Time  Nick       Message
17:02 sfan5      In case anyone was wondering what it looks like when you give 10 players with random_input random items to mess around in creative https://0x0.st/oAnb.png
17:02 sfan5      https://0x0.st/oAnc.png but wait there's more
17:06 Krock      the outcome is also pretty random
17:13 erle       sfan5 what was your goal there, finding crashes or something?
17:13 erle       or just playing around
17:14 sfan5      load testing
17:14 celeron55  random_input with random items is a surprisingly good fuzz test
17:14 celeron55  something more organized would probably be much more efficient though
17:46 sfan5      if appears 35% of time in LuaEntitySAO::step() is spent inside l_get_properties, that's pretty bad
17:46 sfan5      the item entity calls this every step
17:46 sfan5      s/^if/it/
17:47 sfan5      to retrieve a single property, of course
17:48 MTDiscord  <luatic> the fundamental problem here is the Lua - C struct duality
17:48 Krock      seems logical to me that it spends 33% of the time when there's only like 3 functions to process
17:48 MTDiscord  <luatic> the data structures are constantly converted back and forth
17:48 MTDiscord  <luatic> Krock: the problem is that it has to fill the entire property table each time it's called
17:48 sfan5      okay true
17:48 MTDiscord  <luatic> even if usually only a single property is needed
17:48 Krock      measure the effective time, and not percentages.
17:48 MTDiscord  <luatic> true
17:49 MTDiscord  <luatic> by measuring percentages Festus managed to create a virtual performance problem with player_api that just isn't there
17:49 sfan5      3.764E+08 (35.8%) aggregated sample costs in ObjectRef::l_get_properties(lua_State*) (minetest) and below
17:49 sfan5      so how do I convert that to seconds
17:50 Krock      you'd need a reference average execution time of the entire function
17:51 Krock      de facto a ScopeProfiler within the cpp file
17:52 erle       <celeron55> something more organized would probably be much more efficient though
17:52 erle       celeron55 which for me leads directly to the question of “which CSMs from waspsaliva would help debugging in minetest”
17:53 erle       sfan5 how do you profile, like with what tooling?
17:53 sfan5      perf
17:56 erle       thx
18:05 erle       sfan5 which perf subcommand shows function time? can't seem to find any
18:05 sfan5      tell me when you find out
18:13 erle       perf record --call-graph fp $COMMAND; perf report
18:30 erle       if i am interpreting this correctly, starting minetest, loading mineclonia, then exiting the game after walking around a tiny bit and noticing perf does not make it slow down to a crawl results in about 2% of the entire execution time was spent doing useless SHA1 stuff
18:46 celeron55  useless?
18:46 celeron55  i guess it's checking that the cache is consistent
18:46 celeron55  could of course make that optional
18:47 celeron55  or shortcut it altogether in singleplayer
19:17 sfan5      sounds like a good idea
19:17 sfan5      will you send a PR? ;)
19:56 MTDiscord  <wwar> #12002 looks confirmed to me
19:56 ShadowBot  https://github.com/minetest/minetest/issues/12002 -- Enabling fixed_virtual_joystick doesn't fix virtual joystick
20:01 celeron55  sfan5: yes just give me... a decade
21:00 sfan5      apparently the client spends roughly identical time on rendering as on deallocating MapBlockMesh objects
21:00 sfan5      (in my stress test conditions)
21:01 sfan5      std::map<std::pair<u8, u32>, std::map<u32, video::SColor > > this does not look efficient
21:01 erle       wsor i made a new tiny mod! https://content.minetest.net/packages/erlehmann/mcl_polished_stone_stairs/
21:02 erle       (polished stone stairs are the only stairs missing from mcl* games … because they are not in minecraft)
21:14 sfan5      eh it's not that bad, peak I've seen is 3% of time spent in ~MapBlockMesh()
21:14 sfan5      might make a different on low-end devices though?
21:38 erle       sfan5 how to stress test this? fly in a straight line?
21:38 sfan5      produce the situation in my screenshots earlier
22:05 sfan5      after refactoring the code I have managed to decrease performance
22:05 sfan5      this must mean the measuring is wrong
22:22 Eytirth    I am having errors with books_rx, safety deposit  and related mods, my errors: https://pastebin.com/iNkUmDhP
23:28 settl3r[m] Is there a tool with which i can locate single trees (such as rubber tree) in my local world?
23:28 settl3r[m] (i already used fly+fast+noclip mode, and searched through acres of jungle, to no avail :/ )
23:30 settl3r[m] (i only use that bit of cheating, to locate some resources which are very hard to find; normally i play this map in survival mode)
23:30 settl3r[m] (and i don't want to outright cheat these resources out of nothing, as that would totally kill my (rest of) immersion)
23:32 settl3r[m] can i locate rubber trees with WorldEdit? (without changing anything)
23:32 settl3r[m] * changing anything in my survival world)
23:33 settl3r[m] * bit of (information-gathering-)cheating, to
23:34 settl3r[m] so i only need a way to get information in creative mode, not ways to summon that stuff
23:34 settl3r[m] * information in (the creative mode, * creative mode copy of) that map, not
23:34 settl3r[m] * so i only need a way to get information in (the creative-mode copy of) that map, not ways to summon that stuff
23:34 settl3r[m] * so i only need a way to get information in (the creative-mode-copy of) that map, not ways to summon that stuff
23:40 settl3r[m] * so i only need a way to get information in (the creative-mode of) that map, not ways to summon that stuff
23:40 settl3r[m] * information in (the creative mode, * creative mode of) that map, not
23:42 settl3r[m] (I need that rubber, to craft MV cables, using technic mod)