA2192844@nvidia.com> (raw)
In-Reply-To: <20240124192634.GJ1455070@nvidia.com>
On Wed, Jan 24, 2024 at 03:26:34PM -0400, Jason Gunthorpe wrote:
> The suggestion that it should not have any interleaving instructions
> and use STP came from our CPU architecture team.
I got some more details here.
They point to the ARM publication about write combining
https://community.arm.com/cfs-file/__key/telligent-evolution-components-attachments/13-150-00-00-00-00-10-12/Understanding_5F00_Write_5F00_Combining_5F00_on_5F00_Arm_5F00_V.1.0.pdf
specifically to the example code using 4x 128 bit NEON stores.
They point at the actual CPU design and say it is optimized for 128
bit stores (STP and ST4 included, it seems).
64 bit stores trigger some different behavior.
I have no way to know if it will be OK for other drivers that expect
this to be a performance path in the kernel.
Are you *sure* you want to do this str version? If it works for mlx5 I
will send the patch and the other companies can come later with
performance data.
Jason
next prev parent reply other threads:[~2024-01-25 17:43 UTC|newest]
Thread overview: 68+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-11-23 19:04 [PATCH rdma-next 0/2] Add and use memcpy_toio_64() Leon Romanovsky
2023-11-23 19:04 ` [PATCH rdma-next 1/2] arm64/io: add memcpy_toio_64 Leon Romanovsky
2023-11-24 10:16 ` Mark Rutland
2023-11-24 12:23 ` Jason Gunthorpe
2023-11-27 12:42 ` Catalin Marinas
2023-11-27 13:45 ` Jason Gunthorpe
2023-12-04 17:31 ` Catalin Marinas
2023-12-04 18:23 ` Jason Gunthorpe
2023-12-05 17:21 ` Catalin Marinas
2023-12-05 17:51 ` Jason Gunthorpe
2023-12-05 19:34 ` Catalin Marinas
2023-12-05 19:51 ` Jason Gunthorpe
2023-12-06 11:09 ` Catalin Marinas
2023-12-06 12:59 ` Jason Gunthorpe
2024-01-16 18:51 ` Jason Gunthorpe
2024-01-17 12:30 ` Mark Rutland
2024-01-17 12:36 ` Jason Gunthorpe
2024-01-17 12:41 ` Jason Gunthorpe
2024-01-17 13:29 ` Mark Rutland
2024-01-23 20:38 ` Catalin Marinas
2024-01-24 1:27 ` Jason Gunthorpe
2024-01-24 8:26 ` Marc Zyngier
2024-01-24 13:06 ` Jason Gunthorpe
2024-01-24 13:32 ` Marc Zyngier
2024-01-24 15:52 ` Jason Gunthorpe
2024-01-24 17:54 ` Catalin Marinas
2024-01-25 1:29 ` Jason Gunthorpe
2024-01-26 16:15 ` Catalin Marinas
2024-01-26 17:09 ` Jason Gunthorpe
2024-01-24 11:38 ` Mark Rutland
2024-01-24 12:40 ` Catalin Marinas
2024-01-24 13:27 ` Jason Gunthorpe
2024-01-24 17:22 ` Catalin Marinas
2024-01-24 19:26 ` Jason Gunthorpe
2024-01-25 17:43 ` Jason Gunthorpe [this message]
2024-01-26 14:56 ` Catalin Marinas
2024-01-26 15:24 ` Jason Gunthorpe
2024-01-17 14:07 ` Mark Rutland
2024-01-17 15:28 ` Jason Gunthorpe
2024-01-17 16:05 ` Will Deacon
2024-01-18 16:18 ` Jason Gunthorpe
2024-01-24 11:31 ` Mark Rutland
2023-11-24 12:58 ` Robin Murphy
2023-11-24 13:45 ` Jason Gunthorpe
2023-11-24 15:32 ` Robin Murphy
2023-11-24 14:10 ` Niklas Schnelle
2023-11-24 14:20 ` Jason Gunthorpe
2023-11-24 14:48 ` Niklas Schnelle
2023-11-24 14:53 ` Niklas Schnelle
2023-11-24 14:55 ` Jason Gunthorpe
2023-11-24 15:59 ` Niklas Schnelle
2023-11-24 16:06 ` Jason Gunthorpe
2023-11-27 17:43 ` Niklas Schnelle
2023-11-27 17:51 ` Jason Gunthorpe
2023-11-28 16:28 ` Niklas Schnelle
2024-01-16 17:33 ` Jason Gunthorpe
2024-01-17 13:20 ` Niklas Schnelle
2024-01-17 13:26 ` Jason Gunthorpe
2024-01-17 17:55 ` Jason Gunthorpe
2024-01-18 13:46 ` Niklas Schnelle
2024-01-18 14:00 ` Jason Gunthorpe
2024-01-18 15:59 ` Niklas Schnelle
2024-01-18 16:21 ` Jason Gunthorpe
2024-01-18 16:25 ` Niklas Schnelle
2024-01-19 11:52 ` Niklas Schnelle
2024-02-16 12:09 ` Niklas Schnelle
2024-02-16 12:39 ` Jason Gunthorpe
2023-11-23 19:04 ` [PATCH rdma-next 2/2] IB/mlx5: Use memcpy_toio_64() for write combining stores Leon Romanovsky
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=20240125174333.GA2192844@nvidia.com \
--to=jgg@nvidia.com \
--cc=arnd@arndb.de \
--cc=catalin.marinas@arm.com \
--cc=leon@kernel.org \
--cc=linux-arch@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-rdma@vger.kernel.org \
--cc=llvm@lists.linux.dev \
--cc=mark.rutland@arm.com \
--cc=maz@kernel.org \
--cc=michaelgur@mellanox.com \
--cc=nathan@kernel.org \
--cc=ndesaulniers@google.com \
--cc=schnelle@linux.ibm.com \
--cc=will@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;
as well as URLs for NNTP newsgroup(s).