- Mar 28, 2018
-
-
Boris Feld authored
-
- Jan 31, 2018
-
-
Philippe Pepiot authored
In asv development version goal_time is now sample_time: https://github.com/airspeed-velocity/asv/pull/575
-
Philippe Pepiot authored
In ASV setup is run for each repeat but benchmark function is run `number` of time within the same repetition (calling it a "sample"). So for push / pull benchmarks, number should be set to 1 (call setup before each call of the benchmark method), also setup isn't called during "warmup" time, so disable it. Set repeat to 20 (instead of the default 10) this should be enough to have stable results while keeping benchmark time reasonable. Also move benchmark params (number, timer) to class variable.
-
- Mar 02, 2018
-
-
Boris Feld authored
-
- Feb 20, 2018
-
-
Boris Feld authored
-
- Jan 24, 2018
-
-
Boris Feld authored
https://www.mercurial-scm.org/repo/hg/rev/33c8a683 was a big regression that was later put behind a config knob.
-
Philippe Pepiot authored
-
Boris Feld authored
Add https://www.mercurial-scm.org/repo/hg/rev/b5613bda for bookmarks, it mainly affects bookmarks-free repositories.
-
- Jan 23, 2018
-
-
Philippe Pepiot authored
We don't need bundles backups
-
Philippe Pepiot authored
Using --pull to avoid hardlinks (most frequent use case). Only benchmark mercurial-2017 repo for now, others are too big and will timeout.
-
Philippe Pepiot authored
-
- Jan 19, 2018
-
-
Philippe Pepiot authored
expect a return code of 1 for incoming, outgoing and push in this case.
-
Philippe Pepiot authored
Practically this change nothing
-
- Jan 17, 2018
-
-
Philippe Pepiot authored
Implement a new base benchmark class that is shared with incoming/outgoing benchmarks. push benchmark use rsync with hardlinks to "reset" the clone of the partial clone at each repeat. Use a timeout of 300 which is enough for me on mozilla repo.
-
Philippe Pepiot authored
We may want use multiple partial clones in the future, set an unique name for a given partial clone to partial-$repo_name-$stripped_revset(b64 encoded)
-
Philippe Pepiot authored
Use prepare_repos.py to create partial benchmarks which are cloned, then stripped. This use debugupdatecahce which require hg >= 4.3 (we use the system installed hg for that). Partial clones are then used in incoming/outgoing benchmark which can then be switched to "time" benchmark allowing ASV to control the number of loop etc. Since we're using subprocesses use timeit.default_timer as timer.
-
Philippe Pepiot authored
By merging check_repositories() in clone_repositories
-
- Jan 15, 2018
-
-
Boris Feld authored
The test makes the IO/WAIT go up to 50% on the benchmark machine, we need to find another way like maybe snapshotting the repository.
-
Boris Feld authored
The way of cloning repositories when there is some differences is not realist as the resulting repositories will contains only one head.
-
- Jan 11, 2018
-
-
Boris Feld authored
It would be easier to update them this way
-
Boris Feld authored
-
Boris Feld authored
The clone test don't seems to finish at all, comment it to not impact the whole suite.
-
Boris Feld authored
Also extract clone test in their own class to avoid cloning several times
-
- Jan 10, 2018
-
-
Boris Feld authored
-
- Jan 04, 2018
-
-
Boris Feld authored
-
- Dec 14, 2017
-
-
Philippe Pepiot authored
This work by creating local partial clones and executing commands between the clone and the main repo. The variant 'single_rev' is used to compare behavior of commands when a particular revision is given or not using the -r option. Please take care that by default cloned repo are created in a temporary directory inside /tmp, to have a clone using hardlinks (which is faster) repositories must be on the same partition. Default temporary base directory (/tmp) could be overrwrite with the ASVTEMPDIR environment variable.
-
Philippe Pepiot authored
If the first measurement is > 30, the function returned median([]) which return 0.0. Include the first measurement in timings to avoid such case.
-
- Dec 13, 2017
-
-
Philippe Pepiot authored
When our benchmark is a single hg execute but with pre/post setup it's convenient to wrap call to _single_execute().
-
Philippe Pepiot authored
This avoid reading local hg config that may interfer with benchmarks results. Also we don't need the alias trick anymore when running hg commands.
-
- Nov 27, 2017
-
-
Boris Feld authored
Extract the common variants in a separate class. Also forces the benchmarks to run in a clean environment.
-
Boris Feld authored
-
Boris Feld authored
-
Boris Feld authored
Extract common methods of basic commands benchmarks to reduce code duplication and ease modification.
-
- Nov 01, 2017
-
-
Boris Feld authored
-
- Oct 25, 2017
-
-
Boris Feld authored
-
Boris Feld authored
-
Boris Feld authored
-
- Oct 24, 2017
-
-
Boris Feld authored
-
Boris Feld authored
-
- Oct 10, 2017
-
-
Boris Feld authored
We are using perf in launch.sh and might soon use it in benchmarks.
-