Minetest logo

IRC log for #minetest-dev, 2015-12-10

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

All times shown according to UTC.

Time Nick Message
00:12 alket joined #minetest-dev
02:01 nolsen joined #minetest-dev
02:03 sofar where is the default plantlike mesh defined?
02:04 Supertan1er2 joined #minetest-dev
04:18 hmmmm content_mapblock.cpp line 1109
04:26 Player2 joined #minetest-dev
05:53 Supertanker2 joined #minetest-dev
06:00 Supertanker2 joined #minetest-dev
06:26 sofar hm, now I have to figure out what an S3DVertex is :)
06:39 Hunterz joined #minetest-dev
07:03 sofar hmmmm: https://github.com/minetest/minetest/issues/1589
07:58 sofar I think we can make especially grass look a lot better. Even if this could be done just for default:grass_*, it would be a major visual improvement
08:01 troller joined #minetest-dev
08:39 jin_xi joined #minetest-dev
09:06 nrzkt joined #minetest-dev
09:39 RealBadAngel joined #minetest-dev
09:42 RealBadAngel sofar, here?
10:23 kilbith joined #minetest-dev
10:23 kilbith paramat, look what i found : https://lut.im/ePUpE4a7Y9/KCu0cNmnHLnBOTvm.png
10:26 Calinou joined #minetest-dev
10:37 RealBadAngel acaccia zipped? ;)
11:08 nrzkt kilbith, great
11:09 Hunterz it is monument of aliens
11:20 kilbith "monument of aliens" is a good excuse for all mapgen craps...
11:34 Hunterz hehe
11:53 Fixer joined #minetest-dev
12:30 behalebabo joined #minetest-dev
13:45 Taoki RealBadAngel: Any updated news on minimap and realtime lighting BTW, now with holidays approaching and all? Not expecting things to happen overnight of course, just curious if you have any plans expected on a timelmine :)
13:45 RealBadAngel rather not any soon, those are things i cant handle alone
13:46 RealBadAngel i keep asking for help when i see the wall but theres no response
13:46 RealBadAngel ive chosen to go back to the field i know good enough to be the only one there
13:47 RealBadAngel instead of sitting 2-3 weeks in front of rather trivial problem (but not in my area) i can code something else
13:48 RealBadAngel those who could help me chosen to sit and watch
13:54 RealBadAngel taoki, anyway, you will love next PR, thats a promise
13:55 RealBadAngel its gonna be kinda compilation of the few ideas that were here already
14:11 RealBadAngel https://github.com/minetest/minetest/pull/3442
14:12 Taoki Awesome... can't wait :)
14:12 RealBadAngel can somebody give it one more +1 and merge it?
14:14 zat joined #minetest-dev
14:14 * Taoki hopes the speedup means there won't be FPS hiccups with relief mapping enabled any more. IIRC that was a problem in the past
14:15 RealBadAngel this should speed up relief mapping quite a bit
14:15 RealBadAngel theres less iterations
14:15 Taoki Awesomeness.
14:15 RealBadAngel not to mention end effect is way nicer
14:20 RealBadAngel https://github.com/minetest/minetest/pull/3391 , any more comments on this PR?
14:22 RealBadAngel without 3391 in i cannot make PRs with liquid shaders (or any other ones)
14:25 RealBadAngel for 3391 i need more folks to test sample shaders...
14:28 Fixer any windows binaries?
14:34 RealBadAngel ask sfan5
14:35 RealBadAngel http://i.imgur.com/smyD4M8.png
14:35 RealBadAngel (playin with relief mapping to the max, checkin how far we can get now)
15:09 jin_xi joined #minetest-dev
15:12 younishd joined #minetest-dev
15:15 sofar RealBadAngel: was asleep...
15:27 RealBadAngel sofar, see my comnment on plants
15:31 sofar I saw
15:31 sofar mapgen should certainly do that
15:32 sofar but the offset random x,z works even better
15:49 Obani joined #minetest-dev
15:58 CraigyDavi joined #minetest-dev
15:59 Hunterz joined #minetest-dev
16:05 Darcidride joined #minetest-dev
16:11 Darcidride joined #minetest-dev
16:20 younishd_ joined #minetest-dev
16:20 hmmmm joined #minetest-dev
16:33 alket joined #minetest-dev
16:33 younishd joined #minetest-dev
16:39 RealBadAngel hmmmm, can you vote #3442?
16:41 RealBadAngel sofar, such modifications are very specific (i mean those you have proposed) while whats already in the engine is generic
16:42 CraigyDavi joined #minetest-dev
16:44 est31 joined #minetest-dev
16:45 est31 RealBadAngel, I've read through the pr to use binary search
16:45 est31 I now wonder: why make a linear search before the binary one?
16:45 RealBadAngel and?
16:45 est31 why not one single binary search?
16:45 est31 it would spare 12 more iterations
16:45 est31 or 11
16:46 RealBadAngel linear is to find something thats near a hit
16:46 RealBadAngel binary is to get to it as close as possible
16:47 RealBadAngel more linear steps you get better accuracy
16:47 est31 what precisely should the algorithm do
16:48 RealBadAngel find exact interesction eye ray and depth field at texture space
16:48 RealBadAngel but we dont need perfect one
16:48 RealBadAngel something close is enough
16:49 RealBadAngel do you want to read papers on it?
16:49 proller joined #minetest-dev
16:50 est31 yes
16:50 RealBadAngel hold on
16:52 RealBadAngel est31, ur email plz
16:53 est31 RealBadAngel, its in info tab in the client
16:54 est31 Credits tab*
16:54 RealBadAngel ok
16:54 zupoman joined #minetest-dev
16:54 zupoman joined #minetest-dev
16:55 est31 so I wonder, why not something like this instead? https://gist.github.com/anonymous/beed54f9e6d64c99e389
16:56 RealBadAngel sent
16:57 RealBadAngel code you highlighted will create 8 visible depth layers
16:58 RealBadAngel those could be eliminated with filtering
16:58 RealBadAngel but thats not the point, it will be fuzzy and setting dependent
16:59 RealBadAngel and will look just WRONG
16:59 est31 but the depth can have more than just 8 values
16:59 RealBadAngel yes
16:59 * est31 tries his code
16:59 est31 q
16:59 est31 wrong window lol
17:00 RealBadAngel you can divide it into lets say 256
17:00 RealBadAngel you will jamm your gpu
17:00 RealBadAngel wiser is to get close and look for most accurate one
17:02 RealBadAngel old code was using 25 steps (value picked from testing)
17:02 RealBadAngel new one has 16 + 6 steps
17:02 est31 no
17:02 RealBadAngel and is WAY more accurate
17:02 est31 it has 15 + 6
17:02 est31 not 16+6
17:04 RealBadAngel preli,inary step is not counted in, right
17:04 RealBadAngel but math is made for such steps
17:05 RealBadAngel thats why decrement is first thingy in the loop
17:08 RealBadAngel anyway, binary search eliminates visible displacement layers
17:09 RealBadAngel higher the scale of the effect, u notice it more
17:09 Fixer RealBadAngel, sfan5 does not have any win binaries with your shaders
17:10 RealBadAngel with the change you can try enormous scales and still get correct effect
17:10 est31 Fixer, you dont need to change any binaries
17:10 est31 Fixer, the shaders are not in the binary, but "in the open"
17:10 est31 just apply the patch manually
17:10 RealBadAngel est31, for last PR yes
17:11 RealBadAngel but i do have many waitin in line
17:11 RealBadAngel custom shaders patch is just happily hanging around :P
17:13 Fixer but it says Grab the branch with PR and compile:
17:13 Fixer https://github.com/RealBadAngel/minetest/tree/new_shaders
17:14 RealBadAngel Fixer, that branch need compiling, yeah
17:14 RealBadAngel relief mapping patch doesnt, since its altering only GLSL shaders code
17:14 est31 That's what I'm getting with your PR RBA http://pasteboard.co/2LXQwEXa.png
17:15 RealBadAngel est31, scale?
17:16 est31 RealBadAngel, what scale should I set?
17:16 RealBadAngel 0.08 is default
17:16 RealBadAngel make it 0.8
17:16 est31 is it "parallax occlusion scale"?
17:17 RealBadAngel yes
17:18 RealBadAngel its overall scale of the effect
17:18 est31 this is the scene with my "whole" binary search http://pasteboard.co/2LY66Cvq.png
17:19 RealBadAngel look at bricks edges
17:20 RealBadAngel you have fucked it up
17:20 est31 yeah
17:20 RealBadAngel you made layers
17:20 est31 you are right
17:20 est31 +1 to the pr
17:20 est31 can I merge it or do you want to change sth?
17:20 est31 also, the paper recommends to do linear search firt
17:20 RealBadAngel with your vote it has 2
17:20 est31 first
17:21 RealBadAngel i do
17:21 est31 yeah, therefore I can merge
17:21 RealBadAngel linear search is simple -0.1
17:21 est31 so, you want to change nothing anymore, I'll merge it then
17:21 RealBadAngel or whatever step
17:21 RealBadAngel funny name for just -
17:22 RealBadAngel no changes requsted
17:22 RealBadAngel its rdy
17:23 proller joined #minetest-dev
17:24 Krock joined #minetest-dev
17:24 Fixer congrats
17:25 est31 merge
17:25 est31 d
17:26 est31 will push commit like this in 10 minutes https://github.com/est31/minetest/commit/e8053dd781b2c8da40fc8cf5c786947dbfaa3ff6
17:26 est31 copy paste mistake
17:27 Fixer can developers consider changing autorange view system? it is crappy :(
17:28 celeron55 joined #minetest-dev
17:28 Fixer when you walk somewhere it constantly adjusts and trees or other objects in front off you show or dissapear all the time, it is noticable and annoying
17:38 celeron55 joined #minetest-dev
17:42 rubenwardy joined #minetest-dev
17:45 * Fixer looks at http://dev.minetest.net/TODO
17:51 nrzkt joined #minetest-dev
17:51 Fixer RealBadAngel, "RealBadAngel is currently working on shader-based lighting. " you are?
17:56 sofar RealBadAngel: if I can choose shaders for nodes in lua, can I use that to fix this: http://i.imgur.com/BgPdJ16.png - note the color/light difference between the two plants - same texture one is just a mesh node other plantlike
17:57 Obani joined #minetest-dev
17:57 RealBadAngel Fixer, i do have hw lights workin, main problem is that irrlicht can have up to 8 light sources
17:58 RealBadAngel so we have to combine our system with irr
17:58 RealBadAngel mt has to prepare pure lightmap for the whole scene
17:58 RealBadAngel then irrlicht can lit it and combine
17:58 Fixer RealBadAngel, 8 light sources? I think that was forever and everywhere, or I'm wrong?
17:59 RealBadAngel in mt you can have each node in mapblock be the light source
17:59 RealBadAngel whole mt world is single irrlicht scene node
18:00 RealBadAngel so while its easy to do sun, its impossible to do torches
18:00 Fixer from stackoverflow "8 lights is the limitation of fixed GL pipeline, where you enable each of them, set mode, parameters, etc. Now you have pixel shaders, and lighting is done within shader. There you can use large number of dynamic (not backed into textures) lights. You only need to supply all these lights' parameters sufficiently (maybe, in a texture), and test how many lights your shader is able to process. Also, in shader you can
18:00 Fixer cull too weak lights (contributing too little into pixel value) or just too distant ones."
18:01 RealBadAngel yeah
18:01 RealBadAngel mt has to prepare lightmap texture for the scene
18:01 RealBadAngel with own multiple lighht sources
18:01 Fixer RealBadAngel, oh wait, TODO has this: "Irrlicht does support occlusion culling. If enabled, it culls scene nodes which are entirely out of the camera's view fulcrum.
18:01 Fixer Problem: The entire map is an Irrlicht scene node. Solution: Make scene nodes per MapBlock."
18:02 Fixer hmm
18:02 RealBadAngel could be done
18:02 RealBadAngel but limit of light sources still applies
18:03 RealBadAngel so no point to rework it
18:04 RealBadAngel for the world like ours only solution is to prepare lightmaps and then count in sun
18:04 Fixer except for culing?
18:04 RealBadAngel idc bout culling
18:04 RealBadAngel not my interest
18:04 Fixer RealBadAngel, btw why it is not complete darkness in caves?
18:05 RealBadAngel gamma levels
18:05 RealBadAngel see light.cpp
18:05 RealBadAngel we do have many fancy gimmicks bout light
18:06 RealBadAngel atm mt lighting is per vertex lighting
18:06 RealBadAngel see irr docs about that
18:06 RealBadAngel irr doesnt calculate it, we set it
18:10 nrzkt RealBadAngel, if i remember 8 light source is a hardware limitation not irrlicht, it's normal
18:11 Calinou nrzkt, lots of games have > 8 light sources in view…
18:11 Calinou (real-time)
18:12 Calinou all with shadows
18:12 Calinou but yeah that's with shaders
18:12 Calinou nobody should use fixed-function anymore ( ͡° ͜ʖ ͡°)
18:13 nrzkt Calinou, it's not real time it uses deferred shading
18:15 RealBadAngel or lightmaps
18:15 RealBadAngel which are used to bake in lights into rendered scene
18:16 RealBadAngel we are not mature enough for such tricks yet ;)
18:19 Calinou don't lightmaps take seconds to bake?
18:21 kahrl I remember creating lightmaps for quake2 levels, that process took many hours
18:21 nrzkt joined #minetest-dev
18:22 Calinou yeah, depends on games
18:22 Calinou Cube 2 can do it in seconds
18:22 Calinou q3map2 is more like hours :)
18:27 sofar RealBadAngel: any idea about my comment on mesh vs. plantlike shader choice fixing that issue I posted earlier?
18:27 sofar RealBadAngel: also, I assume placing a torch would cause the engine to regenerate the lightmap then
18:29 RealBadAngel sofar. with first please hold on for a few days, i already did 3d plants before, im doing now something WAY better
18:29 RealBadAngel on 2nd, yes
18:30 RealBadAngel old code have to do lightmaps
18:30 RealBadAngel then we need rendering passes
18:31 RealBadAngel and combine hw and our artificial lights,, aswell other effects (like reflections, refractions, and post process ones)
18:31 RealBadAngel i guess first step should be introduce rendering phases
18:32 RealBadAngel and turn off our lights ;)
18:32 RealBadAngel then, fiat lux
18:32 sofar RealBadAngel: (about the plants placement) oh, I'm not in a hurry, I was just excited to see that I could make a very cool visual change that is highly impactful
18:33 sofar will be happy to test what you are cooking
18:35 sofar I was thinking of adding a flag to nodedef that would force randomized x,z placement, e.g. placement = "randomxz", or just a boolean
18:36 sofar but since scene updates require plants to not move randomly (like I implemented) I'd use some sort of x,y,z based fake noise to vary plant placement
18:37 sofar RealBadAngel: what about the stretch part? I do think that's really badly needed
18:38 est31 ShadowNinja, congrats to commit #200 btw
18:49 RealBadAngel sofar. please do try setting param2 of plants
18:49 RealBadAngel it works pretty nice imho
18:49 sofar RealBadAngel: I do think that the plantlike mesh being 0.7x1.0x0.7 is incorrect
18:50 RealBadAngel sofar, what im coding right now is totally different approach
18:50 sofar I'm listening
18:50 RealBadAngel imagine moretrees one, l-system plant
18:50 RealBadAngel one node sized
18:50 sofar (and I will certainly use the param2 setting - I should modify at least mgv7 to use it)
18:51 RealBadAngel no more flat textures
18:51 RealBadAngel instead real 3d plant
18:51 sofar oh the pixel blocky thing you posted earlier?
18:51 RealBadAngel kinda like extruded one
18:51 sofar right
18:52 RealBadAngel but 16x16x16 texture
18:52 sofar as long as you don't make the size of it 0.7x1.0x0.7
18:52 sofar ;^)
18:53 sofar 16 bitmaps stacked?
18:53 sofar e.g. vertically like animations?
18:53 sofar or, can I even animate it?
18:54 sofar 16x horizontal tiles for each "pane", vertical for animations?
19:00 RealBadAngel 16x16x16 texture
19:00 RealBadAngel stored in meta
19:00 zupoman joined #minetest-dev
19:00 zupoman joined #minetest-dev
19:01 RealBadAngel and each "pixel" of it will be a node, just 1/16th of the size
19:01 RealBadAngel hardcoded for best possible performance
19:02 RealBadAngel then waving 3d plants wont be any problem
19:02 RealBadAngel i will just move 3d pixels
19:02 RealBadAngel that was the main reason is stopped developing extruded plants
19:03 RealBadAngel i have stopped
19:03 sofar IC
19:03 RealBadAngel each pixel gonna become a cube
19:04 RealBadAngel so i wont be moving planes, just those 3d pixels
19:04 RealBadAngel and that will look fucking great. i know that
19:06 PenguinDad joined #minetest-dev
19:06 luizrpgluiz joined #minetest-dev
19:10 Darcidride joined #minetest-dev
19:13 luizrpgluiz news which will have the new version of minetest? Version 0.4.14?
19:14 Krock We aren't a company with fixed release termins
19:14 VanessaE luizrpgluiz: it'll be ready when it's ready
19:15 RealBadAngel lol
19:15 VanessaE and PLEASE work on your english grammar!
19:15 VanessaE half the time, we can't understand a fucking thing you're saying :P
19:15 RealBadAngel "when its done" is reserved and trademarked
19:15 Krock VanessaE, are you mad or just rude? :(
19:16 VanessaE Krock: both :P
19:16 Krock RealBadAngel, lol serisously?
19:16 nrzkt what else is also trademarked, stop use it
19:16 RealBadAngel yeah, and she bites, so watch out :P
19:16 VanessaE RealBadAngel: at least that comes sooner than "Real Soon Now"â„¢  :)
19:16 RealBadAngel Krock, see DukeNukem Forever story
19:17 H-H-H joined #minetest-dev
19:17 * sofar quietly steps backwards
19:17 VanessaE RAWWWR
19:17 VanessaE *chomp*
19:17 * sofar throws a raw steak
19:17 VanessaE lol
19:18 * RealBadAngel can see a steak flyin back - medium roasted please
19:19 sofar my charcoal grill is in winter sleep, sorry
19:21 VanessaE sofar: at least you don't use propane. :)
19:21 Darcidride joined #minetest-dev
19:21 sofar duh, I'm not an idiot. Even an oven is better than propane
19:21 VanessaE heh
19:22 RealBadAngel nothing like propane lighter, and piece of meant fried in seconds
19:22 RealBadAngel that smell ;)
19:22 zupoman joined #minetest-dev
19:22 zupoman joined #minetest-dev
19:22 RealBadAngel *meat
19:22 sofar you haven't had my slow-smoked tri-tip yet
19:23 RealBadAngel folks at construction places are having weird ideas
19:23 sofar haha, yeah
19:23 RealBadAngel ive seen a lot, believe me
19:24 sofar I did some odd jobs over summers, I believe you
19:24 RealBadAngel you know what is the best glue for example?
19:24 sofar not wood putty
19:24 RealBadAngel better than composite ones, harder than concrete
19:25 RealBadAngel milk with cornflakes
19:25 sofar that what I eat for breakfast
19:25 Krock You eat glue for breakfast
19:25 RealBadAngel make a test
19:25 sofar glue my bowels shut so I can stand to last until lunch
19:25 RealBadAngel leave it for a week or two in a dish
19:26 RealBadAngel you will never take the spoon out of it
19:26 sofar does it polymerize?
19:26 RealBadAngel there was a tv programme, myth busters i think about it
19:27 RealBadAngel they tried to blow it out with dynamite i think
19:30 VanessaE seriously?
19:31 RealBadAngel look for "myth busters" series
19:31 VanessaE we're all familiar with Mythbusters
19:31 Calinou systemd busters!
19:32 VanessaE Calinou: no, that needs something a lot better than Mythbusters  :)
19:32 VanessaE like "this week in high explosives" :P
19:33 Kray joined #minetest-dev
19:34 RealBadAngel anyway, im off to coding, still have lotsa to do
19:46 zupoman joined #minetest-dev
19:46 zupoman joined #minetest-dev
20:36 luizrpgluiz left #minetest-dev
20:49 zupoman joined #minetest-dev
20:49 zupoman joined #minetest-dev
20:57 sofar Calinou: I've contributed significantly to systemd... lol. blowing that up? over my dead... well just go easy on it
21:00 * sofar logs in on a 40-core machine, and goes "wheeeeeee!"
21:08 VanessaE it's systemd..... lots of geeks hate it :)  (I don't care too much either way, actually, unless the maintainers start enforcing the idea of binary logs etc)
21:08 VanessaE (at which point, it crosses the line)
21:09 exio4 VanessaE: they're already doing so, anyway, 'cause they love the idea of some kind of unification of the linux core packages into systemd, I guess that is what keeps their NIH syndrome going on
21:10 VanessaE *groan*
21:10 VanessaE what we had worked well enough ....
21:10 est31 yet another systemd discussion on the internet
21:23 Robert_Zenz joined #minetest-dev
21:24 VanessaE est31: well it IS systemd, so it's a fair target for jokes :)
21:25 est31 sofar, what did you contribute?
21:26 nolsen joined #minetest-dev
21:26 sofar est31: various fixes, but most importantly systemd-bootchartd in it's entirety
21:26 Kray joined #minetest-dev
21:27 sofar VanessaE: you're entirely right that "it worked well enough", the problem is that it only worked well enough for the people in the ecosystem that had been around for 20 years
21:28 VanessaE I suppose so
21:28 sofar lack of progress == deterioration
21:28 sofar just look at US politics
21:28 sofar ;^)
21:28 VanessaE heh
21:29 sofar btw, the same goes for minetest. if minetest development would stagnate, then that's going to be the end of it, really fast
21:33 sofar I'm not the doomtalker kind. I'm the "let's prevent this" type who hopes people can abandon their fear of change type
21:45 est31 minetest development does stagnate
21:46 est31 tons of stuff gets proposed all the time
21:46 est31 and lots of it cant be merged
21:46 est31 or just people missing to look at it
21:46 est31 and we perform quite badly
21:46 est31 server is essentially single threaded
21:47 est31 emerge threads have huge bureocracy overhead
21:47 sofar nod
21:47 est31 map that gets sent is re-compressed for every single player
21:47 est31 if there are 20 players, it recompresses the same map for all 20 of them
21:47 sofar even if they're in the same area?
21:47 est31 and if its a mapblock filled with all air, it still gets recompressed
21:47 est31 yes
21:47 est31 even then
21:47 sofar can you compress in a separate thread?
21:48 est31 at least I think so
21:48 est31 you could yes
21:48 est31 but that would require to copy the map over
21:48 sofar mmap?
21:48 est31 or to have smarter locking mechanisms
21:48 sofar copy on write?
21:48 est31 can mmap do COW stuff
21:48 sofar linux kernel can handle that
21:49 est31 ohhhh
21:49 est31 cooool!
21:49 sofar I have smart colleagues that do stuff like that
21:49 est31 is that unix thing, or torvalds one?
21:49 sofar not sure, actually
21:50 sofar man 7 mmap-> MAP_PRIVATE
21:50 Fixer that is offtopic, please fix liquids over ignore :}
21:50 sofar Fixer: very on-topic. go chase your fluids for a bit :P
21:51 Fixer sofar, there is minetest for a offtopic disscussion
21:51 Fixer #minetest
21:52 est31 Fixer, we do talk about minetest development.
21:52 sofar Fixer: cow for map updates would be higly beneficial for minetest-core
21:54 sofar est31: you can do CoW stuff in C++, it wouldn't be too hard either, not sure how the data model looks whether how hard it would be
21:55 sofar BSD mmap() has supported MAP_PRIVATE for decades
21:57 nrzkt sofar and what about windows ?
21:57 sofar est31: since we're talking about in-memory map data, I think you'll end up having to mmap() it to shmem, and from there you can re-mmap() with MAP_PRIVATE to all consumers
21:57 nrzkt and compression on a separate thread is useless if you spawn thread only for it. You will generate overhead for nothing. Better is to switch from zlib to lz4
21:57 sofar nrzkt: they'd fall back to the memcpy() code that is slow, likely
21:58 nrzkt i changed my mapblock database compression from zlib to lz4 and i don't see any compression overhead now
21:58 sofar nrzkt: not if that thread is already running and ready to go, and mmap()'s the result back to the sender
21:58 sofar plus minetest's zlib probably needs cpu acceleration
21:58 Fixer help RBA to code last bit of 3199 that can reduce bandwidth usage so you will not need to change anything :}
21:59 nrzkt Fixer, this PR with meta changes ? I had it on my server it's good for clients, server don't care about it
21:59 Fixer yeah
21:59 Fixer clients care :}
21:59 sofar https://github.com/jtkukunas/zlib/commit/d306c75d3bb36cba73aec9b3b3ca378e31d1799e
21:59 nrzkt my client thanks me about this fix
21:59 sofar stuff like that
21:59 Fixer i mean people who play it
21:59 sofar (jkukunas is one of my colleagues)
22:00 nrzkt using LZ4 is better than added patched to patched to patches to zlib
22:00 nrzkt lz4 CPU usage is nothing
22:00 sofar but breaks protocol
22:00 Fixer but is it a main problem?
22:00 est31 yes
22:01 est31 protocol breaks should be avoided
22:01 nrzkt no need to break the protocol, just bump
22:01 est31 but we can bump it
22:01 nrzkt mapblocks are versionned
22:01 est31 exactly
22:01 est31 just have to keep old code around for legacy clients
22:01 nrzkt exact
22:02 est31 soooo wikipedia says lz4 is in fact slower than gzip
22:02 est31 not faster?
22:02 sofar compression algorithms are all the rage, but the question is fast vs. small. and zlib still beats everything overall
22:03 sofar you want fast? you get less compression
22:03 sofar you want small? you get less speed
22:03 sofar you want the best mix of fast and small? for all different types of data? you'll end up with zlib as the best choice when you look at the data
22:03 est31 usual tradeoff of information technology
22:03 sofar (I'm quoting Jim here)
22:04 sofar jim = jkukunas
22:04 est31 ok
22:04 nrzkt est31 for linux kernel and ZFS datasets LZ4 is better than GZIP
22:04 est31 sofar, this his blog? http://jkukunas.blogspot.com/
22:04 sofar yes, it's a specific data set
22:04 sofar lz4 for linux kernels is good because of high decompression speed
22:04 est31 you work at intel?
22:05 sofar since you compress once, and thousands of people decompress
22:05 sofar est31: yes. auke-jan.h.kok@intel.com
22:05 nrzkt in MT you do less writes than read
22:05 est31 O.o
22:05 sofar I just had pizza for lunch with Jim. He's moving states.
22:06 sofar est31: well, I should be working right ... well now... cough
22:06 est31 but still the count of writes is comparatively large nrzkt. compare minetest with youtube video that gets encoded at upload, then gets watched millions of times
22:06 nrzkt est31 in singleplayuer read ~= write, with 20 players it's not the same case.
22:07 sofar well you're symmetrical - each mapblock gets compressed and decompressed once
22:07 nrzkt i should go
22:07 * sofar waves... thinking he should go back to work as well ;^)
22:07 * est31 should not waste time on irc either ... :)
22:08 * est31 presses the button with the little cross
22:08 sofar studies? wife? work? jail time?
22:08 sofar lol
22:18 nolsen_ joined #minetest-dev
22:40 Terusthebird joined #minetest-dev
22:47 nolsen joined #minetest-dev
23:03 DFeniks joined #minetest-dev
23:23 Terusthebird joined #minetest-dev

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