DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 4AEBE3858CDA
MIME-Version: 1.0
References: <CANH4o6Mxai4_C=d1P93Prrimb8_H=trTwm-Eg+WBwpomN3tNJw@mail.gmail.com>
 <ZgwFNde2z804koS_@calimero.vinschen.de>
 <CANH4o6P8cts9TJgpdjR4mi+sj2YvuDa=d49XLcEVvYnRB81KRw@mail.gmail.com>
In-Reply-To: <CANH4o6P8cts9TJgpdjR4mi+sj2YvuDa=d49XLcEVvYnRB81KRw@mail.gmail.com>
Date: Wed, 3 Apr 2024 07:21:00 +0200
Message-ID: <CALXu0UdzRZ-7e2Vc=z1ABKYPK8X8xg6uea41Up3VZqJZ3w=Pzg@mail.gmail.com>
Subject: Re: Cygwin&Win32 file prefetch, block sizes?
To: cygwin@cygwin.com
From: Cedric Blancher via Cygwin <cygwin@cygwin.com>
Reply-To: Cedric Blancher <cedric.blancher@gmail.com>
Content-Type: text/plain; charset="utf-8"
Sender: "Cygwin" <cygwin-bounces+archive-cygwin=delorie.com@cygwin.com>
Content-Transfer-Encoding: 8bit

On Wed, 3 Apr 2024 at 00:36, Martin Wege via Cygwin <cygwin@cygwin.com> wrote:
>
> On Tue, Apr 2, 2024 at 3:17 PM Corinna Vinschen via Cygwin
> <cygwin@cygwin.com> wrote:
> >
> > On Apr  2 02:04, Martin Wege via Cygwin wrote:
> > > Hello,
> > >
> > > Is there any document which describes how Cygwin and Win32 file
> > > prefetch and readahead work, and which sizes are used (e.g. always
> > > read one full page even if only 16 bytes are requested?)?
> >
> > I'm not aware of any docs, but again, keep in mind that Cygwin is a
> > usersapce DLL. We basically do what Windows does for low-level file
> > access.
> >
> > > Quick /usr/bin/stat /etc/profile returns "IO Block: 65536". Does that
> > > mean the file's block size is really 64k? Is this info per filesystem,
> > > or hardcoded in Cygwin?
> >
> > Hardcoded in Cygwin since 2017, based on a discussion in terms of
> > file access performance, especially when using stdio.h functions:
> >
> >   https://cygwin.com/cgit/newlib-cygwin/commit/?id=7bef7db5ccd9c
>
> OUCH.
>
> While I can understand the motivation, FAT32 on multi-GB-devices
> having 64k block size, and Win32 API on Win95/98/ME/Win7 being
> optimized to that insane block size, it is absolutely WRONG with
> today's NTFS and even more so with ReFS. This only works if you stream
> files, but as soon as you are doing random read/writes the performance
> is terrible due to cache thrashing. That could explain the many
> complaints about Cygwin's IO performance.
>

It can also explain why Cygwin is so slow on SMB filesystems. If it
really works on 64k blocks, then this would pull all small files over
the wire, even if only a bit gets touched.

Thinking more about this, this basically defeats every
driver/tcp/ip/ethernet/net/switch optimization like jumbo frames et
al.
/usr/bin/stat %o (optimum IO block size hint) also returns 65536 on
SMB, which is NOT GOOD:

This needs to be confirmed, and if this is really true, then it should
be fixed for SMB.

I need a coffee to think about this...

> So, what can be done? I'm not a benchmarking guru, so I'd like to
> propose to add a tunable called EXPERIMENTAL_PREFERRED_IO_BLKSIZE to
> the CYGWIN env variable (marked as "experimental"), so the
> benchmarking guys can do performance testing without recompiling
> everything, get perf results for Cygwin 3.6, and decide what to do for
> Cygwin 3.7.

I would agree, but I would also clamp the minimum at page size (4096
bytes on x86) and 8M (4x 2M hugepage size) to prevent abuse.

Does Cygwin partake in the GSOC programm (Google Summer Of Code)? This
would IMO be a high priority item.

Ced
-- 
Cedric Blancher <cedric.blancher@gmail.com>
[https://plus.google.com/u/0/+CedricBlancher/]
Institute Pasteur

-- 
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple