| Time |
Nick |
Message |
| 00:07 |
|
bdju joined #luanti |
| 01:08 |
ireallyhateirc2 |
I'l trying to build the latest master branch and get this in tests: |
| 01:08 |
ireallyhateirc2 |
(server) 2025-04-13 01:04:16: ERROR[Server]: ...40.drv-0/source/games/devtest/mods/unittests/version.lua:11: assertion failed! |
| 01:08 |
ireallyhateirc2 |
(server) [FAIL] test_get_version - 0ms |
| 01:09 |
ireallyhateirc |
any clues why this happens? |
| 01:29 |
|
SFENCE joined #luanti |
| 01:42 |
|
operagame joined #luanti |
| 02:42 |
|
SFENCE joined #luanti |
| 03:03 |
|
Verticen joined #luanti |
| 03:32 |
|
shinbet joined #luanti |
| 03:45 |
|
sinvet joined #luanti |
| 03:51 |
|
shinbet joined #luanti |
| 04:00 |
|
MTDiscord joined #luanti |
| 04:05 |
|
Glorfindel joined #luanti |
| 05:05 |
|
fluxionary joined #luanti |
| 05:21 |
|
SFENCE joined #luanti |
| 05:54 |
|
SFENCE joined #luanti |
| 06:26 |
|
SFENCE joined #luanti |
| 06:37 |
|
SFENCE joined #luanti |
| 06:49 |
|
SFENCE joined #luanti |
| 08:03 |
|
SliverFlowCipher joined #luanti |
| 08:22 |
|
Warr1024 joined #luanti |
| 08:23 |
|
jemius joined #luanti |
| 08:30 |
|
Glaedr joined #luanti |
| 08:41 |
|
Noisytoot joined #luanti |
| 08:42 |
|
YuGiOhJCJ joined #luanti |
| 08:46 |
|
Warr1024 joined #luanti |
| 08:52 |
|
Talkless joined #luanti |
| 09:10 |
|
jaca122 joined #luanti |
| 09:14 |
[MatrxMT] |
<Bracket> seems like the version hash is not a string |
| 09:14 |
[MatrxMT] |
<Bracket> why didn't you look at the code? |
| 09:15 |
[MatrxMT] |
<Bracket> nevermind |
| 09:20 |
[MatrxMT] |
<Bracket> please ignore what i said |
| 09:33 |
|
Talkless joined #luanti |
| 09:46 |
|
Talkless joined #luanti |
| 09:53 |
|
Talkless joined #luanti |
| 09:54 |
|
SFENCE joined #luanti |
| 10:01 |
|
Talkless joined #luanti |
| 10:14 |
|
Talkless joined #luanti |
| 11:41 |
|
ireallyhateirc joined #luanti |
| 12:05 |
ireallyhateirc |
I tried the new spatial cache thing with Exile. I didn't look at any numbers, mostly only checked if I can visually feel the difference. |
| 12:06 |
ireallyhateirc |
And the answer is mostly no. Entities usually lag on mapblock updates and our game has a lot of them. |
| 12:11 |
ireallyhateirc |
surprisingly the legboat on mtg appears to be smoother than last time I checked but I also bought an SSD since then |
| 12:11 |
|
mrkubax10 joined #luanti |
| 12:12 |
ireallyhateirc |
but there's still that noticeable and periodic lag where the legs of the legboat lag behind, I believe it happens when the server generates a mapchunk and then emerges blocks |
| 12:27 |
ireallyhateirc |
The thing is probably either https://github.com/luanti-org/luanti/issues/15125 or https://github.com/luanti-org/luanti/issues/14018 |
| 12:29 |
ireallyhateirc |
Or is it because there's no time synchronization between client and server? In Exile for example entities get teleported back and forth when the engine appears to be busy doing something (most likely block emerging) |
| 12:30 |
sfan5 |
block emerging does not happen on the main thread so you need to ask yourself what is blocking it |
| 12:33 |
ireallyhateirc |
are ABMs working on the main thread? |
| 12:34 |
sfan5 |
yes |
| 12:36 |
ireallyhateirc |
well we're processing huge VMs on the main thread and our entities constantly read nodes around them. As for the reduced but still noticeable lags with legboat + mtg then idk what could be blocking there |
| 12:37 |
sfan5 |
the engine can't fix lag sources of your own making |
| 12:38 |
ireallyhateirc |
sfan5, yeah I'm aware. I did not know block emerging should not affect the main thread. |
| 12:39 |
sfan5 |
what the spatial cache PR is about is preventing (additional) server lag when you have lots of entities. it will do literally nothing for a simple "jump into a boat in singleplayer and see if it's smooth" case |
| 12:40 |
ireallyhateirc |
maybe it's just the HDD -> SSD upgrade that made it smooth |
| 12:40 |
ireallyhateirc |
or something else |
| 12:42 |
ireallyhateirc |
My reasoning was that since such a simple mod (legboat + mtg) produces similar lags to Exile then the underlying cause would be engine-related |
| 12:46 |
ireallyhateirc |
I do need to fix the crappy code I wrote though, that's a sure thing. |
| 12:55 |
ireallyhateirc |
sfan5, I started the very same legboat world on the HDD and the boat lags more |
| 12:56 |
sfan5 |
i didn't say that map loading would not affect the main thread, just that it (mostly) doesn't run in it |
| 13:00 |
ireallyhateirc |
so it's mostly "Busy Disks Cause Lag #15125" then? I don't recall playing other games that would lag so much on a HDD |
| 13:00 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/15125 -- Busy Disks Cause Lag |
| 13:00 |
ireallyhateirc |
I doubt the inverse kinematics calculations for the boat are so heavy that we can call it bad code |
| 13:01 |
MTDiscord |
<bastrabun> Try to exhaust your disk IOPS and see if it lags more? |
| 13:01 |
MTDiscord |
<luatic> we need some numbers |
| 13:02 |
sfan5 |
there's far from enough information to even make a guess |
| 13:03 |
ireallyhateirc |
how do I get the info you need? |
| 13:04 |
MTDiscord |
<luatic> some basic things you can do: what is max_lag? could also use the lagometer (https://gitlab.com/sztest/szutilpack/-/tree/master/szutil_lagometer) |
| 13:05 |
MTDiscord |
<luatic> then, you can profile. limitation of that is that you just see the aggregate of everything, buf if something is consistently bogging it down, it should be visible in the flamegraph. |
| 13:06 |
MTDiscord |
<luatic> for profiling the engine, see https://github.com/luanti-org/luanti/blob/master/doc/developing/misc.md#profiling-luanti-on-linux-with-perf. you should ideally profile client and server separately. i think in your case only the server needs profiling. |
| 13:07 |
MTDiscord |
<luatic> for profiling lua, see https://content.minetest.net/packages/jwmhjwmh/jitprofiler/. |
| 13:07 |
ireallyhateirc |
as for the charts I get when pressing F5 - for the FPS chart the upper number is max FPS and the lower is min FPS ? |
| 13:07 |
ireallyhateirc |
and the dips mean lagspikes? |
| 13:08 |
MTDiscord |
<luatic> well is this a client problem or a server problem? |
| 13:08 |
MTDiscord |
<luatic> is your FPS dipping or not? FPS should be essentially independent of the server. |
| 13:08 |
MTDiscord |
<luatic> independent of server lag, that is |
| 13:08 |
ireallyhateirc |
The client has usually 60 FPS then the boat lags and it causes a dip to 30-40 just for a fraction of a second |
| 13:12 |
ireallyhateirc |
and the dips usually happen when there's a spike in packets received by the client |
| 13:13 |
MTDiscord |
<luatic> so it is a client lag problem |
| 13:14 |
ireallyhateirc |
I'm installing the lagometer |
| 13:14 |
MTDiscord |
<luatic> note that the lagometer is for server lag |
| 13:16 |
ireallyhateirc |
if there's no server lag but the boat still lags then it should confirm it's a client-side problem |
| 13:29 |
ireallyhateirc |
idk if I'm reading this properly but: SSD - max lag 0.216, HDD - max lag 0.613 |
| 13:30 |
ireallyhateirc |
(from the lagometer) |
| 13:30 |
ireallyhateirc |
I'd like to note that Luanti is installed on the SSD and I only move the world to the HDD |
| 13:44 |
MTDiscord |
<luatic> what are the non-max-lag statistics (that's kind of the point of the lagometer, the engine already reports max lag) |
| 13:45 |
ireallyhateirc |
I can send a screenshot |
| 13:46 |
ireallyhateirc |
SSD: https://litter.catbox.moe/qp62du.png |
| 13:46 |
ireallyhateirc |
HDD: https://litter.catbox.moe/ulgy9z.png |
| 13:47 |
ireallyhateirc |
idk how to read all that |
| 13:49 |
ireallyhateirc |
Oh it's explained in the documentation |
| 13:51 |
ireallyhateirc |
mean time between steps appears to be higher on HDD: 0.029 (SSD) 0.059 (HDD) |
| 14:00 |
ireallyhateirc |
I love spending sundays on walking around in my legboat |
| 14:14 |
Krock |
ireallyhateirc: your boat is kinda cursed |
| 14:14 |
|
sys4 joined #luanti |
| 14:14 |
ireallyhateirc |
Krock, that's the point of the legboat |
| 14:14 |
[MatrxMT] |
<Blockhead256> that boat got bitten by a radioactive spider |
| 14:21 |
|
jaca122 joined #luanti |
| 15:07 |
MinetestBot |
[git] rubenwardy -> luanti-org/luanti: ContentDB: Add reviews tab (#15254) 75862e3 https://github.com/luanti-org/luanti/commit/75862e33b6bfe6b3805310152c66af1430ac3666 (2025-04-13T15:07:01Z) |
| 15:16 |
ireallyhateirc |
I'll come back with the profiler's output later. My computer is not a potato and I've played more fancy games on it without such lags with HDDs so I believe there's something indeed lurking in the engine. |
| 15:21 |
|
Glaedr_ joined #luanti |
| 15:26 |
ireallyhateirc |
Should I start an issue for the unconfirmed bug? |
| 15:32 |
|
bodiccea joined #luanti |
| 15:54 |
|
SFENCE joined #luanti |
| 16:07 |
ireallyhateirc |
I know it's like comparing apples to oranges but in Supertuxkart I don't get any lags like that. If there's a Luanti game with moving entities that don't lag then send me a link and I'll test it |
| 16:12 |
sfan5 |
the sleep times in the graphs are suspicious, or did you put the game into the background before making the screenshot? |
| 16:13 |
ireallyhateirc |
sfan5, I used the system's screenshot thing instead of luanti's one |
| 16:13 |
ireallyhateirc |
I can redo it with luanti's screenshot thing |
| 16:13 |
sfan5 |
is that a yes or no |
| 16:14 |
ireallyhateirc |
I'm not sure. When I press the "print screen" a window pops up and defocuses the window |
| 16:14 |
ireallyhateirc |
so probably yes? |
| 16:15 |
ireallyhateirc |
Luanti switches to idle mode when that happens so yes, I'll redo the test |
| 16:17 |
sfan5 |
then that's unrelated |
| 16:17 |
sfan5 |
either way I can hop into a MTG boat and drive to newly generated parts without a single hitch |
| 16:19 |
ireallyhateirc |
I'll test the boat without legs first then |
| 16:28 |
|
SwissalpS joined #luanti |
| 16:29 |
ireallyhateirc |
Ok so with a regular boat on a HDD the lags are less noticeable, but I still do get them. Max lag is at ~0.500 and FPS are dropping from time to time to 30-50 |
| 16:30 |
sfan5 |
aside from "lag" being too vague you're mixing things up again: max_lag is server-side, fps is client-side |
| 16:33 |
ireallyhateirc |
I know, I'm just giving the numbers I can see. On the SSD: max lag ~0.100, FPS sometimes drop to 50 |
| 16:38 |
ireallyhateirc |
so whatever it is, it appears to be affected by the storage device's parameters the world files are on |
| 16:38 |
ireallyhateirc |
and I'm not mining crypto or anything, Luanti is the sole user of the HDD at a given moment |
| 16:39 |
ireallyhateirc |
I can provide better numbers and a video of both cases, should I start an issue for this? |
| 16:44 |
sfan5 |
a video is good in any case so people can precisely tell what the lag looks like |
| 16:46 |
|
SFENCE joined #luanti |
| 16:57 |
ireallyhateirc |
Here's one: https://files.catbox.moe/cpiap4.mp4 |
| 16:57 |
ireallyhateirc |
the first lag is around 00:18 |
| 16:58 |
ireallyhateirc |
it looks as if the boat was moved back and forth |
| 17:00 |
ireallyhateirc |
Then another saucy one is around 00:43 |
| 17:02 |
cheapie |
Max lag of 0.4 seems kinda high if you're expecting entities to be smooth |
| 17:03 |
cheapie |
Does the server profiler point at some specific mod as being a problem? |
| 17:08 |
whosit |
if you're using sqlite backends, try setting sqlite_synchronous = 0 |
| 17:08 |
ireallyhateirc |
cheapie, that's vanilla MTG |
| 17:09 |
cheapie |
Being vanilla MTG doesn't exclude the possibility of a bug in MTG |
| 17:10 |
ireallyhateirc |
whosit, where do I put that setting? |
| 17:11 |
whosit |
in minetest.conf |
| 17:11 |
whosit |
or, luanti.conf, idk which one you're using :p |
| 17:12 |
sfan5 |
the boat jumping back to a past position for exactly one frame is unlike anything a server lag should cause |
| 17:13 |
whosit |
can it be caused by packets arriving out of order somehow? or those are reliable and ordered? |
| 17:13 |
whosit |
(or being processed out of order? :p) |
| 17:14 |
sfan5 |
it can |
| 17:17 |
ireallyhateirc |
whosit, appears to be way smoother |
| 17:18 |
whosit |
if it does, you can also try this: https://phiresky.github.io/blog/2020/sqlite-performance-tuning/ |
| 17:18 |
ireallyhateirc |
which is weird because max lag is the same and the client's FPS chart also shows FPS dips to 50 or 40 |
| 17:19 |
whosit |
saw someone link it, somewhere, but I think main issue is the disk sync |
| 17:19 |
ireallyhateirc |
or not? This appears to happen when new chunks are generated |
| 17:20 |
|
jemius joined #luanti |
| 17:20 |
ireallyhateirc |
nah, same story - new chunks new lags |
| 17:20 |
ireallyhateirc |
Hope that the video at least shows that I did not make this up lol |
| 17:21 |
sfan5 |
is this singleplayer or networked? |
| 17:22 |
ireallyhateirc |
singleplayer |
| 17:28 |
ireallyhateirc |
I'll be back later and I'll create an issue for this then |
| 17:28 |
|
sinvet joined #luanti |
| 17:29 |
ireallyhateirc |
the HDD is "2TB WDC WD20EZRZ-22Z" and SSD is "1024GB SSDPR-CX400-01T-" btw if you're curious about the specs |
| 17:35 |
cheapie |
Searching for "SSDPR-CX400-01T-" just gets me those SSDs people buy off Aliexpress for like $20, the ones that usually work but manage to be slower than a hard drive |
| 17:43 |
|
fluxionary joined #luanti |
| 18:04 |
|
fluxionary joined #luanti |
| 18:21 |
ireallyhateirc |
it wasn't that cheap lol |
| 18:21 |
ireallyhateirc |
https://www.goodram.com/produkty/ssd-cx400-sata-iii-25/ |
| 18:21 |
ireallyhateirc |
It was 50$ or so? idk |
| 18:38 |
|
Verticen joined #luanti |
| 18:49 |
ireallyhateirc |
chasing elusive bugs is my passion |
| 19:04 |
|
exoticalexo joined #luanti |
| 19:15 |
|
___nick___ joined #luanti |
| 19:49 |
|
___nick___ joined #luanti |
| 19:52 |
|
___nick___ joined #luanti |
| 20:01 |
|
turtleman joined #luanti |
| 20:43 |
|
YuGiOhJCJ joined #luanti |
| 20:50 |
ireallyhateirc |
Could be better: https://github.com/luanti-org/luanti/issues/16019 |
| 20:50 |
ireallyhateirc |
I'll provide profiler output later once I know how to do that |
| 21:02 |
tsadok |
I have been avoiding SSDs due to some really bad experiences early on, that are *probably* out of date at this point because new tech gets better over time, but eh, spinning-platter drives have become *so* much more reliable over the last three decades, that I no longer have really any interest in trusting my data to a different storage technology. |
| 21:04 |
ireallyhateirc |
that particular brand/model I use is decent and hasn't caused me any problems so far |
| 21:05 |
ireallyhateirc |
I have some ugly experience with NVMe SSDs. One became read-only after my computer experienced unexpected shutdown |
| 21:12 |
cheapie |
Random nonsense with replacing nodes with others that shouldn't be there: https://cheapiesystems.com/media/2025-04-13%2016-07-32.webm |
| 21:14 |
ireallyhateirc |
the ultimate elevator |
| 21:22 |
|
bdju joined #luanti |
| 21:22 |
MTDiscord |
<luatic> the ultimate elevator is a 2 wide shaft that goes straight down, with one column for water, the other for air, and some air pockets every 20 nodes or so. then you can use a boat to go up fast, and drop down and use the water to slow down even faster. and it's very cheap to build. |
| 21:23 |
cheapie |
luatic: Yeah, but can that do this? https://cheapiesystems.com/media/2025-03-21%2003-23-07.webm |
| 21:24 |
MTDiscord |
<luatic> i must concede |
| 21:25 |
cheapie |
Or have keyswitches that just confuse everyone since hardly anyone knows what they do IRL either? :P https://cheapiesystems.com/media/images/screenshot_20250404_225700.png |
| 21:32 |
ireallyhateirc |
can someone send me an example of a sane flamegraph from the jit profiler? |
| 21:32 |
ireallyhateirc |
I get weird and broken results |
| 22:10 |
|
Kimapr joined #luanti |
| 22:26 |
|
liceDibrarian joined #luanti |
| 22:32 |
|
panwolfram joined #luanti |
