Minetest logo

IRC log for #minetest-dev, 2016-03-30

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:18 rubenwardy The rendering of formspecs appears to have changed: my label which is defined after an image is being drawn under that image
00:29 paramat https://github.com/minetest/minetes​t/blob/master/doc/lua_api.txt#L544 surely axisdir is facedir / 4 not modulo? think i'll correct this in my rotation PR
00:30 paramat rotation around the axis is facedir modulo 4
00:47 paramat will merge #3914 #3915 in a moment
00:47 ShadowBot https://github.com/minetest/minetest/issues/3914 -- Mgv7: Decrease cliff steepness by paramat
00:47 ShadowBot https://github.com/minetest/minetest/issues/3915 -- Nodes shader: Decrease amplitude of waving leaves and plants by paramat
00:50 paramat now mergng
00:51 paramat (merging)
00:56 paramat done
01:27 ssieb paramat: I mentioned that div/modulo thing a while ago and was told that I was wrong :-)
01:29 Icedream joined #minetest-dev
01:30 Zeitgeist_ joined #minetest-dev
01:30 paramat yes i got confused
01:30 paramat i checked it because you mentioned it, so thanks
01:52 Void7 joined #minetest-dev
02:12 proller joined #minetest-dev
03:26 sofar wow this boat model... lol
03:26 * sofar uvmaps
03:51 hmmmm ahhhh!  i'm losing it
03:54 hmmmm actually nevermind, i literally just figured it out right now
03:54 hmmmm heh.  it's really subtle.  it's an 'optimization' that i made in spreadLight that turned out to be subtly incorrect
03:54 hmmmm https://github.com/minetest/minetest/commi​t/0338c2eb91ca2ba19551aaa23bbfb46552cee9ac​#diff-f4dada645aa2c6343e16a20b047c07c5L245
03:55 hmmmm "oh, I don't want to repeat the same vmanip index calculation twice, so i'll calc the index first and then use that index to check if the position is in the vmanip!"
03:55 hmmmm wrong part:  "in the vmanip"
03:56 hmmmm the original code checked if the position was contained in the voxelarea 'a', not the vmanip's area
03:56 hmmmm 'a' is a subset of the voxelmanip's voxelarea
03:56 hmmmm oops
04:20 hmmmm https://github.com/kwolekr/minetest/commit​/0115da1d6348a08388238d8728daaa048a603522
04:20 hmmmm pushing in 15 minutes
04:21 AnotherBrick joined #minetest-dev
04:34 sofar http://i.imgur.com/8njhr4l.png
04:34 sofar almost done with the boat I think ^^
04:43 sofar http://i.imgur.com/stiVzsa.jpg
04:43 sofar result in-game appearance of the boat
05:11 Hunterz joined #minetest-dev
07:19 Darcidride joined #minetest-dev
07:19 Darcidride_ joined #minetest-dev
07:41 nrzkt joined #minetest-dev
08:01 Amaz joined #minetest-dev
08:46 davisonio joined #minetest-dev
09:47 jin_xi joined #minetest-dev
10:13 Fixer joined #minetest-dev
10:58 iangp joined #minetest-dev
11:17 Bro666 joined #minetest-dev
11:17 Bro666 Hello!
11:17 Bro666 I noticed that in version 4.10 there is a "Online mod repository" button in the "Mods" tab.
11:17 Bro666 In the 4.13dev version, there is no such button.
11:17 Bro666 [13:06] <Bro666> The button in 4.10 doesn't seem to do anything.
11:25 Fixer https://github.com/minetest/minetest/pull/3926 can be merged i guess
11:34 nore Bro666: Yes, mmdb (Minetest mod database) was dropped
11:34 nore it didn't work
11:38 nrzkt nore can you look at #3924
11:38 ShadowBot https://github.com/minetest/minetest/issues/3924 -- Colored item descriptions
11:38 nrzkt err
11:38 nrzkt #3934
11:38 ShadowBot https://github.com/minetest/minetest/issues/3934 -- Fix ParticleManager::handleParticleEvent coding style by nerzhul
11:40 nore nrzkt: looking
11:41 Fixer Bro666, i think this it (for android): http://rubenwardy.com/mtmods4android/
11:42 Fixer this is it*
11:42 nore nrzkt: looks good
11:43 nrzkt thanks, can you vote ? :)
11:43 nrzkt sfan5, kahrl if you get time too
11:43 sfan5 nrzkt: it's just code style changes... you have my +1
11:50 Bro666 Fixer: checking it out. Thanks for the pointer.
11:51 Bro666 Oh... I see. It is an independent app. Cool.
11:52 Obani joined #minetest-dev
12:02 Obani #3926 has three approvals
12:02 ShadowBot https://github.com/minetest/minetest/issues/3926 -- mg_schematic: fix leak-like behaviour, and small cleanup by est31
12:06 STHGOM joined #minetest-dev
12:06 proller joined #minetest-dev
12:27 nore sfan5: you're moderator on the forum, right?
12:27 sfan5 yes
12:27 nore could you delete this user please? https://forum.minetest.net/memberli​st.php?mode=viewprofile&amp;u=19195
12:27 nore (and the 12 spam posts he made)
12:29 sfan5 tfw someone replied to the spam
12:31 sfan5 500 Internal Server Error
12:31 sfan5 nice
12:34 nore sfan5: I got that one too :/
12:34 sfan5 just try again a few times
12:34 nore reloading fixed it though
12:35 celeron55 what
12:36 gregorycu joined #minetest-dev
12:36 gregorycu Sup cool cats
12:36 nore hi gregorycu :)
12:38 gregorycu 12 hour working day, time for some minetest
12:38 sfan5 working days are overrated
12:38 sfan5 i prefer holidays :)
12:39 gregorycu Well, you're not stupid that's for sure
12:39 gregorycu I've jumped on to look at two bugs, and people have yanked them out from under me
12:40 gregorycu Last couple of days
12:40 celeron55 looks like the server is capping out on php-fpm worker connections
12:40 celeron55 did the 500 page indicate anything like this or was it just a completely generic 500 error page?
12:41 sfan5 generic
12:41 sfan5 the standard nginx one
12:43 Darcidride joined #minetest-dev
12:44 gregorycu c55: Quick question, we don't seem to use the assignee feature of github much
12:44 gregorycu Is that a deliberate decision
12:45 celeron55 usage of it is neither endorsed nor disallowed
12:45 gregorycu Cool
12:45 celeron55 it's not used much due to the way development tends to be organized
12:46 gregorycu Or not organised ;)
12:47 celeron55 basically if someone wants to assign an issue to themselves to indicate to others that they are working on it, they can
12:47 celeron55 other uses don't make sense
12:48 gregorycu Sounds good to me
12:54 celeron55 i doubled the worker connections for the forum and the wiki (which seemed to be slightly limited by them); now everything should run 100% properly again
12:54 gregorycu I'm going to look at the (suspected) performance regression mentioned here: https://github.com/minetest/minetest/issues/3931
12:56 celeron55 uh... or not; what the hell is this request spam
13:01 gregorycu Dearest paramat, what is up with your cave generation
13:01 gregorycu It's 15x slower than it used to be
13:02 gregorycu He ain't here
13:04 Fixer gregorycu, since when? he increased frequency of caves in v7
13:04 gregorycu Nov 10 2015
13:05 Fixer you test in mapgen 7?
13:05 gregorycu Yes
13:07 Fixer gregorycu, https://github.com/minetest/minetest/pull/3413 here he increased the frequency, so it works more to generate probably, but I'm not sure if it is the problem
13:09 gregorycu Looking at #3931 what slowdown do you think he has identified?
13:09 ShadowBot https://github.com/minetest/minetest/issues/3931 -- Block loading feels much more sluggish as of current dev version
13:09 gregorycu Jerkiness, or just slow generation?
13:09 celeron55 it could be due to the mapblock memory limit of 5000
13:10 Fixer gregorycu, i'm not sure, annoying jerkiness at mesh gen was forever, and usual block loading looks good to me (hard to say without actual tests)
13:10 celeron55 it will effectively limit the loading range to something like 140
13:10 celeron55 (140 nodes)
13:10 celeron55 after that, the client just starts deleting them from memory at the rate the server sends them
13:11 Fixer gregorycu, subjective test is try to outfly it %)
13:12 gregorycu What is %?
13:12 Fixer gregorycu, bad smile
13:12 gregorycu ahh ok
13:12 gregorycu lol, of course
13:12 Fixer let me try to outfly it
13:15 Fixer gregorycu, can't outfly it in mgv7, so looks good at least to me, jerkiness is present of course
13:15 gregorycu We really need to fix that :(
13:15 Fixer gregorycu, you better ask paramat about caves
13:15 Fixer send him a message
13:15 Obani !tell
13:15 ShadowBot Obani: tell <nick> <text>
13:16 gregorycu !tell paramat Caves are very slow, you bastard!
13:16 ShadowBot gregorycu: O.K.
13:17 Fixer gregorycu, about jerkiness, you can mitigate it by using max_simultaneous_block_sends_server_total = 1 in singleplayer
13:17 Fixer gregorycu, it is much more smoother that way
13:17 PilzAdam joined #minetest-dev
13:17 gregorycu !tell paramat A more helpful comment is that generating caves are 15x slower than 4 months ago
13:17 ShadowBot gregorycu: O.K.
13:18 Fixer gregorycu, iirc there are 2x more caves in mgv7 now, not sure how much it will cost in performance
13:18 Fixer gregorycu, since Nov 30
13:22 Obani celeron55, I think you should remove the "Unconfirmed bug" label btw
13:22 celeron55 me? why?
13:22 celeron55 and where
13:22 Obani https://github.com/minetest/minetest/issues/3931
13:23 Obani Because you're not away :p
13:23 celeron55 somebody else clearly is more involved with that issue than me
13:23 celeron55 so no
13:24 gregorycu I'm not too sure
13:24 gregorycu I think it's still unconfirmed
13:24 gregorycu Someone optimised the shit out of the perlin noise generator
13:31 gregorycu The reverse has happened for transform liquids
13:34 gregorycu Hey kids, it's nice to use references
13:35 Fixer gregorycu, liquid activation, interested? %)
13:35 gregorycu ContentFeatures cf = nodemgr->get(n0);
13:35 gregorycu This is not very efficient according to the profiler
13:37 dfelinto joined #minetest-dev
13:38 gregorycu We can blame our good friend paramat for this one
13:40 gregorycu I'll recompile with more aggressive inlining to see what the compiler can do
13:53 gregorycu Hooray for me, I made transform liquids fast again, at least on windows
14:13 kaadmy joined #minetest-dev
14:15 gregorycu ##3937
14:15 gregorycu Double trouble
14:15 gregorycu #3937
14:15 ShadowBot https://github.com/minetest/minetest/issues/3937 -- Remove expensive copy of ContentFeatures by gregorycu
14:17 Fixer make minetest great again
14:17 sfan5 ^
14:19 damiel joined #minetest-dev
14:20 Fixer gregorycu, will it be noticable ingame, or it is subtle change? do i need to try it just in play?
14:23 gregorycu Somewhat hard to tell, on the server side there are 3 CPU heavy theads
14:24 gregorycu Maybe only two
14:24 gregorycu Anyway, you have the Server thread, and the Emerge thread
14:25 gregorycu When it does the liquid transformations, the server thread holds a lock that the emerge thread needs to do work
14:25 gregorycu So, it's possible, indirectly, there is some noticeable performance improvement
14:25 gregorycu It certainly will be using less CPU cycles
14:26 kahrl here's another fun one (or two) https://github.com/minetest/minetest/blo​b/master/src/content_mapblock.cpp#L1379
14:27 Fixer oh well, i will compile with it just for fun
14:27 davisonio joined #minetest-dev
14:27 gregorycu It's a commit from Greggy Boy, you know you're going to have a fun time
14:28 gregorycu kahrl: They didn't pop up on my profiling
14:28 gregorycu Probably because I don't have fences or rails :)
14:29 Player_2 joined #minetest-dev
14:29 gregorycu const ContentFeatures *f2 = &nodedef->get(n2);
14:29 gregorycu lol
14:31 kahrl o_O
14:31 kahrl oh, nevermind, that should be fine
14:31 kahrl (referring to the f2 thing)
14:32 gregorycu Yeah, perfectly safe, just a bit silly
14:33 gregorycu Do you mind if i leave the other two instances, as they are not in a hotpath
14:33 gregorycu We can come through when we have C++11 and make a whole bunch of stuff uncopyable when we have move
14:34 gregorycu I'll add the macro and see what happens
14:35 STHGOM joined #minetest-dev
14:37 kahrl the main problem is CNodeDefManager which needs to be able to resize its std::vector<ContentFeatures>
14:38 gregorycu Hence move operator :(
14:38 est31 joined #minetest-dev
14:38 gregorycu I don't think this is going to be trivial
14:38 est31 whats the move operator
14:39 gregorycu I'm more than happy to cover the hotpath for the moment, and come back to this with C++11
14:39 gregorycu It's a C++11 feature
14:40 est31 aha
14:40 est31 i've thought it was just done automatically by the compiler
14:40 est31 seems it really is a new operator
14:40 gregorycu It is, in some cases
14:40 est31 but how should we benefit here
14:41 est31 i mean, ndef->get shouldn't move it out of the vector, no?
14:41 gregorycu no
14:41 gregorycu std::vector has a requirement that the type is movable
14:41 gregorycu Sorry
14:41 gregorycu That the type is COPYABLE
14:42 gregorycu std::vector<ContentFeature> will probably not compile
14:42 nrzkt est31 movable isn't done by compiler and no you should use movable in this case, movable = switch refs
14:42 gregorycu The compiler can generate move operators for you
14:42 est31 so, what would c++11 do better then?
14:42 est31 in this case
14:42 nrzkt it's useful for example if you have void test() { std::string node = "default:dirt"; digNode(node); }
14:43 gregorycu In C++11, with std::vector, the type needs to EITHER be copyable, or movable
14:43 nrzkt because you don't need node in another place than this
14:43 gregorycu (or both)
14:44 gregorycu It's hard to explain the application here
14:45 gregorycu Maybe this angle will work. Move allows you to distinguish cases where you want to actually copy, vs cases where you are forced to copy (as is the current case with std::vector)
14:46 nrzkt est31, http://blog.smartbear.com/c-plus-plus/c​11-tutorial-introducing-the-move-constr​uctor-and-the-move-assignment-operator/
14:47 gregorycu Anyway, it's good stuff for code where you want options when you want to be performant
14:47 kahrl how about adding this: explicit ContentFeatures(const ContentFeatures&) = default;
14:47 kahrl that'll catch all the accidental copies
14:48 gregorycu I'd have to check the standard to see if that will work for std::vector
14:48 est31 kahrl, i think explicit is c++11, no?
14:48 kahrl est31: yes
14:48 kahrl wrap it in #if __cplusplus >= 201103L
14:48 hmmmm joined #minetest-dev
14:49 gregorycu Now we are in trouble, hmmmm  is here
14:49 est31 hehe
14:49 gregorycu I actually wasn't aware explicit was a c++11 feature
14:50 est31 http://en.cppreference.com/w/cpp/language/explicit
14:50 nrzkt est31 if you are okay could you merge #3934 please ?
14:50 ShadowBot nrzkt: Error: Delimiter not found in "Page is too big or the server took too much time to answer the request."
14:50 nrzkt gregorycu, your PR with the bad CF copy is good :)
14:50 nrzkt #3934 ShadowBot !
14:50 ShadowBot https://github.com/minetest/minetest/issues/3934 -- Fix ParticleManager::handleParticleEvent coding style by nerzhul
14:50 est31 nrzkt, okay
14:51 kahrl gregorycu: actually, I don't think it is, but the "= default" is
14:51 est31 nrzkt, but put the space after the for, okay?
14:52 gregorycu nrzkt: thanks :) it really stood out after I profiled
14:52 gregorycu Found it just in time for the next release... hopefully
14:53 gregorycu Just gotta add a space or two
14:54 kahrl got it to compile under -std=c++11 with this: https://gist.github.com/kahrl/e​2f12da8f6dd62b3be6221c79bf62f8c
14:55 gregorycu I guess explicit copy ctor does work with vector
14:56 est31 perhaps that would be a good moment to add c++11 to travis
14:56 est31 kahrl, nice
14:57 nrzkt est31, i don't get time to fix it atm, if you can fix it go, else you should wait some days :s
14:57 gregorycu Most thumbs up in an hour award goes to me
14:58 est31 :)
15:01 Elinvention joined #minetest-dev
15:02 est31 what about doing the following: we let the manager hold a vector of BaseContentFeatures
15:02 est31 and ContentFeatures is a child of that class, with DISABLE_CLASS_COPY
15:02 hmmmm heh that's a cute workaround
15:02 hmmmm i like it
15:04 gregorycu lol
15:04 gregorycu I don't know if we hate the other developers that much
15:05 gregorycu You can also just simple befriend whoever you want to own the vector
15:05 gregorycu (or rather, whatever class does the operations that need copy)
15:05 gregorycu I've fixed that style issue by the way
15:07 gregorycu Good work hmmmm in fixing that lighting issue
15:07 nrzkt joined #minetest-dev
15:07 gregorycu If est31 fixed that memory leak, the author of that mod is going to be super happy
15:07 Fixer active block generation, fps goes from 120 to 60 and back
15:07 gregorycu Or the thing that looks like a memory leak
15:08 gregorycu Fixer: I have no idea if that's good or bad
15:08 Fixer gregorycu, bad ofc
15:08 gregorycu With my change?
15:08 Fixer gregorycu, you change don't fix jerkiness anyway, cpu usage is 20%
15:08 gregorycu *sigh*
15:09 gregorycu Server is probably generating chunks too fast for the client now
15:09 Fixer gregorycu, no, i think problem is this, chunk loading should be spread out in time, so instead of 50 meshes spike, it will process them by 5 for longer time
15:10 gregorycu I'm sick of this jerkiness issue
15:10 Fixer gregorycu, player will not see server performance change you did anyway, thats can be seen in profile, we have no benchmarks anyway
15:11 Fixer except profiler
15:11 gregorycu I was hoping it would speed up block generation
15:11 Fixer gregorycu, you have them?
15:11 Fixer gregorycu, speeding up makes things worse :)
15:11 gregorycu Doesn't everybody have the stuttering issue?
15:11 gregorycu We can add an artificial sleep in the server thread now, instead
15:12 gregorycu Which is better than actually doing CPU work
15:12 hmmmm yup
15:12 gregorycu That wasn't a serious suggestion
15:13 gregorycu What's best is if we fix the underlying issue
15:13 gregorycu It's actually hard for my profiler to detect
15:13 gregorycu Cause it takes an average over time
15:14 Fixer gregorycu, again, try max_simultaneous_block_sends_server_total = 1
15:14 Fixer gregorycu, in singleplayer
15:14 gregorycu Wont that hide the issue?
15:14 Fixer gregorycu, you will understand what i mean by smoothing out
15:15 gregorycu I want to do the opposite of hiding :)
15:15 Fixer gregorycu, you are right
15:15 Fixer gregorycu, it just shows how much smoother it can be
15:15 celeron55 gregorycu: as far as i know, the jerkiness comes at the point when a new mesh is drawn the first time
15:16 celeron55 or, i mean, the slowness of creating a new mapblock mesh and taking it into use
15:16 celeron55 i might be completely wrong but this is what i remember from when i looked into it like years ago
15:16 hmmmm no i'm not talking about that
15:17 hmmmm this is happening only recently, i.e. in the last 2 months or so
15:17 Fixer gregorycu, jerkinness -> https://imgur.com/3EdBMkg, better way: https://imgur.com/bxQRWU0 (somewhat slower)
15:17 Fixer i have jerkiness since 0.3 at least
15:17 hmmmm i'm just trying to move around and test lighting and block loading is super slow
15:18 gregorycu hmmmm: Are you able to try my change?
15:18 hmmmm like i'd fly around in a certain area and the blocks still haven't been there yet
15:18 hmmmm why don't you push it?
15:18 Fixer weird, my blocks load fast
15:18 Fixer can't outfly them
15:18 hmmmm i use vague language because i don't know what the problem is
15:19 Fixer maybe you have debug build?
15:19 gregorycu The emerge thread and server thread share a lock, speed up server = speed up emerge
15:19 hmmmm i always use a debug build
15:19 hmmmm i can still not outrun block loading
15:19 Fixer ahhh
15:19 hmmmm and it's not a mapgen slowness issue like somebody alluded to earlier because this was on v6
15:19 hmmmm v6 isn't futzed with
15:19 gregorycu Yeah, debug config for me sucks
15:19 gregorycu Um...
15:19 gregorycu I can't push cause I'm not a core dev
15:19 Fixer gregorycu, i doubt it, speed up, more blocks will be processed, even worse jerkinness
15:20 hmmmm oh, I'll push and test it for you at the same time then
15:20 gregorycu Fixer: I'm talking about hmmmm's issue
15:20 Fixer gregorycu, problem is it processes big banch of them in short time, oh nvnd
15:20 hmmmm i hate having to cherry pick from other peoples branches because i need to copy and paste commit ids
15:20 Fixer nevermind
15:20 gregorycu All good bro
15:20 hmmmm and find them and stuff
15:20 hmmmm there should be a script for this... grr
15:21 hmmmm lol right now i should be working and doing real work
15:21 gregorycu it's 2:30am for me, I need to be sleeping
15:22 est31 bye I'm going too
15:26 hmmmm awesome
15:26 hmmmm thanks gregory, that fixed the regression
15:26 gregorycu I'm a little surprised
15:26 hmmmm i can now run and jump around as i please and the map always shows up
15:27 gregorycu Awesome
15:27 hmmmm like
15:27 Fixer nice
15:27 hmmmm it'd be a lot clearer with a video
15:27 hmmmm but idk how to make a video of it and i don't even have anywhere to upload it if i did
15:29 gregorycu I trust you in any case
15:29 gregorycu Performance can be a hard thing to quantify sometimes
15:30 hmmmm it's somewhat frightening that a user's experience can be made or broken with a single &
15:30 hmmmm but at the same time i'm grateful that such performance issues are obvious
15:30 Fixer australian + europe time mix badly
15:30 hmmmm everything in australia is upside down
15:30 hmmmm including their clocks
15:31 Fixer europe goes back from work and gregorycu goes to sleep -_-
15:31 gregorycu (:
15:31 Fixer only literally two hours in same chat
15:31 Fixer same time*
15:31 hmmmm lol
15:31 gregorycu I can't even use freenode at work cause they fucking block port != 80
15:31 gregorycu And #minetest-dev has blocked freenode webchat
15:32 hmmmm fixer, before, were you talking about the microstuttering?
15:32 hmmmm that's happened for the longest time but i kind of just accepted it
15:32 hmmmm that has to do with the rendering
15:32 gregorycu I'm not sure it does
15:33 Fixer hmmmm, micro? are you kidding? %)
15:33 hmmmm for kicks, one time i downloaded and ran a version of minetest 0.2 or something and it was buttery smooth
15:33 Fixer hmmmm, yes, and block loading was slow
15:33 hmmmm not bad given how inefficient the code was at that time
15:33 gregorycu I'm pretty sure it's not rending related
15:33 Fixer map load was slow
15:33 hmmmm we're talking about different things then
15:34 Fixer in 0.2 i have constantly bumped into nonloaded blocks
15:34 Fixer hmmmm, it is macro %)
15:34 Fixer hmmmm, micro + macrostuttering
15:34 hmmmm i am always able to reproduce the rendering microstuttering
15:34 hmmmm just turn fly mode on and run forward
15:34 hmmmm every second and a half or so there will be a slight delay
15:35 gregorycu Fixer: That setting
15:35 gregorycu Does it make the stuttering disappear?
15:35 gregorycu Or get dimisnished
15:35 gregorycu I know it's hard to tell
15:35 Fixer gregorycu, only diminished, some are still present, but much less, also block load slower
15:35 gregorycu That's to be expected
15:35 Fixer gregorycu, but experience is much much better
15:36 Fixer hmmmm, hold on, i will search for few gifs i made with it
15:36 gregorycu I've isolated the GUI thread with my profiling data
15:36 gregorycu Drawing is 75%
15:36 hmmmm gregory, you mind if i fix a very minor spacing issue with your commit before i push it to upstream?
15:37 gregorycu Of course not
15:37 gregorycu I think I fixed the issue didn't I?
15:38 gregorycu const ContentFeatures & cf = nodemgr->get(n0);
15:38 gregorycu I have a PR, is that what you're looking at?
15:38 gregorycu #3937
15:38 ShadowBot https://github.com/minetest/minetest/issues/3937 -- Remove expensive copy of ContentFeatures by gregorycu
15:39 Fixer hmmmm, stutter on flat mapgen -> https://imgur.com/mg7UypX
15:40 gregorycu Sorry, 83% drawing, handling block data is 12%
15:40 Fixer gregorycu, with that option set i get this -> https://imgur.com/UW6TIRD much less spikes present, draw time curve is more straight with fewer spikes
15:40 gregorycu That's 12% average
15:40 gregorycu Not 12% of the bad frame
15:41 Fixer gregorycu, interesting
15:41 hmmmm right
15:41 gregorycu First one looks like a fucking EKG
15:41 hmmmm to be honest i always found those profiler graphs to be non-informative
15:41 hmmmm maybe i just don't know how to read them correctly
15:42 Fixer i just know that most of stutters happen when num_processed_meshes goes higher than 3-5 or so
15:42 gregorycu Thank you hmmmm
15:43 Fixer there are some rare stutters not linked to num_processed_meshes but to mainloop_other (on Vanessas servers)
15:43 Fixer that can be as big as 0.7 sec
15:43 hmmmm i think a better way to profile things like minetest is to keep a histogram of recent frame times
15:43 gregorycu That's no good
15:43 Fixer but that far from default game
15:43 hmmmm and then do some statistics like showing how many % of frames are outside of a threshhold
15:43 gregorycu Still
15:43 Fixer hmmmm, i did it anyway
15:44 Fixer hmmmm, you mean this? https://forum.minetest.net/v​iewtopic.php?f=6&amp;t=13360
15:44 Fixer hmmmm, look graphs below
15:44 Fixer with spikes
15:44 hmmmm in an ideal world, minetest would have a reasonable average frame time but a low standard deviation
15:44 Fixer right
15:44 hmmmm yea
15:44 hmmmm those graphs are nice
15:44 Fixer just look through my link and pictures
15:44 hmmmm something like that except realtime
15:45 gregorycu I actually think I have enough data
15:45 gregorycu I'll share what I have, one moment
15:45 Fixer hmmmm, notice, how i compared when mapgen is idle and when working
15:45 Fixer those graphs are interesting
15:46 hmmmm are those all single player or multiplayer/
15:47 Fixer some singleplayer, some multiplayer, read descriptions
15:47 Fixer freeminer graphs impressed me, they are much better
15:47 hmmmm what about the first graphs
15:47 gregorycu https://i.imgur.com/015MltV.png
15:47 Fixer with percentages?
15:47 hmmmm yeah
15:48 Fixer hmmmm, singleplayer ofc
15:48 gregorycu Client::step is probably the thing that kills perf
15:48 hmmmm hmmm ok
15:49 hmmmm is that the visual studio performance profiler?
15:49 gregorycu Yeah, some output from it
15:49 gregorycu Just the GUI thread
15:49 hmmmm that actually looks nice
15:49 hmmmm what, are you too good for us to use valgrind?
15:50 gregorycu These tools are expensive, may as well use them as much as i can :)
15:51 gregorycu Like $3k or something
15:51 hmmmm what does it do better than valgrind though
15:51 gregorycu Runs on windows? lol?
15:51 hmmmm also why are you using your work computer for minetest
15:52 gregorycu I'm not
15:52 Fixer RDP!11
15:52 gregorycu Nah, just installed it at home
15:52 gregorycu I have like a multi-machine licence or something
15:53 gregorycu All legit
15:53 hmmmm oh
15:53 hmmmm i can probably do the same with my msdn subscription
15:53 Hunterz joined #minetest-dev
15:53 gregorycu Just can't share with other people
15:53 hmmmm i can get up to 5 licenses of any microsoft product
15:53 hmmmm *any*!
15:53 Fixer msdn subscription?
15:53 sfan5 windows me too?
15:53 gregorycu Especially windows me
15:53 hmmmm but who actually wants to use windows?
15:54 kahrl minecraft licenses? :P
15:54 Fixer well, i use it, historically, linux is nice though
15:54 hmmmm kahrl:  heh i never checked
15:54 gregorycu Why would you play minecraft if you have minetest
15:54 hmmmm testing to see how they do things
15:54 gregorycu Stop being silly kahrl
15:54 Fixer i played it just to compare performance
15:54 Fixer and it is damn impressive
15:54 gregorycu Don't do that if you want to remain happy
15:54 Fixer super smooth, big view range, constant 80-100 fps
15:55 gregorycu Give me a chance bro
15:56 sfan5 Fixer: you clearly haven't tried minecraft 1.9 on my old laptop
15:56 Fixer it was 1.8.8 + optifine ofc, default performance maybe lower
15:56 sfan5 runs like shit
15:57 Fixer oh, i need to test VBO on my old Pentium M + Intel Integrated notebook, mem was replaced and it works again
15:57 Fixer wonder what is speed gain over there
15:57 Fixer gregorycu, i'm curious, can you expand Game::updateFrame?
15:59 gregorycu Sure can, one moment
16:01 gregorycu https://i.imgur.com/fdAOJEI.png
16:01 Fixer thanks
16:02 gregorycu You know what's under Irrlicht.dll? A whole bunch of other Irrlicht stuff, and also ShaderSource::onSetConstants (19%)
16:02 Fixer whats up with clouds?
16:02 gregorycu Apparently they are pretty intensive
16:03 Fixer gregorycu, superflous shaders should lower it a bit, right?
16:03 Fixer i mean irrlicht part
16:03 gregorycu Yeah, should halve it
16:03 gregorycu Approx
16:03 Fixer yeh, that one gives me very big fps boost
16:03 gregorycu (It varies greatly)
16:04 gregorycu I don't want to go to bed
16:04 kahrl why does it take almost twice as long to draw the crosshair than to draw the hotbar?
16:04 Fixer but there are some problem with lighting sometimes, it gets stuck, not sure why, can be reproduced after restart
16:04 Fixer restart many times, dark dark dark fixed blabla dark looks random
16:05 gregorycu kahrl: Not sure
16:07 Obani joined #minetest-dev
16:11 Fixer disabled clouds, got +3% fps boost %)
16:11 Fixer not much
16:11 gregorycu hmmm...
16:12 gregorycu I wonder...
16:12 gregorycu Sorry, that wasn't a noun
16:19 Elinvention joined #minetest-dev
16:20 Etzos joined #minetest-dev
16:24 Elinvention joined #minetest-dev
16:25 Krock joined #minetest-dev
16:32 * gregorycu &
16:36 Dragonop joined #minetest-dev
16:47 * Krock &
16:47 Elinvention_ joined #minetest-dev
17:09 Calinou joined #minetest-dev
17:35 damiel joined #minetest-dev
17:36 Samson1 joined #minetest-dev
17:41 Elinvention joined #minetest-dev
17:41 Elinvention joined #minetest-dev
18:01 damiel joined #minetest-dev
18:05 hmmmm agh too bad gregory left.
18:05 hmmmm was going to ask if https://github.com/minetest/minetest/pull/3534 is still relevant or if it's been superceded by SN's branch
18:07 sofar ping him in the PR?
18:07 hmmmm do people actually read the notifications?
18:07 sofar depends on your statistical sample size of "people"
18:10 Fixer he is sleeping 100%
18:10 Fixer it is deep night in Australia
18:11 sofar a! the mate is down under!
18:11 Fixer Australia feels like Mars, so far and forgotten, forever alone
18:12 Fixer you need some kind of developer clock here, to align your work
18:13 VanessaE we have one.  it's called US Eastern Daylight Time.  ;)
18:13 hmmmm NO!  it should be equally inconvenient for everybody!  UTC
18:13 sofar UTC and UTC foreveeeeerrrrrr
18:13 Fixer in fact i will look at contributors and were they live, interesting to see
18:14 VanessaE nonono that would be convenient for folks in france and britain :P
18:14 VanessaE Alutian time :)
18:14 sofar france is not at UTC
18:14 VanessaE no but the name is french :)
18:14 Fixer https://github.com/minetest​/minetest/graphs/punch-card
18:15 VanessaE fuck it, sync the clock to ..what was it, Centaurian time?  :)
18:17 Fixer looks like crudely 50/50 Europe-USA and gregorycu
18:19 sofar I'm split - I'm both!
18:19 sofar living in USA but Dutch
18:21 Fixer with USA situation is better, if people on east coast of USA, than work ends in ideal at 18:00 and that will be 22-24:00 in Europe
18:21 Fixer oh well, crappy anyway
18:23 Warr1024 joined #minetest-dev
18:41 paramat joined #minetest-dev
18:50 Elinvention joined #minetest-dev
18:57 paramat good work, sorry for slowing transformliquids
18:58 Dragonop joined #minetest-dev
19:02 paramat !tell gregorycu could the slowing of mgv7 cavegen be due to my slowing of transformliquids?
19:02 ShadowBot paramat: O.K.
19:08 paramat !tell gregorycu https://github.com/minetest/minetest/commi​t/0a8af8814787654dcbe0459a28255586fbfa3bd1 3D perlinmap noise calculations were moved into generateCaves() perhaps that is it
19:08 ShadowBot paramat: O.K.
19:11 paramat !tell gregorycu 3D noise calculation is very heavy so if you're timing generateCaves() a x15 slowdown is possible
19:11 ShadowBot paramat: O.K.
19:31 proller joined #minetest-dev
19:34 paramat i will merge games 974 977 980 989 later
19:35 paramat i'll test the boatmodel and merge if good
19:37 DFeniks joined #minetest-dev
20:09 Darcidride joined #minetest-dev
20:15 proller joined #minetest-dev
20:16 ssieb joined #minetest-dev
20:18 Obani hmmmm, who are you on Github ?
20:19 kahrl Obani: kwolekr
20:19 Obani Yes !
20:20 Obani I was right :D
20:23 Fixer that was slow
20:25 nrzkt joined #minetest-dev
20:28 paramat Fixer is light in tunnels fixed now?
20:28 Fixer paramat, yes
20:28 paramat thanks
20:28 Fixer \o/
20:33 damiel joined #minetest-dev
20:46 kahrl is this test failure already reported as an issue? https://gist.github.com/kahrl/2​5945518347b5732e89a806721e1bddd
20:46 kahrl (in testAtomicSemaphoreThread)
20:46 kahrl can't find any issue for it
20:47 hmmmm that is a new one
20:47 hmmmm and a bad one too
20:47 kahrl I think it appeared one I switched to -std=c++11
20:47 kahrl but let me test that hypothesis
20:47 kahrl s/one/once
20:50 hmmmm well uh.. sounds like a problem with Atomic<T>
20:50 hmmmm sorry SN
20:50 kahrl yeah
21:01 kahrl yup, always passes without -std=c++11, always fails with -std=c++11
21:08 nore paramat: why did you close game#988? Forest biomes can be re-registered at a high altitude to enable them for floatlands
21:08 ShadowBot https://github.com/minetes​t/minetest_game/issues/988 -- Default/mapgen: Add upper limits to forest biomes and trees by paramat
21:12 kahrl #3940
21:12 ShadowBot https://github.com/minetest/minetest/issues/3940 -- testAtomicSemaphoreThread failures (C++11)
21:22 kahrl oh, hah
21:23 kahrl I think it's just that std::atomic doesn't initialize itself
21:23 kahrl so it starts with some weird value, not 0
21:25 nrzkt joined #minetest-dev
21:27 rubenwardy joined #minetest-dev
21:34 proller joined #minetest-dev
21:40 paramat nore hm good point, i'll consider
21:43 paramat re-opened
22:04 paramat will merge games 974 977 980 989 990 in a moment
22:04 sfan5 ame#974 game#977 game#980 game#989 game#990
22:04 ShadowBot https://github.com/minetes​t/minetest_game/issues/977 -- Add recipe to craft clay block back into lumps by tenplus1
22:05 ShadowBot https://github.com/minetes​t/minetest_game/issues/980 -- Start timer as well when moving items around. by sofar
22:05 ShadowBot https://github.com/minetes​t/minetest_game/issues/989 -- Fire: Use get_item_group instead of get_node_group by Rui914
22:05 ShadowBot https://github.com/minetes​t/minetest_game/issues/990 -- Boats: Optimize, re-UV unwrap. by sofar
22:05 sfan5 game #974
22:05 sfan5 game#974
22:05 ShadowBot https://github.com/minetest/minetest/issues/974 -- Add more checks to vector functions by ShadowNinja
22:05 ShadowBot https://github.com/minetes​t/minetest_game/issues/974 -- Shelves: Fix listring functionality + code cleaning by kilbith2
22:05 sfan5 i see
22:08 paramat now merging
22:10 Warr1024 joined #minetest-dev
22:21 paramat merged
22:30 turtleman joined #minetest-dev
22:37 rubenwardy joined #minetest-dev
22:41 proller joined #minetest-dev
22:57 paramat closed #1898
22:57 ShadowBot https://github.com/minetest/minetest/issues/1898 -- Numblock does not work in text entry fields, left/right instead of 4/5
23:08 sofar I think I should re-UV unwrap stairs_stair.obj as well
23:08 sofar I don't think it should be almost 3k large while the mesh doors are not even 1k
23:09 sofar I bet that there are invisible faces in there
23:10 paramat please do
23:10 sofar it also occurred to me that visually the stair block doesn't appear to line up with the wooden block well
23:10 sofar at least texture wise
23:12 sofar hey, while I'm touching every model in the game, might as well touch the last one, right?
23:18 kilbith joined #minetest-dev
23:20 paramat closed #2027 as is same as #3936
23:20 kilbith sofar, the stairs model has a minimal number of vertices but each face is contain in their own mesh object to support up to 6 different tiles (not like the boat)
23:20 ShadowBot https://github.com/minetest/minetest/issues/2027 -- Stack overflow during burning
23:20 ShadowBot https://github.com/minetest/minetest/issues/3936 -- stackoverflow risk with too many attached nodes.
23:20 sofar that would explain the size
23:20 kilbith yes exactly
23:22 kilbith originally it weighted ~0.5k before i added the support for 6-tiles
23:23 yang2003 joined #minetest-dev
23:24 paramat closed #2034
23:24 ShadowBot https://github.com/minetest/minetest/issues/2034 -- Can walk through blocks if placed where player is standing
23:26 Fixer will that famous "shift when falling" trick will be fixed? %)
23:26 Fixer or it is a feature?
23:33 iangp joined #minetest-dev
23:33 paramat closed #2063
23:33 ShadowBot https://github.com/minetest/minetest/issues/2063 -- Crouch has no visual feedback, which is confusing (IIRC there's a PR/patch for this or something like it)
23:35 Fixer books in bookshelfs in protected areas are protected? i've heared that not
23:36 kilbith nope
23:36 kilbith well it depends how good your protection mod is
23:39 paramat closed #2088
23:39 ShadowBot https://github.com/minetest/minetest/issues/2088 -- Android Fps dips:
23:40 Fixer fps dips? is not it is insenely low anyway?
23:41 Fixer paramat, just a little more and will be 600 %)
23:44 paramat closed #2113
23:44 ShadowBot https://github.com/minetest/minetest/issues/2113 -- Mouse & keyboard on Android

| Channels | #minetest-dev index | Today | | Google Search | Plaintext