| Time |
Nick |
Message |
| 01:52 |
MTDiscord |
<exe_virus> Okay, after some decent testing, it's most definitely the ABM code base. Especially mesecons, there's just too many (21 relatively often ones for my game pack) for this old computer to handle haha. More incentive for me to make a more efficient ABM implementation. |
| 03:25 |
|
olliy joined #minetest-dev |
| 03:30 |
|
cheapie joined #minetest-dev |
| 03:53 |
|
v-rob joined #minetest-dev |
| 08:00 |
|
specing_ joined #minetest-dev |
| 08:38 |
|
Alias joined #minetest-dev |
| 08:48 |
|
tech_exorcist joined #minetest-dev |
| 09:27 |
|
Fixer joined #minetest-dev |
| 09:31 |
|
calcul0n_ joined #minetest-dev |
| 10:25 |
|
entuland joined #minetest-dev |
| 10:37 |
|
appguru joined #minetest-dev |
| 11:51 |
|
olliy joined #minetest-dev |
| 12:02 |
|
tech_exorcist joined #minetest-dev |
| 15:33 |
|
specing joined #minetest-dev |
| 15:41 |
|
Extex joined #minetest-dev |
| 17:42 |
nrz |
yep mesecons uses massively ABMs |
| 18:03 |
|
Extex joined #minetest-dev |
| 18:36 |
MTDiscord |
<exe_virus> Why not node timers? |
| 18:46 |
|
v-rob joined #minetest-dev |
| 19:26 |
|
Lone_Wolf joined #minetest-dev |
| 20:00 |
|
specing_ joined #minetest-dev |
| 20:11 |
nrz |
it cost more |
| 20:11 |
nrz |
ABM is the fastest for wide node timer checks |
| 20:12 |
nrz |
node timer cost many memory also, if i remember it's instancited by node, and abm is global and just scan loaded nodes |
| 20:12 |
specing |
why does it need ABMs? Blinkys? node detectors? |
| 20:14 |
|
v-rob joined #minetest-dev |
| 20:14 |
|
entuland_ joined #minetest-dev |
| 20:24 |
sfan5 |
register_abm is used from mesecons_detector, mesecons_hydroturbine, mesecons_solarpanel and mesecons_torch |
| 20:29 |
specing |
I think MT would get a serious speed-up if one could register neighbour-node-change callbacks |
| 20:31 |
v-rob |
It would make custom drawtypes pretty easy, I think, especially with a #9193 for models. |
| 20:31 |
ShadowBot |
https://github.com/minetest/minetest/issues/9193 -- Add paramtype2 for node texture variants |
| 20:36 |
ShadowNinja |
Pushed my server list changes: sl#45 and #11342 |
| 20:36 |
ShadowBot |
https://github.com/minetest/minetest/issues/11342 -- Serverlist enhancements by ShadowNinja |
| 20:36 |
ShadowBot |
https://github.com/minetest/serverlist/issues/45 -- Persist servers in separate database by ShadowNinja |
| 20:37 |
|
Extex joined #minetest-dev |
| 20:38 |
sfan5 |
will look soon |
| 20:49 |
MTDiscord |
<exe_virus> Those callbacks would be expensive too. Each node change would result in 26 callbacks |
| 20:50 |
v-rob |
That assumes that every neighboring node would have a callback. |
| 20:50 |
v-rob |
However, I think a bigger problem would be infinite loops where two nodes update forever |
| 20:53 |
MTDiscord |
<exe_virus> Oof, that too. Even if it doesn't have a callback, you will have to check what those 26 nodes are and see if any of those type have callbacks |
| 20:54 |
MTDiscord |
<exe_virus> Before not doing the callback, similar to how ABMs have a non zero cost for doing nothing |
| 20:58 |
sfan5 |
you'd cache presence of a callback on the C++ side obviously |
| 20:59 |
specing |
Why 26 callbacks? |
| 20:59 |
v-rob |
There are 26 nodes around each node |
| 20:59 |
MTDiscord |
<IhrFussel> haven't ABMs already been optimized like 2-3 years ago? IIRC they also got some kind of cache now |
| 20:59 |
specing |
There'd be different callbacks, one for face adjecency and another for vertex adjecency |
| 20:59 |
specing |
and also, nodes aren't dug/placed as fast as ABMs run |
| 21:00 |
specing |
so it'd be a net reduction of work |
| 21:00 |
v-rob |
Face adjacency would be cheaper at 6, while vertex would still be 26. Edge would be 18. |
| 21:04 |
|
entuland_ joined #minetest-dev |
| 21:05 |
|
Fixer_ joined #minetest-dev |
| 21:05 |
|
v_rob joined #minetest-dev |
| 21:06 |
|
freshreplicant[m joined #minetest-dev |
| 21:06 |
|
appguru1 joined #minetest-dev |
| 21:07 |
|
specing joined #minetest-dev |
| 21:08 |
|
v-rob joined #minetest-dev |
| 21:08 |
|
BuckarooBanzai2 joined #minetest-dev |
| 21:09 |
|
Fixer joined #minetest-dev |
| 21:11 |
|
entuland__ joined #minetest-dev |
| 21:12 |
|
wsor4035 joined #minetest-dev |
| 21:14 |
v-rob |
oh, sfan5: could I get that auto-voice thing back? (I don't know what it does, but I suppose I'm supposed to have it) |
| 21:14 |
|
celeron55_ joined #minetest-dev |
| 21:14 |
sfan5 |
sure |
| 21:14 |
|
olliy joined #minetest-dev |
| 21:14 |
sfan5 |
you'll need to register an account first |
| 21:15 |
|
v_rob joined #minetest-dev |
| 21:15 |
|
Fixer_ joined #minetest-dev |
| 21:15 |
v_rob |
Dang, why do I keep disconnecting and reconnecting with a different name? |
| 21:15 |
|
nore joined #minetest-dev |
| 21:20 |
|
celeron55 joined #minetest-dev |
| 21:22 |
|
specing_ joined #minetest-dev |
| 21:23 |
|
v_rob joined #minetest-dev |
| 21:24 |
|
v_rob joined #minetest-dev |
| 21:25 |
|
v-rob joined #minetest-dev |
| 21:26 |
|
olliy joined #minetest-dev |
| 21:28 |
v-rob |
OK, done. I don't understand how I got auto-voice on freenode then since I didn't have an account there |
| 21:29 |
sfan5 |
there |
| 21:29 |
v-rob |
thanks |
| 21:30 |
|
olliy joined #minetest-dev |
| 21:36 |
sfan5 |
oh and apart from voicing you I think it allows you to edit the topic |
| 22:17 |
|
Extex joined #minetest-dev |
| 22:18 |
|
v-rob joined #minetest-dev |
| 23:20 |
|
queria joined #minetest-dev |
| 23:20 |
|
three joined #minetest-dev |
