Minetest logo

IRC log for #minetest-dev, 2013-02-21

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

All times shown according to UTC.

Time Nick Message
02:06 deivid anyone here?
02:18 RealBadAngel whats up?
02:23 deivid I'm trying to make the blocks draw partially based on distance, but I don't know how to do it, currently the renderMap gets all the meshes from the block and draws them
02:23 deivid s/blocks/nodes
02:23 RealBadAngel theres a bit of code called farmesh
02:24 RealBadAngel maybe you should look for it
02:24 deivid thanks
02:25 RealBadAngel it was supposed to do somethin like you mentioned but im not sure of its state right now
02:25 VanessaE joined #minetest-dev
02:25 RealBadAngel wb VanessaE
02:25 VanessaE thanks.
02:54 hmmmm hey vanessa, have you been looking at my branch at all?
02:55 hmmmm i need real-world testing
02:55 hmmmm lots of it
02:55 hmmmm specifically, single player
02:56 VanessaE I've been waiting for you to tell me to try it :-)
02:57 VanessaE didn't want to pester you about it
02:57 hmmmm oh..
02:57 hmmmm sorry
02:57 hmmmm you see, i'm having a pretty big problem i've been unable to resolve
02:57 VanessaE oh?
02:58 hmmmm and i don't think there's a solution at all for it
02:58 hmmmm so i need to see if people think it's actually a problem or not
02:58 hmmmm or maybe it's just i see a problem with hyperthreading
02:58 hmmmm it's just that i*
02:59 VanessaE does it matter if I'm using minimal?
03:00 hmmmm not at all
03:00 VanessaE ok, it's up, in singleplayer mode, with a new map
03:00 hmmmm wait wait
03:00 hmmmm gotta change some configuration settings....
03:00 * VanessaE waits
03:00 VanessaE ok, shoot
03:00 hmmmm first off, set num_emerge_threads = 4, then emergequeue_limit_generate = 4, emergequeue_limit_diskonly = 20
03:01 VanessaE ok, done
03:01 hmmmm so tell me with those settings, if singleplayer is too choppy
03:02 hmmmm and by choppy i don't mean the block loading really, i mean the responsiveness of the gameplay
03:02 VanessaE it seems respectable at this setting actually.
03:02 VanessaE an occasional frame hiccup, but I've always seen those every so often
03:02 VanessaE nothing I'd call "choppy" by a long stretch
03:02 hmmmm ever since 0.4.4, anyway :p
03:03 hmmmm okay, you sure about that?
03:03 VanessaE lemme test more
03:03 hmmmm then it must be because of my hyperthreading
03:04 VanessaE well, there is a weird glitch when I jump, but it might be an optical illusion.
03:04 hmmmm jump?  not related to my stuff
03:04 VanessaE movement seems...  mm, ...  sharp?
03:04 VanessaE but no, it seems smooth enough to me
03:06 hmmmm when sapier asked you to try out his fix for the hangups, you had to run minetestserver as root, right?
03:06 VanessaE nope, I was running it as a normal user, as always
03:06 VanessaE same here
03:06 hmmmm really?
03:06 VanessaE yup, I never run things as root unless I have to
03:06 hmmmm if that's the case, then it was a placebo
03:06 VanessaE heh, one hell of a placebo then :-)
03:07 VanessaE still, it was only a marginal effect
03:07 hmmmm under Linux, thread priorities have no effect unless it uses a real-time scheduling policy, and you need to be root in order to switch to a real-time scheduling policy
03:08 VanessaE bit of mapgen lag there (new terrain)
03:09 hmmmm if that's all, then that's fine... what i'm concerned is stuttering in the rendering
03:09 VanessaE naw, the rendering quality seems good
03:09 VanessaE certainly no regression that I can see
03:10 VanessaE in fact, you know...I'm starting to wonder if this is maybe a little smoother than I'm used to
03:10 VanessaE lemme torture it a little --> HDX
03:10 hmmmm what's HDX?
03:10 VanessaE my HD texture pack
03:10 hmmmm oh
03:11 VanessaE there's a slight stutter
03:11 VanessaE nothing serious
03:11 VanessaE not enough to worry about anyway
03:11 hmmmm alright
03:12 VanessaE I'd put it at "bottom of the To-do list:  tune rendering smoothness"
03:12 VanessaE as in, practically lowest priority
03:13 * VanessaE turns on some extra gfx settings
03:13 hmmmm i just want to make sure it's perfect
03:14 VanessaE oops
03:14 VanessaE got it to dump core
03:14 VanessaE http://pastebin.com/Pu1Tv2Pb
03:15 VanessaE started a new map after installing minetest_game and turning on my usual settings.
03:15 VanessaE hm, it didn't actually generate a core file though
03:15 VanessaE *shrug*
03:16 hmmmm hm
03:16 hmmmm now why doesn't it say which thread made it crash
03:16 hmmmm oh i'm just blind
03:16 VanessaE lol
03:17 hmmmm yeah i dunno, the mesh update thread is totally not my area
03:17 VanessaE unrelated glitch then
03:17 hmmmm i don't know anything about it, and i'm totally unfamiliar with that code
03:17 VanessaE there we go
03:17 VanessaE 256px HDX, minetest_game, with all my usual settings
03:17 hmmmm doesn't have anything to do with the textures
03:18 hmmmm it just generates the mesh
03:18 VanessaE well no, but bigger textures will slow the video card down some, which would make any stuttering perhaps a little more apparent.  Still, it seems perfectly acceptable
03:18 hmmmm would you have any problem with this being merged to master?
03:19 VanessaE so far it seems reasonable to me; perhaps I should deploy this on the server for a while maybe?
03:19 hmmmm well, sure if you'd like, i'm not really concerned about that though
03:19 hmmmm i'm mostly concerned with single player stuttering, which doesn't seem to be bad
03:20 VanessaE woof! look at those cores go
03:20 hmmmm oh yeah
03:20 VanessaE this is able to max out four of 6, plus 50% of the other two
03:20 hmmmm you might want to crank it down to 3 cores or something, since you do other things on your computer while you host that
03:21 VanessaE this is smooth as cotton (not quite silk ;-)  )
03:21 VanessaE roll it out.
03:21 hmmmm well i've gotta ask more people
03:21 hmmmm i mean i don't want to screw things up
03:21 VanessaE right, that's fair
03:22 VanessaE in that case, let's just say "no, I see no problem with merging it to master" then.
03:23 VanessaE hm, a little stutter there, probably frequency scaling lag
03:26 RealBadAngel hey hmmm
03:26 RealBadAngel im testing it too and noticed a glitch
03:26 hmmmm what's that
03:27 RealBadAngel hold on for screenshot
03:27 VanessaE huh.  <offtopic> interesting cavegen bug.  Bunch of caves cut off abruptly at z=-672.5 </offtopic>
03:27 hmmmm agh the cut-off cave bug is what you're talkign about RBA?
03:29 VanessaE http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/screenshot_4207625953.png
03:29 VanessaE in case you were wondering.
03:29 hmmmm i know
03:29 hmmmm that's caused by my fix for the CONTENT_IGNORE blocks
03:30 VanessaE anywho, I think we're good from the standpoint of your latest updates.
03:30 hmmmm with one emergethread this wasn't a problem
03:30 RealBadAngel http://i.imgur.com/8NtVY4Q.jpg
03:30 hmmmm woah.
03:30 VanessaE RealBadAngel: shame on you for using jpg for screenshots :-)
03:30 hmmmm now that is a problem
03:30 VanessaE hmmmm: I see than with the current upstream mapgen from time to time also, though usually it's a horizontal cut, not vertical
03:30 RealBadAngel tree is cut like on mapblock border
03:30 hmmmm it is
03:31 VanessaE the top of a moretrees tree may fail to appear when spawning from a planted sapling, f.ex
03:31 VanessaE that*
03:31 VanessaE left #minetest-dev
03:32 VanessaE joined #minetest-dev
03:32 VanessaE oops/
03:33 VanessaE whoa.  New problem
03:34 VanessaE we don't care about map loader/mapgen lag at all?  got a sudden, extreme spike here while the game was minimized to handle that screenshot - it's still stuck.
03:34 VanessaE one core is maxed, but the game's responding otherwise.  /status, movement, etc.
03:34 hmmmm hmmm
03:35 VanessaE no mods - just minetest_game
03:35 hmmmm i think i've seen something like that before
03:35 hmmmm it un-freezes though
03:35 hmmmm i need a good way to dump the current status of the queues
03:35 VanessaE oh no.
03:35 RealBadAngel i must say game is workin way faster imho
03:36 VanessaE are we pushing into tens of thousands again? :-)
03:36 hmmmm no
03:36 VanessaE hah ok
03:36 hmmmm more like the 3s or the 4s
03:36 RealBadAngel got 25fps with max view distance
03:37 VanessaE oh fuck
03:37 VanessaE camera update got turned off.
03:37 hmmmm haha..
03:38 VanessaE (how?  I didn't hit that key?)
03:38 hmmmm it's good to know that's not my problem :P!
03:38 VanessaE lol
03:38 hmmmm i still think it's a good idea to have a way to reset queues manually perhaps..
03:39 RealBadAngel well
03:39 RealBadAngel and it is starting in no time with preloaded items visuals
03:39 VanessaE that said, it is starting to lag a little, but no appreciable effect on FPS (as if that were likely anyway)
03:40 RealBadAngel wow
03:40 RealBadAngel 2-3s to launch the game
03:41 RealBadAngel shit
03:41 RealBadAngel i had this off, sorry :)
03:42 VanessaE let's see what happens when I put moretrees on this now.
03:42 hmmmm hrmm
03:43 hmmmm well the client rendering thread isn't stuttering as much as i expected
03:43 hmmmm but i definitely need to fix the caves/trees getting cut off
03:43 VanessaE hoo... this is decidedly more laggy than before
03:45 VanessaE one core spiked (client/rendering), with occasional spikes of 3 or more cores right when the generated world appears
03:46 hmmmm sounds normal enough
03:46 VanessaE but it's taking several seconds at a time without any appreciable CPU usage, then huge CPU spikes as the new land appears
03:46 hmmmm hmm
03:47 hmmmm okay, thanks.  i'll figure something out
03:48 VanessaE lemme guess, that locking issue again?
03:48 hmmmm no, it's completely different
03:48 VanessaE oh ok
03:49 VanessaE this is new:
03:49 VanessaE during those lags, the server isn't responding to commands
03:49 VanessaE (usually it continues to respond even if the terrain is being withheld)
03:51 hmmmm weird.. i have no explanation for that
03:51 hmmmm what respond did you give it exactly?
03:52 VanessaE /time 6000
03:52 VanessaE usually I just do /status, but it was getting dark, so..
03:52 VanessaE by "respond" I mean the server's talkback in the chat area, not the terrain/light level/etc
04:29 thexyz joined #minetest-dev
04:30 sfan5|OFF joined #minetest-dev
04:32 TB`oFF|Vibe-X joined #minetest-dev
04:40 hmmmm pfffffft.... race condition in the thread logging thing
04:40 VanessaE oops!
04:40 VanessaE haha
04:42 hmmmm log_register_thread is also not threadsafe
04:42 hmmmm i would like to avoid a mutex..
04:43 VanessaE I'm headed off to bed, I'll read/respond/test again when I return.
04:43 VanessaE Night :-)
04:54 ffoxin joined #minetest-dev
04:54 deivid hmmm are you still here?
04:54 hmmmm hello..?
04:54 deivid hi
04:55 deivid I'm going to start working on minetest, because I like the idea
04:55 deivid I work as a programmer so I'm ready to start; should I just check the github issues?
04:56 deivid I'm going to start with something simple until I get used to the structures/etc
04:57 hmmmm a lot of the issues are either already being taken care of, or aren't really issues, or are really hard to fix the right way and the fixes that are offered are cruddy
04:57 hmmmm like i guess you could start by doing that
04:58 deivid uhm, ok
05:00 hmmmm but don't be disappointed if you can't find anything that's decent to work on
05:54 deivid I tested minetest on a window manager without reparenting (awesomewm) and it won't draw anything at all unless I minimize it and then restore it
06:07 hmmmm sounds like an awesomewm problem or an Irrlicht problem...
06:07 hmmmm either way, we can't fix that
06:07 celeron55 F00: i agree with convenience over reality; i believe it's the all-around design principle in a game like this
06:09 F00 I think perhaps the dev builds will allow for some more community feedback before 0.4.5
06:13 hmmmm celeron we need to talk about multiple mapvoxelmanipulators with overlapping areas of the map
06:13 F00 My feeling is that hitting water should result in some sort of exponential decel, whereas climbable nodes should do the usual magic instant stop
09:18 Calinou joined #minetest-dev
09:43 jin_xi joined #minetest-dev
10:40 darkrose joined #minetest-dev
11:02 rubenwardy joined #minetest-dev
11:04 SpeedProg joined #minetest-dev
12:56 troller joined #minetest-dev
13:28 troller joined #minetest-dev
13:53 proller joined #minetest-dev
14:10 hmmmm joined #minetest-dev
14:27 rubenwardy joined #minetest-dev
14:29 DpEpsilon joined #minetest-dev
14:47 DpEpsilon joined #minetest-dev
15:10 rubenwardy joined #minetest-dev
15:11 Jontte joined #minetest-dev
15:51 PilzAdam joined #minetest-dev
16:09 Calinou joined #minetest-dev
16:46 hmmmm sooo
16:47 hmmmm looks like celeron never got back to me about that
16:55 zeruel joined #minetest-dev
17:00 david_asd joined #minetest-dev
17:02 iqualfragile joined #minetest-dev
17:08 troller joined #minetest-dev
17:19 troller celeron55, https://github.com/celeron55/minetest/pull/492
17:26 rubenwardy joined #minetest-dev
17:31 celeron55 hmmmm: i didn't bother saying anything as you went away
17:31 hmmmm anyway do you have any ideas on how to handle that
17:31 celeron55 troller: you don't need to touch it for now; it will be merged once somebody is unlazy enough
17:31 celeron55 troller: remind after a week if that hasn't happened
17:32 hmmmm i worked around the problem it causes but people are complaining about the side effects enough, so there needs to be some better way
17:32 celeron55 what is the workaround?
17:33 hmmmm if a block that came from initialEmerge was created 'blank' (filled with all CONTENT_IGNORE) and it is a border block, it won't be written in the blit back
17:34 hmmmm the side effects of this are obviously going to be cut-off caves and trees
17:35 celeron55 well, a good solution probably requires some kind of a merge strategy?
17:35 hmmmm no, a good solution requires a mutex
17:35 hmmmm but that's not a smart solution
17:35 celeron55 you could disallow starting to generate a chunk adjacent to a chunk that is being generated
17:36 celeron55 that will probably create large latencies in some cases
17:36 hmmmm that's true, but it would complicate emergethread logic by a bit
17:37 hmmmm plus there would still be a problem since the chunks being generated won't be close enough
17:37 hmmmm if it's a checkerboard pattern of generation you'd still have two chunk borders sharing the same block
17:38 celeron55 well, then there is the MC way
17:38 celeron55 that will surely work
17:38 hmmmm what is the MC way again?
17:39 celeron55 except for content generated by mods, because there is no well established framework for them to generate the needed kinds of predictability
17:40 celeron55 it generates chunks in parallel so that it over-generates every discrete thing in such an area that everything that might touch the chunk gets generated, and discards the parts that aren't relevant
17:40 troller celeron55, squash it again?
17:41 celeron55 troller: feel free to
17:45 celeron55 this really is a hard problem 8)
17:45 celeron55 and mojang once again isn't as dumb as people think
17:47 celeron55 i think just merging the results so that "most interesting" things are preserved would work just fine
17:48 celeron55 that is, trees and caves preferred over ground, and so
17:48 celeron55 but it's pretty much up to whoever (=hmmmm) who implements it
18:01 hmmmm i don't like mojang's idea, and i did consider keeping only the most interesting parts
18:05 hmmmm so anyway here's what i was thinking
18:05 hmmmm if a node is content_ignore, just don't write anything
18:05 hmmmm leave it the way it is
18:06 hmmmm the obvious downside to that is now we can't use memcpy
18:13 hmmmm so you go to generate a chunk, you grab the borders, a couple of which are blank cignore blocks.  you leave the envlock.  another emergethread goes to grab blocks, it gets a blank block as well, your second thread finishes first and writes the terrain to the block, the first thread finishes and writes all of the nodes that are not content_ignore to that block
18:14 hmmmm in the same manner, your terrain generator grabs a block that's mostly content_ignore but has some caves in it or whatever.  it generates its terrain, then it goes to blit it back, but only overwrites the content_ignore nodes, leaving the caves that were generated intact
18:17 hmmmm hmmmmmmmmmmmmmmmmmmmmmm... i'll just do this, and then when i go to finally add in the SSE noise things, i'll add in an SSE optimized version of this as well
18:18 BackupCoder joined #minetest-dev
18:18 celeron55 umm...
18:18 celeron55 how are you going to handle air?
18:19 hmmmm air is just as valid of a block i'd say
18:19 celeron55 what about trees then?
18:19 hmmmm trees are the same as caves
18:19 celeron55 you didn't really solve anything as far as i can see, what you said is roughly what happens already
18:20 celeron55 oh well whatever, i guess you'll figure it out
18:20 hmmmm well the difference is that i do all of this in the blit back
18:20 hmmmm so i'm comparing with the most recent version of the block
18:21 celeron55 hmm, now i get it
18:22 celeron55 umm... maybe
18:22 hmmmm by the way, why did you do that stuff with the vmanip/mapblock "copyTo" and "copyFrom"?  i found that confusing and verbose
18:22 celeron55 dunno
18:23 hmmmm and a lot of stuff in map.cpp is repetitive
18:23 hmmmm i want to clean that up someday
18:28 proller__ joined #minetest-dev
18:49 doserj joined #minetest-dev
19:12 rubenwardy1 joined #minetest-dev
22:08 hmmmm hey if you're still around i need to ask you about the profiler things
22:08 hmmmm what is the difference between add and avg
22:17 Taoki joined #minetest-dev

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