public inbox for qemu-rust@nongnu.org
 help / color / mirror / Atom feed
From: "Daniel P. Berrangé" <berrange@redhat.com>
To: qemu-devel@nongnu.org
Cc: "Markus Armbruster" <armbru@redhat.com>,
	"Richard Henderson" <richard.henderson@linaro.org>,
	"Philippe Mathieu-Daudé" <philmd@linaro.org>,
	devel@lists.libvirt.org, qemu-rust@nongnu.org,
	"Dr. David Alan Gilbert" <dave@treblig.org>,
	"Gerd Hoffmann" <kraxel@redhat.com>,
	"Christian Schoenebeck" <qemu_oss@crudebyte.com>,
	"Daniel P. Berrangé" <berrange@redhat.com>,
	"Marc-André Lureau" <marcandre.lureau@redhat.com>,
	"Manos Pitsidianakis" <manos.pitsidianakis@linaro.org>,
	"Eduardo Habkost" <eduardo@habkost.net>,
	"Pierrick Bouvier" <pierrick.bouvier@linaro.org>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Stefan Weil" <sw@weilnetz.de>, "Jon Kohler" <jon@nutanix.com>
Subject: [PULL 27/27] util/oslib-posix: increase memprealloc thread count to 32
Date: Thu,  5 Mar 2026 17:47:43 +0000	[thread overview]
Message-ID: <20260305174743.3084606-28-berrange@redhat.com> (raw)
In-Reply-To: <20260305174743.3084606-1-berrange@redhat.com>

From: Jon Kohler <jon@nutanix.com>

Increase MAX_MEM_PREALLOC_THREAD_COUNT from 16 to 32. This was last
touched in 2017 [1] and, since then, physical machine sizes and VMs
therein have continue to get even bigger, both on average and on the
extremes.

For very large VMs, using 16 threads to preallocate memory can be a
non-trivial bottleneck during VM start-up and migration. Increasing
this limit to 32 threads reduces the time taken for these operations.

Test results from quad socket Intel 8490H (4x 60 cores) show a fairly
linear gain of 50% with the 2x thread count increase.

---------------------------------------------
Idle Guest w/ 2M HugePages   | Start-up time
---------------------------------------------
240 vCPU, 7.5TB (16 threads) | 2m41.955s
---------------------------------------------
240 vCPU, 7.5TB (32 threads) | 1m19.404s
---------------------------------------------

Note: Going above 32 threads appears to have diminishing returns at
the point where the memory bandwidth and context switching costs
appear to be a limiting factor to linear scaling. For posterity, on
the same system as above:
- 32 threads: 1m19s
- 48 threads: 1m4s
- 64 threads: 59s
- 240 threads: 50s

Additional thread counts also get less interesting as the amount of
memory is to be preallocated is smaller. Putting that all together,
32 threads appears to be a sane number with a solid speedup on fairly
modern hardware. To go faster, we'd either need to improve the hardware
(CPU/memory) itself or improve clear_pages_*() on the kernel side to
be more efficient.

[1] 1e356fc14bea ("mem-prealloc: reduce large guest start-up and migration time.")

Signed-off-by: Jon Kohler <jon@nutanix.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
---
 util/oslib-posix.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/util/oslib-posix.c b/util/oslib-posix.c
index 3c14b72665..dc001da66d 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -61,7 +61,7 @@
 #include "qemu/memalign.h"
 #include "qemu/mmap-alloc.h"
 
-#define MAX_MEM_PREALLOC_THREAD_COUNT 16
+#define MAX_MEM_PREALLOC_THREAD_COUNT 32
 
 struct MemsetThread;
 
-- 
2.53.0



  parent reply	other threads:[~2026-03-05 17:53 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-05 17:47 [PULL v2 00/27] Misc patches queue Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 01/27] scripts: detect another GPL license boilerplate variant Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 02/27] io: separate freeing of tasks from marking them as complete Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 03/27] io: fix cleanup for TLS I/O source data on cancellation Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 04/27] io: fix cleanup for websock " Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 05/27] docs: simplify DiamondRapids CPU docs Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 06/27] qemu-options: remove extraneous [] around arg values Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 07/27] include: define constant for early constructor priority Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 08/27] monitor: initialize global data from a constructor Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 09/27] system: unconditionally enable thread naming Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 10/27] util: fix race setting thread name on Win32 Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 11/27] util: expose qemu_thread_set_name Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 12/27] audio: make jackaudio use qemu_thread_set_name Daniel P. Berrangé
2026-03-07 11:37   ` Philippe Mathieu-Daudé
2026-03-05 17:47 ` [PULL 13/27] util: set the name for the 'main' thread on Windows Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 14/27] util: add API to fetch the current thread name Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 15/27] util: introduce some API docs for logging APIs Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 16/27] util: avoid repeated prefix on incremental qemu_log calls Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 17/27] util/log: add missing error reporting in qemu_log_trylock_with_err Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 18/27] ui: add proper error reporting for password changes Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 19/27] ui: remove redundant use of error_printf_unless_qmp() Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 20/27] monitor: remove redundant error_[v]printf_unless_qmp Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 21/27] monitor: refactor error_vprintf() Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 22/27] monitor: move error_vprintf back to error-report.c Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 23/27] util: fix interleaving of error & trace output Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 24/27] util: don't skip error prefixes when QMP is active Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 25/27] util: fix interleaving of error prefixes Daniel P. Berrangé
2026-03-05 17:47 ` [PULL 26/27] scripts/checkpatch: Fix MAINTAINERS update warning with --terse Daniel P. Berrangé
2026-03-05 17:47 ` Daniel P. Berrangé [this message]
2026-03-06  9:49 ` [PULL v2 00/27] Misc patches queue Peter Maydell

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=20260305174743.3084606-28-berrange@redhat.com \
    --to=berrange@redhat.com \
    --cc=armbru@redhat.com \
    --cc=dave@treblig.org \
    --cc=devel@lists.libvirt.org \
    --cc=eduardo@habkost.net \
    --cc=jon@nutanix.com \
    --cc=kraxel@redhat.com \
    --cc=manos.pitsidianakis@linaro.org \
    --cc=marcandre.lureau@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=philmd@linaro.org \
    --cc=pierrick.bouvier@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=qemu-rust@nongnu.org \
    --cc=qemu_oss@crudebyte.com \
    --cc=richard.henderson@linaro.org \
    --cc=sw@weilnetz.de \
    /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