--- Log opened Sat Oct 17 00:00:54 2015 00:10 -!- xenos1984 [~xenos1984@127-166-191-90.dyn.estpak.ee] has joined #navit 02:53 -!- _rd [~rd@p5B2C7315.dip0.t-ipconnect.de] has joined #navit 03:06 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has joined #navit 03:11 -!- _rd [~rd@p5B2C7315.dip0.t-ipconnect.de] has quit [] 03:29 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has quit [] 05:15 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has joined #navit 05:20 #navit: < RobertP> Hi all. 05:21 #navit: < RobertP> I have a question: Where can i find an example where a trigger source of a callback is defined? 06:49 -!- tryagain [~quassel@178.216.76.92] has joined #navit 06:52 #navit: < tryagain> RobertP hi 06:52 #navit: < RobertP> hi 06:53 #navit: < tryagain> is this https://github.com/navit-gps/navit/blob/trunk/navit/graphics/android/graphics_android.c#L496 what you're looking for? 06:54 #navit: < RobertP> looks like. thanks 07:36 #navit: < RobertP> FYI: Routing from Freiburg/Breisgau (south west Germany) to Sassnitz (north east Germany) takes 6min on the RPI2 with profile CAR 07:57 #navit: < RobertP> erm forgot to say: on raspbian jessie. 08:04 #navit: < tryagain> i have somewhat the same time on my desktop (maybe 3 minutes, did not meausred it exactly) with binfile accessed over over internet over sshfs. 08:07 #navit: < tryagain> so this might be related to disk throughput. What sd card do you use? Maybe switching to higher class would help? 08:12 #navit: < tryagain> if i switch the route destination to Stuttgart and then back to Sassnitz, it takes only a few seconds. So bottleneck seems to be really in file read/write operations speed. 08:15 #navit: < tryagain> when i restart navit, and rebuild the route, it works fast again. So I guess it's operating system filesystem cache what makes the difference. 08:18 -!- jandegr [5741674b@gateway/web/freenode/ip.87.65.103.75] has joined #navit 08:18 #navit: < tryagain> RobertP you have way less RAM, but trying my experiments may shed some light too. 08:21 #navit: < RobertP> the map is on a usb stick 08:22 #navit: < tryagain> well usb sticks have different throughput too... 08:23 #navit: < jandegr> roberp it seems the code in your repo is outdated, update and it should route in less than half the time 08:24 #navit: < RobertP> ok. ill do so. thank you very much. 08:24 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has quit [Quit: bye] 08:33 #navit: < jandegr> correction, routing is not touched but the time between setting destination and instructions finished is shortened by a tweak in navigation 08:33 #navit: < jandegr> but most users see that timeinterval as the time it takes to route 08:35 #navit: < jandegr> robertp : please report time needed before and after code update for the statistics department :) 08:35 -!- jandegr [5741674b@gateway/web/freenode/ip.87.65.103.75] has quit [Quit: Page closed] 08:37 #navit: < tryagain> jandegr, RobertP i would suggest to reboot the OS between separate test runs. 10:39 -!- zoff99 [~chatzilla@unaffiliated/zoff99] has joined #navit 10:46 -!- zoff99 [~chatzilla@unaffiliated/zoff99] has quit [Quit: ChatZilla 0.9.86-rdmsoft [XULRunner 1.9.1/20090707221522]] 13:06 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has joined #navit 13:16 #navit: < RobertP> jandegr & tryagain: I am impressed! With updated code (from navit-gps/trunk) it only needs 101s (1:41). 13:44 -!- jandegr [5741674b@gateway/web/freenode/ip.87.65.103.75] has joined #navit 13:46 #navit: < jandegr> RobertP thanks for the report, it seems like my promise to do it in less than half of the time is fulfilled 13:48 #navit: < RobertP> But why is Android so much faster? Because of the faster memory-access? 13:50 #navit: < jandegr> I am searching in the logs for the report I did on that route a while ago 13:53 #navit: < jandegr> found it, sept 29 I reported 22 seconds on asus memopad me173 13:53 #navit: < jandegr> is that the start and endpoint ? 13:53 #navit: < jandegr> https://dl.dropboxusercontent.com/u/93775123/Navit/routing/Screenshot_2015-09-29-08-25-11.jpg 13:54 #navit: < RobertP> yes 13:55 #navit: < jandegr> but I must admit I have more changes to speed it up than what is commited to github a while ago 14:08 #navit: < jandegr> I am installing vanilla navit now so in a while I can report for an A/B memopad/RPI with same navit version 14:16 #navit: < RobertP> what's vanilla navit? 14:26 #navit: < jandegr> vanilla navit is the one you can find in github, without some of my other changes 14:30 #navit: < RobertP> Ok, and after your changes its chocolate> :D 14:30 #navit: < jandegr> 80 seconds, but with a map that includes street_4 in all routing 14:31 #navit: < RobertP> on which device? 14:33 #navit: < jandegr> asus memopad, map on SDcard that does 20MB/s read in the benchmark test 14:52 -!- jandegr [5741674b@gateway/web/freenode/ip.87.65.103.75] has quit [Quit: Page closed] 15:04 -!- jjelen [~jjelen@ip4-95-82-163-69.cust.nbox.cz] has joined #navit 15:10 -!- jjelen [~jjelen@ip4-95-82-163-69.cust.nbox.cz] has quit [Ping timeout: 250 seconds] 15:17 #navit: < tryagain> RobertP so your usb drive benchmarks could be useful. Maybe simplest way would be to run sudo hdparm -t /dev/sdX 15:19 #navit: < RobertP> the USB drive might be slow. Its a USB drive, that is almost inside the connector. 15:19 #navit: < RobertP> Those are mostly slow. 15:23 #navit: < RobertP> i get 18MB/s 15:23 #navit: < RobertP> sd shows 16,1 MB/s 15:24 #navit: < tryagain> ha, looks like almost the same as jandegr' test. i appears there's a different bottleneck on rpi than disk. 15:26 #navit: < RobertP> thats measured when navit was running. i have to reboot soon, then i will try it again. 15:28 #navit: < tryagain> we have compile-time options to limit usage of floating point arithmetics. I think we have it's disabled by default for android. Do you know your actual setting? 15:29 #navit: < RobertP> no. 15:32 #navit: < tryagain> what do you have for "grep AVOID_FLOAT config.h" when runned in the build dir? 15:33 #navit: < tryagain> btw do you have swap enabled on your rpi? (cat /proc/swaps) 15:34 #navit: < RobertP> "grep AVOID_FLOAT config.h" sais /* #undef AVOID_FLOAT */ 15:35 #navit: < RobertP> top and /proc/swaps say that there is a swap file 15:36 #navit: < RobertP> with about 100MB 15:37 #navit: < tryagain> quite small swap for your memory configuration... Unsure if it could make any sense anyway... But maybe switching it off worths trying. 15:38 #navit: < tryagain> my android build has -DAVOID_FLOAT=1 flag passed to cmake. So maybe that's another thing to try. 15:39 #navit: < tryagain> afk 16:08 #navit: < RobertP> avoid_float bought another 10s 16:09 #navit: < RobertP> *brought 16:12 #navit: < RobertP> Where can i find better espeak voices? 16:39 #navit: < tryagain> well espeak site has some references for mbrola voices for selected languages. Though never tried them. 16:42 #navit: < tryagain> there are other synthesers available. I think festival is reported to have better results. 16:42 #navit: < tryagain> maybe 1G of ram and quadcore is enough to run it. 16:43 #navit: < RobertP> ihave horrible jitter with espeak 16:43 #navit: < tryagain> also, KaZeR had some success using online google speech synthesizis and capturing it for offline use 16:44 #navit: < tryagain> do you have jitter only when you run it by navit or always? 16:45 #navit: < tryagain> s/capturing/caching/ 16:45 #navit: < RobertP> i dont need espaek without navit 16:45 #navit: < RobertP> :D 16:46 #navit: < tryagain> but you still can try it to see if it's espeak problem itself or it arises when one core is heavy loaded... 16:47 #navit: < RobertP> its espeak 16:47 #navit: < RobertP> from commandline its the same 16:48 #navit: < tryagain> iirc it has wav file outupt. what if you write to wav and then play the file? 16:52 #navit: < tryagain> btw we probably should document somewhere that for ARM architectures it's suggested to use AVOID_FLOAT... 16:52 #navit: < tryagain> ...or even better to code it into CmakeFiles... 16:53 #navit: < RobertP> wav output is ways better, but the voices are still horrible 16:53 #navit: < RobertP> wavs 16:56 #navit: < tryagain> i had problems with cyrillic letters on android port of espeak. It was considering some letters to always mark word ends. So it was pronouncing some words by letter. 16:57 #navit: < tryagain> I would say espeak is imperfect but usable. 16:58 #navit: < RobertP> and what is perfect and usable? festival? 17:01 #navit: < tryagain> iv'e never actually tried it very much. I think it misses Russian completely. 17:02 #navit: < tryagain> i keep with espeak on my device. 17:10 #navit: < tryagain> btw we also have a feature to play prerecorded speech fragments. But you'll miss street names because you can't record them all. 17:10 #navit: < tryagain> And you'll need someone (or yourself) to record these phrases. 17:11 #navit: < tryagain> Also, there's no predefined set of phrases defined anywhere. 17:34 -!- RobertP [5cd9e500@gateway/web/freenode/ip.92.217.229.0] has quit [Quit: gn] 18:03 -!- tryagain [~quassel@178.216.76.92] has quit [Remote host closed the connection] 19:34 -!- xenos1984 [~xenos1984@127-166-191-90.dyn.estpak.ee] has quit [Quit: Leaving.] 21:18 -!- KaZeR [~KaZeR@c-67-161-64-186.hsd1.ca.comcast.net] has quit [Ping timeout: 250 seconds] --- Log closed Sun Oct 18 00:00:56 2015