Ultimate space simulation software

 
User avatar
Mouthwash
Explorer
Explorer
Topic Author
Posts: 163
Joined: 22 May 2017

Scanning every star/planet in the SE universe?

23 Nov 2017 13:21

The star browser doesn't seem capable of much, with a search for all stars within a mere 100 light-years taking twenty seconds on my excellent laptop. But would it be possible for a supercomputer to scan every single celestial body in Space Engine within a reasonable time frame? It would be great fun, perhaps useful for finding bugs, and most importantly would let us find the One True Largest Star (before size was capped).
 
User avatar
SpaceEngineer
Author of SpaceEngine
Author of SpaceEngine
Posts: 677
Joined: 17 May 2016
Location: Saint-Petersburg
Contact:

Scanning every star/planet in the SE universe?

23 Nov 2017 13:42

This can be easily calculated. Take your scan time, divide on number of cores and Ghz, and you'll see the scanning performance. Then multiply on supercomputer specs... I had answered this several times on this forum and on the old one. Scanning the single galaxy would take millions of years even for supercomputer.
 
User avatar
Mouthwash
Explorer
Explorer
Topic Author
Posts: 163
Joined: 22 May 2017

Scanning every star/planet in the SE universe?

23 Nov 2017 14:17

SpaceEngineer wrote:
This can be easily calculated. Take your scan time, divide on number of cores and Ghz, and you'll see the scanning performance. Then multiply on supercomputer specs... I had answered this several times on this forum and on the old one. Scanning the single galaxy would take millions of years even for supercomputer.

That's ridiculous... the Milky Way is only about thirty million times the scale of the star browser search, so it would take my laptop fifteen years. Against a supercomputer with millions of cores?
 
User avatar
XBrain130
Explorer
Explorer
Posts: 284
Joined: 26 Nov 2016
Location: Italy
Contact:

Scanning every star/planet in the SE universe?

23 Nov 2017 15:39

Mouthwash wrote:
SpaceEngineer wrote:
This can be easily calculated. Take your scan time, divide on number of cores and Ghz, and you'll see the scanning performance. Then multiply on supercomputer specs... I had answered this several times on this forum and on the old one. Scanning the single galaxy would take millions of years even for supercomputer.

That's ridiculous... the Milky Way is only about thirty million times the scale of the star browser search, so it would take my laptop fifteen years. Against a supercomputer with millions of cores?

Keep in mind that whatever size you specify, SE currently won't scan more than 10.000 stars in one go, as you can note by keeping an eye on the counter. Milky Way is listed as having 270 billion ones as of the last version given to us beta testers.
SpaceEngine's Italian Discord server: https://discord.gg/NhQbEbC
 
User avatar
Mouthwash
Explorer
Explorer
Topic Author
Posts: 163
Joined: 22 May 2017

Scanning every star/planet in the SE universe?

24 Nov 2017 00:13

XBrain130 wrote:
Mouthwash wrote:
SpaceEngineer wrote:
This can be easily calculated. Take your scan time, divide on number of cores and Ghz, and you'll see the scanning performance. Then multiply on supercomputer specs... I had answered this several times on this forum and on the old one. Scanning the single galaxy would take millions of years even for supercomputer.

That's ridiculous... the Milky Way is only about thirty million times the scale of the star browser search, so it would take my laptop fifteen years. Against a supercomputer with millions of cores?

Keep in mind that whatever size you specify, SE currently won't scan more than 10.000 stars in one go, as you can note by keeping an eye on the counter.

That's just an ingame limitation. I imagine a serious attempt at scanning everything would proceed differently.
 
User avatar
XBrain130
Explorer
Explorer
Posts: 284
Joined: 26 Nov 2016
Location: Italy
Contact:

Scanning every star/planet in the SE universe?

24 Nov 2017 04:29

Mouthwash wrote:
Source of the post I imagine a serious attempt at scanning everything would proceed differently.

That's right. Your PC would overload and crash, if not outright melt :P
Come on dude, stop complaining about everything. If there is a mistake or limitation, it has a reason to be there.
SpaceEngine's Italian Discord server: https://discord.gg/NhQbEbC
 
User avatar
Mouthwash
Explorer
Explorer
Topic Author
Posts: 163
Joined: 22 May 2017

Scanning every star/planet in the SE universe?

24 Nov 2017 06:09

XBrain130 wrote:
Mouthwash wrote:
Source of the post I imagine a serious attempt at scanning everything would proceed differently.

That's right. Your PC would overload and crash, if not outright melt :P
Come on dude, stop complaining about everything. If there is a mistake or limitation, it has a reason to be there.

Who's complaining? Going over 10k wouldn't be much use for most people.
 
User avatar
jadestar
Observer
Observer
Posts: 19
Joined: 01 Dec 2016
Location: Gig Harbor, WA - USA

Scanning every star/planet in the SE universe?

13 Jan 2018 08:42

Mouthwash wrote:
XBrain130 wrote:
Mouthwash wrote:
Source of the post I imagine a serious attempt at scanning everything would proceed differently.

That's right. Your PC would overload and crash, if not outright melt :P
Come on dude, stop complaining about everything. If there is a mistake or limitation, it has a reason to be there.

Who's complaining? Going over 10k wouldn't be much use for most people.

For most people no but I'd love to have the option to go over that amount. It doesn't even have to be something in the GUI but perhaps something we could specify in a .cfg? I wonder if Space Engineer would consider this at some point? Computers are always getting more powerful so it would be a "nice to have" option for some of us who have Space Engine running on large multicore servers.
My specs: Gigabyte X99-UD4 - CPU E5-2658V3, RAM 32gb, GPU AMD RX480 8GB VRAM
 
User avatar
SpaceEngineer
Author of SpaceEngine
Author of SpaceEngine
Posts: 677
Joined: 17 May 2016
Location: Saint-Petersburg
Contact:

Scanning every star/planet in the SE universe?

13 Jan 2018 10:55

Actually it is a GUI limitation. SE generates a huge table, 10k lines is too large already.

Another limitation is how stars are get "scanned". Engine have to generate ALL stars within a specified radius. It is not a problem is radius is <50 ly, there are ~10k stars out there. But if you increase it twice, there will be 8x more stars. Increase it 4 times, and you will have 64 more stars (6.4 million). And so on. Star generator is not multi-threaded (unlike plantary system generation), generating all octree blocks containing those 6.4 million stars would take several minutes, during which SE will not respond. Users will complain that SE is hunging during star search. This is the main reason why I introduced the 10k stars limit. After star list is filled, SE starts to generate the planetary system for each one, and compare it with the planet/moon filter. This stage is multi-threaded, so engine is not hungs you can see a gradual update of the table (untick the "Filter" box in the lower left corner).

Other way to limit could be a search radius, but it may be different in different parts of the galaxy, depending on a local star density. And there is a filter by main system's star class, SE can filter out systems by this parameter prior to generation of planets. Use this if you want to search the Earth twins near G stars (I suppose in 90% cases users doing this in the star browser). This way you may increase the search radius and still be within 10k limit. But be patient, you computer may hung during generation of stars for several minutes.
 
User avatar
tornadotodd
Space Pilot
Space Pilot
Posts: 119
Joined: 20 Dec 2017
Location: Oklahoma

Scanning every star/planet in the SE universe?

14 Jan 2018 02:24

SpaceEngineer wrote:
This can be easily calculated. Take your scan time, divide on number of cores and Ghz, and you'll see the scanning performance. Then multiply on supercomputer specs... I had answered this several times on this forum and on the old one. Scanning the single galaxy would take millions of years even for supercomputer.

That boggles the mind wow.
 
User avatar
jadestar
Observer
Observer
Posts: 19
Joined: 01 Dec 2016
Location: Gig Harbor, WA - USA

Scanning every star/planet in the SE universe?

14 Jan 2018 06:35

SpaceEngineer wrote:
Actually it is a GUI limitation. SE generates a huge table, 10k lines is too large already.

Another limitation is how stars are get "scanned". Engine have to generate ALL stars within a specified radius. It is not a problem is radius is <50 ly, there are ~10k stars out there. But if you increase it twice, there will be 8x more stars. Increase it 4 times, and you will have 64 more stars (6.4 million). And so on. Star generator is not multi-threaded (unlike plantary system generation), generating all octree blocks containing those 6.4 million stars would take several minutes, during which SE will not respond. Users will complain that SE is hunging during star search. This is the main reason why I introduced the 10k stars limit. After star list is filled, SE starts to generate the planetary system for each one, and compare it with the planet/moon filter. This stage is multi-threaded, so engine is not hungs you can see a gradual update of the table (untick the "Filter" box in the lower left corner).

Other way to limit could be a search radius, but it may be different in different parts of the galaxy, depending on a local star density. And there is a filter by main system's star class, SE can filter out systems by this parameter prior to generation of planets. Use this if you want to search the Earth twins near G stars (I suppose in 90% cases users doing this in the star browser). This way you may increase the search radius and still be within 10k limit. But be patient, you computer may hung during generation of stars for several minutes.

I just wanted to thank you SO much for this highly detailed explanation and possible workarounds. It means a lot that you took the time to post this. Onward and upward!
My specs: Gigabyte X99-UD4 - CPU E5-2658V3, RAM 32gb, GPU AMD RX480 8GB VRAM
 
A-L-E-X
World Builder
World Builder
Posts: 838
Joined: 06 Mar 2017

Scanning every star/planet in the SE universe?

26 Jan 2018 15:09

jadestar wrote:
Mouthwash wrote:
XBrain130 wrote:
That's right. Your PC would overload and crash, if not outright melt :P
Come on dude, stop complaining about everything. If there is a mistake or limitation, it has a reason to be there.

Who's complaining? Going over 10k wouldn't be much use for most people.

For most people no but I'd love to have the option to go over that amount. It doesn't even have to be something in the GUI but perhaps something we could specify in a .cfg? I wonder if Space Engineer would consider this at some point? Computers are always getting more powerful so it would be a "nice to have" option for some of us who have Space Engine running on large multicore servers.

You own a server- wow.  I just did a new build with new components (the reason I was absent for so long), but I thought shelling out 2K for all the components was a splurge on my part, but didn't even dream of buying one of those servers.
 
vlad01
Astronaut
Astronaut
Posts: 54
Joined: 02 May 2018

Scanning every star/planet in the SE universe?

26 May 2018 06:17

Is there a reason why the star browser only seems to find results on the surface of the sphere its searching and rarely anything in the volume withing the search radius?

For example. If I search for a certain type of object within 25 Ly I get a few certain results and nearly always they will be 20-24.xx Ly away and nothing closer. If I expand the same search to 50 Ly while still within the 10k limit the results change to the 40-50 Ly range but will no longer have the previous results of the 25 Ly radius.

So my work around atm is search 10, then 30, 50, 75, 100, 150 etc... and work through the results that way one by one.  Is this a limitation of SE? or mine is doing something weird? 
 
User avatar
FastFourierTransform
Explorer
Explorer
Posts: 208
Joined: 17 Nov 2016

Scanning every star/planet in the SE universe?

26 May 2018 09:10

vlad01 wrote:
Source of the post If I expand the same search to 50 Ly while still within the 10k limit the results change to the 40-50 Ly range but will no longer have the previous results of the 25 Ly radius.

I don't know about this (I can't check it now either), but I think this:
vlad01 wrote:
Source of the post  If I search for a certain type of object within 25 Ly I get a few certain results and nearly always they will be 20-24.xx Ly away and nothing closer.

Could be explained in part by an observational bias and not any real bug or problem in SE. You see, when you search for stars in a radius of 25 ly what you are asking is for stars in a volume of 65.450 ly3 and when you search for stars in a radius of 20 ly what you are asking is stars in a volume of 33.510 ly3. This means that the spherical shell where stars between 20 ly and 25 ly distance fall has a volume of 31.940 ly3 (65.450 ly3- 33.510 ly3).
Look closely at that. If stars are homogeneusly spaced in the volume of space (this is more or less true both in SE and real life), for example 1 star each cubic light year (much more crowded than the solar neighbourhood but just to pick a value), then you would expect to have 65.450 stars inside the sphere of radius 25 ly and of those you would expect 31.940 to be in the carcass between 20 ly and 25 ly. Expect then that a 25 ly radius search yields half of the stars (48,8%) in the last 5 ly of your sphere.
This effect is produced by the mathematical concept of volume been proportional to the cube of the radius, or to the common sense principle that there are more objects farther away from you than closer to you. The fact that SE shows this is another demonstration of its hability to show reality with respectable accuracy. So bare this in mind. Maybe there's a bug at the same time but this effect could make you think that (that's all what I'm saying here).
The high proportion of stars in a layer between r and R with respect to the overall stellar content of a sphere with radius R can be calculated as 1-(r/R)^3. For r=20 ly and R=25 ly is 48.8% and for the r=40 ly and R=50 ly is exactly the same value.
The part I can't explain is the absence of the stars in the 25 ly radius in the 50 ly search. Are you sure about this?
 
vlad01
Astronaut
Astronaut
Posts: 54
Joined: 02 May 2018

Scanning every star/planet in the SE universe?

26 May 2018 23:53

No, I have tested this like 100 times over. What I mean is a larger radius search from the same point always fails to find previous results from a smaller radius in the previous otherwise exact same search. So If I just search a larger radius hoping to find all results I actually miss out on just about everything in that volume. The results only seem to be around the surface of the sphere give or take several light years.

See this example.

Exact same search repeated with only the radius slowly increased.

Note the distances of the matches and failure to find previously found matches of the next smaller radius.  This happens more than and also less than 10k limit, 10k limit has 0 influence on this effect.

On some rare occasions it will find previous results but normally no more than 1 as it has in this example.

scr00043.jpg

scr00044.jpg

scr00045.jpg

scr00046.jpg

Who is online

Users browsing this forum: No registered users and 1 guest