Posts: 372
Threads: 29
Joined: Mar 2011
Thanks: 1
Given 21 thank(s) in 18 post(s)
If you want me to run certain builds for 4mil or even higher just upload the build and tell me to run it on the server :p.
the server does not have the tools to compile.
I just use it to test then :p.
Can always make a survival server out of it later, havent even setup the proper subdomain yet so.
Posts: 4,634
Threads: 115
Joined: Dec 2011
Thanks: 695
Given 495 thank(s) in 424 post(s)
I wonder why the speed suddenly drops like that. Though it wouldn't happen with allot of people, since this was a server generating non-stop for 22 hours, it could be possible that it would happen when a large server is online for a long time. Of course it's also possible that this can only happen when a server is preparing the spawn.
Posts: 6,485
Threads: 176
Joined: Jan 2012
Thanks: 131
Given 1075 thank(s) in 852 post(s)
I believe it's mostly due to the caches.
Take the biome cache, for example. By default it stores up to 4 * 64 chunks worth of biomes. When the spawn generator goes from left to right for the first time, it primes the cache. When it goes back from right to left, it re-uses most of the values in the cache, adding only those that are one chunk below. However, that only works as long as the values from the first pass stay in the cache. When the left-to-right pass is too long, the values from the start of the run get pushed off out of the cache before the left-to-right pass is completed, so the following right-to-left pass has to generate them again. In my opinion that's what happens at the first drop in the graph, at around 40 % of the X axis.
First the biome cache gets depleted like this, because it's the most heavily used one. Then the composition cache gets depleted, and that's when the line starts going all wiggly.
Posts: 6,485
Threads: 176
Joined: Jan 2012
Thanks: 131
Given 1075 thank(s) in 852 post(s)
Hmm, wait, no, that's not right. The spawn preparation runs the entire length of the spawn area from the beginning. I had thought it was making con-centric squares. The cache theory is all wrong now. And I have no idea as for the slowdown.
Posts: 513
Threads: 10
Joined: May 2014
Thanks: 138
Given 89 thank(s) in 75 post(s)
10-06-2015, 09:36 PM
(This post was last modified: 10-06-2015, 09:50 PM by Seadragon91.)
I created a world with PregenerateDistance=200.
Result 3 GB RAM and on generation 160 - 180 chunks/second.
Tested this on another PC.
Result 1,5 GB RAM and on generation 60 - 65 chunks/second.
Anyone has a idea why the RAM here has only 50 % less then the another one? It's the same seed.
Posts: 6,485
Threads: 176
Joined: Jan 2012
Thanks: 131
Given 1075 thank(s) in 852 post(s)
The spawn chunks aren't kept in memory, they are unloaded after a while.
Posts: 372
Threads: 29
Joined: Mar 2011
Thanks: 1
Given 21 thank(s) in 18 post(s)
Xoft want teamviewer access to the system ?
As long as you dont close random windows that are open in the background its fine, no private data on there at all so.
Posts: 372
Threads: 29
Joined: Mar 2011
Thanks: 1
Given 21 thank(s) in 18 post(s)
Yeah i would guess thats true, and it doesn't really bother me even if i had to run it for a week to generate it :p.
Its online 24/7 so meh.
I was more like why and is this known :p.