jeirmeister/Cockatrice
1
0
Fork 0
mirror of https://github.com/Cockatrice/Cockatrice.git synced 2026-11-11 20:43:25 -08:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

[CI] Keep the measured ccache bounds instead of raising them

Browse source 
Steady-state master shows Arch at 90.1% and Debian 13 at 96.75% per-run
hits on the existing 850M/800M bounds. The 75%/82% that motivated the
increase were warming runs right after the -g1 switch invalidated the
cache, not thrashing. Keep the bounds, keep the macOS 14 cache drop.
This commit is contained in:
Lukas Brübach 2026-11-07 10:12:45 +01:00
parent 08eb791ad7
commit 9b542cd360
1 changed files with 22 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

43
.github/workflows/desktop-build.yml vendored
View file

@ -150,7 +150,7 @@ jobs:
allow-failure: yes allow-failure: yes
package: skip # We are packaged in Arch already package: skip # We are packaged in Arch already
debug_test: yes # see the note on Fedora 45 below debug_test: yes # see the note on Fedora 45 below
ccache_size: 1200M # Debug plus the suite: 680 MB working set at -g1; 850M thrashed ccache_size: 850M # Debug plus the suite: 680 MB working set at -g1
- distro: Servatrice_Debian - distro: Servatrice_Debian
version: 12 version: 12
@ -165,7 +165,7 @@ jobs:
version: 13 version: 13
package: DEB package: DEB
ccache_size: 1050M # Release plus the suite: 625 MB working set at -g1; 800M thrashed ccache_size: 800M # Release plus the suite: 625 MB working set at -g1
# The single Fedora leg used to build Debug for the suite and Release for the # The single Fedora leg used to build Debug for the suite and Release for the
# package in one job (measured on Fedora 44: 30.8 min, the only leg slower than # package in one job (measured on Fedora 44: 30.8 min, the only leg slower than
@ -225,33 +225,34 @@ jobs:
# whole configuration therefore buys almost nothing, and the quota goes to fewer legs # whole configuration therefore buys almost nothing, and the quota goes to fewer legs
# at a size that holds one: # at a size that holds one:
# #
# Arch Debug + suite 680 MB working set 1200M # Arch Debug + suite 680 MB working set 850M
# Debian 13 Release + suite 625 MB 1050M # Debian 13 Release + suite 625 MB 800M
# Fedora 45 debug Debug + suite 660 MB 850M # Fedora 45 debug Debug + suite 660 MB 850M
# Fedora 45 release Release + package 614 MB 800M # Fedora 45 release Release + package 614 MB 800M
# #
# That is about 3.9 GB of bounds, and with the fixed entries above (about 4.4 GB) it # That is 3.3 GB of bounds, and with the fixed entries above (about 4.4 GB) it puts
# puts the quota at roughly 8.3 GB of 10 - so only legs worth their share get one. A # the quota at roughly 7.7 GB of 10 - so only legs worth their share get one. A leg
# leg keeps no cache when it duplicates another leg's coverage (Ubuntu 26.04 behind # keeps no cache when it duplicates another leg's coverage (Ubuntu 26.04 behind
# Debian 13's Release + suite; macOS 15 and macOS 27 Debug behind macOS 14) or # Debian 13's Release + suite; the macOS legs behind the Linux ones) or finishes well
# finishes well inside the Windows leg (Servatrice covers the server alone in 4.7 # inside the Windows leg (Servatrice covers the server alone in 4.7 min). A macOS
# min). All three macOS legs run cold: macOS 14's 800M cache served 42% and saved # cache served 42% and saved about 32s on a 20 min leg that is already shorter than
# about 32s on a 20 min leg that is already shorter than Windows, so that bound is # Windows, so that bound is better spent here. The four Linux legs that stay compile
# better spent here. The four Linux legs that stay compile the client on the # the client on the toolchains a change can break and are the ones a cache shortens
# toolchains a change can break and are the ones a cache shortens by the most. # by the most.
# Measured job totals on master 2026-11-04 also show the ceiling this leaves: # Measured job totals on master 2026-11-04 also show the ceiling this leaves:
# Windows 10 is 29.5 min and compiles for 23.7 of it with no compiler cache at all, so # Windows 10 is 29.5 min and compiles for 23.7 of it with no compiler cache at all, so
# Windows still sets the pull-request wall clock no matter how warm these legs are # Windows still sets the pull-request wall clock no matter how warm these legs are
# (see build-vcpkg). # (see build-vcpkg).
# #
# Measured on master 2026-11-06: Fedora 45 debug/release hit 99.5% at 80% full, while # Steady-state master data corrected this on 2026-11-07: Arch 90.1% and Debian 13
# Arch served 75% at 96% full and Debian 13 82% at 96% - both evict entries they need. # 96.75% per-run hits at the 850M/800M bounds, Fedora 45 debug 99.18% and release
# A cold 3G run on 2026-11-07 measured one configuration at 680 MB (Arch) and 625 MB # 99.2%. The 75%/82% seen the day before were warming runs - the -g1 switch had
# (Debian 13), so the old 850M/800M bounds already exceeded the working set. They # invalidated the cache one commit earlier - not thrashing at the bound. A cold 3G
# thrashed because commit-to-commit churn filled the ~170 MB of slack and evicted live # run on the same day measured one configuration at 680 MB (Arch) and 625 MB
# entries. The bounds above leave 520/425 MB for that churn, so the working set stays # (Debian 13), so those bounds already hold the working set with ~170 MB of slack for
# resident across compiler and image updates. The sanitize leg's 1.2G cache is # commit-to-commit churn, and raising them buys hit rate the numbers do not show a
# nightly-only now, and macOS 14's 800M is dropped; that is what funds the increase. # need for. macOS 14's 800M entry is dropped because persistence buys 32s there; the
# quota stays free for the legs whose bounds would actually miss.
# #
# Re-add a leg only with a bound that holds its whole configuration: a smaller one is # Re-add a leg only with a bound that holds its whole configuration: a smaller one is
# worse than none. # worse than none.