IRC logs for #openttd on OFTC at 2015-05-20
⏴ go to previous day
02:04:58 *** supermop has joined #openttd
03:02:40 *** Gantradies has joined #openttd
03:04:48 <Gantradies> bit of an odd question
03:05:07 <Gantradies> is there somewhere where i can nab the 1.4.4 Winx64 stable installer srtill?
03:05:50 <Gantradies> our server's running 1.4.4 and i dont know linuix well enoguh ot update it wityhout the fear of screwing something up ( our admin insisted on a linuix console session, something aobut less load)
03:06:53 <Gantradies> first look on the site only has donwloads for the new version available, and im still sleepy right now
03:07:09 <Supercheese> or you can choose a closer mirror if you aren't in the US
03:07:27 <Gantradies> any major changes in the new version i sohudl know aobut? oyuk now, anything big?
03:08:02 <glx> it should pick the closest one if you remove the us. part
03:08:31 <Supercheese> yeah was copypasted before I could edit
04:16:59 *** Biolunar_ has joined #openttd
05:19:23 *** DarkenMoon has joined #openttd
06:39:53 <supermop> anyone have an oculus rift?
06:56:16 *** Eddi|zuHause has joined #openttd
07:10:32 *** zwamkat has joined #openttd
07:12:54 *** zwamkat has joined #openttd
07:18:39 *** zwamkat_ has joined #openttd
07:24:54 *** zwamkat has joined #openttd
07:31:59 *** andythenorth has joined #openttd
07:32:01 *** zwamkat has joined #openttd
07:33:59 <supermop> not sure how wide this thing is from pictures
07:34:06 <supermop> dont think i'll model it
07:54:41 *** andythenorth has joined #openttd
08:00:56 <supermop> just modeled a building across the street from my project for no real reason
08:01:45 <supermop> 17x36m though so a bit awkward to shoehorn into ottd tiles
08:03:13 <supermop> 1.5x3 tiles, maybe i can scale it by 2/3 without the windows looking bizarre
08:22:53 <andythenorth> nml spritesets have identifiers
08:23:05 <andythenorth> which nml will expand to hex IDs at compile time
08:25:38 *** andythenorth has left #openttd
08:30:19 <Eddi|zuHause> random ramblings are random?
08:49:06 *** Johnnei has joined #openttd
09:22:15 <supermop> modelling buildings on this block:
09:22:46 <supermop> first the deco tower, now the terrace shophouses next to it
09:23:31 <supermop> when can we pay pineapples for pineapples?
09:29:11 <Pikka> late next month probably :)
09:30:00 <Pikka> nice block, trams and everything
09:30:55 <supermop> i was doing work work for something in one of the buildings, and thought maybe i should model a few across the street for context
09:31:45 <supermop> a melb house set really need those awnings over the sidewalk in some cases
09:35:15 <supermop> ok going to run to coles back in a few
09:53:03 <V453000> the latest nightly causes desyncs even with paused game :0
09:53:23 <V453000> k maybe not but desyncs :0
10:10:38 *** andythenorth has joined #openttd
10:11:01 <Pikka> herr thenorth I presume
10:21:59 <andythenorth> what is 0 || 0 || 1
10:22:58 <^Spike^> atleast if it is normal programming
10:28:15 <supermop> and tram had to reboot
10:31:42 <Pikka> is adjusting cargo decay rate a bad feature?
10:32:14 <V453000> I do it, I justify it as "slugs are eating cargo"
10:32:24 <V453000> in reward they get increased power :D
10:33:19 <Pikka> on the one hand, hidden stats are bad.
10:33:57 <V453000> aint hidden, red text :>
10:34:06 <Pikka> on the other, it seems necessary to widen the performance gap between buses and coaches...
10:34:46 <V453000> makes sense actually yes
10:34:56 <V453000> but I would just increase load speed
10:35:13 <V453000> long distance busses with long load speed makes them stupidly unfit for city centers
10:35:28 <V453000> then you can make them very fast and with ok capacity
10:35:30 <Pikka> but the problem is not that coaches are too good at short distance
10:35:37 <Pikka> the problem is the buses are too good at long distance :P
10:35:58 <V453000> if they have stupidly bad loading speed, a normal bus with good loading speed and half the speed will win
10:37:09 <Pikka> I'm thinking buses (and trams) should have faster decay, while coaches keep the same decay as other vehicle types
10:40:01 <andythenorth> I do it for some stuff
10:43:22 <supermop> i think it makes sense
10:43:53 <supermop> otherwise there is really no point in a coach/intercity train/etc
10:44:39 <V453000> many people wont care about profits
10:45:01 <V453000> and the difference will be so small in the global scope of a bigger network that it isnt very strong argument
10:45:18 <Pikka> there are other distinctions
10:45:27 <Pikka> but I'm trying to make a distinction in profits also :)
10:45:55 <andythenorth> hide_sprite: (construction_state != 3) || LOAD_TEMP(13) || LOAD_TEMP(128);
10:46:08 <andythenorth> even when LOAD_TEMP(128) = 1
10:46:19 <andythenorth> what do I do wrong :(
10:54:16 <andythenorth> in time-honoured fashion
10:55:42 <andythenorth> I store 1, but it’s read as 0
10:56:23 <Eddi|zuHause> <Pikka> I'm thinking buses (and trams) should have faster decay, while coaches keep the same decay as other vehicle types <-- that was basically the intention behind the feature
10:56:45 <Pikka> doesn't necessarily make it a good feature, though ;)
10:57:01 <Pikka> it works for me, I'll stick with it for now. :) We'll see what testers think, one day.
10:59:11 <supermop> i'd be happy making no money on most bus legs if it meant getting people around smoothly to other legs
10:59:29 <Eddi|zuHause> and you know you can un-hide a feature with CB23? :p
11:00:16 * andythenorth is so crap at bool
11:00:34 <andythenorth> but (1 == 1) is true
11:00:55 <Eddi|zuHause> maybe it's not actually 1?
11:01:26 <andythenorth> (LOAD_TEMP(128)) is false
11:01:31 <Eddi|zuHause> or maybe the specs are weird
11:01:37 <andythenorth> but (LOAD_TEMP(128) == 1) is true
11:01:45 <Eddi|zuHause> or maybe nml does invalid operations
11:01:58 <andythenorth> I suspect that nml evaluates 1 == True in some places, but not others
11:02:50 <andythenorth> or more likely, I’m doing it wrong
11:03:10 *** Myhorta has joined #openttd
11:12:10 <__ln__> anyone been to Gibraltar and/or Ceuta?
11:13:01 <andythenorth> does using LOAD_TEMP() destroy the register contents?
11:31:29 *** andythenorth has joined #openttd
11:34:57 <andythenorth> reading a register twice produces different values
12:04:59 *** AbsoluteVeritas has joined #openttd
12:10:21 *** Klanticus has joined #openttd
12:21:28 *** shirish has joined #openttd
12:47:27 * andythenorth had multiple switches
12:47:36 <andythenorth> scribbling over the same register
12:47:42 <andythenorth> took a while to find that :P
13:03:07 <andythenorth> but compiles slower :)
13:03:12 <andythenorth> that was 6 hours well spent :)
13:07:59 <andythenorth> but eh, I learnt some things about registers
13:17:18 *** supermop_ has joined #openttd
13:28:08 <andythenorth> now actually faster
13:28:13 <andythenorth> probably about 5s, averaged
13:28:31 <V453000> what are the new features? :P
13:29:12 <andythenorth> hide_sprite calculations are down once instead of (count of building sprites) * (num date variations) * ([snow, non-snow])
13:29:30 <andythenorth> so wastes less OpenTTD time also
13:30:11 <andythenorth> @calc 16 * 4 * 4 * 2
13:30:19 <andythenorth> once is better than 512 times
13:30:53 <Eddi|zuHause> andythenorth: you should probably give the registers names, instead of using the numbers. helps in keeping things unique
13:31:08 <andythenorth> I am going to do that
13:31:39 <andythenorth> afaict, nml doesn’t provide any shorthand for that
13:31:45 <andythenorth> I can do it in the compile though
13:31:53 <Eddi|zuHause> no, but that's what you have macros for
13:32:11 *** Supercheese is now known as Guest5788
13:32:14 *** tokai|noir has joined #openttd
13:32:14 *** ChanServ sets mode: +v tokai|noir
13:32:16 *** Supercheese has joined #openttd
13:33:21 * andythenorth considers *not* removing all the CPP
13:33:33 <andythenorth> it’s remarkably convenient for trivial identifiers
13:38:38 <andythenorth> eh 50s compile with primed caches
13:38:51 <andythenorth> pretty poor, but better than 71s
14:03:57 *** roidal_ has joined #openttd
14:04:12 *** Myhorta has joined #openttd
14:51:26 *** sla_ro|master has joined #openttd
15:01:06 *** JacobD88 has joined #openttd
15:30:49 *** andythenorth has joined #openttd
15:35:21 <andythenorth> how can I make it faster?
15:46:48 <andythenorth> spritelayouts are slow
15:52:46 *** shirish has joined #openttd
16:01:15 *** Celestar has joined #openttd
17:05:21 *** HerzogDeXtEr has joined #openttd
17:29:33 *** Supercheese is now known as Guest5814
17:29:38 *** Supercheese has joined #openttd
17:33:42 *** Alberth has joined #openttd
17:33:42 *** ChanServ sets mode: +o Alberth
17:34:14 *** shirish_ has joined #openttd
17:35:54 *** Progman has joined #openttd
18:07:22 *** smoke_fumus has joined #openttd
18:13:02 *** Myhorta[1] has joined #openttd
18:18:09 *** TheMask96 has joined #openttd
18:35:07 *** Supercheese has joined #openttd
19:00:24 *** supermop has joined #openttd
19:31:41 *** frosch123 has joined #openttd
19:55:48 <Alberth> doesn't seem terribly busy here, today
20:08:30 *** andythenorth has joined #openttd
20:10:25 <frosch123> someone knows a 32bpp vehicle grf that uses the complete 0-255 brightness range in company colours?
20:12:20 <andythenorth> so I managed to make FIRS compile about 6 seconds faster
20:12:50 <andythenorth> took me about 8 hours
20:14:21 <frosch123> how many irc lines?
20:14:29 <andythenorth> couple of hundred?
20:15:20 <andythenorth> a full compile with primed caches is now about 52s
20:15:40 <andythenorth> firs.nml is 10.9MB instead of 14.1MB
20:16:00 <andythenorth> the resulting grf is 3.2MB instead of 5.2MB
20:16:10 <andythenorth> previous compile was about 71s
20:17:23 <Eddi|zuHause> just remember that for every hour you spend writing code you spend 9 hours maintaining that code
20:17:24 <frosch123> so you deleted lots of stuff :p
20:17:36 <andythenorth> I deleted lots of generated code
20:17:55 <andythenorth> look up the snow tile once per spritelayout, not ~512 times
20:17:58 <DorpsGek> Commit by rubidium :: r27289 trunk/src/3rdparty/squirrel/squirrel/sqcompiler.cpp (2015-05-20 20:17:50 +0200 )
20:17:59 <DorpsGek> -Fix: some uninitialised instance variables
20:18:17 <andythenorth> also no longer runs out of registers trivially in the graphics chain
20:18:32 <DorpsGek> Commit by rubidium :: r27290 /trunk/src/os (3 files in 3 dirs) (2015-05-20 20:18:26 +0200 )
20:18:33 <DorpsGek> -Fix: sanitize the command line arguments before doing anything complex with them
20:21:19 *** Celestar has joined #openttd
20:24:09 * andythenorth has run out of obvious ways to make FIRS faster now
20:24:35 <andythenorth> been wondering about one text cb switch chain instead of 66
20:32:17 <andythenorth> commenting the text switches out makes no difference to compile time
20:39:34 <argoneus> good morning train friends
20:40:02 *** Klanticus_ has joined #openttd
20:41:47 <V453000> I can make you such vehicle frosch123
20:42:00 <V453000> try nuts, idk how brightness goes there
20:42:07 <V453000> slugs/maglev has a lot of CC
20:44:38 <argoneus> V453000: can you into 3d graphics in general?
20:45:37 <argoneus> like bump maps, specular/diffusion lighting etc
20:45:42 <V453000> is that a sentence? :D
20:45:56 <V453000> yes I know some stuff about it
20:45:58 <frosch123> his sentence a verb :)
20:46:37 *** JacobD88 has joined #openttd
20:47:09 <frosch123> he, it's one of my favorite responses to broken sentences
20:47:13 *** Klanticus has joined #openttd
20:47:39 <frosch123> not sure where i learned it
20:47:42 <frosch123> maybe from dalestan
20:48:00 <argoneus> V453000: do you know what anisotropic filtering is?
20:48:41 <argoneus> im trying to into 3d
20:48:44 <argoneus> but there's a lot of stuff to learn
20:49:18 <V453000> well, anisotropy is when the "thing" like reflection is like transformed to sides
20:49:30 <V453000> in short it is used to simulate brushed materials, typically brushed metal
20:50:29 <Rubidium> isn't that just a fancy word for polaroid filters?
20:52:57 <Rubidium> interesting how abbreviations get abbreviated even more, so the text could be interpreted in two completely different ways
20:53:00 <andythenorth> looks to me like anistropic deals with anti-alias blur of a texture that is mapped onto a receeding plane
20:53:15 <argoneus> the way I understood it
20:53:23 <argoneus> textures are usually mapped to a rectangular area
20:53:31 <argoneus> except like in a game you see the textures under an angle
20:53:40 <argoneus> and anisotropic filtering is supposed to map the texture properly to the angled texture
20:53:51 <argoneus> to the angled model*
20:53:52 * Rubidium thought 3D works only in triangles
20:53:56 <V453000> I guess you could explain it that way
20:54:06 <argoneus> there's tris and quads
20:54:08 <andythenorth> looks like edge enhancement to me
20:54:16 <V453000> it kind of does Rubi but yeah :)
20:54:33 <V453000> quads always have a triangulating diagonal which splits it to triangles
20:54:49 <V453000> or any other polygons larger than quads or whatnot
20:54:51 <andythenorth> instead of averaging neighbouring pixels (resulting in mush), it seems to preserve contrast and hue whilst reducing the pixel dimensions
20:55:18 * andythenorth only read wikipedia page about it, probably lies
20:55:38 <argoneus> isn't anisotropic filtering a different layer from nearest neighbour / bilinear / trilinear filtering?
20:55:54 <Rubidium> you shouldn't have said that... now we can't use your talk about it here as source for the wikipedia page
20:56:01 <V453000> probably not argoneus
20:56:21 <V453000> idk what exactly bilinear and whatnot filtering does but I think that is more related to sampling of the render
20:56:33 <V453000> (how pixels of the result are approximated etc)
20:56:44 <argoneus> anisotropic filtering is a better trilinear filtering
20:56:48 <andythenorth> nearest neighbour = smush
20:56:49 <argoneus> that works for angled things
20:57:04 <V453000> while anisotropy is a property of the material where the shader behaves that way
20:57:12 <andythenorth> photoshop has similar filters when resizing
20:57:13 <V453000> I think at least, that is how I know it
20:57:16 <andythenorth> it’s easy to see the results there
21:10:01 <andythenorth> so FIRS coal mine
21:10:06 <andythenorth> needs re-implemented for reasons
21:10:24 <andythenorth> currently uses TTD default tiles and layouts, swaps the sprites
21:10:35 <andythenorth> should I keep the layouts, or do new ones?
21:10:56 * andythenorth outsources decisions :P
21:17:14 <Alberth> assuming it's not going to be very much different, I'd just keep what you have
21:18:55 * andythenorth would consider building materials
21:19:08 <Alberth> as input, so it can be burned :p
21:19:26 <andythenorth> hmm, Michael gets so much right imho
21:19:30 <andythenorth> shame he won’t play nicely :P
21:21:43 <andythenorth> FIRS is pretty much in-line with original ECS idea
21:22:33 <Alberth> but given the source (some standard classification of industries), it's not that surprising
21:23:04 <Alberth> you've been thinking about such things too for a long time
21:31:29 <Eddi|zuHause> <argoneus> there's tris and quads <-- shouldn't it be "tetras"?
21:31:42 <argoneus> I think they're called quads specifically Eddi|zuHause
21:31:49 <argoneus> but I could be wrong
21:38:56 <frosch123> ten years ago there were triangle fans an strips
21:39:28 <frosch123> all with the purpose to draw as many triangles as possibly with as little individual vertices being transfered to the gpu
21:44:38 <Eddi|zuHause> i would assume there's little optimisation potential there, given that you're working with mostly planar graphs
21:54:33 <frosch123> you mean "little left"?
21:54:53 <frosch123> because triangle fans/stips safe 2/3 of data compared to single triangles
21:55:20 <Eddi|zuHause> maybe i'm just not completely understanding the context
21:56:14 <Eddi|zuHause> of course, submitting individual triangles basically triples the data...
22:01:50 <michi_cc> The big bottleneck of today's GPUs (and driver architectures) is submitting commands to the GPU. Whether you need one or three vertices per triangle doesn't matter much when one options means you can draw them in a single batch.
22:18:06 *** Supercheese is now known as Guest5847
22:18:10 *** Supercheese has joined #openttd
22:40:17 * andythenorth needs a random start / stop animation
22:40:32 <andythenorth> or I cheat, and just fill out a lot of spacer frames :P
22:42:47 <Eddi|zuHause> what's the difficulty?
22:43:16 <Eddi|zuHause> default industries randomly start and stop their animations, right?
22:46:09 <andythenorth> I tried implementing it a few weeks ago, but failed
22:52:24 <andythenorth> seems to need switching spritelayouts
22:52:32 <andythenorth> rather than relying on the animation
22:53:09 <andythenorth> can’t be sensible
22:55:42 <andythenorth> must be a better way :P
23:19:06 *** JacobD88 has joined #openttd
23:54:41 *** shirish has joined #openttd
continue to next day ⏵