Fr0stbyte's Development Log

Thanks, at least now I know!

tonyri wrote:

Dr. Mackeroth wrote:

Well, you should all be very pleased to know that the video doesn't work on my phone. Could someone explain what happened?

The terrain loaded very fast, much faster than with optifine or any let's player's computer. But random chunks would disappear and reappear in no particular order, so it would be like seeing chunk errors in a checkerboard, if you have ever seen chunk errors on multi-player.

P.S. When I say randomly, I mean randomly for me. Frostbyte probably knows why some render and others don't.

Not really.
The ones on the surface are supposed to render. The ones below ground are not. Something is terribly wrong, though I am not sure what. It's just going to take a ton of code tracing, but I have a suspicion it isn't screwed up as badly as it looks.

fr0stbyte124 wrote:

tonyri wrote:

Dr. Mackeroth wrote:

Well, you should all be very pleased to know that the video doesn't work on my phone. Could someone explain what happened?

The terrain loaded very fast, much faster than with optifine or any let's player's computer. But random chunks would disappear and reappear in no particular order, so it would be like seeing chunk errors in a checkerboard, if you have ever seen chunk errors on multi-player.

P.S. When I say randomly, I mean randomly for me. Frostbyte probably knows why some render and others don't.

Not really.
The ones on the surface are supposed to render. The ones below ground are not. Something is terribly wrong, though I am not sure what. It's just going to take a ton of code tracing, but I have a suspicion it isn't screwed up as badly as it looks.

Are the cave systems suppose to render? Or did they only render because the surface became transparent.

P.S. I'm feeling pretty epic right now. I just got the Album of the most popular video game songs performed by the London Philharmonic Orchestra. Battlefield Theme makes me feel like a total badass. Same with Legend of Zelda.

Made some headway on the problem. As far as I can tell the base algorithm seems to be working properly, which is best-case scenario. Instead, it seems that chunks being registered in the set are all in the wrong places. This is why there are random missing chunks in the video: sampling rays only pass through chunks through certain directions depending on the location of the camera. If it thinks that chunk in the north is actually in the south for instance, it will declare that chunk a dead-end because no south-bound rays passed through it.

There's just so much messed up with this data structure, it might be easier just to tear it out and try a new approach, something more self-contained and automated, and with some built-in error checking so I don't have to do this wild goose chase again.

I think I fixed the bugs. I'll have to do more testing but it's looking really good. Got some unexpected results though.

If everything is correct, then according to my very unscientific eyeball test, it looks like this occlusion culling gets between 40% to 75% of the chunks that advanced opengl gets. Unfortunately, 50% is not good enough to beat advanced opengl, at least on my computer. I'm thinking about enhancing the resolution by calculating 8^3 mini-chunks, plus I haven't verified this is entirely correct (it never culls a visible chunk, at least).

But here's the part I didn't anticipate: if you run advanced opengl and this culling method together, it's actually faster than either separately. In retrospect this makes sense, assuming advanced opengl is the faster technique. It has to go over all the geometry to make its calculations. Since my technique reduces the geometry, though not as much as advanced opengl, it can finish faster.

I would also be interested in seeing how the two compare on weaker systems.
The plan now is spend the weekend testing and optimizing, maybe seeing about setting up the higher resolution scan, setting up some sort of way to gather statistics, and then release the alpha on Monday.

fr0stbyte124 wrote:

I think I fixed the bugs. I'll have to do more testing but it's looking really good. Got some unexpected results though.

If everything is correct, then according to my very unscientific eyeball test, it looks like this occlusion culling gets between 40% to 75% of the chunks that advanced opengl gets. Unfortunately, 50% is not good enough to beat advanced opengl, at least on my computer. I'm thinking about enhancing the resolution by calculating 8^3 mini-chunks, plus I haven't verified this is entirely correct (it never culls a visible chunk, at least).

But here's the part I didn't anticipate: if you run advanced opengl and this culling method together, it's actually faster than either separately. In retrospect this makes sense, assuming advanced opengl is the faster technique. It has to go over all the geometry to make its calculations. Since my technique reduces the geometry, though not as much as advanced opengl, it can finish faster.

I would also be interested in seeing how the two compare on weaker systems.
The plan now is spend the weekend testing and optimizing, maybe seeing about setting up the higher resolution scan, setting up some sort of way to gather statistics, and then release the alpha on Monday.

Made a video. This is what the culling does right now.


It's a little bit disappointing it is not more selective, but I believe this is simply a result of the chunk resolution being too high. If I shrink the chunk test regions to 8^3 or 4^3 I think it may work a lot better, not that it's doing all that bad right now. I'd also like to look into getting a better method for determining whether a chunk has proper line-of-sight.

Can I post this video on the thread page over on the minecraft forums with an explanation? They would probably be happy to see some progress being made.

(or I could look into a time machine and see you already posted it, then feel stupid because I didn't check first)

Something decidedly odd I discovered today while playing minecraft with IC2 and some addons installed. My FPS is generally not great - unless I'm in the fusion reactor chamber, with the buzz of the reactor and beeping of the computers in the background. Then my movement, everything, is flawless and superfast. When I move out of hearing range, it drops again. Why is this?

I ask because the answer might help with our FPS boosting as well.

Need more information. What kind of rig are you using? Also, is "advanced opengl" enabled or disabled? What about fast graphics and render distance?

I suspect you have advanced opengl enabled, which will dramatically reduce the drawn geometry when you are in small, enclosed spaces. There is pretty significant overhead to this, though, so it has to cull quite a number of chunks just to break even.

Also need this: Pull up the debug info from F3, and give me all the values on the second line (C: ?/? F: ? O: ? E: ?) when you are inside that room.

Dell Inspiron 1525 laptop. Nothing special, with a 2 ghz dual core.

C: 7/648 , F: 55, O:76, E: 538

AOpenGL: Fast

Sorry for hijacking your thread for a moment. I just thought that the FPS boost was intriguing and worth a mention.

Shiva wrote:

AOpenGL: Fast

You're using optimine, aren't you. Well in any case, your card is only drawing 7 chunks, though somehow 83% of the chunks around you are considered empty, which just seems odd. I wonder if that's another trick of optimine or it it's coincidence...

fr0stbyte124 wrote:

Shiva wrote:

AOpenGL: Fast

You're using optimine, aren't you. Well in any case, your card is only drawing 7 chunks, though somehow 83% of the chunks around you are considered empty, which just seems odd. I wonder if that's another trick of optimine or it it's coincidence...

Cubic Chunks? Wait no, CC actively deletes any chunks that are empty.

I didn't say cubic chunks. Also when I say chunks I mean world render chunks, which are the 16x16x16 ones and not the 16x16x128.

I heard you say that you were releasing the occlusion culling mod on Monday. Is this going to be in Area 51? Or is it going to be on the main Futurecraft forums? I know you won't release it on the Minecraft forums. You said so yourself a while ago. It would completely flood the forums.










P.S. How does this "level up stars" thing work above the profile picture? Is it number of posts? Or does a admin/moderator do it?

daman200 wrote:

I heard you say that you were releasing the occlusion culling mod on Monday. Is this going to be in Area 51? Or is it going to be on the main Futurecraft forums? I know you won't release it on the Minecraft forums. You said so yourself a while ago. It would completely flood the forums.










P.S. How does this "level up stars" thing work above the profile picture? Is it number of posts? Or does a admin/moderator do it?

I think he ran into some issues with merging it with minecraft code, so it might be a bit delayed. Also, I think the stars correspond to post count.

fr0stbyte124 wrote:

I didn't say cubic chunks. Also when I say chunks I mean world render chunks, which are the 16x16x16 ones and not the 16x16x128.

I thought it was normal chunks, so thats why I said I thought it was CC- wait, IC2 FUSION reactor? Theres no IC2 addon that adds a fusion reactor you can walk into, I should know, I literally looked through all IC2 addons hours ago for addons to install with IC2 when I helped my friend install IC2 & RP2 4b on our server. Only one had fusion, and it was a single-block reactor.

Yes I am using Optimine.

The stars correspond to post count.

I can't wait for alpha!

It was the Rocket Science add on. And it's a single block - I simply have a roomful of them.

I have not run into any new problems with implementation, but I did run into some problems with being invited to too many new year's activities, so I am now behind schedule. I will still try to release today, but it's going to be ugly. Also it will almost certainly break optifine since I haven't even looked at how that all works.

If it doesn't I'll be shocked and pleasantly surprised.

Let me rephrase that then. It will break optifine until such a time as I can merge the two.

fr0stbyte124 wrote:

Let me rephrase that then. It will break optifine until such a time as I can merge the two.

Is there a reason to? I mean, what can Optifine do that this can't?

tonyri wrote:

fr0stbyte124 wrote:

Let me rephrase that then. It will break optifine until such a time as I can merge the two.

Is there a reason to? I mean, what can Optifine do that this can't?

Some of the different options, such as eliminating bedrock fog, selectively changing the quality of some visuals, etc.

Performance-wise, probably not much.

Performance-wise, Optimine is going to win handily. Optimine is a lot more encompassing and aggressive in improving framerate, where I am going highly tuned for minimal footprint. However, when merged the two will compliment one another quite nicely.

Also I decided I wanted to try 8^3 chunks before releasing, so first release is going to take a tad longer. Also it will look kinda glitchy because I'm just focusing on benchmarking on various rigs for this release and realtime updating messes with the statistics. Second release will be nicer.

Then it's finally back to work on ships.

Right, I don't even know why I wrote all that, it's just
meaningless. So is this. I think I'm going crazy.