IRC logs for #openttd on OFTC at 2020-05-23
⏴ go to previous day
00:00:15 <DorpsGek_III> [OpenTTD/OpenTTD] michicc commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2bi
00:02:08 <DorpsGek_III> [OpenTTD/OpenTTD] michicc commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2bM
00:07:54 <FLHerne> michi_cc: The ideal behaviour for pathfinders would be to avoid duplicate entries, but set the priority to the minimum of the old and new entries
00:09:37 <DorpsGek_III> [OpenTTD/OpenTTD] michicc commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2bA
00:12:14 <DorpsGek_III> [OpenTTD/OpenTTD] michicc commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2Nv
00:26:50 <DorpsGek_III> [OpenTTD/OpenTTD] Yexo commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2Ng
00:27:47 <DorpsGek_III> [OpenTTD/OpenTTD] Yexo commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jf2Na
02:59:21 *** Extrems` has joined #openttd
03:00:00 *** Extrems` is now known as Extrems
04:27:06 *** tokai|noir has joined #openttd
04:27:06 *** ChanServ sets mode: +v tokai|noir
05:17:54 *** WormnestAndroid has quit IRC
05:18:49 *** WormnestAndroid has joined #openttd
05:31:04 *** WormnestAndroid has quit IRC
05:31:58 *** WormnestAndroid has joined #openttd
05:40:00 *** WormnestAndroid has quit IRC
05:40:08 *** WormnestAndroid has joined #openttd
08:52:58 *** andythenorth has joined #openttd
09:10:40 *** sla_ro|master has joined #openttd
09:30:17 *** Progman has joined #openttd
09:37:24 *** andythenorth has joined #openttd
10:08:44 <Samu> the WaterPF requires a different builder than mine
10:24:06 *** Gustavo6046 has joined #openttd
11:58:42 *** WormnestAndroid has quit IRC
11:58:55 *** WormnestAndroid has joined #openttd
12:17:55 <andythenorth> if a newgrf procedure stores to a temp register, will the register value persist after the procedure returns?
12:24:20 <nielsm> the temps are allocated per callback invocation, as I understand it
12:24:44 <nielsm> so when the callback completes, the temps are lost, not before
12:32:56 <andythenorth> that sounds hopeful
12:33:09 <andythenorth> I'm going to try and make FIRS compile faster
12:33:18 <andythenorth> it's only 18 seconds for me from clean, but still
12:33:48 <andythenorth> pypy makes an insane difference to FIRS compile time
12:59:08 <andythenorth> macos won't compile
13:01:17 <andythenorth> older rev compiles
13:02:11 <andythenorth> I haven't done a full bisection, just went to last commit before recent std:: changes
13:11:07 <michi_cc> andythenorth: make clean
13:11:19 <michi_cc> Seems to be a failure in our dep detection.
13:12:06 <andythenorth> I did configure, didn't think about clean
13:15:43 * andythenorth wonders if spritelayout can use procedures
13:15:49 <andythenorth> already uses expressions
13:20:05 <DorpsGek_III> [OpenTTD/OpenTTD] michicc updated pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/JfUub
13:20:08 <andythenorth> they are generated hundreds of times into firs.nml
13:20:11 <DorpsGek_III> [OpenTTD/OpenTTD] michicc commented on pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/Jfash
13:48:36 <andythenorth> the nml expressions here break my brain :)
13:48:39 <andythenorth> I didn't write them
13:57:07 <LordAro> FLHerne: something in the regression test wouldn't go amiss
13:57:25 <FLHerne> LordAro: ...there is?
13:57:46 <LordAro> FLHerne: i should scroll down the whole diff
16:11:24 <andythenorth> is it worth using an intermediate grfcodec step for FIRS
16:12:22 <andythenorth> only saves 5s max
16:12:54 <andythenorth> typical compile is < 14s now
16:13:02 <andythenorth> that is really quite good
16:13:04 <FLHerne> andythenorth: What would make FIRS a lot faster would be if you directly built an NML AST
16:13:44 <FLHerne> Currently you spend about half the time templating all your numbers into a wall-of-text, and then parsing said wall-of-text
16:13:47 <andythenorth> is that a practical thing to do? o_O
16:14:06 <FLHerne> I mean, it's all Python
16:17:05 <FLHerne> I suppose it would be a bit more likely to get broken by nml changes
16:17:25 <FLHerne> But the syntax isn't a stable interface either
16:18:09 <andythenorth> if I was a bit smarter :P
16:18:20 <andythenorth> FIRS *does* have a tree of industry objects, and tile objects
16:18:53 <andythenorth> and some of the callbacks are in a tree structure
16:19:02 <andythenorth> there's a bit too much logic in actual nml right now though
16:20:00 <FLHerne> You'd pretty much go through all the templates and replace them with functions that returned a chunk of AST
16:23:01 * andythenorth pretends he knows what an AST is, instead of just what the definition means :P
16:23:28 *** WormnestAndroid has quit IRC
16:23:41 *** WormnestAndroid has joined #openttd
16:27:04 <FLHerne> andythenorth: My suggestion is to pretty much do what that does, except without the intermediate "write everything and parse it" step
16:27:06 * andythenorth wonders whether to look for more 0.1s to take off FIRS
16:27:12 <andythenorth> or just buy a computer that is not slow ass
16:27:30 <andythenorth> I downgraded my CPU significantly in March :|
16:28:44 <andythenorth> how would we get it into nml to write out the result? pickle? Or import nml modules to FIRS?
16:29:07 <FLHerne> andythenorth: The latter
16:29:26 <nielsm> import the nml modules into FIRS and generate the parse tree directly
16:29:45 <nielsm> then you can have nml output the parse tree to nml format if you want an .nml file
16:32:03 <FLHerne> That might need a bit of improvement if he wants a *nice* nml file :p
16:36:41 <andythenorth> currently I'm favouring the 'buy a not slow computer' theory :P
16:38:11 <FLHerne> glx: Were you planning to post that unused-parameter optimization as a PR?
16:39:48 <andythenorth> hmm should Intel Turbo Boost apply for single-threaded Python?
16:40:35 <FLHerne> They have turbo for any number of cores now, but highest freq if only one loaded
16:40:42 <FLHerne> But only if it's not too hot
16:59:37 <DorpsGek_III> [OpenTTD/OpenTTD] michicc updated pull request #8091: Add: [Script] Native priority queue; useful e.g. for pathfinders. https://git.io/JfUub
17:28:22 *** snail_UES_ has joined #openttd
17:33:37 *** Smedles has joined #openttd
17:37:34 <FLHerne> > return ((input * this->multiplier) + (round && this->shift != 0 ? 1 << (this->shift - 1) : 0)) >> this->shift;
17:38:04 <FLHerne> Given that `round && <anything truthy>` is always 1, I feel like this is broken
17:40:31 <FLHerne> No, I misunderstood the precedence
18:11:04 <michi_cc> planetmaker: Performance is usually used as the reason for not doing "dynamic" track sprites, but if you look at what a crapload of stuff the catenary drawing does to get crossings right, I don't see any performance problem with a few more vars.
18:11:12 <michi_cc> Of course, somebody still has to code it.
18:30:59 *** sla_ro|master has joined #openttd
18:35:25 *** Progman has joined #openttd
19:22:55 <andythenorth> hmm FIRS makefile might be clown shoes
19:29:43 <andythenorth> that's saved 10% of compile time
20:15:48 *** patrick[m]1 has joined #openttd
20:47:01 <DorpsGek_III> [OpenTTD/nml] FLHerne commented on pull request #151: Add: Ignore unreferenced named parameters https://git.io/JfazT
20:53:36 <andythenorth> I cut 1MB off firs.nml
21:12:19 *** gelignite has joined #openttd
21:41:50 *** Progman has joined #openttd
21:43:06 <andythenorth> reduced redundancy
21:43:22 <andythenorth> eliminated repeated spritesets
21:43:58 *** snail_UES_ has joined #openttd
21:44:05 <andythenorth> also a bunch of repeated code for setting terrain tile
21:44:32 <andythenorth> it's snail_UES_ :)
22:16:32 *** snail_UES_ has joined #openttd
22:17:19 *** snail_UES_ has joined #openttd
22:24:50 <andythenorth> sub 10s FIRS compile with primed caches :)
22:24:54 <andythenorth> that's really not bad
22:25:13 <andythenorth> grfcodec would not be much faster
22:25:58 <andythenorth> ok grfcodec is sub 2s, but still
22:52:12 <DorpsGek> Samu: yexo was last seen in #openttd 22 hours, 58 minutes, and 13 seconds ago: <Yexo> GN
23:14:10 <supermop_Home> andythenorth is the hst car like a luxury car?
23:25:44 <andythenorth> supermop_Home intermediate
23:25:59 <andythenorth> it has a small payment decay bonus
23:48:26 <supermop_Home> cant decide whether these 5 tile wyvern hauled luxury cars should become HSTs or something banal
23:49:01 <andythenorth> what's the class 50 thing?
23:49:38 <andythenorth> what version are you playing? :)
23:50:03 <andythenorth> wait for the Onslaught I reckon
23:52:22 <supermop_Home> I've got the onslaught
23:52:30 <supermop_Home> it seems overkill for 5 tiles?
23:53:00 <supermop_Home> vanguard seems best incremental upgrade
23:53:17 <supermop_Home> but i think i might need to switch to regular cars to increase capacity
23:53:17 <andythenorth> but it's deliberately a bit vanilla and dull
23:53:42 <supermop_Home> maybe electrify the line and use the Olympic emu thing?
23:53:55 <supermop_Home> not a super busy line though
23:55:04 <andythenorth> use a snowplough
23:55:10 <andythenorth> silly andythenorth
23:57:18 <supermop_Home> currently DMUs turn back in the center platform
23:57:56 <andythenorth> such road types :)
continue to next day ⏵