Put bluntly I don't believe that the scripted geometry can get anywhere near the performance of a reasonably textured, single mesh model without any dynamic code.
I say this because I've done my best to measure it in other games in the past by using a model that was made of many smaller components vs the same model once it has been welded into a single mesh. That's tricky in Pioneer without having a matching like-for-like model - although I am working on welding the YT-1300 to demonstrate it.
The reason is simple; with a single mesh and single texture and no dynamic geometry/texturing the game only needs to do this:
1) run static lua script
2) set texture
3) set mesh
4) render
However with dynamic or scripted geometry it must do something like this:
1) run static lua script
2) set texture
3) set mesh
4) render
5) run dynamic script
6) begin loops
__6a) set dynamic texture
__6b) set dynamic geometry
__6c) render
__6d) goto 6a
Not only that but each of those dynamic script calls can create the geometry there and then, i.e; it does the same work over and over again every frame.
Also these rendering batches are usually tiny, and by tiny I mean less than 1000 triangles. A lot less!
Rendering works BEST when you take a large number of triangles, with a reasonable number of textures and render them all at once.
Rendering works WORST when you take a large number of triangle, with a reasonable number of textures and render them all in little bits with constantly changing rendering states for alpha, glow, blending and textures.
A lot of our models are closer to the worst state than the best. We don't often see this as a performance problem because we don't render that many of them.
A better way of seeing this is to load up the modelviewer with a single mesh/texture ship and run the performance test, then try one with a lot of scripted geometry.
None of this is meant as criticism of anyones modelling abilities, it's just that because of the scripted system we seem to have ended up with a lot of ships made of small bits with many separate textures and lots of different rendering states.
Andy
Let me go further.
I just tested the Wave: 11080 triangles ran at less than 1 fps
Then tested my "thor"*: 52622 triangles at 10 or 11 fps
Almost 5 times the triangles, but runs more than 10 times faster.
NB: *"thor" is an abandoned model from the Infinity Universe archive that I use for testing because the original model has more vertices than our 16bit indices supports - used without permission which I why I don't release it. It has a single 1024x1024 texture applied and no dynamic lua function.
Thanks for articulating some of the detail. All the reasons you mention are why we're trying to move away from dynamic sections.hat i
Right now the best case is when dynamic only does submodel selection for equipment etc. A good next step is probably arranging for static hardpoints to be declared and let the renderer decided when to render a missile or whatever.
Following that if we can get declared animations of some sort in, then it shouldn't be necessary to have a dynamic section at all.
Not to mention the fact that it'll be much easier to model, at least if you manage to support a popular format... 😀
Me too, actually. And when it comes to a game, there's always two facts to consider: one is FPS, and the other is awesome, and your goal is to get the most awsome per frame. Which is a thing static, well designed meshes are usually best at.
@spike1984, Have you flown to Gates spaceport? If you have, I may have installed this wrong, and should erase it and start over.
I'm going to put this aside for now 'cause I want to get back to upgrading Philbywhizz's pad station. (almost done!)
sorry for the late reply mate i been busy lol ........eeer no not yet .....but i will give it a go for ya lol
EDIT ...just noticed ..you probly fixed the issue lol ...sorry
right on, make those stupid matrix4x4 animations and kill pioneers identity.
the past few models will be my last i guess, you go somewhere i don't like to follow.
really in the beginning i had some reservations about pioneers modeling system.
but after i got the "clue", i started to love it and was confirmed with tomms expression that is more sophisticated rather meshes and matrix4x4 anims.
it WAS why i liked pioneer, somekind of a toy in all aspects. now you like to make something out of it it never was, imho.
it was somekind of skeleton that could be co-developed by everybody.
it was held simple and could be easely understood by anyone.
now a small group of devs is dictating the future of pioneer, instead that it became a community driven toy (toy, because i saw pioneer more as a toy rather game).
[/hr]
my fault to understand "Pioneer" as concept 😉
Anybody can have a hand in Pioneer's future. Our sticky thread has information about how to effectively contribute, including a link to our wiki page about the design process.
This forum is not the place where design decisions are made. Those who only use the forum in their attempts to contribute will go unnoticed. Complaining about it here will also go unnoticed.
There is no cabal. New people are continually managing to get themselves involved.
the past few models will be my last i guess, you go somewhere i don't like to follow.
Hi, I don't know much about modelling, so I'm not quite sure I'm following this discussion. What are matrix4x4 animations, and why does making them kill Pioneer's identity?
Hmm, no-one said that any modelling functionality was to be removed, only added to.
The consensus is simply to make things easier for people, not to remove .lua. Making things easier might involve being able to create animations in blender and export them along with your model. It might also involve telling blender where components go like weapons or radar ect and the script automatically exporting that to a .lua file.
Get on IRC and talk to us so we can understand whats going on here.
Good question 🙂
As for what it is a 4x4 matrix is a way of rotating or translating objects, someone else can fill in the blanks 🙂
Anything that makes it easier to build while retaining functionability is a good thing imo.
easier and...
(i would have liked to place some examples here, but i need to get them from my PC, this will take a little while)
...often wrong (exported animations)
[/hr]
for sure scripting it's not as easy as exporting a animation, there are some limits (deniable ones imo) but smarter? smoother?
whatever, you will have to understand HOW it works and not only "press a button".
i can't simply export a animation and expect that it will work for 100% in a game (it will, if you respect that any mesh format has no (auto)correctional data for the animation, while most CAD progs do this automatic, especially keeping care of mirrored axes and such).
further .obj don't support matrix animations, you would have to change the filetype for the models or accept that you have to load a .obj for each set key (very lumpy).
btw, a matrix animation is almost the same as this,
you load a mesh at each set key and transform it according to the stored values and animate it from key to key (still lumpy, compared to a "real" anim).
(a mesh is stored in general with a top-level frame which contains the pos/rot/scale matrix (alignment and scale of the model), "sub-models" usually have a sub-frame.
animation keys call the "sub-model" and transforms it according to what is stored as pos/rot/scale values in the key.
the "models-world" alignment (top-level frame) is authoritative to all sub-models, in fact such animation values refere always to the centre of the "models-world". that's why animated models have their animated parts either "exploded" or centered when loaded to a CAD).
[/hr]
the disadvantage of a scripted animation is, that you can't change the speed of a animation (part) easy, to correct differences between rotation and
translation of two parts.
each "constraint" you like to give to the animation path has to be scripted, it needs sometimes a little brain-twist.
[/hr]
but i forget sometimes that i'm the (lazy one?)
something off topic
if interested you can get this in full quality from
http://www.archive.org/details/gov.archives.arc.67908
another link to a walk-through of space exploration
Thanks a lot for that link Gernot! There was a cartoon serial that used the animations from this film that I saw as a child. I can't remember the title, but it was about a little boy searching for his lost father throughout the solar system. It was hand drawn animation except for the spaceship shots. I was extremely impressed with them at the time. I've seen a couple of stills from this movie, but never the whole thing. I'm still impressed! 😀
yep, i stumbled over this one too, i guess it was launched in the 60's, but i'm not sure.
i will lookout for it, perhaps gathering all on a different thread here in SSC.
archive.org has many old short movies & series stored.
the movie (Spaceship1 launches) was cut together of two movies that was left unfinished because of WWII (to sad 🙁 well, like one stated, if they only would have invested all in spaceflight instead in war, Mars would have become much closer).
it was such exact that the "reich" feared some could take technical informations about rockets & spaceflight,
because of that it seems it was forbidden then (check the info on http://www.astronautix.com/index.html ).
the main actor (pilot) i've seen once in another german movie, "incident at platform 5" (i'm not sure about the title, i will have to confirm it).
it's somekind of sci-fi ("Technische Fantasie", i like that term) about a oil well in the open sea, many years before such exists.
the interesting thing about this one is, that the platform has been given up by government and all people that worked on it was left to a uncertain future (would die).
[/hr]
hey, wasn't that movie about a boy, saving his father from a abandoned mars exploration?
the ships was called Polaris1 & Polaris2
if it's the right one, it's named "The Space Explorers" (1958) and "The New Adventures of the Space Explorers," (1961).
http://www.fantastic-plastic.com/PolarisIIPage.htm
check this, some nice retro spaceship resin models.
http://www.thespaceexplorers.com/
seems to be a website only about this movie 😮
edit:
the rest of the movie...
That's the one! Thanks again! That's one of the influences that made me a space cadet for life! 😆
all of my models ready for apha16
[/hr]
sub-models
this REPLACES the complete sub-model folder.
missiles have been moved here (since long) and should therefore be disabled in ships.lua
no other sub-models needed!
the old (scripted geometry) pilot has been removed, ships (not mine) which still use the old pilot model must be updated therefore, or the old pilot1 sub-model must be kept.
new sub-models
- additional scanner models
- different cargo models
- "mono rail"
[/hr]
buildings
all buildings of mine, plus "greyoxide" blds.
use it by choice 😉
delete old "city" folder, models which are still in use are included in city3k,
the old factory shouldn't be used before we can separate hostile and breathable atmosphere planets.
[/hr]
stations
use it by choice 😉
changes the groundstations models and the "big crappy"
delete old stations.lua if used
[/hr]
ships
"new" means, not included in alpha16
eagle - revision, new pilot
changed size (orig.), removed some variations
adc - new
adder - revision
asp sparks - revision
cobra - revision
ship can easely be removed from stock...
constrictor - revision, new pilot
courier - revision, new texture, less variations, new pilot
gelios - new
gunboat - new
hullcutter - new
ip-shuttle - revision, new pilot
lanner - revision
lynx - revision, rescaled
rapier - new
shift - new
sidewinder - revision, new pilot
stardust - revision
model can be removed from stock if you like
terra - new
viper coolhand - revision/replacement, different (true) UC animation, new pilot, flipped textures
use it if you like...
viper x - revision, new pilot, fixed zbias
one is missing so far, the "defender", i will add it soon.
Wow! So many new toys that I hardly know which ones to play with! 😀 Are you going to put them into the issuse tracker?
+1 from me too! 🙂 Great work once again Gernot. Please consider adding them to the issue tracker, these models NEED to be in the game.
Yeah! Especially your original ships, in case we someday have to ditch the Elite derivatives.
of course, 30mins before i posted it here 😎
Hey nice work dude. We noticed the issue, but everyone has been extremely busy lately what with the holidays coming and all the extra work thats on.
Since we just entered into the 'freeze' before alpha 17, you should expect something from January, alpha 18.
Edit// You could also help out by explaining why some changes were made and providing a few screenshots for whoever merges it, Its not exactly required but it would help speed the process along, as whoever merges will have to inspect every change and understand why it was made.
The newer ships obviously only need a screenshot, no need to explain things there, unless there is some strange new concept you added 😉
Thanks.
it seems no one noticed the little "bug"?
the pilot and the gelios produced small problem sometimes.
i fixed it the original link(s) contain now the fixed versions (sub-models, gelios).
i haven't noticed it before all was running together, sorry, but i still have a hard time running pioneer on the macbook 🙁
[/hr]
i will see if i can make some new screenshots, why not.
and add more comments to the links on the issue tracker 😉
Back to something off topic, here's another of my earliest space related memories.
need some adequate test results 😉
i started to combine the textures of some models and want to know if there is a difference in performance, as far as i can see there is no (hmm...).
to get proper results of the framerate in the modelviewer you MUST disable "vertical sync" on your gfx card (else it's leveled to 60 fps).
check all three models in the same zoom distance, farer drawn objects have a much better framerate due to the mip-mapping of the texture.
one thing i reached for sure, smaller file size.
but therefore less detailed textures.
here's a model to compare
courier dec2011 1-1 (multiple texture files, same as posted before)
courier dec2011 1-2 (single* texture sheet)
finally
courier_dec2011_1b.zip (multiple texture files, size (almost) same as combined
check out this one! i have best performance with this model. now tell me what you think where that is from?
the textures i used are the same i used for the combined texture sheet.
the total texture size is slightly bigger rather the size of the texture sheet, but performance is better...
[/hr]
*single, with exclusion of texture for scripted geometry.
or show me how to position a scripted geometry on a combinied texture 😆 (would need "endless" tries, a exact calculation is nearly impossible, except you would use a exact grid for it. it's to imagine to combine p.o.t. textures in the same size, i.e. 4x512x512, or 2x512x512 plus 8x256x256 and so on. but still it would be very difficult to find the right U.V. values).
another problem are repeated textures....
i'm still not convinced that a single texture sheet is better in performance...
...make it proof!
i started this after i made several combined texture sheets and never got a real performance gain, even if the combined texture is far smaller rather the original ones.
that brought up the idea to do it vice versa, to use the smaller textures as i used them before each for the corresponding mesh.
and surprise, surprise...
[/hr]
you can't simply compare XNA models with the type used in pioneer.
or you have to imagine that the script reflects the whole XNA mesh, which is build from many parts as well (at least if you have animated parts).
it's imo mostly the same if i have the meshes together in one file, or many to animate them in the script. because even the XNA model contains the same count of meshes (parts) and not only one.
[attachment=995:adder.x.txt] (30 meshes/parts in this model, 23 meshes/parts for my pioneer adder)
for the texture use i have a similar imagination, imo it doesn't matters if a mesh is textured by single textures or a sheet.
but XNA needs sheets, unlike a model which is put together in a script.
even the glow-map is a XNA convention and wouldn't be needed for pioneer, but i guess some can only handle XNA 😉 and have a "microsoft-mind"
(i need a glow-map, else a i can't... and i need this and that... --- i don't need it! i didn't even need a CAD!).
[/hr]
well, some like to change everything, even if there is no real reason for and i'm still wondering what's the goal of it, destroy pioneer?
make it less good? make it the same as all the rest?
*XNA, *texture sheets, *glow-map, *theories, all wrong!
yes, use *XNA style and kill the unique way of modelling for pioneer...
one texture, one material and wrong setup animations with parts loosing orientation 😈
cheap animated undercarriages and more displeasements, like only one animation set supported by default (to use 2, like i did often for ffed3d is not intented by XNA conventions and only possible if the game-engine supports more then one. and only a second alternative animation is possible which stopps when the other starts).
impossible to deform the mesh dynamically,
no dynamic texture use,
no model variations,
more?
[/hr]
but, what i CAN see is, that some can lay better hands on it, if it's made in the "common" way 👿
i guess now i know why "Driver" models are encrypted... (not to avoid spreading).
...yes it's horrible to work on a script that you haven't wrote yourself (especially mine i guess).
but animations need parts to animate...
variations need parts...
ship extras need parts...
hierarchy must be kept...
after all not easy to manage if it should be fairly good in the end,
to make a "stiff" model is easy (i was called lazy, but i haven't seen many models using the possibilities pioneer gives to you. who's lazy? only if i don't make models for all lod's. well you could reach this in another way, leave out some details. the adder (mine) is a good example, the body has only a few hundret polys, but the model with all details has above 5000 polys. why should i model a different body part if it's low poly already? and don't tell me that a complete model for each lod will be better! i didn't believe anything now).
Latest Post: Limit Theory: An Infinite, Procedural Space Game Our newest member: RaymondFah Recent Posts Unread Posts
