Time Nick Message 06:23 nerzhul merging #6587 06:23 ShadowBot https://github.com/minetest/minetest/issues/6587 -- Optimize entity-entity collision by numberZero 06:26 nerzhul i moved 0.4.11 changelog to the old changelog page, i will try to update our current 0.5 changelog (3 months of commits) 06:36 nerzhul i updated the current changelog from 13 jan to 19 feb 06:46 nerzhul Please and i published the announcement for the release https://forum.minetest.net/viewtopic.php?f=18&t=19904 08:54 Krock nerzhul, your linked topic does not exist anymore. Was it superseded by https://forum.minetest.net/viewtopic.php?p=315875#p315875 ? 08:58 Krock Will merge #6520 in 15 minutes 08:59 ShadowBot https://github.com/minetest/minetest/issues/6520 -- Selected ItemStack: Reduce black magic by SmallJoker 09:07 nerzhul Krock: WTF 09:07 nerzhul who moved my message ? 09:07 nerzhul it was an announcement, i'm not sure it was same as this topic which is a discussion about future :p 09:14 Krock merging... 09:15 nerzhul ty 16:08 nerzhul is this me or is this like we said in french "le ménage de printemps" (spring cleaning) 16:08 nerzhul :p 16:10 Krock regarding the refactors in our code base? 16:14 IcyDiamond I advocate for a refactoring regarding textures. They should be able to be defined in the metadata of nodes and items. 16:15 nerzhul no, in the PR and issue list :) 16:15 nerzhul the refactors are quite nice 16:16 nerzhul i push a trivial bunch of fixes (without including missing to cmath) to fix many performance-type-promotion-in-math-fn issues reported by clang (~15 occurences) 16:16 nerzhul see #67a4cb7d8a4461fe7d5206189fd4e9539beb20b7 16:16 nerzhul https://github.com/minetest/minetest/commit/67a4cb7d8a4461fe7d5206189fd4e9539beb20b7 16:19 nerzhul i will continue some little patchs when i get time, hoping i can fix remaining performance-type-promotion-in-math-fn and triggering a vote to promote it in CI blocking :) 16:36 paramat request, please don't remove '//#include "profiler.h" // For TimeTaker' frommapgens, very useful for profiling, and is commented out anyway 16:48 Wuzzy can some1 plz look at ? again? i think its fine now 16:48 Krock #7019 16:48 ShadowBot https://github.com/minetest/minetest/issues/7019 -- Colorize command parameters and privilege names by Wuzzy2 16:58 rubenwardy #4576 16:58 ShadowBot https://github.com/minetest/minetest/issues/4576 -- Add string.regex_search(string, pattern) 17:00 Krock should be kept open IMO regex is quite helpful and Lua's methods are limited 17:01 sfan5 and where do we pull regex from? the C++ standard library? 17:01 sfan5 (is that even available everywhere?) 17:01 sofar it should be, but it's fairly limited afaicr 17:02 sofar I usualy go for pcre, but I wouldn't recommend that in this case 17:02 rubenwardy Right 17:03 Krock C++ has regex, . But I don't know how much it supports 17:26 p_gimeno well, there's this: https://swtch.com/~rsc/regexp/regexp1.html which is an explanation about regexes, how to compile them into a graph and how to trace the graph - step by step with C source 17:26 sfan5 I don't think we want to NIH our own regex library 17:29 p_gimeno it's not a bad idea from the security standpoint, but certainly not as feature rich as other solutions 17:29 nerzhul paramat: noted 17:29 nerzhul sfan5: c++11 has regex embedded 17:31 paramat thanks nerz 17:34 p_gimeno http://seclists.org/fulldisclosure/2014/Aug/1 17:35 p_gimeno that's the kind of security I was talking about 17:38 p_gimeno both linked GCC bugs remain open 17:45 nerzhul it seems to depend on the developed usecase 17:49 p_gimeno the guy who wrote the article about NFAs also wrote a library that is apparently maintained now by Google: https://github.com/google/re2 17:53 p_gimeno I'll add that last link to the issue 17:59 p_gimeno on a different matter, I have a question. Should I rebase my PRs when I notice there's a conflict, or wait until being told to do so? 18:39 Wuzzy What's holding back #6325? 18:39 ShadowBot https://github.com/minetest/minetest/issues/6325 -- Add scripts to find and update translations. by Ekdohibs 19:58 nerzhul pushing another trivial commit to fix performance-type-promotion-in-math-fn problems 21:04 nerzhul and pushing the last commit for this issue (i also added 2 missing copyright notices) 21:43 paramat thanks, looks ok 21:54 nerzhul sorry guys 80 PR :) i just pushed 2 little PR 21:54 paramat partly my fault too 21:57 nerzhul heh 21:58 nerzhul my 2 PR are tiny, one require tidy to pass and will promote a quality rule, other is just a little performance fix 22:41 nerzhul wow i found a nice improvement in mapblock loading 22:41 nerzhul out LBM loading code has a very high cost on each node, but changing loop order & cache some lookups i divided by 20 the LBM def lookup on mapblock loading 22:43 nerzhul merging #7193 (remaining builds will be okay i compiled with clang 6 on my machine) 22:43 ShadowBot https://github.com/minetest/minetest/issues/7193 -- Optimize a little bit isBlockInSight, adjustDist & collisions by nerzhul 22:44 rubenwardy oh wow 22:44 rubenwardy nice! 22:44 nerzhul callgrind report a nice reduction on that part 22:45 nerzhul see #7195 22:45 ShadowBot https://github.com/minetest/minetest/issues/7195 -- Huge LBM lookup performance improvement on mapblock loading by nerzhul 22:46 nerzhul before, for 455 mapblock activations, 3M lookups on LBMContentMapping. After for 351 actations only 142k calls. Lookup reduced from 23% CPU time to 1.53% 22:47 Fixer nice work! 22:47 nerzhul review is needed to ensure my unroll + caching is good 22:47 nerzhul note: this will not reduce the Lua overhead on executing those ABM :) but core take less time to lookup 22:47 rubenwardy nice! 22:48 rubenwardy more of that please! 22:48 nerzhul just take time to analyse with callgrind if you want, sometimes i like to look at this, i didn't thought i could find that today :p 22:48 rubenwardy I suck at callgrind 22:48 nerzhul just be sure my change is good, analyse the caching please :) if i miss a case or something, i don't think but anyway 22:49 nerzhul just call it when needed, use kcachegrind to show report and that's all ^^ 22:49 rubenwardy I did a course which involved serial optimisations, but find analyising kcachegrind hard 22:50 nerzhul our main problem is to have separate bench on various code parts. Generally when i run callgrind i lost many time on mapblock loading, and game is very slow due to the analysis then ... it's hard to test other code parts 22:52 nerzhul nice catch sfan5 :) we don't need to copy iterator anymore heh 22:53 sfan5 first i wanted to suggest `auto iit = it` but then I noticed that the copy is useless 22:54 nerzhul world loading on regular MTG is blazing fast like this, i hope caching algo is good. I go to bed, see you 22:54 nerzhul push your approval when you get time :)