Log for #openttdcoop.devzone on 20th November 2014:
Times are UTC Toggle Colours
10:20:33  *** tycoondemon has quit IRC
10:39:53  *** tycoondemon has joined #openttdcoop.devzone
12:37:36  *** Supercheese has quit IRC
12:47:17  *** gelignite has joined #openttdcoop.devzone
13:09:40  *** gelignite has quit IRC
13:09:40  *** tycoondemon has quit IRC
13:09:40  *** dihedral has quit IRC
13:09:40  *** orudge has quit IRC
13:10:59  *** gelignite has joined #openttdcoop.devzone
13:10:59  *** tycoondemon has joined #openttdcoop.devzone
13:10:59  *** dihedral has joined #openttdcoop.devzone
13:10:59  *** orudge has joined #openttdcoop.devzone
13:15:40  <DevZone> Project World Airliner Set build #635-push: SUCCESS in 3 min 37 sec:
13:42:45  <DevZone> Project World Airliner Set build #636-push: SUCCESS in 3 min 6 sec:
14:15:15  <DevZone> Project World Airliner Set build #637-push: SUCCESS in 3 min 5 sec:
14:33:09  *** tycoondemon2 has joined #openttdcoop.devzone
14:33:19  <DevZone> Project World Airliner Set build #638-push: SUCCESS in 3 min 4 sec:
14:35:48  *** tycoondemon has quit IRC
14:39:48  *** LSky` has joined #openttdcoop.devzone
15:53:59  *** Alberth has joined #openttdcoop.devzone
15:54:52  *** ODM has joined #openttdcoop.devzone
16:51:49  <DevZone> Project xussrset - Trains from Russia build #558-push: SUCCESS in 4 min 16 sec:
16:59:42  *** oskari89 has joined #openttdcoop.devzone
18:08:27  <Alberth> eints needs a some form of notes for standard translation of terms
18:10:04  <planetmaker> you mean to keep notes on that XX should always be translated to YY?
18:16:47  <Alberth> yep
18:17:16  <Alberth> the related strings do help, but not always
18:21:28  <planetmaker> yes, that's true. I wonder though how I would want such help to look like
18:23:22  <Alberth> it could even do somewhat automagic translation of similar sentences :)
18:23:41  <Alberth> most files have a lot of almost duplicate strings
18:23:49  <Alberth> *projects
18:28:04  *** juzza1 has joined #openttdcoop.devzone
18:29:58  <planetmaker> yes. It could recommend a translation based on other translations from same project
18:32:42  <Alberth> V453000: STR_industry_5X_no_input has a {GREEN} colour code in the string, while eg STR_industry_3CC_no_input and STR_industry_3C_no_input  don't have that
18:33:07  <V453000> Sylf: ^ :)
18:33:15  <V453000> Alberth: Sylf now does all the coding, I only do imagez
18:33:29  <V453000> thanks for the report though (:
18:33:39  <Alberth> Sylf:  STR_industry_4X_need_food   has no "." at the end
18:33:47  <Alberth> :)
18:39:52  <Alberth> ok, Dutch translation updated, please wait about 23:30 +1:53 for a build :)
18:40:35  <DevZone> Project road-hog build #482-nightlies: SUCCESS in 34 sec:
19:15:08  *** andythenorth has joined #openttdcoop.devzone
19:22:54  *** Supercheese has joined #openttdcoop.devzone
19:27:37  *** oskari892 has joined #openttdcoop.devzone
19:34:48  *** oskari89 has quit IRC
20:03:51  *** frosch123 has joined #openttdcoop.devzone
20:19:44  <andythenorth> frosch123: so how is faster nml going? o_O
20:27:37  <Rubidium> slowly? ;)
20:28:03  <frosch123> i think i have a working concept
20:28:08  <frosch123> for now only yeti is twice as fast
20:28:49  <frosch123> at least i can drop my plans for writing a grfcodec replacement. it's easier to beat grfcodec with nml :)
20:29:42  <frosch123> other than that "a wild deadline appeared" or so :p
20:32:19  <planetmaker> a wild deadline?
20:32:37  <andythenorth> you think nmlc will go faster than grfcodec?  Or you think nmlc is easier to work on? o_O
20:33:19  <frosch123> andythenorth: for big 32bpp sprites my nml beats grfcodec in encoding speed due to multithreading
20:33:38  <andythenorth> wow
20:33:50  <andythenorth> super :)
20:34:16  <frosch123> turned out you only need to move a single 20 line function into a compiled accelerator module
20:34:29  <frosch123> and then you can use ordinary python threading
20:34:37  * Rubidium wonders whether SSE can improve encoding
20:34:38  <andythenorth> that is neat
20:34:50  * andythenorth likes simple speed ups
20:34:59  <frosch123> the single function takes so much time, that the global interpreter lock is no bottleneck
20:35:52  <andythenorth> been wondering if FIRS would be faster if I simplified the nml
20:36:11  <Rubidium> who 20 line function is that then?
20:36:19  <Rubidium> s/who/what/
20:36:29  <frosch123> the sprite compression
20:36:43  <andythenorth> so thread safe
20:36:44  <frosch123> reading of png and cropping and stuff is maginally in comparison
20:36:46  <DevZone> Project Iron Horse build #1123-push: SUCCESS in 3 min 10 sec:
20:37:35  <frosch123>
20:38:18  <frosch123> for yeti the encode() function takes about 0.5 seconds per sprite
20:38:49  <frosch123> i guess you can utilize something along 8 or 10 cores before the single-threaded python cannot feed the encoder anymore
20:39:42  <frosch123> <- that were the numbers from a simple prototype
20:40:40  <frosch123> it only utilised two cores on average though, since it could not encode stuff from multple source images
20:40:57  <frosch123> it always waited for the encoding to finish before proeceding with the next source png
20:41:19  <frosch123> but as you can see, it only works for big yeti sprites, it completely fails for 8bpp vehicle sets :)
20:42:08  <planetmaker> well, already a big gain, I think!
20:42:17  <frosch123> above _lz77.c can still be improved by 20% btw
20:42:51  <frosch123> i profiled it against grfcodec, and grfcodec's strategy seems to be somewhat more effective on average
20:42:56  <andythenorth> I think I’d make FIRS much faster by removing all the tile checks in spritelayouts
20:43:01  <andythenorth> but I haven’t proven that yet
20:55:07  <andythenorth> binning a stack of hide_sprite calculations saves about 10s
20:55:12  <andythenorth> not significant really
20:55:22  <frosch123> and not constructive :p
20:58:07  <andythenorth> intrigued where the time is spent
20:58:24  <andythenorth> and have never acquired profiling skills, other than ‘delete and time'
21:14:41  *** ODM has quit IRC
21:53:09  *** andythenorth has left #openttdcoop.devzone
22:01:01  *** oskari892 has quit IRC
22:09:18  *** LSky` has quit IRC
22:27:51  *** frosch123 has quit IRC
22:36:18  *** gelignite has quit IRC
23:03:36  *** George|2 has joined #openttdcoop.devzone
23:03:36  *** George is now known as Guest369
23:03:36  *** George|2 is now known as George
23:08:57  *** Guest369 has quit IRC
23:35:33  *** Alberth has left #openttdcoop.devzone

Powered by YARRSTE version: svn-trunk