Age | Commit message (Collapse) | Author |
|
... to give each at least 1 dedicated box.
Change-Id: I7699bdfd51a2438a9d1fb8db377495ebc2c72cae
|
|
Change-Id: I40f5e37fd6d8df5eac3fa65801ad1f8ff8081a8e
|
|
Binutils to binutils-2_33-branch,
LLVM to release/10.x, and
Linux to linux-5.4.y .
Change-Id: I8c46e2cedd7f8919733a8ac0a304674813326b09
|
|
Change-Id: I454e87afd1f6ea56f858dcc354c70584ab428997
|
|
Change-Id: I3f0e2f4e907a6c7033ef3a14c1fe43e5fd589cfd
|
|
Change-Id: I1df043e96f9ecdfc6c3d4628574ff4b3e3ce70ce
|
|
Change-Id: Ida4eecbce81cc69ea7ab3352f1c4c895b980091b
|
|
Change-Id: I77c5dd98ba9f3bf14f7aa8a3cc0f0bbfa9cc3feb
|
|
Change-Id: I988aa610c0b4ceddc8f81ecf8746db490b1f15f7
|
|
Change-Id: I74b595a565a8a01b1ac08acbf6b6f07fbc37a73e
|
|
... to allow scaling the number of benchmarking board types.
Right now we are blocking all tcwg_bmk-* jobs when bisecting, which
causes TX1 (aarch64) jobs to block on TK1 (armhf) regressions and
vice versa. This is bearable now, but with adding SynQuacers
(for SPEC CPU2017 benchmarkins) and/or X15 (for armhf) benchmarking
we will hit the bottleneck.
This also allows benchmarking of AArch32 code on ARMv8 boards.
Change-Id: I4fc9444f8c62d11baee14f364740ca36c7600f3d
|
|
... in a generic way. This replaces "benchmarks" parameter.
Change-Id: Ie7deff7a28df3862df6a3bf30201a79da008a9f4
|
|
Change-Id: Ia99b6e1c84d910a6eb08cdd4be4f3e20e89b0683
|
|
Change-Id: I6af3cc2cd8008d090354439e145777bf63ed5dbe
|
|
Move replay_log to jenkins-scripts, which is cleaned before the build
starts. This prevents us from using replay_log from a previous build.
Change-Id: Ibcc20d6966a5f45d973dda386b95a450c7a49c2f
|
|
Change-Id: Ife280b7e93fa5be31475e2faaccbdc870fdfc810
|
|
Change-Id: Iaa282dffc950c41c768ba3788132dad159534ee7
|
|
This parameter will be added to the automated CI loops after testing.
Change-Id: Icbbc667595521edba4f2671961da87468963ef12
|
|
Change-Id: I2be23e636e4ec794a74c99fe4a6e5819ed778502
|
|
... for now.
Change-Id: Id7e747166a1134dbb71cc18240b3d7f1592eddb6
|
|
Change-Id: Ic5925267329a96a02a92024686b3245782c3acf7
|
|
Assuming 10 steps 3h each we can easily exceed current timeout.
Change-Id: I569c410f190689ff12161269e122727ce4fa2ffa
|
|
Change-Id: Ic99ec051344cea650fc88433728cfe6f719d464e
|
|
... in recent toolchain releases with tcwg_bmk-build and tcwg_bmk-bisect
jobs.
Change-Id: Iaad1b61ffed571f061dd239e63f7d79d78bee568
|
|
Change-Id: If1fd9879dc4be8233740175de3492b94ab54ea9f
|
|
Change-Id: I9cc1874598a1a82c349fcc840661f357c4454d47
|
|
Change-Id: I4698d5e788e2b0abd21e505beb97d79262a8bfd6
|
|
Change-Id: Ib1ce3ffccc70daa32b15de101c2d84e10a0a17c2
|
|
Change-Id: I37cac881783aa8c5840ab45f16e5e5926e8b7d27
|
|
Change-Id: Ic02b0c8404c99c3be793a964336646844fd005bd
|
|
... in preparation for tcwg_bmk-build job definition.
Change-Id: I9667a1eb7bc6825d2288eafcb9a5f4eb3bb12708
|
|
We've seen sporadic failures where Jenkins couldn't get
git-us.linaro.org's certificate and we switched a subset of our jobs
to git.linaro.org instead. There was no such failure since then, so
apply the same change to all our jobs, since we just got that problem
with tcwg-cleanup-stale-workspaces.
Change-Id: Id6452208f93f4a3c3eed90a0b0c02bd9dc0dc0b9
|
|
Change from every 2 or 3 hours to 2 or 3 days.
Change-Id: I6fae9feb54a5b7fb87ed7600239d97a00c4708ff
|
|
Change-Id: I6d44cc5b02cd389ddeb44e5486667eb1aa594560
|
|
It appears that "H H H/1 * *" means once a month, not every day, as intended.
Change-Id: I192489fc4e8537682e0aefbbf54e7090a080b0a0
|
|
Master configurations are still disabled to give bandwidth to update
release branches.
Change-Id: Ic4f54a815bb2abdc71b43a327352d50249ff364a
|
|
Change-Id: I135b4970cbbbb9af2bc548bf1c84737ea5bcb720
|
|
Change-Id: Ia8ef02e1bce95ba857e051633db4c98ab20f1026
|
|
We have introduced an infrastructure bug in perfdatadir2csv.sh just
as we updated release branches for gcc, glibc and llvm, so we got
a bunch of failed builds. Disable builds and revert them back to
previous release versions. Then cleanup/reset baselines, and do
the release update again -- see subsequent commits.
Change-Id: I2418ee05d09bbfdf80e8858716ac2163e61a0830
|
|
Change-Id: Ic31dd26d4234dc962029abd65bef1130ee3dc3f0
|
|
... while we are shaking out instabilities.
Change-Id: Ib6855956d28ac5a1507da66d3723d97cffe91d54
|
|
Change-Id: I2bbc4d25001a4f96044e0a60b6b557407b35d5ce
|
|
Change-Id: Ie9db55bf05e3ac02d6d94be2c531ce8e52f88edf
|
|
... now that csvs2table.py outputs symbols with 60s+ runtime.
Change-Id: Id9f45fcea6763f3d09cd99f9586441762c150137
|
|
... since they have been overloading dev-01 lately.
Change-Id: I31ad6162cc1b65376be9da6734b3f52d1cb27382
|
|
... until we improve regression heuristics.
Change-Id: Idbb7657a915a119ae3cbd1a0418d552238e78d6b
|
|
Change-Id: Iccd9f03d95e73a6535d29416d5b77b4c65bde92c
|
|
Turns out the perf workaround was necessary after all.
Change-Id: I8339239bbce16b88212a493a49fb9ce737f4a555
|
|
Change-Id: Ibd2a988b2201969f05f299bb1081848e466fa33c
|
|
Turns out we have been uploading wrong results (from run id "999"
instead of "1nnn") for a few weeks.
Change-Id: I60fd4d4dfe7c60e52ce2153ecb06cb0b18aefd91
|