Latest Arauna

(L) [2006/12/26] [Phantom] [Latest Arauna] Wayback!

Allright here's a recent build of the Arauna tracer. It's very much work in progress, but (numerical) stability should have increased quite a bit since last attempt. Speed is also a tad higher. And, the kd-tree can be used for dynamic scenes, just like BIH. The scenery still looks like crap, especially now I made almost every triangle normal mapped and ridiculously specular. [SMILEY Wink] Anyway, check it out:

[LINK http://www.bik5.com/arauna.rar]

Sources will be released later this week. It includes some kind of an API and two examples ('walkthrough' and 'bouncing cubes'), both implemented in <100 lines of code.

I actually hit 'submit' before wishing you all a merry Christmas and a good 2007, so this is my first edit. Happy new year! May the Lord bless your coding and research.

- Jacco.
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/26] [Ho Ho] [Latest Arauna] Wayback!

The shadows are kind of jumpy and the back wall doesn't look that good. Other than that, that thing looks impressive!

[IMG #1 ]

Waiting for source and enough free time to finish the Linux port at last. Perhaps I can then have a real border around the window so that I can reactivate it once some other window gets the focus. Currently when running through Wine it is impossible to manouver around once the focus is lost.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

[IMG #1]:Not scraped: https://web.archive.org/web/20071025012116im_/http://img403.imageshack.us/img403/3369/arauna3ip0.jpg

(L) [2006/12/26] [Phantom] [Latest Arauna] Wayback!

Ah crap then I fixed the wrong problem. [SMILEY Smile]

[LINK http://www.zshare.net/download/arauna_exe-rar.html]

The exe in that file has proper 'N.L' diffuse lighting. Resolves the popping shadows. I can't believe I didn't spot that earlier... Thanks for notifying me, even if you didn't intend to. [SMILEY Wink]

By the way, the solution to the noise problem, according to a thread on gamedev.net, is to increase your monitor resolution to extreme levels. Feel free to help greenhybrid a bit over there:

[LINK http://www.gamedev.net/community/forums/topic.asp?topic_id=429607]
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/26] [Ho Ho] [Latest Arauna] Wayback!

Much better now [SMILEY Smile]

Though there is a minor problem of some texture coordinates being a bit off but I think that isn't big enough problem.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2006/12/27] [Phantom] [Latest Arauna] Wayback!

Well now that you put it that way, it could very well be a problem with my importer. I'll investigate.

EDIT: Checked it, but this doesn't seem to be the problem... That doesn't guarantee that my code is correct of course. Could someone else with a tracer capable of running Sponza check it?
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/27] [tbp] [Latest Arauna] Wayback!

I can confirm your specific version/file has issues, just like some of my old versions of that scene. Stuff was lost in translation.
_________________
May you live in interesting times.

[LINK https://gna.org/projects/radius/ radius] | [LINK http://ompf.org/ ompf] | [LINK http://ompf.org/wiki/ WompfKi]

(L) [2006/12/27] [Phantom] [Latest Arauna] Wayback!

Here's a Sponza scene with some dynamic objects. I'm rather proud. [SMILEY Smile]

[LINK http://www.zshare.net/download/arauna_dyn-rar.html]

The reflections in the spinning cubes suck, but it does at least show interaction between the scene and the dynamic objects.

Technique used is a double kd-tree, one for dynamic objects, one for static geometry.
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/27] [Ho Ho] [Latest Arauna] Wayback!

This version segfaults right after building the kdtree, or at least thats what I think. It starts just the same as the older one. First there is a blank black screen for a few seconds and then the rendering begins. Almost immediately after rendering the first frame Wine segfaults. I'm not sure if the problem is in Wine or your program. Wine thinks it's your program doing bad things on stack [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2006/12/27] [Phantom] [Latest Arauna] Wayback!

It must be my program. It's a straight dump of the first semi-correct result. [SMILEY Smile] If I build something more stable tomorrow, I'll post it. It should work fine on windows machines though.
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/28] [Ho Ho] [Latest Arauna] Wayback!

I just ran the latest Arauna on some single core 3GHz P4 box with 32bit XP. At default camera position it gets around 3.3Mray/s. Looking very nice but I think the reflections would look even better if the cube wouldn't be blue but something else.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2006/12/28] [Phantom] [Latest Arauna] Wayback!

In the meantime (late evening / early morning [SMILEY Smile] ) I fixed the double-barrel reflection traversal, so now reflections of the red cubes in the blue faces are correct. I'm very pleased with the fact that the code has almost no overhead; just the static scene performs at nearly 100% of what I had yesterday, same for my reference animated scene. This means that Arauna now handles animation very much like a regular 3D engine: You don't need special provisions and you can still exploit the fact that 99% of the scenery is usually static. Of course, a double kd-tree isn't very good in some cases: 1000 static triangles and 1000 dynamic triangles would require 2 x 10 traversal steps, instead of 11 if they would all have been in a single tree.

I am now working on an MD2 importer, so I can have some simple keyframe anims in the scene.

About the blue reflections: What did you have in mind? Something normal mapped? I just wanted to see if it worked, so I didn't bother making it aesthetically satisfying. [SMILEY Smile]
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/28] [Phantom] [Latest Arauna] Wayback!

Tbp, I totally agree... Apart from the mouse control, that is. I don't like mlook that much when the frame rate drops below 10.

About Quake1: The alternative is a proper scenegraph, as you pointed out, and skeletal animation. That takes some proper test art, and a lot of (non-rtrt-related) code, so for now, I'll just do a quick-and-dirty MD2 thingy. Advantage is that everyone associates that Marine with 'games', so it might help the rtrt discussion in the short term. Just like adding normal mapping: It doesn't prove anything, on the contrary; I consider normal mapping a rasterizing trick that should have been rendered obsolete by the high-poly capacity of ray tracers. So a proper ray tracing demo should convert all bump maps to subdivided geometry right away... But that's again a lot of non-rtrt-related code.

Anyway, MD2 loading and playing is 80% complete right now (after 1.5 hours of coding) so I'll just finish it. [SMILEY Smile]
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/28] [tbp] [Latest Arauna] Wayback!

I'm dropping in that code everytime i need a camera, so yes it also ended up in Radius (where the sampler is also used to shoot rays through the pinhole). As presented it matches OpenGL conventions.
_________________
May you live in interesting times.

[LINK https://gna.org/projects/radius/ radius] | [LINK http://ompf.org/ ompf] | [LINK http://ompf.org/wiki/ WompfKi]

(L) [2006/12/28] [Phantom] [Latest Arauna] Wayback!

Here's another update:

[LINK http://www.zshare.net/download/arauna_dyn2-rar.html]

This time, a modified Quake Marine is running up and down one of the galeries. I'll also post this on gamedev, people were asking for a game scene with some skinned meshes. I know this hardly qualifies, but still.

Added tech: Just an MD2 file loader and a keyframe interpolator for my scene graph.
_________________
--------------------------------------------------------------

Whatever

(L) [2006/12/28] [Ho Ho] [Latest Arauna] Wayback!

Nice and this time it worked in Wine too. Though you might want to make the marine a bit bigger or at least say that it is behind some wall. Took me about three minutes to find him [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2006/12/29] [olliej] [Latest Arauna] Wayback!

Very nice, lowest framerate i got was 14fps, and that was with one blue cube filling the entire screen, and showing a reflection of the other blue cube. Very nice.

For just wandering around the scene it seemed to average ~24-26fps [SMILEY Very Happy]

Still the texture coords seem kind of borked [SMILEY Sad]

(L) [2006/12/29] [olliej] [Latest Arauna] Wayback!

Oh, forgot to mention, the big advantage of an excessive home computer: when testing your retarded linear search raytracer the performance is "good enough" to not cause insanity to strike while you fix things.

*sigh*

My kdtree is just not working at all so it's beginning to get annoying (just doing a spacial binary tree rather than anything sensible) and the tree is borked :-/

(L) [2007/01/10] [Phantom] [Latest Arauna] Wayback!

Here's another update. We are preparing for game development, so it's linked to FMOD, scene graph stuff is coming along nicely, and refraction has been fixed (again). There's a glass somewhere on the ledge, but I am afraid it's even harder to find than the Marine. [SMILEY Smile] The Marine now holds a flashlight by the way.

[LINK http://www.zshare.net/download/arauna_jan_10-zip.html]
_________________
--------------------------------------------------------------

Whatever

(L) [2007/01/10] [Ho Ho] [Latest Arauna] Wayback!

I'd run it but it seems to be segfaulting every single time I run it after rendering the first frame. I tried around ten times and it was exactly the same every time.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/01/10] [Phantom] [Latest Arauna] Wayback!

I have no idea what might be causing that... I'll prepare a minimal version tomorrow, without some of the recent features.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/01/10] [Ho Ho] [Latest Arauna] Wayback!

I have no ideas what it could be but I remember having similar problems with some older version. I guess it is just something that has always been there but shows up based on moon phase at the moment of compiling [SMILEY Smile]

If it takes too much time you don't have to bother making one Linux user happy. I can live without it and hope that it will work next time [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/01/11] [fpsunflower] [Latest Arauna] Wayback!

Works good here. Most of the artifacts I saw before are gone now.

The window has no border and there is no fps counter that i could find. Do I need to press something to bring that up?

Oh and I like the music [SMILEY Smile]

(L) [2007/01/11] [Phantom] [Latest Arauna] Wayback!

Thanks, I indeed think I solved most of the accuracy problems (not all though). You can enable full statistics through scene.txt, just uncomment the 'showstats' line (one of the first lines in the file). It's disabled because the stats are rather intrusive now. I forgot to include one figure in the conditional block, that's the coords that are plotted at the centre; it's the result of a single ray query and I use it to position objects.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/16] [Phantom] [Latest Arauna] Wayback!

Here's a fresh Arauna demo. Students are now working with it (and on it), but there's no original art yet.

[LINK http://www.zshare.net/download/arauna_demo_feb2007-rar.html]

New in this version:

- Pretty fast dynamic objects using a 'double barrel' approach: Kd-tree and BIH;

- Refraction is working again (hence the initial camera view);

- Lots of small things that I forgot (API, mainly).

Instructions:

The initial camera viewpoint is used to test the refraction. You may want to move away for a better (and faster) view.

Controls are: Cursor keys for camera direction; ASDW for camera translation, R/F for up/down (reversed, still didn't bother to fix it).

You may also want to experiment with the settings in scene.txt. Most are obvious; besides those 'hdrsurface' attempts to render to an FP128 render target (including 4 stage post processing pipeline, shaders are not good yet), 'hdrtarget' does the same but only in software (multithreaded though).
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/16] [Shadow007] [Latest Arauna] Wayback!

Hi,

I use an AZERTY keyboard, thus the WASD key binding is absolutely sub-optimal for me ...

Do you think it would be possible to add an alternate binding ? (for example 8456 on the numpad) ?

It will make it a LOT better for me [SMILEY Smile]

Thanks !

(L) [2007/02/16] [Phantom] [Latest Arauna] Wayback!

Here's an azerty executable, quickly whacked together:

[LINK http://www.zshare.net/download/arauna_exe_azerty-rar.html]

Enjoy. [SMILEY Smile]
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/19] [Phantom] [Latest Arauna] Wayback!

Can you give some info on your CPU and OS? I fear that I (once again) compiled a Pentium-M and above version... Regarding the OS: I fixed lots of problems, so I was hoping it would at least run under Wine again.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/19] [Ho Ho] [Latest Arauna] Wayback!

Damn, I wasn't logged in during the few last days so I missed the new demo. I downloaded it at work and upgraded the exe right away but it doesn't work through Wine since it needs d3dx9_32.dll. Wine DX9 support isn't quite there yet. I'll try again later tonight at home where I had a bit tweaked Wine that should have better DX9 support.

Had you used OpenGL there would have been much higher chance of success [SMILEY Smile]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/02/19] [Phantom] [Latest Arauna] Wayback!

I know, but in GameDev, OpenGL is still pretty much not the standard, so I thought I would practice on DirectX instead. [SMILEY Smile]
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/19] [Ho Ho] [Latest Arauna] Wayback!

Well, ID and Epic still use OpenGL quite a lot amongst other developers. Also, it works nicely on pretty much anything there is without much problems and has (far) less overhead according to one NVidia GPU designer/architect with whom I had a pleasure to argue about if RTRT is feasible or not [SMILEY Smile]

If you have too much time and nothing to do with it you can always check out [LINK http://developer.nvidia.com/page/home.html NVidia developers site]. It has loads of information about [LINK http://developer.nvidia.com/page/opengl.html OpenGL] (and [LINK http://developer.nvidia.com/page/directx.html D3D]), general [LINK http://developer.nvidia.com/page/documentation.html documentation] and several good [LINK http://developer.nvidia.com/page/tools.html tools], some are even cross-platform. Btw, first 8800 series demos that were supposed to show the powers of DX10 were also programmed in OpenGL since DX10 was pretty much non-existent then [SMILEY Razz]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/02/19] [Phantom] [Latest Arauna] Wayback!

But on ATi it's the other way round, right? Those differences between ATi and NVidia are anoying anyway, I believe the 7800 was the first to have full 128bit FP, correct?

By the way, what did this NVidia bloke say about RTRT? Do they take it into account for the future, or is it not influencing them at all?
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/19] [toxie] [Latest Arauna] Wayback!

I can tell something about "plans" for RTRT. But the info is soooo old. I talked to a nvidia-hardware-designer bout 3 years ago bout RTRT. And he told me that they had some prototype-stuff (i don't know if it was already "real" hardware or just "emulation" or FPGA) but came to the point that it's not really worth the trouble YET. So i suppose that both nvidia and AMD/ATI are on it, but we won't be knowing about it until the first chip will pop up.
_________________
Eat plutonium death you disgusting alien weirdos!

(L) [2007/02/19] [Ho Ho] [Latest Arauna] Wayback!

Good news is I installed DX9 to under Linux and now it stopped complaining about those DLL's [SMILEY Smile]

Bad news is I get black screen with only the OSD stuff visible. The thing seems to run at aroun 8-9FPS in the default position.

Wine sais this in console:

(L) [2007/02/19] [Phantom] [Latest Arauna] Wayback!

The shaders are from an ATi example program, 'Rendering with natural light'. I changed the code that uses the shaders (heavily) but not the shaders themselves. I am not a DX expert, so I have no idea what Wine is complaining about...
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/19] [Ho Ho] [Latest Arauna] Wayback!

I'm using 6600GT and Wine translates those shaders into opengl shaders on the fly (GLSL). I have no ideas where might be the problem. One possibility is that those shaders are using some ATI specific things that don't work that nicely on NV. Other possibility is that Wine's shader translation isn't all that good.

Btw, would it be too much work to make the shader usage configurable? For start simple boolean would be enough. When true it'll use all the shaders, when false it wouldn't use any of them.
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/02/19] [knallbunt] [Latest Arauna] Wayback!

runs on Vista but only after installing the february sdk

end-user runtimes should be enough though

[LINK http://www.microsoft.com/downloads/details.aspx?familyid=4b1f5d0c-5e44-4864-93cd-464ef59da050&displaylang=en]

(L) [2007/02/19] [Phantom] [Latest Arauna] Wayback!

That's good to know. I was considering to install Vista (64bit) as a second OS on the octcore, but I wasn't sure it would run Arauna at all... By the way, forgot to mention that the latest demo is hardwired for 2 cores. It will perform poor on systems with more than that.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/20] [MishoM] [Latest Arauna] Wayback!

About the crash on my computer. This is my computer at work. At first it complained about missing DirectX dll's so I downloaded the latest DirectX from Microsoft. Then the program started crashing immediately after the output window is displayed. The processor is a Prescott Pentium 4 at 3GHz. The operating system is WinXP Pro.
_________________
--------

MishoM

(L) [2007/02/21] [Shadow007] [Latest Arauna] Wayback!

Still on the milestone code subject, I've got one more question :

In the core, when you compute Color and Normal (getColorAtIP) (ie shading ?), you make a first test to check if all the intersections of 4 rays row point to the same primitive.

If so, you compute in one pass the results for the 4 rays.

If not, you revert to a single ray shading.

First reflexion( a minor one) : wouldn't it be possible to add a coherency mask for the 4 rows ?

Second reflexion : I don't know how heavy the shading is in the tracing time, but wouldn't it be possible to add intermediate modes ?

By that I mean some way to compute the 4 Colors/Normals, and only keep the ones for the rays that have the same primitive.

That way, if prim[0] == prim[1] && prim[2] == prim[3], you'd only have 2 passes ...

I'd guess you could also go and check in the other rows to find additional rays intersecting the same primitive...

In that case, if you had prim[0] == prim[1] == prim[4] == prim[5] && prim[2] == prim[3] == prim[6] == prim[7], you'd still have only 2 passes...

I'd guess it would then depend on the difficulty/computation cost to implement the filtering/computation masks/results dispatches ...

(L) [2007/02/21] [Phantom] [Latest Arauna] Wayback!

GetColorAtIP has indeed been promoted to do the entire shading; it used to be just texture lookup, but I needed some local variables for shading that already where calculated for bilinear filtering. I guess the name should change.

Coherency for four rows won't improve performance, since only four rays can be handled simultaneously (SIMD).

Except for the 'passes' idea that you mention; if e.g. all 16 rays hit the same prim, some data obviously will be the same for each row. I'll look into that. Shading is taking a lot of time, although most is probably spend on cache misses for texture lookups.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/02/22] [Shadow007] [Latest Arauna] Wayback!

It seems you don't get what I wanted to say :

If for example you the intersections data prims are those :

1 2 2 2

3 2 4 4

3 2 2 4

3 3 2 4

In the current code, you'll take that as absoluteloy not coherent and make 16 'single ray' passes.

With what I have in mind, you'd have

1 single ray pass for ray 0 (prim 1)

1 SIMD pass for rays 1 2 3 5 (prim 2)

1 SIMD pass for rays 9 10 14 (prim 2) (1 ray lots, but masked out)

1 SIMD pass for rays 4 8 12 13 (prim 3)

1 SIMD pass for rays 6 7 11 15 (prim 4)

So 5 passes only.

(L) [2007/02/22] [Phantom] [Latest Arauna] Wayback!

Allright, I see what you mean. That indeed sounds interesting. I wonder if the gains will outweight the logic to gather rays this way though. Need to check that. [SMILEY Smile] Thanks for the suggestion, it's quite a fresh idea.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/09/24] [Ho Ho] [Latest Arauna] Wayback!

I believe latest released sources are downloadable [LINK http://rt07.raytracing.nl/ here]
_________________
In theory, there is no difference between theory and practice. But, in practice, there is.

Jan L.A. van de Snepscheut

(L) [2007/09/24] [Phantom] [Latest Arauna] Wayback!

The latest version is available here:

[LINK http://rt07.raytracing.nl/] . This is the last version before I switched to BVH. BVH is not yet done, so it doesn't make sense to release that (you can check a work-in-progress version though, I posted a package somewhere on this forum).

Note that all versions have pretty severe compatibility issues. I badly need a better frame to build this on, since I use very few DX features.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/10/06] [TomasRiker] [Latest Arauna] Wayback!

Hi,

I can't get the latest Arauna demo working.

I start the application, but nothing happens. Seems like it quits right away because I can't find it in the task manager seconds later. The applog.txt is empty.

I'm using Windows Vista, but I think earlier versions worked with Vista.

Bye

David

(L) [2007/10/06] [lycium] [Latest Arauna] Wayback!

today i upgraded my pc to an intel q6600, so now the demo starts (of course, since it's not an amd box ;)) but it still doesn't work :( it just displays the flickering arauna logo, indicating 65fps but i can't see anything. i pressed the i key, no go.

windows xp 64bit, geforce 8800gts 640mb, all the latest drivers etc.

(L) [2007/10/06] [Phantom] [Latest Arauna] Wayback!

I know, and it has been reported by many people: Arauna doesn't seem to work on Vista, 64bit and/or Linux, and apparently several other systems have problems as well. I need to make a way more compatible demo at some point, but this is something that I keep pushing forward since I hate bothering with the windows-specific code... Perhaps I'll make a SDL port or something, just to be done with it.
_________________
--------------------------------------------------------------

Whatever

(L) [2007/10/06] [lycium] [Latest Arauna] Wayback!

sdl is supposedly 64bit compatible, but i think that's only valid on linux because back when i tried it last (when they'd just released the 64bit compatible one) it didn't work for me on msvc8. probably if you compile it from source now it'll work fine :) 64bit is highly recommended btw, usually i see a very good speed improvement with sse heavy code.

oh and since you're opensource, it shouldn't be much of a problem to use tbp's spanky opengl fbo display path.

Latest Arauna back