mehere | Can you be a bit more specific? Tests ran smoothly here... | 00:51 |
---|---|---|
prologic | ahh | 00:52 |
prologic | you're on 2.7 right? | 00:52 |
mehere | No, 2.6.7. I need it for this environment, cannot upgrade in my embedded application... | 01:00 |
prologic | ahh k | 01:01 |
prologic | hmm I'll get back to you then in that case | 01:01 |
prologic | many random tests are failing at tip | 01:01 |
prologic | bear with me :0 | 01:01 |
mehere | Mmh, after running the test several times it looks like I have one of the udp tests failing one out of five times (but that's the only one). Looking into it. | 01:14 |
mehere | I'm unsure whether that pytest.wait_for really works. If you look e.g. at test_udp.py:111, isn't it possible that the (concurrently) running manager has fired the "ready" *before* we start to wait for it? | 01:20 |
mehere | Ah, sorry waits for state, not event, forget it. | 01:21 |
mehere | I've backed out that last optimization. I found something that I may have overlooked. I'll try again. | 01:45 |
prologic | ahh | 03:36 |
prologic | thanks for taking the time! | 03:36 |
prologic | I hadn't had time yet to look into it yet :) | 03:36 |
prologic | baby and all :) | 03:36 |
mehere | I've split things. Could you check (when baby's sleeping ;-) ) if there's any different behaviour between 3417 (should behave just like yesterday) and 3418 (minor change, mustn't really show any difference) when you run the tests? | 03:56 |
*** Osso has joined #circuits-dev | 03:57 | |
*** Ossoleil has joined #circuits-dev | 04:21 | |
*** Osso has quit IRC | 05:51 | |
*** Osso_ has joined #circuits-dev | 06:02 | |
mehere | With 3419 I've got 100% successful test in five consecutive runs. Now heading for the final optimization. | 06:59 |
*** Osso has joined #circuits-dev | 12:23 | |
*** Osso_ has joined #circuits-dev | 12:25 | |
*** Osso_ has quit IRC | 13:31 | |
*** Osso_ has joined #circuits-dev | 13:31 | |
*** Osso_ has quit IRC | 14:32 | |
*** Osso has quit IRC | 14:41 | |
mehere | Pushed all fixes and optimizations for event generation. Ran tests 7 times without problems. Please test latest version. If it shows problems at your site, please go back revision by revision and tell me where problems start. Good night. | 15:02 |
*** Osso has joined #circuits-dev | 15:04 | |
*** Osso has quit IRC | 15:04 | |
prologic | mehere, that revision passes all tests | 17:41 |
prologic | 3418 | 17:41 |
prologic | mehere, latest revision - confirmed | 17:49 |
prologic | nice work | 17:49 |
prologic | $ circuits.bench -e 100000 | 17:49 |
prologic | Setting up Speed Test... | 17:49 |
prologic | Setting up Sender... | 17:49 |
prologic | Setting up Receiver... | 17:49 |
prologic | () | 17:49 |
prologic | Total Events: 100006 (9260/s after 10.80s) | 17:49 |
prologic | oooh wow | 17:49 |
prologic | holy crap :) | 17:49 |
prologic | performance has dropped by a factor of 2-3x | 17:50 |
prologic | yeah wow | 17:50 |
prologic | rev 3418 is at 22k events/s here | 17:51 |
prologic | revision 3421 is where things start to get 2-3 times slower | 17:52 |
prologic | hmmm | 17:53 |
prologic | why are we updating the root on fire? | 17:53 |
prologic | doing so is O(n^2) I believe | 17:53 |
prologic | and should really only be done at registration and deregistration | 17:53 |
prologic | I think that could be the cause of the slowness | 17:54 |
prologic | walking the tree is slow | 17:54 |
*** marky1991 has joined #circuits-dev | 18:36 | |
prologic | I wanted to say... | 18:37 |
prologic | Congratulations guys! | 18:37 |
prologic | We have achieved a velocity of 8 over the last 3 sprints | 18:37 |
prologic | Good work! | 18:37 |
prologic | To compare | 18:37 |
prologic | We have a 4x better productivity than a commercial project (my work) where there are about a dozen project members all getting paid large sums of money working 40hrs/week | 18:37 |
prologic | So again congrats! | 18:37 |
*** marky1991 has quit IRC | 19:27 |
Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!