Minetest logo

IRC log for #minetest-dev, 2014-05-31

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

All times shown according to UTC.

Time Nick Message
00:03 ShadowNinja PilzAdam: How about "LGPL-2.1+ and CC-BY-SA-3.0 and MIT and Apache-2.0".  That's in rough order of significance (Core, core textures, library code, fonts).
00:03 ShadowNinja I don't know if this field is processed though, if not I'd use commas and less "and"s.
00:04 PilzAdam replace the first 2 "and"s with ,
00:04 PilzAdam and what about the <updatecontact> field?
00:08 CraigyDavi joined #minetest-dev
00:32 ShadowNinja [project_license] is not valid: SPDX ID 'LGPL-2.1+, CC-BY-SA-3.0, MIT' unknown
00:32 ShadowNinja ^ Seems the ands are needed.
00:34 ShadowNinja I've made these changes: http://sprunge.us/fNeB?diff
00:37 LemonLake joined #minetest-dev
00:38 VanessaE ShadowNinja: we need to stop calling it "infinite-world" since it's hardly so...
00:38 ShadowNinja VanessaE: I changed it to near-infinite.
00:38 VanessaE that works,
00:38 VanessaE I was about to suggest that.
00:39 VanessaE that reminds me, did nore's super-huge-worlds patch ever get past that last bug?
00:40 ShadowNinja !seenany nore
00:40 ShadowBot ShadowNinja: I saw nore in #minetest-dev 6 hours, 2 minutes, and 28 seconds ago quiting (Ping timeout: 265 seconds)
00:41 ShadowNinja Hmmm, haven't seen much of him recently, but seems like he's been around.
00:41 VanessaE he has, yes
00:41 VanessaE said he's not had much time for coding though
00:41 ShadowNinja I never heard that he fixed it though.
00:53 RealBadAngel VanessaE, ShadowNinja he left (pinged out) just when i asked him about microblocks
00:54 RealBadAngel and thats suspicious
00:54 VanessaE naw
00:54 VanessaE RealBadAngel: ---> bug report #minetest-technic
00:54 RealBadAngel comin
00:54 VanessaE he probably was afk anyways
00:54 RealBadAngel i know
00:55 VanessaE the ping out was probably like with minetest.  you highlighted him, causing his client to finally try to respond, but it ain't there so ping timeout :)
01:40 zat joined #minetest-dev
02:07 ShadowNinja ~tell sapier I can't exit the key change menu on your latest Android build.
02:07 ShadowBot ShadowNinja: O.K.
02:11 RealBadAngel ~tell sapier why esc doesnt work as formspec exit?
02:11 ShadowBot RealBadAngel: O.K.
02:15 ShadowNinja Inventory cubes render properly now, but not for the right item, eg mese renders as lava, and they're upside-down.
02:27 proller joined #minetest-dev
02:32 RealBadAngel they dont render properly in all cases
02:33 RealBadAngel ive seen many items made out out random data noise
02:34 RealBadAngel some of them also on black background
02:34 RealBadAngel and most of them upside down
02:34 VanessaE last build I tried many of the inventory cubes just rendered as a simple, white square
02:35 VanessaE prior to that, when they did render, they were inverted
02:35 VanessaE rendered properly, but inverted and on some sort of solid color background (similar to RBA)
02:35 RealBadAngel i have to make it all same way as wielded
02:36 RealBadAngel rendered by engine, not us
02:36 RealBadAngel it seems we cannot handle correctly all the cases
02:37 RealBadAngel and we will get rid of that fuckin preload item visuals once and for all
02:37 VanessaE get rid of it and use the G*d damned in-world actual objects.
02:38 VanessaE what reason is there to render to texture at all anyway?
02:38 RealBadAngel next to none
02:39 RealBadAngel and i want selected item to spin :)
02:39 RealBadAngel that will be cool
02:39 VanessaE what will you do with objects that provide a wield/inventory image? :)
02:40 RealBadAngel pass them to extrude
02:41 VanessaE ew.
02:41 VanessaE the same extrude routine I disable on my client because it's so slow when using an HD texture pack
02:41 RealBadAngel it has to be fixed
02:42 * VanessaE pokes kahrl
02:42 RealBadAngel or moved outta engine
02:42 VanessaE I don't know that it can be; disabling it altogether should be offered in the menu or .conf
02:42 RealBadAngel which could be even better
02:42 VanessaE out of the engine and into what?
02:42 VanessaE can't do it in a shader, not everyone has them
02:42 RealBadAngel script?
02:43 VanessaE eh?
02:43 VanessaE you're not suggesting pre-caching them
02:43 RealBadAngel pre making them
02:43 VanessaE well actually,
02:43 VanessaE that's not such a bad idea really
02:43 VanessaE *can* they be cached?
02:43 VanessaE generate them once, load the generated results next time around instead of re-doing them every damn time
02:44 * VanessaE wonders how big HDX's "mushroom spores" item image (512px) would be as a mesh :P
02:44 VanessaE that one is far and away the WORST ever image I've passed to the extrude routine
02:45 VanessaE https://github.com/VanessaE/hdx-5​12/blob/master/mushroom_spore.png
02:45 VanessaE this.
02:45 VanessaE go ahead.  I triple-dog-dare you to let Minetest try to extrude that.
02:59 Exio4 joined #minetest-dev
03:46 kahrl joined #minetest-dev
03:46 dulrich joined #minetest-dev
03:46 deltib joined #minetest-dev
03:47 Exio4 joined #minetest-dev
03:53 dulrich_ joined #minetest-dev
04:12 Miner_48er joined #minetest-dev
05:58 Hunterz joined #minetest-dev
06:18 darkrose joined #minetest-dev
07:21 CheapSeth joined #minetest-dev
07:40 whiskers75 joined #minetest-dev
07:46 Fresh_m__ joined #minetest-dev
07:50 tomreyn joined #minetest-dev
08:49 LemonLake joined #minetest-dev
09:35 sapier joined #minetest-dev
09:37 sapier ~tell RealBadAngel Ahh the key change menu ... I almost forgot about it ... yea thats an open issue as key events don't reach it correct. Same for escape and menus, it's lost somewhere within stack or doesn't even exist for most android devices.
09:37 ShadowBot sapier: O.K.
09:54 restcoser joined #minetest-dev
10:14 vifino joined #minetest-dev
10:20 PenguinDad joined #minetest-dev
10:45 ImQ009 joined #minetest-dev
10:59 grrk-bzzt joined #minetest-dev
11:01 PilzAdam joined #minetest-dev
11:08 Jordach joined #minetest-dev
11:17 kahrl joined #minetest-dev
11:40 NakedFury joined #minetest-dev
11:46 kaeza joined #minetest-dev
11:58 grrk-bzzt joined #minetest-dev
12:31 iqualfragile joined #minetest-dev
12:34 Zeitgeist_ joined #minetest-dev
12:46 Gethiox joined #minetest-dev
13:15 nore joined #minetest-dev
13:38 Exio4 joined #minetest-dev
13:54 werwerwer_ joined #minetest-dev
14:14 proller joined #minetest-dev
14:15 Megaf joined #minetest-dev
14:25 hmmmm joined #minetest-dev
14:27 Megaf #1345
14:27 ShadowBot https://github.com/minetest/minetest/issues/1345 -- Lua feature to hide player nametags
14:27 LemonLake thanks megaf
14:28 Megaf ;)
14:28 LemonLake Yeah, can you give me your opinion on that please?
14:29 Exio4 lots of opinions, noone is coding it though
14:31 Megaf Now, let's see whos faster to compile minetest server. 4 ARMv7 core at 2,1 GHz running make -j 4 or One AMD Athlon XP 2800+ at 2067MHz running make -j 1
14:34 grrk-bzzt joined #minetest-dev
14:39 Piggybear87 joined #minetest-dev
15:08 kaeza joined #minetest-dev
15:09 SEx2 joined #minetest-dev
15:09 kaeza joined #minetest-dev
15:12 kaeza joined #minetest-dev
15:16 ImQ009 joined #minetest-dev
15:23 SoniEx2 joined #minetest-dev
15:32 Megaf_ joined #minetest-dev
15:42 sapier joined #minetest-dev
15:43 sapier megaf, I'd expect the athlon xp to be faster, what's the real result?
15:44 MichaelRpdx joined #minetest-dev
15:45 Megaf sapier; Athlon XP http://paste.debian.net/102693/
15:45 Megaf sapier; ODroid U3 http://paste.debian.net/102694/
15:45 Megaf sapier; when running 4 jobs at once ARM is really faster
15:46 Megaf but with a single job, Athlon is waaay faster
15:46 sapier that's interesting, I'd expected arm to be slower even on 4 core ... guess I need to rethink a little bit
15:46 Megaf So, minetest, at the moment, runs better on freaking old and energy not eficient hardware that on new and energy efficient hardware
15:46 sapier wait it's a xp 1800 so 1.5ghz?
15:47 Megaf now, ARM will use 5 Watts and Athlon close to 150 Watts
15:47 Megaf sapier; 2800+
15:47 sapier the difference is arm is risc while athlon is xisc
15:47 Megaf 2083 MHz
15:47 sapier cisc
15:47 sapier ok so almost same frequency
15:47 Megaf sapier; actually, all x86 CPUs are RISC internaly
15:47 Megaf they have a translation layer
15:48 sapier no they aren't
15:48 sapier transmeta tried that once ... with little success
15:49 Megaf sapier; so, can you see how faster and more efficient minetest server would be if it hada modern design?
15:49 sapier of course modern cpu's have parts beeing similar to risc and cisc architecture in their design but still x86 cpu's are optimized for cisc command set
15:49 Megaf it could run faster burning 90% less energy
15:50 sapier optimizing minetest towards more parallelism will take a lot of time ... it wont be done tomorrow
15:51 Megaf It should be done from the bery begining
15:51 Megaf very*
15:51 sapier while this is our overall direction don't expect it to happen within a year
15:51 Megaf That's my poing
15:51 Megaf the time it will take to change, it shorter than the time it will take to start from scrach
15:52 sapier Almost all projects I saw that started from the beginning did to exactly same mistakes that have been done before
15:53 sapier if you start from the beginning you'll be where minetest is now in about two years
15:53 sapier but minetest will have evolved for 2 years, so you're competing to that future minetest version by the time
15:56 Calinou joined #minetest-dev
15:58 sapier but megaf, arm devices ARE limited, that's why "cloud" services are advertised that loud nowadays
15:58 sapier all things that require cpu power are done in cloud ... where those energy hungry but fast x86 cpu's are most common
16:01 Megaf they are limited yes, but they have some pretty neat extensions too
16:01 Megaf things that x86 need lots of cycles and software to do, ARM do in a single cycle, via hardware
16:01 sapier extensions not usefull if you're just doing generic control flow
16:02 sapier usually it's the other way round cisc is doing things in one step while risc requires several
16:02 sapier and x86 processors have extended api features today too
16:03 Megaf RISC can do several little things faster than CISC
16:03 Megaf heh
16:03 Megaf but CISC do a huge thing faster than RISC
16:03 sapier I'd not sign that as a general statement
16:04 Megaf anyway, the compiler should do a better job too
16:04 sapier yes if minetest was compiled for intel using icc it'd most likely run way faster
16:05 sapier and of course there might be a lot of suboptimal algorithms in too
16:05 Megaf I wonder what the best compiler and compiling options for ARM would be
16:06 sapier arm recently switched to llvm as default compiler so maybe latest version of clang might be optimal
16:06 sapier that's a difference to gcc on x86 too ... intel doesn't really improve gcc for their architecture
16:07 sfan5 I just read a small part of the conversation... But have we come to blaming the compiler for Minetest's slowness now?
16:08 sapier no
16:08 Calinou_ joined #minetest-dev
16:09 sapier we're just talking about the influence of compiler on comparing arm to x86
16:09 sapier of course comparing those is quite difficult
16:15 hmmmm not only that, but how do you know how agressive and time-consuming the optimizations for each architecture are
16:16 hmmmm so i wouldn't say they're comparable unless you're cross compiling to the same architecture
16:17 sapier true using compile time of minetest to benchmark the architecture is even less telling
16:39 proller joined #minetest-dev
16:39 Exio4 joined #minetest-dev
16:51 zat joined #minetest-dev
16:52 Exio4 that is what i told him
17:34 sapier great we need to update all png files
17:35 sapier "PNG warning: iCCP: Not recognizing known sRGB profile that has been edited" is the message shown on recent libpng implementations
17:36 VanessaE PNGcrush them while you're at it
17:37 VanessaE (or optipng or whatever your favorite shrink tool is)
17:37 sapier well I wont do that right now but it's something to put on todo list
17:38 sapier I'm looking for the reason why inventory items are still broken ... having wrong textures shown may be a hint about some memory handling issues
17:39 sapier I found a few error cases that aren't handled properly within minetest, but none of those cases have been this problem
17:44 kaeza joined #minetest-dev
18:06 Garmine joined #minetest-dev
18:40 grrk-bzzt joined #minetest-dev
18:47 Piggybear87 joined #minetest-dev
18:55 Calinou joined #minetest-dev
19:04 diemartin joined #minetest-dev
19:06 grrk-bzzt joined #minetest-dev
19:13 sapier Thats crazy android emulator claims to support GL_BGRA textures but actually doesn't ... resulting in broken textures ... I wonder if those broken textures some ppl experience are same issue
19:13 sapier hmm most likely not as on emulator there's no texture at all
19:27 proller joined #minetest-dev
19:43 VanessaE G*d damn will someone PLEASE FIX THE FUCKING WATER OPACITY?
19:43 VanessaE (where it randomly switches between opaque and translucent for no discernible reason)
19:44 Gethiox joined #minetest-dev
19:45 sapier *g* I haven't even seen that one yet
19:48 sapier VanessaE you do have that inventory item bug do you?
19:48 VanessaE sapier: yeah, with that May 18 build at least
19:49 VanessaE prior to that one when you still called it Minetest-Debug.apk, it was fine
19:50 proller joined #minetest-dev
19:50 sapier may 18? that's quite old :-) but I havent done anything supposed to fix it till today. I'm compiling right now. It's actually a fix for emulator not showing correct textures. but error message was same you have
19:51 VanessaE I haven't seen a new build link from you since then
19:51 VanessaE you don't exactly public a forum page with your builds :
19:51 VanessaE :P
19:51 VanessaE publish*
19:52 VanessaE I gotta run, bbl
19:52 sapier https://forum.minetest.net/v​iewtopic.php?f=3&amp;t=9389
19:52 VanessaE oh, so you DO publish one :P
19:52 VanessaE fine, *bookmarked*
19:52 VanessaE bbl now :)
20:06 RealBadAngel joined #minetest-dev
20:29 sapier if there aren't additional comments I'm gonna merge 1332 in a few minutes
20:45 RealBadAngel #1332
20:45 ShadowBot https://github.com/minetest/minetest/issues/1332 -- Small cleanup of hud add/remove code by sapier
20:46 RealBadAngel ^^ created clickable link to read it ;)
20:46 RealBadAngel im ok with it
20:51 PilzAdam https://github.com/minetest/minetest/commi​t/18fe277d949a82fcc6fbc47c734df514d3ea1f52
20:51 PilzAdam there was a reason why the items didn't have "wielditem" visuals before
20:52 sapier :-) if you tell the reason we others would know it too ;-)
20:52 sapier pushing 1332 now
20:52 PilzAdam inventory_image is supposed to override the texture of dropped items
20:52 sapier maybe we should fix it and add a comment telling this
20:53 PilzAdam also why did RealBadAngel change the size of the items?
20:54 sapier don't items now "grow" if they're more?
20:54 RealBadAngel PilzAdam, item entity size now represents size of a stack
20:54 PilzAdam thats stupid
20:54 RealBadAngel no its not
20:54 PilzAdam do you use the bows and arrows mod?
20:55 RealBadAngel no i dont
20:55 RealBadAngel its badly coded
20:56 RealBadAngel and i asked you many times about your opinion but lately you have chosen not to respond at all
20:56 RealBadAngel so please do not comment with something like that
20:56 PilzAdam you asked for my opinion and don't want to hear it now?
20:57 RealBadAngel now? i keep asking weeks ago
20:57 RealBadAngel but ok, go on
20:57 sapier comments on #1302
20:57 ShadowBot https://github.com/minetest/minetest/issues/1302 -- Add daemon support for linux like operating systems by sapier
20:58 PilzAdam your way of stacking items on the ground only works for nodes
20:58 sapier what's missing?
20:59 sapier I don't want mobs to automerge ... just to be sure ;-)
20:59 PilzAdam 2 torches don't merge together into one big
20:59 sapier well they shouldn't
20:59 sapier at least not if they're placed
21:01 RealBadAngel PilzAdam, they cant merge because of max items per stack
21:01 RealBadAngel if max = 1 they simply cant
21:01 PilzAdam stack_max of torches is 99
21:02 RealBadAngel ouch
21:02 RealBadAngel lemme check it out
21:03 RealBadAngel they do merge
21:04 RealBadAngel and grow the size
21:04 PilzAdam thats the problem
21:04 RealBadAngel please explain
21:05 PilzAdam merging items doesn't make sense for items that aren't nodes
21:06 RealBadAngel we are not talking nodes here but item entities
21:07 PilzAdam if you throws 2 sticks together IRL they don't merge into one big stick
21:07 PilzAdam do you know how minecraft handles the merging of dropped items?
21:08 RealBadAngel i dont care bout mc
21:08 PilzAdam they display the amount of items in the stack visually, but limit it to something like 4 or so
21:08 PilzAdam it's way more intuitive and realistic
21:09 RealBadAngel since when IRL mergeing is limited to 4?
21:09 RealBadAngel have i missed some EU paper on that?
21:09 PilzAdam the visual representation is limited to 4
21:10 PilzAdam I don't  have anything against the merging itself, just the visual representation you chose is bad
21:10 RealBadAngel by now visual size can range from 0.15 -> minimal size of an item
21:10 RealBadAngel to 0.3 - maximum size
21:11 RealBadAngel if it is a full stack (max_count) then size = 0.3
21:11 RealBadAngel there are no steps (like that 4 from mc)
21:12 RealBadAngel stack of 5 will be slightly bigger than of 4
21:13 PilzAdam I'm not sure if you understood what I said
21:20 RealBadAngel you said that possible sizes are 4
21:21 RealBadAngel at least i get it that way
21:22 PilzAdam then you got it wrong
21:23 PilzAdam when there are more than 1 item in the item_entity stack, it should display several items, and not a big one
21:27 RealBadAngel how?
21:28 RealBadAngel a few items spinning at same pos? we had that before
21:28 RealBadAngel and THAT was wrong
21:34 PilzAdam http://hydra-media.cursecdn.com/minecraf​t.gamepedia.com/6/6f/Item_entities.png?v​ersion=b9efd484f545042faed5f251511e8d05
21:34 PilzAdam see the sand and torches
21:35 RealBadAngel hmmm
21:36 RealBadAngel thats a good idea, but slightly different way
21:36 RealBadAngel if theres more than one stack make them offset
21:37 RealBadAngel i will code that
21:37 RealBadAngel that will eliminate flickering
21:37 RealBadAngel thx for the idea
21:37 PilzAdam also don't use the "wielditem" visual; it doesn't respect the inventory_image
21:38 PilzAdam either use the old way of displaying items with an inventory_image, or code a new visual similar to "wielditem" that uses inventory_image
21:40 RealBadAngel thats gonna change, i will change it to use real time rendering of the item/node
21:40 RealBadAngel for all the elements
21:41 RealBadAngel so inventory_image will become obsolete
21:41 RealBadAngel for that extrude routine have to improved in the first place
21:42 PilzAdam no, inventory_image is supposed to define the texture of the item, in all it representations
21:42 PilzAdam don't break the API
21:42 RealBadAngel thats unaceptable that node/item looks different way wielded/placed
21:42 RealBadAngel shaders have to be applied
21:43 RealBadAngel it have to be rendered
21:43 RealBadAngel and not in faulty code like preload item visuals
21:44 RealBadAngel engine is capable of doing that without any problem
21:44 RealBadAngel problem was just our implementation
21:45 PilzAdam don't break the API
21:45 RealBadAngel it will be broken sooner or later
21:45 RealBadAngel and it wont be me to break it
21:45 RealBadAngel so dont worry
21:46 RealBadAngel im a modder too, i can take care of my own business
21:47 RealBadAngel in fact all the engine changes i do, i do for modding
21:48 PilzAdam not using inventory_image is breaking the API
21:48 PilzAdam you already broke it
21:48 RealBadAngel yup
21:48 RealBadAngel it is using wielded
21:48 RealBadAngel real 3d objects
21:49 sapier could we stop accusing ppl and just fix the issue it doesn't seem to be a big one
21:49 RealBadAngel so no more sprites staring at you in a magical way
21:50 RealBadAngel which was imho one of biggest mt fails all the time
21:50 sapier btw RealBadAngel shaders are actively disabled for inventory items, if you want them to look same as nodes the fix would be minor
21:50 PilzAdam sapier, just reverting RealBadAngel's commit without explaining why isn't good either
21:50 sapier for what I read about the discussion that wielditem thingy is just a small part of it ... am I wrong about that?
21:51 RealBadAngel sapier, shaders will apply to everything, but not yet
21:51 RealBadAngel code is not ready for it
21:51 sapier and the multiple entities instead of one is better then what rba implemented too yes, but what was added is still better then it was before (at least in my oppinion)
21:52 RealBadAngel multiple entities can refer to many stacks at same pos
21:52 RealBadAngel which should be offseted to avoid rendering textures at same coords
21:52 sapier I meant display as multiple entities not really multiple entities
21:53 RealBadAngel make a poll about it
21:54 RealBadAngel 2 or 3 votes are not enough to decide whats best
21:54 RealBadAngel especially if 1 is pro, 2nd con and 3rd not decided
21:54 RealBadAngel ;)
21:55 sapier I'm pushing #1302 in a few minutes
21:55 ShadowBot https://github.com/minetest/minetest/issues/1302 -- Add daemon support for linux like operating systems by sapier
22:05 ImQ009 joined #minetest-dev
22:11 RealBadAngel so, PilzAdam, the conclusion, if you want the items to behave different way, just ask ppl about it
22:24 vifino joined #minetest-dev
22:24 vifino joined #minetest-dev
22:27 hmmmm hmmm
22:28 hmmmm closing stderr and stdout will make any attempts to write to them fail which might cause bugs
22:28 hmmmm why wouldn't you redirect them to /dev/null instead?  what about stdin?
22:29 hmmmm why not just use daemon()?
22:29 hmmmm man 3 daemon
22:30 sapier *smile* you're about 4 minutes to late but O'll check if those things really cause issues
22:30 hmmmm at the very least, you should s/fclose(stdout);/freopen("/dev/null", "w", stdout);
22:30 sapier while this is not supposed to happen I'll still write the fix for it
22:31 hmmmm although
22:32 hmmmm /dev/null is common on unix-like systems, it's not a part of the POSIX standard
22:32 sapier seems like daemon would already do everything for us
22:32 sapier I'll just change it to use that function ... couldn't you have mentioned that one the last weeks ;-)
22:33 hmmmm sorry, I haven't been looking at what's being said in the channel
22:33 sapier :-) no problem
22:33 hmmmm eh, all of this reaks of platform dependent stuff
22:33 hmmmm "if it's not Windows, then it's unix-like!"
22:33 sapier yes so lack of /dev/null isn't an issue at all
22:33 hmmmm it's as much of an issue as depending on /var
22:34 hmmmm or is it
22:34 sapier we don't depend as you can specify your own pidfile
22:36 RealBadAngel btw, are we heading for next stable release anytime soon?
22:37 sapier hmm daemon has a problem ... I can't create a pidfile on using it
22:38 ImQ009 joined #minetest-dev
22:42 sapier I fixed the redirections
22:45 RealBadAngel ^^ ?
22:45 RealBadAngel i would like to merge shaders code
22:45 sapier did someone test it yet?
22:46 RealBadAngel there were 2 issues found thx to testing, fixed them
22:46 sapier if it just a small bugfix I guess it's always fine to merge
22:47 RealBadAngel no, its not small bugfix
22:47 RealBadAngel its huge rewrite
22:47 sapier I'd feel better if it was tested by a little bit more then two ppl
22:48 RealBadAngel it is already tested
22:48 RealBadAngel code is frozen since one month already
22:48 sapier ok, do you have time the next week to fix bugs? ;-)
22:49 RealBadAngel ive made bugfixes for the frozen state already
22:49 RealBadAngel nothing serious tho
22:49 RealBadAngel just issues with water animations (ive disabled them for water surface shaders)
22:49 sapier I just want to avoid same thing to happen as with glowing water ;-)
22:50 sapier while Imho that was one of the more nice bugs :-)
22:50 RealBadAngel hehe
22:50 RealBadAngel btw, that code improves performance greatly when using shaders
22:51 RealBadAngel before shaders were compiled and assigned on each refresh
22:51 sapier ok so you don't have a 120h work week next week? ;-)
22:51 RealBadAngel now it takes place only on startup
22:51 RealBadAngel and whats most important you can have a shader per tile
22:52 LemonLake joined #minetest-dev
22:52 RealBadAngel lets say water surface tile
22:52 RealBadAngel which simply leads to water with reflections and refraction
22:53 sapier you don't have to persuade me your fixes usually are better :-) but as any major changes they contain bugs
22:54 LemonLake left #minetest-dev
22:54 RealBadAngel ofc i can avoid that
22:55 RealBadAngel cant
22:55 RealBadAngel glsl is a bitch
22:55 RealBadAngel it works different on each setup
22:55 sapier yes we know how different different cards react
22:56 RealBadAngel but on the other hand i cant find those bugs without making folks use it
22:56 sapier so if you've time to fix occuring bugs no problem, but don't merge something like that if you already know you wont have time to fix the next weeks ;-)
22:56 RealBadAngel last days im rather active
22:57 RealBadAngel and that wont change rather
22:58 Megaf folks, I'd like to use the -O3 flag for GCC when running make
22:58 Megaf how do I do that?
22:59 RealBadAngel idk, sapier?
22:59 sapier edit src/CMakeLists.txt
22:59 Megaf Thanks
23:00 sapier be carefull about what you change if you do a release build changing debug flags wont have an effect of course ;-)
23:00 RealBadAngel btw, somebody should add statement to the channel that Minecraft is not a guideline for us
23:00 sapier we already know that ;-)
23:00 RealBadAngel PA dont
23:01 sapier he knows too
23:01 Megaf Well, I still have no idea where I insert a flag
23:01 sapier not beeing a guideline doesn't disallow having a look at how things have been solved there
23:01 sapier megaf look for -O2
23:01 RealBadAngel ofc, i do have mc server in my git
23:01 RealBadAngel and i do read it
23:02 Megaf sapier; could not find
23:02 RealBadAngel but i dont want mt to act exactly as mc does
23:03 sapier because release is already built -O3
23:03 Megaf hm
23:04 Megaf so how can I make it optmize even more? I'm taking a look at this sapier http://gcc.gnu.org/onlinedoc​s/gcc/Optimize-Options.html
23:04 sapier more then O3?
23:04 Megaf nope, 3 is max
23:04 sapier I'd not recommend that unless you really know what you're doing
23:05 Megaf -Ofast
23:05 Megaf That's the deal
23:06 sapier just because it's called fast it doesn't have to be faster then -O3
23:10 sapier but you can check in docs what's difference both are aliases only
23:11 Exio4 -Ofast is -O3 + unsafe math callz
23:12 sapier not a good idea
23:12 Exio4 unsafe math callz in x86 are complaint
23:12 sapier hmm
23:12 sapier we already do that so it's most likely irrelevant for what we do
23:14 Megaf The compiler should be able to tell what could be parellized too
23:14 sapier I agree with "it should" ... but for real life examples compiler cant tell that one
23:15 sapier that's not a thing about implementation but a theoretical impossibility
23:15 Megaf interesting doc right here http://www.cmake.org/Wiki/CMake_Performance_Tip​s#Build_CMake_binary_with_optimization_enabled
23:16 Exio4 sapier: functional programming
23:16 sapier still a compiler can't tell what can be parallelized and what can't ... at least not in general ... for some special cases that might work automaticaly. but that's what parallel languages are used for to explicitly tell compiler what is safe to paralellize
23:17 Exio4 what do you define by "safe"
23:17 Exio4 without side-effects?
23:17 sapier save as in doesn't cause data or control races
23:17 Exio4 basically something purely functional
23:18 sapier even there it's not possible to tell for all cases
23:19 sapier and as usuall if compiler can't be sure about its supposed to not use parallelism
23:21 Megaf I was thinking, I know it's not possible, but... In the case of my ODroid U3, it's CPU have to do the USB thing, the TCP thing, the MapGen, ServerThread. All that are running on a single core. If was possible to tell the software to use one core for USB thing, one core of TCP thing, one core for ServerThread and another core for MapGen, we would have a much more efficient system, and fast
23:22 Exio4 that is the job of the kernel's scheduler
23:22 sapier actually linux is supposed to do load balancing and it works for me. if that doesn't work for you you should find out why
23:24 Megaf well, It's not doing that very well in my case
23:27 sapier maybe wrong kernel configuration?
23:27 Megaf maybe
23:29 Exio4 do you see just a single core maxed out and all but that core idling?
23:37 sapier hmm if it's not a configuration bug could it be 3 cores waiting for IO while only one remains for mt?
23:39 restcoser joined #minetest-dev
23:44 ImQ009 joined #minetest-dev
23:57 ImQ009 joined #minetest-dev
23:57 Megaf Exio4; 99.9% on a single core and at most 11% on any other core
23:57 us`0gb joined #minetest-dev
23:57 Megaf [20:37:55] <sapier> hmm if it's not a configuration bug could it be 3 cores waiting for IO while only one remains for mt?
23:57 Megaf Yep
23:57 Megaf IO wait is a nightmare
23:57 Exio4 the main problem probably is locked lua
23:58 Exio4 more like all the stuff waiting for lua
23:58 sapier if those cores are really waiting for io nothing we de would change anything
23:58 OldCoder joined #minetest-dev
23:58 Exio4 his server has a hell lot of mods, doesn't it?
23:58 Megaf Exio4; nope
23:59 sapier no he's just using a arm device with crapy drivers
23:59 sapier e.g. if your network driver does busy waiting for a packet to be received that core is locked no matter what we do

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