Devicetree
 help / color / mirror / Atom feed
From: David Laight <david.laight.linux@gmail.com>
To: Doug Anderson <dianders@chromium.org>
Cc: Devicetree-compiler@vger.kernel.org,
	"Rob Herring" <robh@kernel.org>,
	"Krzysztof Kozlowski" <krzk+dt@kernel.org>,
	"Conor Dooley" <conor+dt@kernel.org>,
	"Konrad Dybcio" <konradybcio@kernel.org>,
	Devicetree@vger.kernel.org,
	"Tudor Ambarus" <tudor.ambarus@linaro.org>,
	"Chen-Yu Tsai" <wenst@chromium.org>,
	"Saravana Kannan" <saravanak@kernel.org>,
	"David Gibson" <david@gibson.dropbear.id.au>,
	"André Draszik" <andre.draszik@linaro.org>,
	"Simon Glass" <sjg@chromium.org>,
	linux-kernel@vger.kernel.org,
	"Brian Norris" <briannorris@chromium.org>,
	"Peter Griffin" <peter.griffin@linaro.org>,
	"Julius Werner" <jwerner@chromium.org>
Subject: Re: [PATCH v4 08/14] libfdt: Speed up fdt_find_string_len_() with memchr()
Date: Wed, 18 Nov 2026 22:15:38 +0000	[thread overview]
Message-ID: <20261118221538.4de223fd@pumpkin> (raw)
In-Reply-To: <CAD=FV=U+KX9-XMK0xAzeciAapjFecfTS_DjyVDMGwypVmEE7vg@mail.gmail.com>

On Wed, 18 Nov 2026 12:39:59 -0800
Doug Anderson <dianders@chromium.org> wrote:

> Hi David,
> 
> On Wed, Nov 18, 2026 at 11:15 AM David Laight
> <david.laight.linux@gmail.com> wrote:
> >  
> > > 2. Every stored string ends with `'\0'`, whereas `s[0]` only appears in
> > >    a smaller fraction of bytes. On a real upstream Qualcomm base DTB's
> > >    `dt_strings` block (855 strings in 11,259 bytes, average 13.2 bytes
> > >    per string), `'\0'` occurs on 7.6% of bytes while `s[0]` averages
> > >    3.9%. Replaying those 855 insertions makes 362k `memchr()` calls when
> > >    scanning for `'\0'` vs. 187k `memchr()` calls when scanning for
> > >    `s[0]`.  
> >
> > You need to count the number of times that a letter that is the first
> > character of a string you are going to lookup appears in the entire buffer.
> > There might only be one 'j' in the entire buffer - but you never look for it.
> > (But I might still be wrong)  
> 
> That 3.9% (`187,315` `memchr()` calls across ~4.77M scanned bytes) was
> actually measured directly by replaying those 855 real lookups (`s[0]`
> of each string being inserted), not an unweighted 1/26 average across
> the alphabet. Even though common prefixes like `'p'`, `'u'`, `'q'`, and
> `'s'` get looked up a lot, 26.6% of the `dt_strings` block (`2,139`
> bytes of `'_'`, `'-'`, `'0'-'9'`, and `'#'`, plus `855` NUL bytes) are
> characters that almost never appear as `s[0]`, which pulls the
> lookup-weighted hit rate back down to 3.9%.
> 
> Intuitively, `'\0'` hits once per string (`1 / L`), so `s[0]` wins as
> long as average string length `L` stays below the ~22-26 byte crossover
> -- and even on plain non-`-@` DTBs without `/__symbols__`, property
> names average ~12-14 bytes (`'\0'` still made 1.6x-2.2x more `memchr()`
> calls than `s[0]` across plain DTB `fdt_sw` builds, `/chosen` fixups,
> and `.dtbo` overlay application).
> 
> > Have you tried an open coded scans instead of calling memchr() and memcmp()?
> > With an average of 13 bytes per string (26 for your memchr()) the code loop
> > might be faster.
> > But beware, gcc is likely to convert the code back to a call to memcmp()
> > - it certainly does for strlen and memcpy.
> >
> > Oh - and if this code runs in kernel the timings will be significantly
> > different to glibc - it uses some magic avx instructions for strlen()
> > and memchr() is effectively the same.
> > The kernel functions are likely to be out of line byte loops.  
> 
> Out of curiosity I timed all of those across 200 replays of the same
> 855-string (11,259-byte) Qualcomm `dt_strings` block with `gcc 15.2` on
> x86_64 (`gcc -O2` kept the open-coded loops inline and didn't turn them
> back into `memcmp()`):

Once you are doing timings, adding a few more doesn't take long.

> * glibc `memchr` + `!p[slen]` + glibc `memcmp`:
>   0.25 s (`-O0`), 0.23 s (`-Os`), 0.22 s (`-O2`)
> * glibc `memchr` + `!p[slen]` + open-coded inner compare:
>   0.23 s (`-O0`), 0.21 s (`-Os`), 0.21 s (`-O2`)

So that one is a marginal win.

> * Fully open-coded (`p[0] == s[0] && !p[slen]` + inner byte loop):
>   0.91 s (`-O0`), 0.62 s (`-Os`), 0.40 s (`-O2`)
> * Out-of-line byte-loop `memchr` + `!p[slen]` + byte-loop `memcmp`:
>   0.86 s (`-O0`), 0.90 s (`-Os`), 0.50 s (`-O2`)
> * Original per-byte out-of-line byte-loop `memcmp`:
>   3.31 s (`-O0`), 1.87 s (`-Os`), 1.86 s (`-O2`)
> 
> Even with a ~25-byte average stride, glibc's AVX2 `memchr()` handles
> almost an entire stride in a single 32-byte vector load + compare, so it
> beats the open-coded scalar loop by ~1.9x at `-O2` (and ~2.7x at `-Os`).

Using avx2 is 'cheating' ...
Someone at Intel realised that strlen() does matter and threw some silicon at it.
They finally did the same for integer divide (but I think AMD did it first)
and did a divide that generates move than one bit per clock. 

> Open-coding just the inner compare barely moves the needle since
> `!p[slen]` already cut `memcmp()` down to 14.3k calls across the 855
> insertions.
> 
> On the kernel / bootloader side, `fdt_find_string_len_()` is only called
> by `fdt_sw` and `fdt_rw` when adding new property names to `dt_strings`
> -- normal read-only DTB parsing (`fdt_ro.c`) never calls it. And even
> with a naive out-of-line byte-loop `memchr()`, `memchr()` + `!p[slen]` +
> `memcmp()` is still 3.7x faster at `-O2` (`1.86 s -> 0.50 s`) and 2.1x
> faster at `-Os` (`1.87 s -> 0.90 s`) than the old code, so keeping the
> simpler `memchr()` + `!p[slen]` + `memcmp()` seems like the right
> tradeoff.

Seems reasonable.

David

> 
> -Doug


  reply	other threads:[~2026-11-18 22:15 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-16 20:52 [PATCH v4 00/14] fdtfactor: Introduce fdtfactor to auto-factor overlays Douglas Anderson
2026-11-16 20:52 ` [PATCH v4 01/14] README: Add overview blurbs for dtc and ancillary tools Douglas Anderson
2026-11-16 20:52 ` [PATCH v4 02/14] pylibfdt: Make Property a subclass of bytes instead of bytearray Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 03/14] pylibfdt: Trim Fdt bytearray to totalsize() in FdtSw.as_fdt() Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 04/14] pylibfdt: Add Property.from_*() creation methods Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 05/14] pylibfdt: Unpack Property.as_list() in a single struct.unpack() call Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 06/14] pylibfdt: Speed up FdtRo.properties() with a C helper Douglas Anderson
2026-11-17  3:47   ` David Gibson
2026-11-17 16:17     ` Doug Anderson
2026-11-16 20:53 ` [PATCH v4 07/14] libpydt: Add DeviceTree live tree representation Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 08/14] libfdt: Speed up fdt_find_string_len_() with memchr() Douglas Anderson
2026-11-17  4:00   ` David Gibson
2026-11-17  4:07     ` David Gibson
2026-11-17 16:21       ` Doug Anderson
2026-11-18  9:54   ` David Laight
2026-11-18 17:24     ` Doug Anderson
2026-11-18 19:15       ` David Laight
2026-11-18 20:39         ` Doug Anderson
2026-11-18 22:15           ` David Laight [this message]
2026-11-16 20:53 ` [PATCH v4 09/14] libfdt: overlay: Update base /__local_fixups__ on apply Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 10/14] fdtfactor: Add tool to split DTBs into a base DTB and overlays Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 11/14] fdtfactor: Add -S/--strip-local-fixups option Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 12/14] fdtfactor: Support compressed DTB inputs and outputs Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 13/14] fdtfactor: Add compiled FIT factoring support Douglas Anderson
2026-11-16 20:53 ` [PATCH v4 14/14] fdtfactor: Parallelize blob compression and decompression Douglas Anderson
2026-11-17  3:48 ` [PATCH v4 00/14] fdtfactor: Introduce fdtfactor to auto-factor overlays David Gibson
2026-11-17  4:01   ` David Gibson

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=20261118221538.4de223fd@pumpkin \
    --to=david.laight.linux@gmail.com \
    --cc=Devicetree-compiler@vger.kernel.org \
    --cc=Devicetree@vger.kernel.org \
    --cc=andre.draszik@linaro.org \
    --cc=briannorris@chromium.org \
    --cc=conor+dt@kernel.org \
    --cc=david@gibson.dropbear.id.au \
    --cc=dianders@chromium.org \
    --cc=jwerner@chromium.org \
    --cc=konradybcio@kernel.org \
    --cc=krzk+dt@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=peter.griffin@linaro.org \
    --cc=robh@kernel.org \
    --cc=saravanak@kernel.org \
    --cc=sjg@chromium.org \
    --cc=tudor.ambarus@linaro.org \
    --cc=wenst@chromium.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