[QUOTE=Lexic;35964697]Depends on the shape of oscillation he wants. Pure sinewaves don't produce smooth deltas, especially at the peaks.[/QUOTE]
I'm not sure what you mean. Would one cosine and one divide be slower to compute than the square root, exponents and divisions used in a bell curve?
If you knew that frame counts were always powers of 2 you could build a high resolution lookup using either method (say 256 entries) and get values from there.
[QUOTE=DrLuke;35964746]the E has to be a non-capital e, or else it's an exponent as in 10^x[/QUOTE]
it's mathematica, all built-ins are pascal cased
E, Sin, ContourPlot, etc
I tried using sin, that was the first thing I tried. I don't think it works properly because of the frame numbers. Although thinking about it again - it seems like it should work, because the complete curve balances out - so as long as the frames are spaced out evenly it should work?
a 'nice' solution would probably take more effort and be slower at runtime than an ugly hardcoded lookup table
don't bother
[QUOTE=Ziks;35964755]I'm not sure what you mean. Would one cosine and one divide be slower to compute than the square root, exponents and divisions used in a bell curve?
If you knew that frame counts were always powers of 2 you could build a high resolution lookup using either method (say 256 entries) and get values from there.[/QUOTE]
[IMG]http://i.imgur.com/ehuJm.png[/IMG]
(Dotted line = sinewaves, red = deltas)
I was referring more to the distribution of changes in value, probably using the wrong words.
I don't know what Garry's using these for, but depending on the curve used to generate the values, you can get completely different effects while leaving the overall result the same.
I assumed this would be for an oscillating animation, and using a bell curve would be weighted too much towards 0 whereas a sine has equal weighting between 0 and the max (2).
[QUOTE=swift and shift;35964863]a 'nice' solution would probably take more effort and be slower at runtime than an ugly hardcoded lookup table
don't bother[/QUOTE]
I think you are right
Im trying to make a voxel rederer ala minecraft, but I was wondering, how should I manage the optimisation? I thought of: add each cube's triangles to a chunks VBO, then go through all cubes and check if there are other cubes on/under/next to the current one, if so go on to check the next one. if not add the index of that given side of that cube.
Then ofcourse only update when the player has deleted/created a new cube
Extended my heating model for the aerospace simulator.
[IMG]http://i.imgur.com/TfLiy.png[/IMG]
[IMG]http://i.imgur.com/TmPBr.png[/IMG]
A bit nicer, but low poly:
[img]http://i.imgur.com/kV1Xq.png[/img]
It has basic simulation of shockwave boundary layer (at least the shockwaves formed by leading/trailing edges of wings, also shockwaves formed by fuselage). On the picture that's the large green area/streaks on the fuselage side (increased heating due to shockwaves iteracting with fuselage).
Ignore the very hot tail - that's a limitation of calculations (it doesn't work well for surfaces which cast shockwaves on themselves yet).
[B]The heating model works in realtime[/B]. It is not CFD (as in solving equations of flow), but it's an approximation of CFD (it accounts for geometry and gives sufficiently nice results).
Compare to real result (some heating components like thermal radiation from plasma around the spacecraft are not accounted for but general trend is OK). [B]This is Buran. The vehicle on my pictures is longer than this[/B]:
[IMG]http://buran.ru/images/jpg/ter261.jpg[/IMG]
[QUOTE=Richy19;35966701]Im trying to make a voxel rederer ala minecraft, but I was wondering, how should I manage the optimisation? I thought of: add each cube's triangles to a chunks VBO, then go through all cubes and check if there are other cubes on/under/next to the current one, if so go on to check the next one. if not add the index of that given side of that cube.
Then ofcourse only update when the player has deleted/created a new cube[/QUOTE]
Something like that, yeah. In my code each chunk has an updateMesh() method, which rebuilds the mesh (as an index buffer) and reuploads it to video memory. The vertex buffer (with positions and normals and stuff) is left unchanged; in fact, I only have one buffer which is shared by all of the chunks, but that's not really an important optimisation.
In terms of culling faces, which I think you were asking about, you need to run through each block in the chunk, and check for a neighbouring block in each direction - if not, push that face to your index buffer. For faces on the edge of the chunk you're best off assuming that the adjacent block is unoccupied, until you add cross-chunk culling; otherwise you'll have missing faces at chunk boundaries.
Another thing to think about is putting each set of faces (for each direction) into a separate buffer: this means you can render only faces pointing towards the player, which cuts the transform cost significantly. That's more of a fiddly thing though, you're better off getting it mostly working first before you go all-out on the optimization.
[editline]15th May 2012[/editline]
It's bloody messy, but if you like you can search for "bloxelcraft" on github, my code's all public.
I've got a spare Pi, so I've decided to give it to a good (developer) home.
Tweet me a reason why you should get a Pi, and I'll pick one within the next day or two and get a Pi out to you within the next week or so: [url]https://twitter.com/#!/hexxeh/status/202444390046777344[/url]
[vid]http://dl.dropbox.com/u/4093439/projectiles.webm[/vid]
Working on Minor Destruction again. I've teamed up with a seasoned game designer who worked on Homefront (and he quit because they wouldn't listen to him when he said it was going to be shit :v:)
This is projectile networking that I've just finished. The game can handle 256 projectiles at any one time (due to the projectile ID being a byte for network speed), and the projectiles' positions are predicted clientside; the server doesn't tell the client where the projectiles are at all.
[QUOTE=Hexxeh;35967372]I've got a spare Pi, so I've decided to give it to a good (developer) home.
Tweet me a reason why you should get a Pi, and I'll pick one within the next day or two and get a Pi out to you within the next week or so: [url]https://twitter.com/#!/hexxeh/status/202444390046777344[/url][/QUOTE]
Turb should port JSOS to it.
[QUOTE=Jookia;35964000]Fuck off. You changed the line. Don't act like I didn't read it.[/QUOTE]
I had changed it prior to your post though, which was why I said that, it isn't my fault you didn't refresh, I just thought you had read it quickly and legitimately missed it.
Sorry for the misunderstanding. I'm not a prick, or at least I wasn't trying to be :(
-snip-
[QUOTE=BlackPhoenix;35966752]Extended my heating model for the aerospace simulator.
[url]http://i.imgur.com/TfLiy.png[/url]
[url]http://i.imgur.com/TmPBr.png[/url]
A bit nicer, but low poly:
[url]http://i.imgur.com/kV1Xq.png[/url]
It has basic simulation of shockwave boundary layer (at least the shockwaves formed by leading/trailing edges of wings, also shockwaves formed by fuselage). On the picture that's the large green area/streaks on the fuselage side (increased heating due to shockwaves iteracting with fuselage).
Ignore the very hot tail - that's a limitation of calculations (it doesn't work well for surfaces which cast shockwaves on themselves yet).
[B]The heating model works in realtime[/B]. It is not CFD (as in solving equations of flow), but it's an approximation of CFD (it accounts for geometry and gives sufficiently nice results).
Compare to real result (some heating components like thermal radiation from plasma around the spacecraft are not accounted for but general trend is OK). [B]This is Buran. The vehicle on my pictures is longer than this[/B]:
[url]http://buran.ru/images/jpg/ter261.jpg[/url][/QUOTE]
When can I play this, and where do I send all my money to?
[QUOTE=DrLuke;35968247]When can I play this, and where do I send all my money to?[/QUOTE]
Right now it comes as a plugin for X-Plane: [URL]http://forums.x-plane.org/index.php?app=downloads&showfile=9115[/URL]
That version only has a capsule, not this shuttle, and not the more advanced heating model. Just a rocket that flies, and a little bit of stuff to make own designs (with X-Plane Plane Maker).
If you want to help... I want to write my own flight dynamics and make my own free aerospace flight simulator. I'm thinking of Outerra for rendering (photo-realism/HDR is required). So I guess I could use money to buy a license from them or something - but they didn't really give any detailed information on licensing yet.
I mean if you're serious about the money and the Outerra guys are serious about selling their engine, then I could build an aerospace simulator akin to X-Plane, though more oriented on making own spacecraft.
[b]Edit:[/b]
Think this, but with a GUI [i]that fits me better[/i]:
[img]http://soportealainformatica.files.wordpress.com/2012/02/vomf.png[/img]
This is amazing, I wish I had xplane to try it out.
When you really move this to your own engine, will you support some sort of scripting similar to Orbiter? Will you implement docking? What about some sort of plug and play rocketry like kerbal space program?
[QUOTE=DrLuke;35968758]This is amazing, I wish I had xplane to try it out.
When you really move this to your own engine, will you support some sort of scripting similar to Orbiter? Will you implement docking? What about some sort of plug and play rocketry like kerbal space program?[/QUOTE]
No plug and play rocketry. The editor lets you specify any sort of spacecraft/rocket parametrically (this means that you specify the fuselage shape by giving it a few crossections, and you specify wings by telling it what airfoil you want).
This data is used by simulator for both physics and rendering (the idea similar to X-Plane).
Docking will be supported. Please take note of this project too: [url]http://vsfl.wireos.com/[/url] Essentially that's an RPG for aerospace engineers - highly realistic one. It's a bit private - you gotta design at least one spacecraft capable of autonomous flight to enter. But this is built in X-Space networking (which can be reused for just multiplayer).
I was planning to port X-Space (or rather networking part) to Orbiter, so Orbiter could be used with the VSFL network too. Of course the most "native" support would be in my simulator.
There will be Lua scripting - my engine will be written in C and Lua (excluding Outerra, which is C++ and a bunch of other things). All high-level code is in Lua, so scripting is nearly trivial to implement (X-Space plugin is already built this way).
IVSS (internal vessels systems simulator) is another big project, which will be part of this. In fact, it's the second most interesting one - it's for simulating complex systems in realtime by specifying them as simple blocks. Essentially a rapid prototyping tool. Say I want a hydraulic system of a nuclear reactor:
[img]http://brain.wireos.com/wp-content/uploads/reactor-400x230.png[/img]
I enter that into simulator, and get a realtime model which describes all transient processes and whatever that happens in my hydraulic system:
[img]http://brain.wireos.com/wp-content/uploads/tr1-400x191.png[/img]
[img]http://brain.wireos.com/wp-content/uploads/tr2-400x191.png[/img]
This entire thing should come as a single modular package (simulator, aerodynamics editor, systems editor, and links to other software, e.g. CFD software for getting better aerodynamic solutions). [b]Target goal[/b]: should be possible to fully design an aircraft on the preliminary level within months of work. After this first step is done and the design is sufficiently finalized the main engineering work can start.
[img]http://i.imgur.com/KmhJ9.png[/img]
I don't think my phong shader is working top notch :v:
Clamp?
[QUOTE=Darkwater124;35969377]Clamp?[/QUOTE]
[IMG]http://i.imgur.com/jIcnK.png[/IMG]
Thanks, added more balls and clamped - but now they're full bright, gonna check out the code for additions.
[B]
Edit:
[/B][img]http://i.imgur.com/wlFMz.png[/img]
Yeah! It works! (Somewhat)
[vid]http://dl.dropbox.com/u/41041550/videos/LatestPhysics.webm[/vid]
I've come a long way since the last video I showed.
The collision is done "per area" which means that the physboxes only check for collisions in their immediate area. Enjoy!
I just generated my first [url=http://dl.dropbox.com/u/19826351/Facepunch/packages/index.html]Javadoc[/url] from a class I've been working on recently. It's an object that handles IPv4 and IPv6 addresses.
[QUOTE=MrTilepy;35969686]I just generated my first [url=http://dl.dropbox.com/u/19826351/Facepunch/packages/index.html]Javadoc[/url] from a class I've been working on recently. It's an object that handles IPv4 and IPv6 addresses.[/QUOTE]
Looks very organized.
Although... Am I the only one who likes the look of [URL="http://docs.oracle.com/javase/1.4.2/docs/api/java/lang/System.html"]this type[/URL] of Java documentation?
Not strictly what I'm working on, but If anyone else has been having the annoying Nvidia driver bug where the graphics card underclocks itself (and refuses to put the clocks back again), try this:
[url]http://dl.dropbox.com/u/33076954/clockFixer.rar[/url]
Essentially it just crashes the driver by timing out the maximum length for clEnqueueNDRangeKernel under opencl
If anyone particularly wants the source:
[url]http://dl.dropbox.com/u/33076954/GF.rar[/url]
FP Programmers group has been compromised.
:D
[QUOTE=MrTilepy;35969970]FP Programmers group has been compromised.[/QUOTE]
Austech becomes chat-mod
[IMG]http://puu.sh/uVqm[/IMG]
Invites everyone to chat.
[IMG]http://puu.sh/uVBf[/IMG]
Suddenly thousands of announcements.
Conclusion: this is all Overv's fault.
PS:
[img]http://puu.sh/uVww[/img]
I can post and delete announcements. Is that the same for everyone else?
Sorry, you need to Log In to post a reply to this thread.
