IRC logs for #openttd on OFTC at 2007-11-05
⏴ go to previous day
00:21:17 *** LeviathNL has joined #openttd
00:28:59 <glx> Ammller: I found the problem :)
00:41:23 *** bruce89 has joined #openttd
01:30:40 *** Eddi|zuHause3 has joined #openttd
01:39:20 <CIA-1> OpenTTD: glx * r11384 /trunk/src/table/build_industry.h: -Fix: second and third accepted cargo were inverted in original industry tiles
01:50:35 *** Netghost has joined #openttd
03:07:34 *** Diabolic-Angel has quit IRC
06:49:08 *** Deathmaker has joined #openttd
06:55:57 *** Frostregen_ has joined #openttd
07:03:07 *** Frostregen_ is now known as Frostregen
07:21:00 *** XeryusTC has joined #openttd
07:49:59 *** XeryusTC has joined #openttd
08:36:17 *** paula35 has joined #openttd
08:47:14 *** Farden123 has joined #openttd
08:48:02 *** lolman is now known as John
08:48:08 *** John is now known as lolman
08:49:27 *** Farden123 is now known as Farden
08:53:04 *** stillunknown has joined #openttd
08:53:57 *** Greyscale has joined #openttd
08:55:51 *** blathijs has joined #openttd
09:13:05 *** Bastiaan has joined #openttd
09:15:36 *** blathijs has joined #openttd
09:16:55 *** XeryusTC has joined #openttd
09:18:27 *** Bastiaan is now known as Guest499
09:21:32 *** mucht_work has joined #openttd
09:33:37 *** XeryusTC has joined #openttd
09:57:41 *** dihedral has joined #openttd
10:14:56 *** Diabolic-Angel has joined #openttd
10:19:34 *** XeryusTC has joined #openttd
10:37:26 <Eddi|zuHause3> you missed the "good" ;)
10:37:44 <TrueBrain> why would it be good?
10:37:47 <Eddi|zuHause3> well, technically, it's monday morning, it can't be good ;)
10:37:49 <TrueBrain> they started redecorating nextdoor
10:37:53 <TrueBrain> at 730 in the morning..
10:38:14 <Eddi|zuHause3> i'm sure that is fun :p
11:13:51 *** Vikthor has joined #openttd
11:20:28 *** blathijs has joined #openttd
11:42:10 *** mikk36|work has joined #openttd
11:42:25 <mikk36|work> does UKRS work in desert ?
11:44:44 *** Progman has joined #openttd
11:45:51 *** ChanServ sets mode: +v tokai
11:49:00 <Ammler> mikk36|work: look for UKRS in the sun
11:49:09 <Ammler> should be available on Pikkas Wikki
11:49:24 <Ammler> yes,its a modded ukrs for tropical
11:49:48 <mikk36|work> a search for pikkas wikki gives me nothing :)
11:50:57 <mikk36|work> and do i have to remove normal ukrs from the gcf list ?
11:52:02 <Ammler> I think so, but you will see it fast ;)
11:54:29 <mikk36|work> and what about that "add-ons" grf ?
11:54:43 <Ammler> hmm, should work, I guess
11:55:21 <mikk36|work> what about the parameters for that "in the sun" ?
11:55:28 <mikk36|work> or just default is enough ?
11:55:47 <mikk36|work> aha, default is ok :)
11:56:23 <mikk36|work> no more questions then :)
12:02:53 <mikk36|work> one problem still remains in the nightly
12:02:57 <mikk36|work> at least in 11368
12:03:06 <mikk36|work> add grf window won't stay in the middle
12:03:20 <mikk36|work> and if i drag the window too small, i can't close it any more
12:03:30 <mikk36|work> cause the x button is not visible any more
12:16:45 *** dihedral has joined #openttd
12:16:57 *** dihedral has joined #openttd
12:19:04 *** Tino|R152 has joined #openttd
12:23:41 <Ailure> what's the IP of tt-forums
12:36:19 <mikk36|work> question: why is Available vehicles button greyed out for spectators ?
12:36:49 <Ailure> and the one for openTTD?
12:36:58 <Ailure> I need to get a alternate DNS at this point
12:36:59 <mikk36|work> or, there is no button at all :P
12:37:13 <mikk36|work> i'd like to see that button though :)
12:37:28 <Ailure> I think that's a coding leftover
12:37:30 <mikk36|work> so i don't have to be a player to see the train elements info etc
12:37:33 <Ailure> from when that was a "create vehicle" button
12:37:47 <Ailure> it was replaced by something
12:37:57 <Ailure> then people wanted it back and it came back as avaible vehicles
12:55:41 *** Brianetta has joined #openttd
13:19:46 <mst> Hey. I have a problem. I have developed monrail in TTD so I wanted to change old railways to new one. OK - done. But when new train reach one station it is stoped there and have no option to move to another (though plan). What should I do?
13:20:37 <Ammler> send them to depot -> sell it -> buy on monorail depot a new train
13:21:35 <mst> I have three station - A, B, C
13:22:56 <mst> When the first (or second; whatever) reach one station i have to send to depot and then click once again to cancel sending to depot
13:23:33 <mst> It has full passengers onboard and standing there waiting for nothing
13:25:50 <mst> Maybe station should be same length as train? I have no idea... And you?
13:26:08 <Eddi|zuHause3> a station should be longer than the train
13:26:28 <Eddi|zuHause3> for every wagon that sticks out of the station, the loading time increases drastically
13:30:00 <mst> Another quest: is in openttd any option to notice me about reliability decrease or i have to manual watch all the trains or trucks?
13:31:13 *** Diabolic-Angel has quit IRC
13:31:20 <Eddi|zuHause3> mst: you can enable the option to automatically service if reliability drops below a certain %
13:31:29 * TrueBrain is now a FON member :)
13:31:32 <Eddi|zuHause3> instead of after a certain number of days
13:45:12 <mst> Eddi|zuHause3 where is it?
13:45:35 <Eddi|zuHause3> in patches -> vehicle section
13:45:42 <Eddi|zuHause3> "service intervals in %"
13:47:28 *** Vikthor has joined #openttd
13:59:44 *** HerzogDeXtEr has joined #openttd
14:04:52 *** DaleStan_ has joined #openttd
14:23:29 *** dihedral has joined #openttd
14:26:53 *** glx|away has joined #openttd
14:26:53 *** ChanServ sets mode: +v glx|away
14:28:03 *** glx is now known as Guest511
14:28:03 *** glx|away is now known as glx
14:36:22 <Eddi|zuHause3> soylent green is people joining and quitting IRC
14:39:45 *** frosch123 has joined #openttd
14:45:36 *** Diabolic-Angel has joined #openttd
14:50:16 *** |Bastiaan| has joined #openttd
14:54:37 *** |Bastiaan| has joined #openttd
14:58:56 *** Tino|R152 has joined #openttd
15:07:13 *** DaleStan_ is now known as DaleStan
15:07:38 *** |Bastiaan| has joined #openttd
15:10:00 *** skidd13 has joined #openttd
15:11:20 <skidd13> Anyone in here who uses pidgin?
15:12:09 <skidd13> I hate it to authentificate manualy every time
15:12:59 <Eddi|zuHause3> no "remember this fucking input next time" button? yeah, i hate such situations, too
15:13:49 <skidd13> Rubidium: I don't want to waste my memory One app for all the messenger stuff is enough
15:15:48 <TrueBrain> skidd13: I only have IRC ;)
15:16:51 <Rubidium> skidd13: waste memory?
15:17:05 <blathijs> skidd13: Bitlbee will only save you memory :-)
15:17:25 <Rubidium> bitlbee does the same as pidgin/gaim (IIRC). The only difference is that bitlbee fakes an IRC server to which you can connect with your IRC client.
15:17:47 <Rubidium> it also allows you to run it on a server making you always online and able to receive PMs and such.
15:18:45 <skidd13> somehow I prefer a GUI istead of terminal ;)
15:19:08 <Rubidium> GUIs kinda fail to work with putty/ssh ;)
15:19:30 <skidd13> And pidgin shows hidden ICQ users. I don't want to miss that neat bug/feature/whatever :D
15:19:59 <skidd13> Rubidium: x over ssh?
15:21:10 <Rubidium> kinda unuseable on PCs where you can't install anything.
15:26:53 *** skidd13 has joined #openttd
15:27:42 <skidd13> Is there a command to chgeck if I'm authentificated?
15:40:16 <blathijs> skidd13: I think there is a stray space in your vim command :-)
15:40:18 *** |Bastiaan| has joined #openttd
15:40:34 <blathijs> skidd13: Authenticated against what?
15:42:20 <skidd13> blathijs: Never mind! I found a plugin in the recent two minute which does the job for me now. Damn google... I should have formulated my request better ;)
15:47:35 *** Deathmaker has joined #openttd
15:56:27 *** |Bastiaan| has joined #openttd
16:14:51 *** XeryusTC has joined #openttd
16:41:28 *** Farden123 has joined #openttd
16:52:30 <fjb> Is it just me who is too stupid or can anybody confirm the road vehicles don't find their route in r 11384?
16:53:23 <TrueBrain> fjb: would be funny if they didn't :p
16:53:26 <TrueBrain> show me a savegame :)
16:54:22 <fjb> TrueBrain: Where can I put it to shown you? I don't have web space right now.
16:55:18 <fjb> The drive in the wrong direction, turn around just ar corner away from theirr destination, etc..
16:55:29 *** Farden123 has joined #openttd
16:56:54 *** LeviathNL has joined #openttd
16:58:23 <skidd13> fjb: I can reproduce it
16:59:22 <skidd13> TrueBrain, Rubidium: Might be the KillBits thing :(
17:01:34 *** Farden123 is now known as Farden
17:02:31 <glx> <TrueBrain> I am sure it is :) <-- same for me :)
17:04:06 <fjb> Why do I have to find all the bugs? (Not only in Openttd...)
17:04:42 <skidd13> So the | 0x... is needed for pathfinding?
17:04:43 <skidd13> fjb: Cause it has been late last night ;)
17:07:11 <TrueBrain> skidd13: more the question is: why
17:07:18 <TrueBrain> we ruled out that it could cause problems, or so we thought :p
17:07:21 <TrueBrain> fjb: do trains still route okay?
17:07:39 <glx> fjb: because devs can't think of all stupoid things a user can do ;)
17:08:21 <fjb> TrueBrain: I think trains are okay, but I will have a look.
17:09:49 <fjb> >Looks like trains are ok, or my network is too small...
17:12:02 *** |Jeroen| has joined #openttd
17:12:09 *** dihedral has joined #openttd
17:12:11 <fjb> I had some other strange effects with trains in the last days, but it is nothing that I could reproduce. I had trains trying to enter busy platforms at a station, wwhen a free platform was availlable. But that happened already days ago.
17:12:54 <TrueBrain> if it happeened before yesterday, it aint our fault :p
17:13:22 <fjb> Trains still find their way.
17:13:47 <fjb> Yeah, the train thing is older, and I found no way to reproduce it.
17:14:07 <skidd13> Hmm the 0x... value seems not to fix it :(
17:14:23 <fjb> But somebody has to implement new signals anyway, some day...
17:14:41 <Ailure> fun thing about the train pathfinder
17:14:53 <Ailure> if it dosen't find a path between X and Y
17:15:34 <Ailure> it won't try avoid red signals at all
17:15:44 <Ailure> this have caused me a few deadlocks at some terminal stations for me :/
17:16:12 <blathijs> I think it simply chooses random directions if it can't find it's way
17:16:14 <fjb> Ailure: Maybe that is the bug i was talking about.
17:16:31 <Ailure> I don't think they're related at all
17:16:39 <Ailure> I dunno if I should call it a bug even
17:16:48 <Ailure> since it happens in situations that shouldn't happen in first place anyway
17:17:27 <Ailure> and that the pathfinder (for understandable reasons) don't understand that trains can turn around at end of the line
17:18:07 <fjb> I had trains insiting to go to platforms that were occupied when a free platform was next to them and reachable. But that is unrelated to the road vehicles not finding their way.
17:19:07 <Ailure> it caused a deadlock as the train trying to enter the occupied station was surrounded by a one-way signal and a station exit-signal
17:19:14 <Ailure> and the station itself was occupied too
17:19:25 <Ailure> I think I managed to cause a accident crash in a such situation too :)
17:24:27 <fjb> Is there a way to debug the pathfinder?
17:28:43 <Belugas> of course, fjb. the sources are there, just compile in debug mode and follow the functions and the flow of the PF
17:29:47 <fjb> Belugas: Thank you. I will try that tmorrow.
17:30:18 <fjb> I have to go now anyway. Have fun hunting bugs. Maybe I will be here again later this avening.
17:30:42 * Belugas does not hunt bugs. He can barely type...
17:39:12 *** skidd13 has joined #openttd
17:42:00 *** Osai is now known as Osai^Kendo
17:42:17 <ln-> what does a cheapish flight to the USA cost?
17:43:55 <Belugas> the price of a car gaz tank for me ;)
17:44:51 <skidd13> ln-: Oneway nothing: Call the CIA and tell them someting ;)
17:46:43 <Belugas> but, ln-, may i point out that your question by itself is totally invalid either?
17:46:52 <Belugas> it lacks the point of origin :P
17:47:21 <Belugas> so, for sure, a ticket bough in australia does not cost the same as one from Great Britain ^_^
17:48:50 <ln-> Belugas: virtually everyone is from europe
17:49:35 <ln-> skidd13: that doesn't tell me what's the optimal starting point in europe for cheap flights.
17:50:10 <TrueBrain> there you go skidd13 :)
17:50:15 <skidd13> ln-: define cheap! Cause you need to add the costs for driving there too ;)
17:50:32 * Belugas is not from europ, altough born there, but that's beside he point...
17:50:38 <CIA-1> OpenTTD: truelight * r11385 /trunk/src/roadveh_cmd.cpp: -Fix r11383: bool operation can be hard to translate for some people ;) (tnx fjb for noticing this bug ;))
17:50:51 <Belugas> there are more non-europeans than you think in here :D
17:51:29 <ln-> skidd13: cheap := cheap if we assume one could get to the starting point at no cost.
17:51:43 <skidd13> TrueBrain: Bah. What a damned mistake. And 2 people overlooked it :(
17:55:30 *** Wezz6400 has joined #openttd
18:08:04 <CIA-1> OpenTTD: miham * r11386 /trunk/src/lang/ (6 files): (log message trimmed)
18:08:04 <CIA-1> OpenTTD: -Update: WebTranslator2 update to 2007-11-05 19:07:09
18:08:04 <CIA-1> OpenTTD: afrikaans - 24 changed by TrueTenacity (24)
18:08:04 <CIA-1> OpenTTD: brazilian_portuguese - 1 fixed by tucalipe (1)
18:08:04 <CIA-1> OpenTTD: czech - 1 fixed by Hadez (1)
18:08:04 <CIA-1> OpenTTD: dutch - 1 fixed by habell (1)
18:08:06 <CIA-1> OpenTTD: norwegian_bokmal - 4 fixed, 43 changed by jhsoby (47)
18:11:11 *** Brianetta has joined #openttd
18:21:10 *** DorpsGek has joined #openttd
18:27:27 *** Ammller has joined #openttd
18:37:39 *** HerzogDeXtEr has joined #openttd
18:40:01 *** DJ_Mirage has joined #openttd
18:41:11 *** Greyscale has joined #openttd
19:00:03 *** Unknown has joined #openttd
19:04:03 *** Tlustoch has joined #openttd
19:05:23 *** ChanServ sets mode: +v orudge
19:05:32 *** Wezz6400 is now known as Guest2
19:11:43 *** Wolf01 is now known as Wolf01|AWAY
19:19:15 *** Diabolic-Angel has quit IRC
19:35:33 *** Wolf01|AWAY is now known as Wolf01
19:58:25 *** Diabolic-Angel has joined #openttd
20:26:16 <hylje> mm.. delicious train imageboard
20:31:46 *** Dark_Link^ has joined #openttd
21:05:21 *** Desolator has joined #openttd
21:05:25 <Rubidium> is he going to get (kinda) slashdotted?
21:05:55 <Vikthor> Rubidium: Exactly my thought :)
21:06:40 <Vikthor> The other one was something about exhibitionism :P
21:07:59 <Rubidium> question is whether the strain is so much that he's getting kicked (read timed out) from IRC
21:08:00 <Desolator> is Chris around here?
21:08:16 <Rubidium> THE Chris has never been here
21:08:36 <Rubidium> or at least he didn't publicly announce himself as THE Chris
21:08:41 <Desolator> I mean the guy who made ChrisIN
21:09:16 <_42_> Desolator, if you can't see ludde here right now, you probably need new glasses. ^_^
21:09:29 <Desolator> we need an new bot...
21:09:29 <Vikthor> Desolator: Isn't he in the ChriIN channel?
21:10:40 <Prof_Frink> Rubidium: He's tycooned ;)
21:12:17 *** skidd13 has joined #openttd
21:15:38 *** Osai^2 is now known as Osai
21:16:23 <skidd13> TrueBrain: Did you fixed the patch-list attatched thing?
21:17:15 *** Ben_1 is now known as _Ben_
21:31:32 *** XeryusTC has joined #openttd
21:34:41 *** glx|away has joined #openttd
21:34:41 *** ChanServ sets mode: +v glx|away
21:37:47 <Smoovious> <Desolator> I mean the guy who made ChrisIN <--- he pokes his head in IRC occasionally, but mostly sticks to the forum
21:42:18 <mikk36> is it possible to change difficulty settings with rcon ?
21:44:45 *** glx|away is now known as glx
22:00:45 *** DJ_Mirage has joined #openttd
22:01:51 <ln-> who has watched the tv series "professor poopsnaggle and his flying zeppelin" as a kid?
22:03:58 <Eddi|zuHause3> i think that ran under the name "mister snuggles" here
22:04:28 <Eddi|zuHause3> i don't think i watched more than two episodes ever...
22:04:41 <ln-> doctor snuggles was a completely different thing.
22:05:13 <ln-> the former was not a cartoon.
22:05:54 <Eddi|zuHause3> hm, then i probably have no idea what you are talking about
22:06:14 <Eddi|zuHause3> and the picture above is definitely photoshopped :p
22:08:44 <ln-> (that was not a youtube link, you all saw wrong)
22:09:00 <ln-> (besides, it was relevant to discussion)
22:09:10 <Eddi|zuHause3> you forgot the handwaving ;)
22:13:17 <mikk36> never seen that series :)
22:20:46 <SmatZ> I see skidd13's updates are already commited
22:20:53 <SmatZ> so there is no need to profile them
22:21:03 <SmatZ> anyway, I did a look into generated ASM code
22:21:27 <SmatZ> I didn't look into all functions, but it looks the same
22:22:15 <SmatZ> KillFirstBit2x64 is not used anywhere in the code :)
22:23:01 <skidd13> SmatZ: not all stuff is committed. The define -> template stuff hasn't been commited
22:23:57 <skidd13> But thanks for the checks
22:24:41 <SmatZ> the "FindFirstBit2x64" generated code is not nice :-(
22:25:03 * Rubidium knows more code that is not nice
22:26:00 <skidd13> SmatZ: I'm on this dirty little piece of code. I'll post my fist change in the next minutes ;)
22:26:09 <Eddi|zuHause3> ln-: no, i can definitely say i have never seen that either
22:26:24 <Rubidium> let me say that putting GoogleEarth panels in a .NET application's window does not yield nice code
22:27:23 <SmatZ> about the "rep ret" - it is a gcc bug while using Intel syntax, fixed in newest releases
22:27:34 <SmatZ> skidd13: I am looking forward to see it
22:28:07 <skidd13> I got only rid of GB in this function. This should increase it a bit.
22:28:08 <SmatZ> Rubidium: only the description is enough for me :)
22:28:26 <ln-> Eddi|zuHause3: pity. it was a good series, although i don't remember almost anything about it.
22:28:34 <skidd13> SmatZ: Assigned allready on the patch-mailinglist?
22:29:25 <Eddi|zuHause3> ln-: but in my defense, in east germany you hardly ever had the ability to watch western shows
22:29:34 <SmatZ> too bad my Firefox wants to save all attachments to a file
22:29:44 <SmatZ> so I cannot browse patches in the browser...
22:30:00 <SmatZ> save + open in an editor is boring :-/
22:30:55 <Rubidium> SmatZ: curl + gpm + less maybe?
22:32:47 <SmatZ> Rubidium: nice, now I would only need something with C++ syntax highlighting :)
22:33:01 <SmatZ> something like paste.openttd.org to browse attachments would be nice
22:34:28 <skidd13> SmatZ: If I were AOL I'd say You've got post :D
22:36:36 <SmatZ> skidd13: do you mean some kind of private message?
22:37:26 <TrueBrain> skidd13: did you do some profiling?
22:37:36 <Rubidium> skidd13: I think it would rather be something like 'You've got mail' (trying to translate the German dubbed version of "You've got mail" back to English?)
22:38:09 <skidd13> Rubidium: whatever ;)
22:38:11 <TrueBrain> skidd13: btw, I didn't fix the attachment stuff yet, let me see if it is possible..
22:38:36 <TrueBrain> what should I ask him? :p
22:38:52 <TrueBrain> skidd13: you can profile too, you know ;)
22:38:58 <TrueBrain> too lazy, I ask you :)
22:39:01 <SmatZ> skidd13: now when some parts are commited, what patch should I profile?
22:39:34 <skidd13> SmatZ: The one from the mailinglist ;) But pleas replace the n -= 1; with n--;
22:39:56 <SmatZ> skidd13: it's ok, compiler will handle it the same way
22:40:16 <skidd13> I know, but I prefer cleaner style ;)
22:41:13 <SmatZ> there are a lot of them :D
22:42:08 <skidd13> TrueBrain: I'm at profiling too. But my pc is so damned slow :D
22:42:42 <TrueBrain> slower computers show differences better
22:43:07 <SmatZ> skidd13: you may use ccache or distcc to compile faster :)
22:43:36 <skidd13> TrueBrain: I hate waiting for the next build ;)
22:43:42 *** ThePizzaKing has joined #openttd
22:43:43 <TrueBrain> skidd13: I made some changes to the maillist, it might or might not help with attachments ;)
22:44:36 <TrueBrain> skidd13: I made some changes to the maillist 2 seconds ago, it might or might not help with attachments ;)
22:44:52 <skidd13> checked one second ago ;)
22:45:05 <TrueBrain> skidd13: I made some changes to the maillist 2 seconds ago, it might or might not help with attachments, but will only work for new incoming emails ;)
22:47:39 <SmatZ> skidd13: are you the 'register' attribute is a good idea?
22:48:16 <TrueBrain> SmatZ: shouldn't a compiler decide on that?
22:48:27 <TrueBrain> as using register for such functions might in fact decrease performance
22:49:43 <SmatZ> TrueBrain: actually, I have the same opinion
22:51:39 <SmatZ> +template<typename T> static inline T ALIGN(const T x, register uint n) {
22:51:56 <SmatZ> I hope the compiler won't place the constant in a register :)
22:52:16 <SmatZ> and given that most architectures will have 2 registers for function parameters...
22:53:13 <skidd13> SmatZ: it's not a constant
22:53:31 <SmatZ> it is often called with a constant parameter, isn't it?
22:54:28 <skidd13> it is called often with the same parameter, but the value itself is not constant ;)
22:54:28 <TrueBrain> 'static inlines' are assumed to be optimized by the compiler as optimal as possible, including pushing things via the register
22:54:39 <TrueBrain> but, in fact, I believe register in inline params are ignored ;)
22:55:30 <skidd13> Ok I removed all the register stuff
22:55:40 <SmatZ> + return (T)(x << n | x >> (sizeof(x) * 8 - n));
22:55:52 <SmatZ> you really know the C++ operator priority ;-)
22:56:24 <SmatZ> the same as in the original macro
22:57:08 <skidd13> No obvious possible outsourcing or simplification so keep it ;)
22:57:28 <skidd13> That's what I thought
22:58:35 <skidd13> what about -fast-math ?
22:58:36 <SmatZ> with -O3 -fno-inline -fno-optimize-sibling-calls
22:58:47 <SmatZ> skidd13: it optimizes only floating point operations
22:59:16 <skidd13> SmatZ: Thanks... One thing I didn't knew
23:02:17 <SmatZ> it is OK when inlining is on
23:07:42 *** TrainzStoffe has joined #openttd
23:08:21 *** lolman is now known as John
23:08:29 *** John is now known as lolman
23:12:47 <SmatZ> running 2 copies of ottd with nice -19
23:12:58 <SmatZ> make smy computer really laggy
23:14:59 *** TrainzStoffe is now known as Stoffe
23:18:16 <SmatZ> I don't know where is the performance gain (eg. in which function), but the skidd13's patch makes it faster
23:18:45 <SmatZ> I did 2 tests and both ended with similiar results
23:18:57 <SmatZ> I will recheck openttd.cfg to be the same
23:21:54 <Unknown> TrueBrain: got a second?
23:24:24 <SmatZ> I shouldn't be profiling on a savegame with AI...
23:26:11 <TrueBrain> Unknown: always just speak, maybe someone else have the answer
23:26:19 <TrueBrain> SmatZ: might be not the smartest
23:26:36 *** Unknown is now known as Unkown_Entity
23:27:51 *** liona29 has joined #openttd
23:28:14 <Unkown_Entity> would 320 KB for each _spritecache and _spritecache_ptr be enough?
23:28:33 <TrueBrain> spritecache is a dynamic array nowedays
23:28:57 <TrueBrain> anyway, 320 KiB is kind of low
23:28:58 <SmatZ> patched version : user 1m6.190s original version : user 1m0.650s
23:29:19 <TrueBrain> what would happen, is that on low cache-values, it reads a lot from disk
23:29:22 <TrueBrain> so the game will get pretty slow
23:29:27 <TrueBrain> but I guess it is a matter of trail-and-error
23:29:33 <TrueBrain> Unkown_Entity: why exactly? Is the memory that limited?
23:29:56 <TrueBrain> SmatZ: lol, was that without inlining?
23:30:02 <TrueBrain> haha, no wonder it is slower :)
23:30:05 <TrueBrain> I could tell you that :p
23:30:15 <TrueBrain> a jump and a return
23:30:20 <Unkown_Entity> yes, thge DS only has 4 MB of main memory and for some reason I can only allocate about 2 MB.
23:30:25 <glx> replacing a macro without inlining?
23:30:27 <TrueBrain> depending on your CPU, they are both expensive :)
23:30:37 <TrueBrain> Unkown_Entity: iek, that makes it tricky...
23:30:41 <SmatZ> it was my fault, I forgot to turn off -fno-inline -fno-optimize-sibling-calls
23:30:55 <Unkown_Entity> so I abused the VRAM for the spritecache but I wasn't sure if that would cause much trouble
23:31:04 <TrueBrain> oh, wait, I misread you:p
23:31:11 <Unkown_Entity> an VRAM only has 640 KB, so not much better :(
23:31:12 <TrueBrain> Unkown_Entity: lol, is that possible?
23:31:26 <TrueBrain> (clearly, but okay :))
23:31:36 <Unkown_Entity> TrueBrain: sure, you can basically map a variable to any memory you want to
23:31:42 <TrueBrain> anyway, a cache of 640 KiB should be enough I guess
23:32:08 <TrueBrain> Unkown_Entity: real pcs have virtual memory, which doesn't make that job that easy ;)
23:32:35 <SmatZ> Unkown_Entity: interesting, but given how OTTD is slow on a Pentium class PC - are you able to play it?
23:33:06 <TrueBrain> that is, with old pathfinder and 256x256 maps?
23:33:10 <TrueBrain> it in fact should run perfectly..
23:33:13 * glx should restart his P133 to check :)
23:33:18 <SmatZ> I don't know, I play only HUGE maps
23:33:21 *** Diabolic-Angel has quit IRC
23:33:34 <TrueBrain> SmatZ: see, that is YOUR mistake, not OTTDs
23:33:54 <Unkown_Entity> currently I have based my code on the repository but I might have to fork it from 0.5.3 because the older versions only use about half the memory
23:33:55 <SmatZ> you are right, I am sorry, <3 OTTD
23:34:13 <TrueBrain> Unkown_Entity: huh?
23:34:33 <TrueBrain> Unkown_Entity: tweaked correctly, trunk shouldn't use more than 0.5.3
23:34:40 <TrueBrain> (if you don't load up all the extras, that is ;))
23:35:10 <Unkown_Entity> TrueBrain: I comiled both 0.5.3 and trunk with the same settings and trunk used 4.0MB, 0.5.3 only 2.3MB
23:35:27 <TrueBrain> Unkown_Entity: are you sure you used the same amount of sprite-cache value?
23:35:31 <glx> trunk reduced static array usage
23:35:41 <TrueBrain> in 0.5.3 this was hardcoded, in trunk it is dynamic, but a higher default value
23:36:03 <Unkown_Entity> TrueBrain: ah, that must be the reason. I didn't know that
23:36:09 <SmatZ> the vehicle cache is bigger, too, but I think it went from 64kB to 2x128kB
23:36:10 <TrueBrain> put, for example, the sprite_cache_size at 1
23:36:20 <Unkown_Entity> I only studied the code from trunk I didn't really compare it to 0.5.3
23:37:09 <TrueBrain> SmatZ: if you mean the vheicle pool, it is the same size
23:37:13 <TrueBrain> we don't have a vehicle cache
23:37:20 <TrueBrain> only a pathfinder cache, but that is YAPF
23:37:23 <Unkown_Entity> but still, even if i'd ever get it running on the ds i might have to fork it to make use of both processors
23:37:43 <TrueBrain> Unkown_Entity: OpenTTD and multicore is a no-can-do
23:38:11 <glx> multithread is only used for saving and map generation
23:38:13 <TrueBrain> any 'fork' would result in a direct rewrite from scratch in order to get that working
23:38:39 <Unkown_Entity> TrueBrain: definately sound processing and maybe even the blitter could be done on the second processor i think
23:38:40 <TrueBrain> glx: and map-generation is fake :)
23:38:42 <SmatZ> TrueBrain: I mean static Vehicle *_new_vehicle_position_hash[TOTAL_HASH_SIZE]; and _vehicle_position hash, both have 128kB - in 0.5 is only on hash pool, with 64kB (I think)
23:38:52 <TrueBrain> Unkown_Entity: blitter, no (believe me, I wrote it :p). Sound, yes, possible
23:38:56 <TrueBrain> but little gain I guess
23:39:08 <TrueBrain> SmatZ: ah, position hash, yes, indeed
23:39:27 <SmatZ> TrueBrain: finally, working comparison of skidds13's updates: trunk - user 0m41.550s ... patched - user 0m41.520s
23:39:39 <TrueBrain> SmatZ: but, if you don't use maps bigger than 256x256, it is safe to say you can reduce the HASH_SIZE :)
23:39:47 <TrueBrain> SmatZ: good, that is promising :)
23:39:49 <glx> SmatZ: right, but 0.5.x has bugs in tunnels because of that IIRC
23:39:51 <TrueBrain> you are signed on to the maillist?
23:40:00 <TrueBrain> glx: you are correct :)
23:40:16 <Unkown_Entity> TrueBrain: but couldn't the second processor check for dirty sprites and redraw them on it's own?
23:40:25 <TrueBrain> Unkown_Entity: no, as it needs _m access
23:40:37 <TrueBrain> SmatZ: please post the profile result and how you did it (how many runs etc etc) to the maillist
23:40:52 <Unkown_Entity> maybe i could map _m to shared memory ^^
23:40:56 <TrueBrain> Unkown_Entity: you want a longer explanation? (I dunno how much you know already :))
23:41:16 <TrueBrain> no no, the share problem isn't the issue at all
23:41:24 <Unkown_Entity> TrueBrain: not yet, i'll spare you the time. first i have to get it running on one preocessor ;)
23:41:35 <TrueBrain> well, let me tell you anyway :p
23:41:44 <glx> the problem is to synchronise read and write to _m
23:41:46 <TrueBrain> OpenTTD works around _m, our map
23:41:54 <TrueBrain> reading, writing, everything is done via wrappers, but in no common order
23:42:03 <TrueBrain> if you would make threads, you would need to make read/write locking
23:42:11 <TrueBrain> as else one thread might be writing in _m, while the other is reading
23:42:17 <TrueBrain> so, drawing bytes means _m access
23:42:26 <glx> and every action should ALWAYS happen in the same order
23:42:26 <TrueBrain> locking would be dead slow in its current form
23:42:38 <TrueBrain> glx: the blitter doesn't, so that can be worked around
23:42:48 <SmatZ> huh, it would be impossible to make it network-safe
23:42:52 <TrueBrain> Unkown_Entity: above that, the blitter only consumes a little 2% CPU time in the last profile I had
23:43:15 <Unkown_Entity> ok, I agree, sharing _m would be impossible
23:43:26 <exe> guys how about doing vehicle ticks 50% less often? and interpolating their positions on screen view.
23:43:33 <TrueBrain> so, the only way would be, to 'cache' all the tiles you want to draw to blitter in a temporary cache
23:43:39 <Unkown_Entity> what other parts took enough cpu time to be worth the effort to move them to the other cpu?
23:43:39 <TrueBrain> which is a thread-safe-stack
23:43:41 <glx> path finding is the most consuming task I think, and it's impossible to do it in a separate thread
23:43:51 <TrueBrain> exe: costs CPU too ;)
23:44:13 <TrueBrain> Unkown_Entity: the main consumer is either TileLoop, which we just said was impossible to do, or TrainController, whcih depends heavily on _m too
23:44:28 <TrueBrain> exe: depends, not reall
23:44:32 <TrueBrain> it already is at a mimimum
23:44:38 <TrueBrain> in-tile-movement is simple
23:44:43 <TrueBrain> pathfinder, that is your killer :)
23:45:03 <TrueBrain> disabling realistic acceleration helps against CPU load too btw :p
23:45:04 <Unkown_Entity> TrueBrain: too bad then. :( but currently the main issue seems to be memory not cpu time anyway.
23:45:17 <TrueBrain> Unkown_Entity: well.. all options to make things multicore, involves more memory ;)
23:45:21 <exe> do all the pathfinders crawl long straight rail segments one-by-one tile?
23:45:23 <TrueBrain> so I guess you are out of luck for thatone anyway
23:45:40 <TrueBrain> exe: basicly, yes. YAPF does something extra: it caches the result
23:45:45 <TrueBrain> so, one segment, 1 check
23:45:51 <TrueBrain> which increases the performance incredible
23:45:53 <TrueBrain> at cost of memory :)
23:46:09 <TrueBrain> Unkown_Entity: but some tips to reduce memory: don't allow map-sizes > 256 in both directions
23:46:20 <TrueBrain> reduces things like pool-size, hash-size, spritecache-size
23:46:26 <TrueBrain> and you should be able to fit it in 2 MiB just fine
23:46:32 <TrueBrain> (at the cost of bigmaps)
23:46:34 <Unkown_Entity> TrueBrain: i already lowered map size to 64x64 :(
23:46:51 <exe> how does it know when to clear cache?
23:46:54 <Jello> how much ram do u have?
23:47:09 <TrueBrain> exe: callbacks at layout change procedures
23:47:16 <Unkown_Entity> theoreticall 4 MB but for some reason i can only allocate 2 MB
23:47:16 <TrueBrain> (placing signal is bye segment cache)
23:47:24 <TrueBrain> Unkown_Entity: 2 MiB per CPU?
23:47:44 <Unkown_Entity> might be because I currently onlywork on an emulator, I'll get the hardware next week hopefully
23:47:53 <Unkown_Entity> TrueBrain: no, they are shared between both
23:47:55 <TrueBrain> what kind of OS does it run?
23:48:16 <Unkown_Entity> TrueBrain: the emulator? i use DeSmuMe on Linux
23:48:36 <Unkown_Entity> no OS at all ^^
23:48:48 <Unkown_Entity> there's a linux for ds in development though
23:48:55 <TrueBrain> it needs something to boot a game
23:49:04 <SmatZ> Bjarni and I (maybe others, I don't know) was doing some experiments about threading - some kind of vehicles can be done in separated threads (air, ships, rail+road), and rendering can be done threaded... but there are synchronization points, most noticeably GetSprite, that make everything really slow
23:49:21 <SmatZ> you get 10% performance increase, but lose 20% because of synchronizing
23:49:26 <TrueBrain> SmatZ: above that, any general solution makes the game DEAD slow on single-core
23:49:36 <Unkown_Entity> TrueBrain: that's done in hardware i believe. it simply has to know the memory location of the binary which is always the same on every cartridge
23:49:47 <SmatZ> TrueBrain: it could be a compile-time option
23:50:03 <TrueBrain> Unkown_Entity: so it loads directly from BIOS.. so you need to call interupts yourself? Or is an API taking care of that?
23:50:13 <TrueBrain> (what I am trying to figure out: how to compile a game? :p)
23:50:23 <Unkown_Entity> TrueBrain: i'm using the open source API devkitPro
23:50:37 <TrueBrain> and it produces a working binary?
23:50:56 <TrueBrain> what kind of 'target' is DS? (gcc is used?)
23:51:05 <Unkown_Entity> TrueBrain: yes, when you have it setup correctly. actually it produces two binaries, one for each cpu
23:52:11 <Unkown_Entity> TrueBrain: it's gcc, you can download it from www.devkitpro.org. easy to set up actually
23:52:29 <TrueBrain> ah, it is an ARM CPU
23:52:33 <Unkown_Entity> but you can't really test your stuff on emulators. they are to slow and not accurate enough yet
23:53:52 <TrueBrain> oh, and the tihng I always wondered about the most: does it have 2 GPUs?
23:53:57 <TrueBrain> (sorry, finally someone I can ask ;))
23:54:09 <TrueBrain> SmatZ: btw, I really don't see any future in multithread and OpenTTD :)
23:54:51 <TrueBrain> (okay, I need to be more exact: parallel threads :p)
23:54:58 <TrueBrain> NoAI uses threads ;) (serial-threads :))
23:55:34 <Unkown_Entity> TrueBrain: 33,6 MHz on one, I'm not sure about the other. and it's only one GPU I think. not sure though
23:55:36 <exe> how about relaxing pathfinding corretness a bit? then train can pathfind in background threads some time before exiting tile on crossing.
23:56:18 <TrueBrain> exe: pathfinding means a lot of _m access, so no thread for mister pathfinder...
23:56:39 <SmatZ> hmm I got an idea about compiling OTTD for some PIC :)
23:57:22 <TrueBrain> the ARM7 really is just there for support
23:57:44 <Unkown_Entity> the ARM9 runs at 67 MHz, the ARM7 at 33 MHz
23:57:57 <TrueBrain> so combined 100 MHz... I expected more
23:58:00 <TrueBrain> the double in fact, was my estimate
23:58:02 <Unkown_Entity> i think most commercial games don't use the ARM7 at all yet
23:58:27 <TrueBrain> you can only run think-routines on it :)
23:58:33 <Unkown_Entity> but it can be put to use for sound processing very effectively
23:58:41 *** Prof_Frink has joined #openttd
23:58:49 <TrueBrain> btw, I don't see a GPU.. I guess the ARM9 _is_ the GPU? :)
23:58:51 <SmatZ> TrueBrain: do you want me to run OTTD with profiling with optimizations turned on?
23:59:10 <SmatZ> including inlining and optimized sibling calls?
23:59:19 <TrueBrain> SmatZ: of course. See, I know by looking at the code skidd13's patch is slower without optimizations
23:59:26 <TrueBrain> so, to profile that gives me a 'duh' moment :)
23:59:35 <TrueBrain> mor eimportant I guess is: what happens in a release version
23:59:38 <TrueBrain> so full optimizations
23:59:44 <TrueBrain> of course, including what happens on a -fno-inline is useful too
23:59:48 <TrueBrain> for the developers under us
23:59:53 <TrueBrain> and damn, I really talk too much :p
continue to next day ⏵
