Ogre WIP and random screenshot thread (for everyone!)

[SomeFusion]

Some time ago I've accidentally spawned lots of Goblins in a town. This is the result:
[youtube]85g7bH7f-ao[/youtube]

This reminds me that Sumwars desperately needs more varying soundeffects. And better guards They died like flies.

[c6burns]

This reminds me that Sumwars desperately needs more varying soundeffects. And better guards They died like flies.

I like it ... sounds like an Aphex Twin song

[fsteff]

@fsteff: looks very nice! Will you integrate these shadows into Ogre?

I am not the one that created this... i only integrated the code into my engine and modified it a bit.
http://www.ogre3d.org/forums/viewtopic.php?f=11&t=71142

[Kojack]

Some time ago I've accidentally spawned lots of Goblins in a town. This is the result:


This reminds me that Sumwars desperately needs more varying soundeffects. And better guards They died like flies.

Hehe, that's got a good rhythm to it.

[Kojack]

Spherical Harmonic Irradiance time.

In these pics I sampled a bunch of pixels (can't remember if it was 200 or 2000) from the skybox texture, treated them as distant light sources and constructed a spherical harmonic from them. I then rendered the scene with a shader that used the spherical harmonic as a directed ambient light. Worked pretty good, and runs very fast.

[bstone]

Nice! I wonder though how does that compare to directly sampling a strategically blurred cube map for ambient diffuse term? Or to an ambient cube lighting?

[Kojack]

The SH would be more math heavy (it's around 9 vector formulas to solve), but only requires 9 floating point colours to store. No texture lookups.
Blurred cube maps will require texture lookups per pixel. Probably only one if it's a pre-blurred texture.
So I guess it comes down to which has the bigger hit on the given gpu, a texture lookup or some math.

The spherical harmonic stuff can also be generated at runtime each frame. I had 200 moving light sources being merged into a single SH on the cpu then passed to the shader as constants. Dynamic lights and precomputed lights from a skybox are easily blended, the shader that renders the result doesn't need to know about that. Although of course SH isn't suited to nearby lights or high detail lighting. You still want actual lights in there too.

I want to give the cube map stuff a try, it sounds interesting.

Something I really want to work on is getting a good Cone Step Map working (looks better than parallax occlusion and runs faster). The patent expired 1 or 2 years ago, but it seems everybody has forgotten about it (except XNormal, which can generate the maps).

[c6burns]

Something I really want to work on is getting a good Cone Step Map working (looks better than parallax occlusion and runs faster). The patent expired 1 or 2 years ago, but it seems everybody has forgotten about it (except XNormal, which can generate the maps).

you are right, it expired in like Dec 2012! Time to make an implementation, thanks Kojack!

[Zonder]

Something I really want to work on is getting a good Cone Step Map working (looks better than parallax occlusion and runs faster). The patent expired 1 or 2 years ago, but it seems everybody has forgotten about it (except XNormal, which can generate the maps).

you are right, it expired in like Dec 2012! Time to make an implementation, thanks Kojack!

http://www.ogre3d.org/forums/viewtopic.php?t=42706

[c6burns]

Something I really want to work on is getting a good Cone Step Map working (looks better than parallax occlusion and runs faster). The patent expired 1 or 2 years ago, but it seems everybody has forgotten about it (except XNormal, which can generate the maps).

you are right, it expired in like Dec 2012! Time to make an implementation, thanks Kojack!

http://www.ogre3d.org/forums/viewtopic.php?t=42706

I've been tinkering with the GLSL one from gamedev ... this is great though as it will cover HLSL down the road without me having to put much thought into it

[Kojack]

Nullsquared's one was cool, but a deferred rendering compatible CSM would be good.

[Zonder]

I've not read the full topic but he says it was designed to run in his deferred pipeline

[Kojack]

Today is the 10th anniversary of me joining the Ogre forum. YAY!

Back in early 2003 my college started using a closed source commercial game engine called Jet, made by the brisbane games company Auran (Dark Reign, Trainz). Jet was partly funded by the local government, Auran got something like $4million to develop it, to put brisbane on the world game dev map. Unfortunately, it kind of sucked. But being a brisbane company with an ex Auran employee as our head of programming, we made a deal with them to give all our students a copy of Jet (normally something like $100 for a license).
I wasn't a fan of Jet, I was sure we could assemble an open source (or at least free) framework that would give more flexibility to students.
So in january of 2004 I found Ogre. It was version 0.13 back then and was nowhere near as powerful as today, but it still kicked ass compared to the alternatives. The other staff were skeptical. So I made a quick little demo: a primitive asteroids game. No shooting, no collision, no real game. But it had engine particles and a moving light, and ran fast with a few objects on screen. I showed that to the other teachers, and they were convinced. I've included a copy of the game (had to dig through my old hard drives to find it) below in case you want to see Ogre 0.13's config screen.

At the end of 2004 a team of our students, using a simple ogre framework I hacked together, won the best indie game award at the australian game developer conference. I guess we chose the right direction.
Another nail in the coffin of Jet was that it was Visual C++ 6 only. We upgraded all of our computers to Visual C++.Net and Auran refused to rebuild Jet to be compatible. They kept saying "Jet 2.0 will support VC++.Net", but it was never released.


My crappy asteroids-ish test from a decade ago:
https://dl.dropboxusercontent.com/u/254 ... teroids.7z
Made one day after joining the forum.

[insider]

Today is the 10th anniversary of me joining the Ogre forum. YAY!

Back in early 2003 my college started using a closed source commercial game engine called Jet, made by the brisbane games company Auran (Dark Reign, Trainz). Jet was partly funded by the local government, Auran got something like $4million to develop it, to put brisbane on the world game dev map. Unfortunately, it kind of sucked. But being a brisbane company with an ex Auran employee as our head of programming, we made a deal with them to give all our students a copy of Jet (normally something like $100 for a license).
I wasn't a fan of Jet, I was sure we could assemble an open source (or at least free) framework that would give more flexibility to students.
So in january of 2004 I found Ogre. It was version 0.13 back then and was nowhere near as powerful as today, but it still kicked ass compared to the alternatives. The other staff were skeptical. So I made a quick little demo: a primitive asteroids game. No shooting, no collision, no real game. But it had engine particles and a moving light, and ran fast with a few objects on screen. I showed that to the other teachers, and they were convinced. I've included a copy of the game (had to dig through my old hard drives to find it) below in case you want to see Ogre 0.13's config screen.

At the end of 2004 a team of our students, using a simple ogre framework I hacked together, won the best indie game award at the australian game developer conference. I guess we chose the right direction.
Another nail in the coffin of Jet was that it was Visual C++ 6 only. We upgraded all of our computers to Visual C++.Net and Auran refused to rebuild Jet to be compatible. They kept saying "Jet 2.0 will support VC++.Net", but it was never released.


My crappy asteroids-ish test from a decade ago:
https://dl.dropboxusercontent.com/u/254 ... teroids.7z
Made one day after joining the forum.

+1
If only all Universities had teachers like Kojack game development would be a breeze.

[Kojack]

Cavalier Projection. This is the same projection style as used in Ultima 7. Cabinet Projection is an alternative that is almost the same, except the y axis projects a shorter distance. It looks more realistic, but the aim of this test was cavalier.

The castle bsp map from Irrlicht being loaded in Ogre. I made a little demo that replicated the camera flythrough of the Irrlicht demo.

I have no idea what I was doing here. The sphere looks like it's running a shader where negative object space position is used as colour (red to the left, green down and blue back).

[nevarim]

nice

[cord]

This is not my work. But I wanted to showcase somebody else's work anyway because, well ...

render2

image55286.jpg (13.45 KiB) Viewed 15773 times

render1

image192605.jpg (31.75 KiB) Viewed 15773 times

https://github.com/D-POWER/vpath

VPath - Physically Based GPU Volumetric Path Tracer

VPath is a small CUDA path tracer that fully solves the radiative transfer equation, which makes it capable of rendering complex lighting effects in an unbiased manner.

Features:

• Full Volumetric Scattering of light in homogeneous media
• Accurate Monte Carlo Sub-surface Scattering
• GPU Acceleration using CUDA architecture
• Diffuse BRDF
• Fresnel Reflection/Refraction
• Sphere primitives

[Crashy]

A quick and dirty implementation of Light Propagation Volume + ambient probes.

For the moment the reflective shadow maps are done in the gpu but all the injection and propagation stuff is done on CPU side, because I wanted to understand all the math and also because I need to add the render to 3D texture support in the D3D11 renderer.

Once the 3D RTT is supported, I will port everything to GPU and add the blocker volume.
I don't have a lot of time to work on this for that right now, I have a game to finish. :p

[fractile]

I have nothing to do with this project, but I noticed it uses Ogre and looks good:

https://www.youtube.com/watch?v=6VD_w9B5lwk

Oh, and it looks fairly similar to my own Ogre game project

[dark_sylinc]

I thought this was the place to post these:

Very WIP pics of what's coming to Ogre 2.0 Final: HLMS (High Level Material System). No need for shader knowledge (but if you have, you can easily extend it), easy shadow mapping setup, many lights, point light shadow mapping using Dual Paraboloid maps, Physically Based (Cook Torrance BRDF by default), linear pipeline, automatic instancing and handling of skeletal animations through vertex shading.

[dark_sylinc]

More shots from the Physically Based Shading system:





The cubemap is from Humus' database Saint Peter's Basilica. Under Creative Commons Attribution 3.0 Unported License.

[Kojack]

Shiny.

[Klaim]

Shiny.

++

[Kojack]

Although it does mess with my head every time I see the debug panel with values like average fps: 15.2, best fps: 759.5, worst fps: 2230.
I'm guessing avg fps is really avg ms, but then best and worst don't make sense (if they are ms, best should be below average. If they are fps, worse should be below best).

[Zonder]

Although it does mess with my head every time I see the debug panel with values like average fps: 15.2, best fps: 759.5, worst fps: 2230.
I'm guessing avg fps is really avg ms, but then best and worst don't make sense (if they are ms, best should be below average. If they are fps, worse should be below best).

I noticed as well but was too busy looking at the shiny shiny.

What are the artifacts from on the edges on the sphere?