Jump to content

[1.8.x-1.12.x]R-T-B's Kopernicus Unified "Bleeding Edge" Branch[REACTIVATED]


R-T-B

Recommended Posts

9 hours ago, Murdabenne said:

Advice to the wise: Do not mock the evil frog god. He will eat you. :ph34r:

Flood's got nothing on the Frog God.  He will consume all.

As for that "next release" I promised like, yesterday or something....  It's been delayed until I can work out one more bug.  Won't be long.

Edited by R-T-B
Link to comment
Share on other sites

40 minutes ago, Not a Spy said:

i have a small question

so i try to load up the game and it just says that the mod is not compatiable with 1.11?

Kopernicus is version locked - it's coded not to run on KSP versions for which it wasn't built.  You can either wait for it to be updated or consider these two other possibilities

 

I know the Bleeding Edge fork has a 1.11 build which you might have to install manually.  I don't know about Kopernicus Continued.

Edited by Brigadier
missing word
Link to comment
Share on other sites

1 hour ago, Brigadier said:

Kopernicus is version locked - it's coded not to run on KSP versions for which it wasn't built.  You can either wait for it to be updated or consider these two other possibilities

 

I know the Bleeding Edge fork has a 1.11 build which you might have to install manually.  I don't know about Kopernicus Continued.

ok im prettey sure this is the bleeding edge fork...

Link to comment
Share on other sites

7 hours ago, Brigadier said:

You're correct, my mistake.  It's been a long day.

Have you downloaded the correct KBE version for your KSP version?  Which Release are you using?

This latest release has changed the file naming scheme to make it less confusing (now if only github could put 1.11.0 on top, it doesn't for some reason).  Also this release has some bugfixes, of course.

Here it is:

@R-T-B R-T-B released this 8 minutes ago

This is R-T-B's "Bleeding Edge" branch of Kopernicus, intended to support the latest features, KSP editions, and also the latest bugs. Please keep in mind this branch may be more buggy than Prestja's mainline Kopernicus branch, but it also supports more KSP versions and has more features implemented for testing reasons. Many features that make it into mainline Kopernicus are born, tested, and trialed by fire here.

This is release 59. It contains the following changes:

1.) The Scatter distance culler has been broken for a few releases (oof). It has been fixed, reenabled, and the cull distance reduced to 6250m (adjustable in config) to hopefully massively improve near-terrain flying performance with land scatters on.

2.) Some hazardous body tweaks, it still may be buggy. Testing needed.

3.) File download names have been modified to be better understandable to end users.

4.) Modular Flight Integrator has been unforked until we can find a better, more reliable way to work around the "distant sinking" issue. The fork we had is not the way to do it.

Known Bugs:

1.) Distant bodies (farther out than Eeloo) sometimes exhibit issues with sinking landing gear.  We are tracking this but don't have a fix just yet.

2.)  At interstellar ranges, heat can sometimes behave strangely.  It is best to turn off part heating when traveling far far away.

3.)  The ring shader is broken in OpenGL environments.  This is a hopefully temporary setback.  OpenGL means anything non windows, or Windows in forced OpenGL mode (if you don't know how to do this, you aren't).

Please download the right output zip for your version. "1.9.1" zips are for 1.9.1, "1.10.1" for 1.10.1, "1.11.0" for 1.11.0, etc.

Thanks and as always, report bugs!

-RTB

Edited by R-T-B
Link to comment
Share on other sites

40 minutes ago, R-T-B said:

This latest release has changed the file naming scheme to make it less confusing (now if only github could put 1.11.0 on top, it doesn't for some reason).  Also this release has some bugfixes, of course.

It doesn't see 10 and 11, it sees 1, then 0 or 1, then 1. It's ordering them character by character. 1.9.1 is below those two because 9 comes after 1.

Also, with my install I'm not sure I'd know if a bug was Kopernicus acting up or something else. I don't imagine that'd be too useful.

Link to comment
Share on other sites

29 minutes ago, RyanRising said:

It doesn't see 10 and 11, it sees 1, then 0 or 1, then 1. It's ordering them character by character. 1.9.1 is below those two because 9 comes after 1.

That makes sense.  Still, the new naming should be easy to understand.

  

29 minutes ago, RyanRising said:

Also, with my install I'm not sure I'd know if a bug was Kopernicus acting up or something else. I don't imagine that'd be too useful.


I treat it like the pigs in Animal Farm.  All reports are useful, but some reports are more useful than others.

In short, if you have a bug, send a report and log file and I'll never yell at you for it, but depending on mod load it may or may not be diagnosable.

Edited by R-T-B
Link to comment
Share on other sites

2 hours ago, R-T-B said:

4.) Modular Flight Integrator has been unforked until we can find a better, more reliable way to work around the "distant sinking" issue. The fork we had is not the way to do it.

If this is the case, can Kopernicus BE stop distributing it itself via CKAN and revert back to depending on Sarbian's pre-existing package?

Link to comment
Share on other sites

1 hour ago, Poodmund said:

If this is the case, can Kopernicus BE stop distributing it itself via CKAN and revert back to depending on Sarbian's pre-existing package?

It already has with all builds except version 1.11.0, which still does so simply because Sarbian has yet to approve a MFI package for 1.11.0.  I need to ping him about that, actually.

For packages that have been unforked in the CKAN tree, it should seamlessly select and install an appropriate mainline MFI package for you.

PS:  I added the current "big bugs" that are known to the release notes, which were previously falsely implying no known bugs right now.  Just want to highlight that edit.  The bugs we know of aren't major for most users, but they exist.

Edited by R-T-B
Link to comment
Share on other sites

I have done some testing with 1.11.0 and have no performance difference between 58 and 59 when it comes to terrain scatters. Keep in mind my testing is done at boomerang launch site because that is where it tanks performance the most.  As I have mentioned before terrain scatters does not have nearly as much impact on other bodies on my system. It is mainly kerbin. 

Edited by dave1904
Link to comment
Share on other sites

3 hours ago, dave1904 said:

I have done some testing with 1.11.0 and have no performance difference between 58 and 59 when it comes to terrain scatters. Keep in mind my testing is done at boomerang launch site because that is where it tanks performance the most.  As I have mentioned before terrain scatters does not have nearly as much impact on other bodies on my system. It is mainly kerbin. 

I'm actually finding similar.  It's like it's still not loading the distance cull value and defaulting to the highest possible value, which is 10,000m (nearly double what I declare as a sensible performance saving "default").

Further investigation is certainly warranted here.  I will attempt to address this today if I can find the origin.

Edited by R-T-B
Link to comment
Share on other sites

This is a novel thing:  The distance culler itself is the origin of the performance loss!

In short, the thing I wrote to save performance, is ruining performance.  It's not rate limited, which means it spends all it's time thinking about how to improve performance by removing things in the distance, until it's literally out of CPU cycles.  This obviously hurts FPS more than it helps.

I'll have a release fixing this shortly...  :o

Link to comment
Share on other sites

This is R-T-B's "Bleeding Edge" branch of Kopernicus, intended to support the latest features, KSP editions, and also the latest bugs. Please keep in mind this branch may be more buggy than Prestja's mainline Kopernicus branch, but it also supports more KSP versions and has more features implemented for testing reasons. Many features that make it into mainline Kopernicus are born, tested, and trialed by fire here.

This is release 60. It contains the following changes:

1.) The Scatter distance culler was still broken (in the sense that it hurt performance rather than helped) and has seen a hotfix to address that. It now works and offers a good performance uplift.

Known Bugs:

1.) Distant bodies (farther out than Eeloo) sometimes exhibit issues with sinking landing gear. We are tracking this but don't have a fix just yet.

2.) At interstellar ranges, heat can sometimes behave strangely. It is best to turn off part heating when traveling far far away.

3.) The ring shader is broken in OpenGL environments. This is a hopefully temporary setback.

Please download the right output zip for your version. "1.9.1" zips are for 1.9.1, "1.10.1" for 1.10.1, "1.11.0" for 1.11.0, etc.

Thanks and as always, report bugs!

-RTB

PS:

18FPS in Woomerang with 100% scatters before, 40+FPS after patch on my rig for reference.  I run a i9 9900k and RTX 2080 Super at 4k.  @dave1904

To quote HBO Chernobyl:  "Not great not terrible."  We will continue to work on this as able.

Edited by R-T-B
Link to comment
Share on other sites

55 minutes ago, R-T-B said:

18FPS in Woomerang with 100% scatters before, 40+FPS after patch on my rig for reference.  I run a i9 9900k and RTX 2080K at 4k. 

FPS are up from 15 to 35 on ultra. Wonder why a 9900k/2080 is faster at 4k than a 10700k/3080 at 3.5k. 

I then went into my heavily modded savegame and FPS on a 103 part shuttle with all the atmos mods and such are at 13. Without terrain scatters its around 25-35 but since its only on kerbin and I hardly ever launch from wommerang I will keep scatters on to keep testing. Scatters are really important for the other bodies imo. 

R8iXkBN.jpg

I wonder if more ram would help. I only have 16. 

Link to comment
Share on other sites

3 hours ago, dave1904 said:

FPS are up from 15 to 35 on ultra. Wonder why a 9900k/2080 is faster at 4k than a 10700k/3080 at 3.5k. 

It's because it's all single threaded logic on the scatter thread (part of the issues honestly).  Multithreading is needed badly here, but that's far easier said than done.  At any rate the GPU is nearly a nonfactor here and on the CPU front my 9900k is running at 5.2GHz single thread clock so it probably slightly edges out yours just by virtue of it being a watercooled hardcore OC situation.

Just so you know I'm not BSing, here's my screenshot with an FPS counter in the small of the upper left:

screenshot14.png

I will explore other optimization ideas (including multithreading) with time, of course.  This is only the beginning of an effort to fix that issue, but I have to balance other bugs too so it will only get certain slices of time unfortunately.  Today was one of them.  Hopefully we'll have another big push soon.  In the meantime using a 50% scatters setting would be the compromise I'd advise for you.

And man, I'd kill for that 3080, good job nabbing one. ;)

 

3 hours ago, dave1904 said:

I wonder if more ram would help. I only have 16. 

What's your mod list?  More ram would certainly help, but there are some other things you can try depending on what you've got going on in your modstack.  KSP loves ram.  I run with 64GBs and it can use nearly every bit of it if I configure it properly to delay garbage collection, which is probably another reason I outperform you.

The big mod to look into is heappadder.  It is usually meant for more than 16GBs of ram but it might help you as well.  It allows you to expand KSPs memory heap to prevent garbage collection from running too frequently, and hurting FPS.  Look it up.

Edited by R-T-B
Link to comment
Share on other sites

Looks like this issue will be getting more time afterall, because the issue where the scatters float away into the ether with time has made a regression and come back.

Not my day.  At least I get one more chance to optimize while fixing it.

Link to comment
Share on other sites

I made a breakthrough.

It appears in stock KSP, the scatterer spawning thread itself is rate limited.  I'm attempting to emulate this logic.  So far, I'm seeing +100FPS uplifts in scatter heavy situations, no joke.  It's quite incredible.  Expect a build with a major performance impact on scatters soon, making them behave nearly if not identical to stock in performance.

Of course bug testing will be needed, but this is very good looking performance wise.

Edited by R-T-B
Link to comment
Share on other sites

This is R-T-B's "Bleeding Edge" branch of Kopernicus, intended to support the latest features, KSP editions, and also the latest bugs. Please keep in mind this branch may be more buggy than Prestja's mainline Kopernicus branch, but it also supports more KSP versions and has more features implemented for testing reasons. Many features that make it into mainline Kopernicus are born, tested, and trialed by fire here.

This is release 61. It contains the following changes:

1.) The Scatter distance-based culler was still broken (in the sense that it both had bugs and still didn't offer much uplift) and has seen yet another hotfix to address that. It now works and offers a massive, super good performance uplift. 100FPS uplifts in scatter heavy scenarios have been realized on select hardware, no joke.

Known Bugs:

1.) Distant bodies (farther out than Eeloo) sometimes exhibit issues with sinking landing gear. We are tracking this but don't have a fix just yet.

2.) At interstellar ranges, heat can sometimes behave strangely. It is best to turn off part heating when traveling far far away.

3.) The ring shader is broken in OpenGL environments. This is a hopefully temporary setback.

Please download the right output zip for your version. "1.9.1" zips are for 1.9.1, "1.10.1" for 1.10.1, "1.11.0" for 1.11.0, etc.

Thanks and as always, report bugs!

-RTB

Edited by R-T-B
Link to comment
Share on other sites

Love the way you say it will take time and then drop the next update 10 min later. At your pace you will end up rewriting the entire code of ksp just to fix this lol. Can't wait to test when I get home. I'll look into the ram tips but ram is so expensive theses days.... 

Link to comment
Share on other sites

2 hours ago, R-T-B said:

I made a breakthrough.

It appears in stock KSP, the scatterer spawning thread itself is rate limited.  I'm attempting to emulate this logic.  So far, I'm seeing +100FPS uplifts in scatter heavy situations, no joke.  It's quite incredible.  Expect a build with a major performance impact on scatters soon, making them behave nearly if not identical to stock in performance.

 

2 hours ago, R-T-B said:

Holy cow, I'm used to SoonTM, not soon. Just installed on a JNSQ save, and yeah, seems on par with what I'd expect from stock even though the scatters have collision there. If you're not on one heck of an accomplishment high right now you should be.

Link to comment
Share on other sites

8 hours ago, RyanRising said:

 

Holy cow, I'm used to SoonTM, not soon. Just installed on a JNSQ save, and yeah, seems on par with what I'd expect from stock even though the scatters have collision there. If you're not on one heck of an accomplishment high right now you should be.

Just to be clear, collisions in JNSQ are working, right?

Yeah I had an epiphany, it wasn't expected but was most welcome.

By the way, not that many people are back on 1.9.1, but if you are the stable branch got this pushed as well.  No sense making the stable people wait for this miracle, lol.

Edited by R-T-B
Link to comment
Share on other sites

3 hours ago, R-T-B said:

Just to be clear, collisions in JNSQ are working, right?

Yeah I had an epiphany, it wasn't expected but was most welcome.

By the way, not that many people are back on 1.9.1, but if you are the stable branch got this pushed as well.  No sense making the stable people wait for this miracle, lol.

Yeah, the collisions still work. That's awesome!

Link to comment
Share on other sites

1 hour ago, RyanRising said:

Yeah, the collisions still work. That's awesome!

Yeah, I literally eeked out very close to 10x the performance at Woomerang launchpad with scatters at 100%...

Uh, wow:

Before:

screenshot15.png

After:

screenshot16.png

Hardware is fairly high end, but you get the point.

Edited by R-T-B
Link to comment
Share on other sites

This thread is quite old. Please consider starting a new thread rather than reviving this one.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...