]> git.ipfire.org Git - thirdparty/haproxy.git/commit
projects / thirdparty / haproxy.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: eca1f90) | patch
OPTIM: ring: always relax in the ring lock and leader wait loop
authorWilly Tarreau <w@1wt.eu>
Thu, 18 Sep 2025 13:01:29 +0000 (15:01 +0200)
committerWilly Tarreau <w@1wt.eu>
Thu, 18 Sep 2025 13:07:56 +0000 (15:07 +0200)
commitd25099b359df8277e9296a1deaa7297b66e95ad0
treee4e6178ddfd02b4b977065c3d7dd1b5b7de53834tree | snapshot
parenteca1f90e16d793fccdf70c3259d3bcfb72c74bbecommit | diff
OPTIM: ring: always relax in the ring lock and leader wait loop

Tests have shown that AMD systems really need to use a cpu_relax()
in these two loops. The performance improves from 10.03 to 10.56M
messages per second (+5%) on a 128-thread system, without affecting
intel nor ARM, so let's do this.
src/ring.c diff | blob | blame | history
Mirror of https://github.com/haproxy/haproxy.git