Time  Nick         Message
01:47 MinetestBot  02[git] 04SmallJoker -> 03minetest/minetest: Menu content tab: Fix clipped description and missing depends (#7703) 136ed9c6f https://git.io/fAzcn (152018-09-07T01:46:58Z)
02:00 Quiark       is it possible to reload Lua mod while minetest is still running? for faster test cycle
02:02 Nic_Wow      hia
02:02 Nic_Wow      Have any of you considered using RIOT instead of IRC for chat??
02:04 Quiark       there is Minetest room on matrix.org
02:14 Nic_Wow      ohhhhhhhhhhhhh
02:16 Nic_Wow      when you click "community" on "minetest.net" it just has "Overview", "Forum", "Wiki", "IRC", "Subreddit", "Contributor" and, "Server list"
02:19 Nic_Wow      @Quiark would you mind sending me the URl?
02:20 Quiark       uh, my client doesn't display it but you should be able to find it easily ... also the room is quiet most of the time
02:21 Nic_Wow      I couldn't fin it :/
02:21 Nic_Wow      Where would I find it?
02:22 Quiark       #minetest-matrix:matrix.org
02:35 ChimneySwift IRC is the official chat for minetest
02:36 ChimneySwift also I don't think it's possible to reload a mod at runtime since they all share the same namespace
02:36 ChimneySwift not at runtime, while running
03:07 aidalgol     Wouldn't it make more sense to just use a Matrix<->IRC bridge instead of creating yet another chat room?
03:08 ChimneySwift it would, but I don't think the devs would want a bridge on the channel
03:09 aidalgol     You already can join this channel via a bridge.  It's completely transparent.
03:09 ChimneySwift it's not like they're official channels anyway
03:09 aidalgol     I'm using IRC right now via Matrix.
03:09 ChimneySwift you could, but sending messages from there to here is the issue
03:10 ChimneySwift you can't change nicks on the fly like that, so it looks ugly
03:10 ChimneySwift and it really isn't hard to just join irc if ya need some help n
03:11 ChimneySwift :p
03:11 aidalgol     I haven't had any issues.
04:18 Quiark       too bad it can't reload the code. Any tips for efficient dev/test cycle?
05:15 ChimneySwift Quiark: I just reload the game a lot lol
05:45 ANAND        Quiark: Using minimal instead of MTG would help with loading the game faster. You'd still have to keep restarting the game though :)
13:51 deltasquared what would be the recommended way to build minetest from sources with debug symbols? I did a build of master but the profiling thing I'm doing (based on perf record) is giving me several blocks of [unknown] still
13:52 deltasquared I did try export CFLAGS="-g" beforehand, not sure if that is sufficient
13:52 calcul0n     i'd try -g -O0
13:53 deltasquared (it's a "flame graph", wanted to know what was being slow for something I was doing - the current output can be seen here: https://ptpb.pw/CCRI.svg )
13:53 calcul0n     ho well, that's the default, not sure it will change anything :)
13:54 deltasquared gcc man page suggests -Og for some optimisations but better debugging, so I'll play around with it a bit.
13:55 deltasquared ...
13:55 deltasquared hmm, why isn't it... bah. rm -rf and retry
13:55 deltasquared there we go.
13:55 deltasquared aaaaand wait another ten minutes.
14:14 deltasquared ok, still having difficulties here, even after compiling with CFLAGS="-g -Og" and verifying the options were set in the build files. https://github.com/brendangregg/FlameGraph is the tool I was using, capture data was via "perf record -F 100 -g -p $minetest_pid". it still seems to be showing blocks of "unknown" - https://ptpb.pw/5Dsk.svg - might this be due to something going on in lua? (but then that doesn't seem right, there's some
14:14 deltasquared irrlicht stuff going on called by the big "[unknown]" block in turn called by the "minetest" block at the bottom)
14:16 deltasquared would anyone be capable of assisting me with debugging of these prolifer attempts? all I wanted to see was what was causing everything to slow down in a particular use case...
14:28 Ingar        dependencies that don't have debug code ?
14:28 Ingar        I was goign to ask nvidia, but I see intel libs ;)
14:33 rubenwardy   The width is time
14:33 rubenwardy   Having tall things isn't an issue as long as they're not wide
14:33 rubenwardy   So you want to reduce the width of things
14:33 deltasquared Ingar: it's possible, I guess. i965_dri.so is def. there... maybe that's in mesa code, though I'd expect there to be intervening stack frames between "minetest" and the calls to the driver module
14:33 deltasquared rubenwardy: yeah I'm aware
14:33 rubenwardy   That being said, tall things run the risk of stack overflow
14:33 rubenwardy   And are probably inefficient maybe
14:33 rubenwardy   But heh
14:34 rubenwardy   Oh wait
14:34 rubenwardy   Nevermind, misunderstood
14:34 deltasquared rubenwardy: can you see the [unknown] blocks? I'm trying to figure out what those might be
14:35 deltasquared apparently this is more to do with the "perf record" invocation, as flamegraph isn't doing anything with those, just leaving them be
14:35 deltasquared I did try doing a record using dwarf debugging info for stack frames, but the resulting data file is now causing "perf script" to core dump :/
14:36 deltasquared ... and "perf report" it looks like.... maybe I'll just turn down the sample interval a bit
14:36 deltasquared (I would have thought 100hz wouldn't have been fatal, but oh well)
14:36 deltasquared (no, it's not an OOM condition AFAICT)
14:38 deltasquared ... ok, that seems to have worked, going to have another look now
14:39 deltasquared ... ah, that's a lot different. https://ptpb.pw/P6yI.svg
14:39 deltasquared LAWL there's a function called the_game
14:42 deltasquared hmm, that doesn't actually tell a lot, still a lot of time spend in the driver. hrm
14:44 deltasquared see that's still a bit annoying because it's not telling me what's going on inside the driver >_>
14:45 deltasquared for disclosure's sake, I'm currently trying to obtain some empirical data showing why drawing lots of entities (literally, only the cube drawtype atm, so only what 12 tris per entity?) would bring my FPS to a crawl
14:46 deltasquared counters and stuff in game show it's somewhere in the region of 600 in that area of the world atm and my FPS drops to about 6. which is unfortunate for me because it's kinda central to my use case...
14:47 deltasquared intuition tells me most systems should handle 7,200 triangles per frame quite easily, so I would assume there is something else going on.
14:47 deltasquared I have a hunch what it might be, but simply looking at the profiler output from outside hasn't really confirmed nor denied it yet
14:51 deltasquared all of this is boiling down to: minetest is slow in this particular case, it is seemingly illogical why it should be, and I want to try and help fix it.
14:51 rubenwardy   also
14:51 rubenwardy   -DCMAKE_BUILD_TYPE=Debug
14:51 rubenwardy   for debug flags
14:52 deltasquared rubenwardy: cheers, I didn't think a release build would be the default, but I will remember that one
14:54 sfan5        deltasquared: are you using luajit? that code cannot be profiled
15:01 rubenwardy   what even is that svg file
15:01 rubenwardy   how is it interactive
15:01 rubenwardy   whaaat
15:01 deltasquared rubenwardy: javascript can go in svg, apparently
15:01 deltasquared don't look at me, I just ran the generator! :P
15:01 deltasquared again, it's https://github.com/brendangregg/FlameGraph
15:02 deltasquared it is cool though, no?
15:04 rubenwardy   do you have a specific issue you're trying to find the cause of?
15:06 deltasquared rubenwardy: yeah, why drawing hundreds of entities with really simple meshes (namely, the cube visual) is so dang S L O W
15:06 deltasquared my intuition and (admittedly limited) 3D graphics knowledge is telling me this should not be so
15:06 rubenwardy   try the entity PR
15:07 deltasquared link?
15:07 deltasquared *opening browser, gis sec*
15:09 rubenwardy   can't find it
15:09 rubenwardy   here we are: https://github.com/minetest/minetest/pull/6803
15:09 rubenwardy   I'm not sure this applies here though
15:09 rubenwardy   as it's a cube
15:13 deltasquared rubenwardy: doesn't look like it - it seems to be more changing the mesh up for items as opposed to idk how they're issued for drawing
15:13 rubenwardy   well, idk
15:13 deltasquared nor do I >_>
15:14 rubenwardy   it's probably doing individual draw calls
15:14 deltasquared OOF
15:14 deltasquared I'd have to look into that. (diving into irrlicht too? how fun)
15:15 deltasquared though from what I've heard around the internets, a drawcall per entity, if that were the case, could be a very viable explaination
15:18 Hijiri       shouldn't hundreds of drawcalls be not crippling either
15:18 Hijiri       especially if it's not doing a lot of state changes between them
15:19 deltasquared Hijiri: the possibility of state changes being done by say irrlicht is bothering me, hence why I feel the need to go on a source code dive.
15:19 deltasquared superfluous state changes, that is
15:19 Hijiri       you could also use apitrace
15:20 deltasquared Hijiri: hmm?
15:20 Hijiri       it's a OpenGL debugging tool that will trace all the openGL library calls that happen during a program's executino
15:20 Hijiri       sometimes it's hard to read but it will show every OpenGL thing that happens
15:21 Hijiri       so if you can see the state switching in there you might be able to confirm it before having to dive in the source code
15:21 deltasquared however, I do feel you are right, it's not like we can currently set custom shaders for entities, therefore it *should* be possible to arrange all the vertices of all visible entities into a single batch and chuck that in one go. (though, it does raise the question... does irrlicht set a translation uniform for instance... yeah I had better have a look at that apitrace thing)
15:21 Hijiri       I didn't say that
15:21 Hijiri       I said that one draw call per entity shouldn't be so bad as to reduce FPS to 6
15:22 Hijiri       as long as context switching isn't happening between
15:22 Hijiri       arranging all the vertices together would probably reduce performance because you would need to send the vertices to the GPU every frame
15:23 Hijiri       deltasquared: are these hundreds of entities collidable?
15:23 Hijiri       I've never had performance problems from using pipeworks with a lot of "item position marker" entities floating around the pipes
15:24 Hijiri       However if it's a lot of collidable entities, it could be slowing Minetest down because of how entity collisions are calculated
15:25 sovetskiy    rubenwardy, do work with entities described in your modding book?
15:28 Hijiri       deltasquared: entity collision checking is very slow if you have lots of entities that can collide with other entities
15:28 Hijiri       If those entities are collidable I would set them to not be collidable and see if they still cause FPS drop
15:29 deltasquared Hijiri: I'll try that actually, they are collide-able (that is somewhat the point of my use case, but very well might as well give it a try)
15:31 deltasquared Hijiri: ... no change. setting entity property defaults to physical = false / collides_with_objects = false, but still the same
15:32 Hijiri       ok, I guess it's not that then
15:32 Hijiri       that's weird though, because I haven't seen the same problems with lots of entities
15:32 MinetestBot  02[git] 04numberZero -> 03minetest-mods/mesecons: Remove tiny (+0.001) selection box oversize 13a4f5ae5 https://git.io/fAgsh (152018-09-07T15:31:49Z)
15:32 MinetestBot  02[git] 04numberZero -> 03minetest-mods/mesecons: Make insulated wires’ selection box fit in the node 139e6eac4 https://git.io/fAgsj (152018-09-07T15:31:49Z)
15:32 Hijiri       you could open a github issue, including your hardware and graphics settings if not the default
15:33 Hijiri       and maybe a test mod
15:34 deltasquared Hijiri: hah, test mod, I can show you if you want, it's actually something I'm working on atm
15:34 deltasquared I have a test server with it on. would you like to see? (just so everyone doesn't know I'm mad >_>)
15:35 deltasquared err s/know/think/
15:35 Hijiri       ok, sure
15:35 deltasquared give me a sec, updating it
15:36 Hijiri       unless the problem is obviously your fault though, I would still open a github issue
15:36 Hijiri       I don't know if I'll be able to tell what's wrong just by looking
15:36 deltasquared Hijiri: daconcepts.com 33000
15:36 deltasquared Hijiri: and yes I will consider that
15:40 Jhalman      ##minetest_ctf
15:40 Jhalman      join ##minetest_ctf
15:46 deltasquared Hijiri: can you come back? I goofed with the lua command. that was slightly *too* many of them.
15:51 Hijiri       deltasquared: sorry, had to leave the game
15:51 Hijiri       I'm at work currently
15:52 deltasquared Hijiri: 's fine
15:55 deltasquared Hijiri: guess I'll prepare a minimal test case for the github issue then
22:52 NicWow       so what I say here shows up in game
22:52 NicWow       (wrong chat))
22:52 sfan5        what
22:53 Hijiri       probably thought this was an IRC channel bridged with a minetest server?
22:54 piesquared   No, he is on a different channel thats bridged.
22:56 Hijiri       I mean that he thought that this channel was that channel
23:18 NicWow       out of curiosity, do can you make an afk fish farm in Minetest (I played minecraft years ago)
23:25 Jordach      no
23:25 Jordach      there is no fishing
23:31 calcul0n     there are some fishing mods, but afaik none let you farm them
23:32 calcul0n     i mean you must fish by hand
23:39 MinetestBot  02[git] 04paramat -> 03minetest/minetest: Particles: Make collision with objects optional (#7682) 13766fb7b https://git.io/fAgx5 (152018-09-07T23:38:35Z)
23:55 paramat      there's no fishing in minetest game, but a mod could do it, or a different game