| *** philn has quit IRC | 01:40 | |
| *** tristan has quit IRC | 04:43 | |
| *** tristan has joined #buildstream | 05:13 | |
| *** ChanServ sets mode: +o tristan | 05:13 | |
| *** mohan43u has quit IRC | 06:49 | |
| *** mohan43u has joined #buildstream | 06:49 | |
| *** nanonyme has quit IRC | 06:52 | |
| *** tristan has quit IRC | 07:29 | |
| *** benschubert has joined #buildstream | 07:33 | |
| *** philn has joined #buildstream | 07:33 | |
| benschubert | douglaswinship: in BST 2.0 there will still be the git plugin yes. It will possibly live in the same repository as git_tag | 07:36 |
|---|---|---|
| *** tristan has joined #buildstream | 07:36 | |
| *** ChanServ sets mode: +o tristan | 07:36 | |
| tristan | benschubert, how do you launch benchmarks for multiple branches ? `tox -- --config normal --show-only branch/one branch/two other_commit_sha` like that ? | 07:37 |
| tristan | benchmark.py doesnt seem to parse the argv in a way that is familiar to me :-/ | 07:38 |
| benschubert | tristan: I don't remember if that works. it's possible. Usually I was only interested in a single branch with few commits | 07:38 |
| benschubert | one sec | 07:38 |
| tristan | Ah but, you were doing it for me last week | 07:38 |
| tristan | Just branch + commit would be equally fine | 07:38 |
| benschubert | I was running it twice and merging the results | 07:39 |
| tristan | Ah I see | 07:39 |
| benschubert | @click.argument("candidate") -> a single commit | 07:39 |
| benschubert | though a MR is welcome if you want to improve this :) | 07:39 |
| benschubert | there is also `--between` to run the benchmark between master and the tip of your branch | 07:39 |
| tristan | I see | 07:40 |
| tristan | tox -- --config normal --show-only tristan/variables-refactor <--- should work right ? | 07:40 |
| benschubert | yup | 07:41 |
| tristan | I'm getting https://gitlab.com/snippets/1996528 :-/ | 07:41 |
| benschubert | oh, with --between? | 07:42 |
| benschubert | I might have forgotten to push a patch | 07:42 |
| benschubert | https://gitlab.com/BuildStream/benchmarking/bst-benchmarks/-/blob/master/benchmark.py#L210 remove the `git_repo` param and use `cwd=BUILDSTREAM_PATH,` in the body | 07:42 |
| tristan | Should I use --between maybe ? | 07:42 |
| benschubert | wait no | 07:43 |
| benschubert | ah! | 07:43 |
| benschubert | you don't have `time` installed | 07:43 |
| tristan | What's that, should I add it to benchmark requirements.txt ? | 07:44 |
| benschubert | https://gitlab.com/BuildStream/benchmarking/bst-benchmarks/-/issues/1 | 07:44 |
| benschubert | apt-get install time? | 07:44 |
| benschubert | No, the `time` utility :) | 07:44 |
| tristan | I certainly have that | 07:45 |
| benschubert | Are you sure you don't have the shell builtin `time` instead? | 07:45 |
| tristan | when I type `type <something>` and it shows me the real/user/sys time ? | 07:45 |
| tristan | I probably do | 07:45 |
| tristan | I see, so I need to have a binary for it | 07:45 |
| tristan | ok | 07:45 |
| benschubert | we basically need GNU time | 07:46 |
| tristan | benschubert, does the benchmarks do the warmup rounds and multiple run averages by itself ? | 07:47 |
| benschubert | yep | 07:47 |
| tristan | Or I need to fiddle and collate reports | 07:47 |
| tristan | ok good :) | 07:47 |
| benschubert | it's specified by the `--config` | 07:47 |
| benschubert | if you look into configs/ you'll see multiple .json file | 07:47 |
| benschubert | they contain the information on what is going to be run | 07:48 |
| tristan | I'm rewriting _variables.pyx to have recursive/non-recursive paths | 07:49 |
| tristan | I think we can also reduce the recursion permissiveness with that too, currently it's set to 1000 recursions, but each recursion technically involves 2 stack depths | 07:50 |
| tristan | We can probably say that if a variable takes more than 200 indirections to resolve, just use the non-recursive variant | 07:52 |
| benschubert | that would seem sensible | 07:52 |
| benschubert | and 200 indirections is probably something that will be very rare | 07:52 |
| tristan | exactly, that is most likely already a circular reference error | 07:53 |
| benschubert | I like this idea | 07:53 |
| tristan | it will sadly be more code | 07:54 |
| tristan | But eh, have to move on from perfectionism into realism and work towards 2.0 items :) | 07:54 |
| tristan | I have an average 0.05 seconds of variance between runs it would seem | 07:58 |
| tristan | Anyway, the act of changing those toplevel for Variable methods instead, doesn't change performance (some runs it's faster, some runs slower) | 07:59 |
| *** tristan has quit IRC | 08:22 | |
| *** tristan has joined #buildstream | 08:30 | |
| *** ChanServ sets mode: +o tristan | 08:30 | |
| *** santi has joined #buildstream | 08:34 | |
| *** jjardon has quit IRC | 12:59 | |
| *** jjardon has joined #buildstream | 13:00 | |
| *** ChanServ sets mode: +o jjardon | 13:00 | |
| *** santi has quit IRC | 14:38 | |
| *** santi has joined #buildstream | 14:41 | |
| *** lchlan has quit IRC | 16:54 | |
| *** lchlan has joined #buildstream | 16:54 | |
| *** santi has quit IRC | 18:03 | |
| *** benschubert has quit IRC | 20:11 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!