The Future of Triangles Part 5: The Technology of the Future

I've rewritten this post about four times now...let me try the brief version.

Basically, X-Plane is not an early adopter of graphics technology. Because of the nature of the rendering we do, we can directly benefit from "more of the same", e.g. if you simply gave me twice as many objects per second or twice as many polygons, we could make the sim look a lot nicer. So we don't need to adopt new graphics technologies until they're proven in games that need them more, like first person shooters. We're a small company with no influence on the industry, so we write the tightest message we can and use new features when the dust settles.

(From a utilization standpoint, we also provide the best graphics to the most people by using card features that are going to become wide spread, so it doesn't make sense for us to gamble on vendor-specific extensions that might not become available to everyone.)

With that in mind, there is some cool stuff that people are talking about that maybe someday we'll get to play with:

• Irregular Shadow Mapping - given a super-programmable card, you can create a rendering scheme that optimizes shadow map creation to remove artifacts.
• Out-of-order blending - the graphics card resorts incoming geometry so that all translucent geometry is drawn back to front. Doing this on the CPU is expensive (and in X-Plane's case, we often just don't get it right at all).
• Multiple dispatch to multiple targets. Even on a big multi-chip GPU (a lot of modern cards are two cards stuck together) the only render to one screen or texture at a time, even if there are a lot of parallel elements. This is good for a few big complex scenes but not good for lots of small scenes. I'd like to see all vendors support dispatch to multiple targets - this will make things like dynamic reflection via environment cube maps potentially a lot faster.
• Voxel Octrees. This is the one I hear a lot about - basically it's a change from 2-d to 3-d data structures on the graphics card to manage fast access to large chunks of graphics data. (Shadow maps, z-buffers, and environment maps are all more or less 2-d data structures.)
  

Will we see this? I don't know. Will Larabbee change everything? Who knows...Intel has to build a high-end graphics card to fight ATI and NV's attempt to get into supercomputing, but if they happen to also build a really nice video card, I can live with that. But I won't hold my breath - the titans need to duke it out without me!

The Future of Triangles Part 4: Pie in the Sky

Per-pixel lighting is something I hope to have in X-Plane soon.  A number of other features will take longer, and quite possibly might never happen.  This is the "pie in the sky" list - with this list, we're looking at higher hardware requirements, a lot of development time, and potential fundamental problems in the rendering algorithm!

High Dynamic Range (HDR) Lighting

HDR is a process whereby a program renders its scene with super bright and super dark regions, using a more detailed frame-buffer to draw.  When it comes time to show the image, some kind of "mapping" algorithm then represents that image using the limited contrast available on a computer monitor.  Typical approaches include:

• Scaling the brightness of the scene to mimic what our eyes do in dark or bright scenes.
• Creating "bloom", or blown out white regions, around very bright areas.

Besides creating more plausible lighting, the mathematics behind an HDR render would also potentially improve the look of lit textures when they are far away.  (Right now, a lit and dark pixel are blended to make semi-lit pixels when far away as the texture scales down.  If a lit pixel can be "super-bright" it will still look bright even after such blending.)

Besides development time, HDR requires serious hardware; the process of drawing to a framebuffer with the range to draw chews up a lot of GPU power, so HDR would be appropriate for a card like the GeForce 8800.

While there aren't any technical hurdles to stop us from implementing HDR, I must point out that, given a number of the "art" features of X-Plane like the sun glare, HDR might not be as noticeable as you'd think.  For example, our sun "glares" when you look at it (similar to an HDR trick), but this is done simply by us detecting the view angle and drawing the glare in.

Reflection Mapped Airplanes

Reflection maps are textures of the environment that are mapped onto the airplane to create the appearance of a shiny reflective surface.  We already have one reflection map: the sky and possibly scenery are mapped onto the water to create water reflections.

Reflection maps are very much possible, but they are also very expensive; we have to go through a drawing pass to prepare each one.  And reflection maps for 3-d objects like airplanes usually have to be done via cube maps, which means six environment maps!

There's a lot of room for cheating when it comes to environment maps.  For example: rendering environment maps with pre-made images or with simplified worlds.

Shadows

Shadows are the biggest missing feature in the sim's rendering path, and they are also by far the hardest to code.  I always hesitate to announce any in-progress code because there is a risk it won't work.  But in this case I can do so safely:

I have already coded global shadow maps, and we are not going to enable it in X-Plane.  The technique just doesn't work.  The code has been ripped out and I am going to have to try again with a different approach.

The problem with shadows is the combination of two unfortunate facts:

• The X-Plane world is very, very big and
• The human eye is very, very picky when it comes to shadows.

For reflections, we can cheat a lot -- if we don't get something quite right, the water waves hide a lot of sins.  (To work on the water, I have to turn the waves completely off to see what I' m doing!)  By comparison, anything less than perfect shadows really sticks out.

Shadow maps fail for X-Plane because it's a technology with limited resolution in a very large world.  At best I could apply shadows to the nearest 500 - 1000 meters, which is nice for an airport, but still pretty useless for most situations.

(Lest someone send the paper to me, I already tried "TSM" - X-Plane is off by about a factor of 10 in shadow map res; TSM gives us about 50% better texture use, which isn't even close.)

A user mentioned stencil shadow volumes, which would be an alternative to shadow maps.  I don't think they're viable for X-Plane; stencil shadow volumes require regenerating the shadow volumes any time the relative orientation of the shadow caster and the light source change; for a plane in flight this is every single plane.  Given the complexity of planes that are being created, I believe that they would perform even worse than shadow maps; where shadow maps run out of resolution, stencil shadow volumes would bury the CPU and PCIe bus with per-frame geometry.  Stencil shadow volumes also have the problem of not shadowing correctly for alpha-based transparent geometry.

(Theoretically geometry shaders could be used to generate stencil shadow volumes; in practice, geometry shaders have their own performance/throughput limitations - see below for more.)

Shadows matter a lot, and I am sure I will burn a lot more of my developer time working on them.  But I can also say that they're about the hardest rendering problem I'm looking at.

Dynamic Tessellation

Finally, I've spent some time looking at graphics-card based tessellation.  This is a process whereby the graphics card splits triangles into more triangles to make curved surfaces look more round.  The advantage of this would be lower triangle counts - the graphics card can split only the triangles that are close to the foreground for super-round surfaces.

The problem with dynamic tessellation is that the performance of the hardware is not yet that good.  I tried implementing tessellation using geometry shaders, and the performance is poor enough that you'd be better off simply using more triangles (which is what everyone does now).

I still have hopes for this; ATI's Radeon HD cards have a hardware tessellator and from what I've heard its performance is very good.  If this kind of functionality ends up in the DirectX 11 specification, we'll see comparable hardware on nVidia's side and an OpenGL extension.

(I will comment more on this later, but: X-Plane does not use DirectX - we use OpenGL.  We have no plans to switch from OpenGL to DirectX, or to drop support for Linux or the Mac.  Do not panic!  I mention DirectX 11 only because ATI and nVidia pay attention to the DirectX specification and thus functionality in DirectX tends to be functionality that is available on all modern cards.  We will use new features when they are available via OpenGL drivers, which usually happens within a few months of the cards being released, if not sooner.)

The Future of Triangles Part 3: X-Plane 9

Before I post anything to my blog saying what might happen, standard disclaimers:

• This blog represents my rambling about the directions I am considering for X-Plane's rendering engine.
• This blog is not a promise or commitment of any kind to deliver any particular feature.
• If I say I am looking at doing feature X, and feature X does not materialize, either in the near or far future, or, like, ever, consider this to be one big fat "I told you so."

With that in mind, I think the direction for lighting in version 9 is to introduce per-pixel lighting.

I don't know what other set of features we'll get with per-pixel lighting, but I am reviewing normal maps, specular maps, and the material attributes.  Per pixel lighting will mean smooth, round, shiny looking surfaces without using a huge number of triangles.

Now there are two sets of hardware that will not be able to support per-pixel lighting:

• Cards without pixel shaders.  (GeForce 2,3,4, Radeon 7000-9200.)  You might know your card does not have pixel shaders because the pixel shader check box is not available in the rendering settings.
• Cards with first generation shaders.  (This is the GeForce FX series and the Radeon 9500-9800 and X300-X600.)  These cards can actually perform per-pixel lighting, but they are so slow that per-pixel lighting will bring them below minimum frame-rate.

So unfortunately, there will be an authoring decision: add more triangles so that per-vertex lighting looks good, or use fewer triangles and rely on per-pixel lighting.  The decision will depend on what hardware you want to target at what performance level.  (For what it's worth, hardware that cannot support per-pixel lighting usually isn't very powerful, so there is something to be said for not having a lot of triangles on these lower end machines.)

The Future of Triangles Part 2: X-Plane 8

X-Plane 8 provides a useful baseline for rendering technology:

• It is finished and unchanging.
• Its use of shaders is very minimal, so even lower-end hardware can show the "X-plane 8 model" of lighting.
• X-Plane 8 rendering is completely supported in X-Plane 9.  (That is, turn off shaders, and OBJs should look the same in X-Plane 8 and 9.)

So what do we have, and is it any good?  Well, we have:

• Per-vertex lighting.  Lighting is calculated per vertex, and interpolated between vertices.
• Very limited materials.  Basically you can use attributes to set emissive lighting (so your day texture stays bright when back-lit, like taxiway signs) and shininess (to induce white specular hilites).  The shininess ratio isn't very flexible, but it does match what the built-in ACF shiny property does.
• Very fast vertex output within a batch.

I looked at some nice third party planes before writing this up, and one thing became clear: X-Plane can output a lot of vertices in an object if they are batched, and authors are using this aggressively.  The advantage of just using a lot of vertices is: curved surfaces look round, the errors that are induced by per-vertex lighting are less ugly, and the object looks the same everywhere (because this path isn't dependent on having pixel shaders).

The big weakness of the current situation is that you have to burn a lot of vertices to get close to per-pixel lighting, particularly for very shiny surfaces.  I saw at least one plane (I do not recall who authored it) that just had more triangles in the engine nacelles than you could imagine.  They look beautiful even in X-Plane 8 - great specular hilites.  But that eats into your vertex budget pretty severely - it's not a technique that you could use for every static airplane on a tarmac at LAX.

Bad Alloc, New Patch, New Installers

The 921 patch is now available. This fixes the bad-alloc problems relating to complex airports.

Please note: if you have a bad-alloc crash, it could be because you are out of memory. Make sure you have virtual memory turned on, your page file is large enough, your disk isn't full, etc. If you have a bad alloc error, try the sim without third party add-ons to see if you really are running out of memory. If you are running Vista or XP, use /3GB or the BCD - see this for more info. Basically when you are running out of memory, you either can crash on "bad alloc" if we need memory for the CPU or "we ran out of video card memory" if we can't map geometry into virtual memory.

If you get a crash with 921 on OS X, please let me know by email! I've seen one of these reports.

I also have cuts of the new installer/updater suite, version 2.05 - if you are going to update, tryo ne of these:

http://dev.x-plane.com/update/installers9/stage/

The main features of the new installer are:

• Support for French, Spanish, Italian, German, and Russian.
• Clearer colors on the world map.
• Scanning the global scenery folder to add/remove scenery is much faster.

I recommedn the new installers for that last point alone.

The Future of Triangles Part 1: Terminology

The triangle is at the heart of 3-d modeling - but before we discuss what might become of the triangle, we need terminology.

• Per-vertex lighting.  This means that the brightness of the model (a function of the sun and camera position, etc.) is calculated for each vertex in the model, and then crudely interpolated between the vertices to light the pixels.
• Per-pixel lighting.  This means that the brightness of the model (a function of the sun and camera position, etc.) is calculated for every pixel on the screen separately.
• Tessellation.  This is the process of splitting a triangle into a number of smaller triangles, increasing the number of vertices in a model.
• Specular lighting.  The specular lighting component is an extra amount of brightness that you get when the angle from the sun to the model to your eye is very small.  (That is, if the model was a mirror and you could see the sun by looking at a certain location, then that location would have a bright "specular hilite".)
• Normal map.  A normal map is a texture that describes the way light bounces off a surface.  This is one way to do "bump mapping".  This tutorial shows a pretty good example of how normal maps work.  (The earth orbit textures in version 9 use normal maps to create "bumpy" mountains when pixel shaders are in use.)
• Specular map.  A specular map is a texture that describes how strong the specular component of the lighting model appears for a given textured location.  Here's another tutorial that explains it.
• Environment Map.  An environment map is a texture that represents the world around an object, used to simulate reflections.  Here's another blender tutorial that explains it better than I.  (The reflective water in X-Plane 9 is effectively using a dynamic environment map created by taking a picture of part of the sim's world every frame,)
• Material attributes.  These are OBJ attributes that change the lighting model.  For example, ATTR_shiny_rat changes the lighting model so that specular hilites appear.
• Batch.  A batch is a single set of triangles sent to the graphics card without any change of mode.  Basically every TRIS command in an OBJ becomes a batch; submitting a batch requires the CPU, but submitting a bigger batch (more triangles) does not require more CPU.

That's enough vocabulary to describe just about everything that is happening now, will be happening in the future, as well as some pie-in-the-sky stuff. :-)

Patches, Patches, Patches

There are a lot of patches coming out soon:

• 921rc1 - this is a patch on 920 to fix the bad-alloc crash.
  
• 921rc2 - same as 921rc1, but with some improvements to the translations.

Why two patches in a row? Basically we did 921 rc1 a while ago, but Austin's net connection got hit by lightning. He's on a, um, "borrowed" wifi signal right now, and it isn't real strong, so the 921 patch is just taking forever to post. It's equally likely that we skip straight to 921 rc2 because we can get the translation improvements done before we even get the patch up on the server.

Now regarding French: first I must say thank you to Daniel, who is helping us with the translation...we've been behind schedule, and his helping us with a big block of strings is a real life saver. I must also thank my wife, who was also up pretty late working on some strings. 921 rc2 will address all other languages; we'll do a new patch once we have complete a French translation. If this is soon, that could be 922 (another patch). If it takes a while, it'll be part of 930 (which will have new features). Time will tell!

There is also an update coming out for the iphone soon - improvements to the UI and some framerate improvements. I believe that the iTunes store should make downloading the update pretty seamless.

After all this patching, we're on to 930. I'm not entirely sure what will be in 930 yet, but I have already made some rendering engine changes for 930 that improve framerate.

EDIT: 921r1 is now up as a beta - so you can get it if you set your installer to "get betas". If you have the bad_alloc crash, try this beta.

I Can't Talk Now, I'm Flying a Plane!

Traditionally, a pilot's priorities are: aviate, navigate, communicate.

But that might not be true for X-Plane for the iPhone.

It's real! And it pretty much is X-Plane - there really are OBJs and DSFs in there, as well as an ACF model, all tuned for the iPhone.

In the next few posts I'll blog a little bit about the impact of doing an iPhone port on scenery development. The iPhone is an embedded device; if you go digging for system specs you'll see that it's a very different beast from the desktop. The porting process really helped me understand the problems of the rendering engine a lot better, and some of the techniques we developed for the iPhone are proving useful for desktop machines as well.

Hardware Buckets

Yesterday I went off on a rant about how the OpenGL 3.0 spec doesn't matter a lot because OpenGL grows via a la carte extension.  And I mentioned that this creates a dilemma for anyone designing a rendering engine that has to run on a wide range of hardware, with very different underlying capabilities.

Back in the old days of X-Plane 6, there wasn't a lot of variety in hardware.  Some cards were faster than others, but the only real difference in capability was how many texture units you had. X-Plane's job was simple...

• First we have a runway texture.
• Still got a second texure unit?  Great!  Skid marks!
• Still got a third texture unit?  Great!  Landing light!
• Got another one?  Etc...

Other than texture stacking, there wasn't much to do.

Since then the rendering engine has become a lot more complex, as have OpenGL cards.  To try to keep the combinations down, I tried to use a "bucketing" strategy for X-Plane 9.  The idea of bucketing is to group cards into major buckets based on whole sets of common functionality, so that we only have to test a few configurations (the "low end" bucket and "high end" bucket), rather than having to test every single card for its own unique set of features.

The obvious bucketing choice was pixel shaders - given a card with shaders and a few other features, we can render all of the new effects.  A card without shaders basically gets X-Plane 8.

So what went wrong?  Driver compatibility, that's what.  Ideally we don't want to allow every single underlying rendering engine feature to be turned on or off individually because the combinations are uncontrollable.*  But in practice, being able to turn features on and off is necessary to trouble-shoot drivers that don't always do what we expect them to.

With the GeForce 8800 and Radeon HD cards, there is a potential third bucket for DirectX-10 capable cards, capable of significantly more advanced pixel shading effects.  But time will tell whether we can actually make a bucket, or we have to look at each feature individually.  My suspicion is that even if we organize the shipping features into buckets, we'll have to support a lot of combinations under the hood just to trouble-shoot our own application.

*Example: cross a standard vs. oversized panel with the presence or absence of advanced framebuffer blending, crossed with whether render-to-texture works. That's 8 possible ways just to render the 3-d cockpit panel texture.  Cross that with panel regions and 3-d cockpits and the new panel spotlighting and you have 64 configurations. Ouch!

OpenGL 3.0

A few people have asked me about OpenGL 3.0 - and if you read some of the news coverage of the OpenGL community, you'd think the sky was falling.  In particular, a bunch of OpenGL developers posted their unhappiness that the spec had prioritized compatibility over new features.  Here's my take on OpenGL 3.0:

First, major revisions to the OpenGL specification simply don't matter that much.  OpenGL grows by extensions - that is, incremental a la carte additions to what OpenGL can do.  Eventually the more important ones become part of a new spec.  But the extensions almost always come before the spec.  So what really matters for OpenGL is: are extensions coming out quickly enough to support new hardware to its fullest capacity?  Are the extensions cross-vendor so that applications don't have to code to specific cards?  Is the real implementation of high quality?

So how are we doing with extensions?  My answer would be: "okay".  When the GeForce 8800 first came out, the OpenGL extensions that provide DirectX 10-like functionality were NVidia-specific.  Since then, it has become clear that all of this functionality will make it into cross-platform extensions, the core spec, or some of each.  But for early adopters there was a difficult point where there was no guarantee that ATI and NVidia's DirectX 10 features would be accessible through the same extensions.

(This was not as much of an issue for DX9-like features, e.g. the first generation of truly programmable cards.  NVidia had a bunch of proprietary additional extensions designed to make the GeForce FX series less slow, but the basic cross-platform shader interface was available everywhere.)

Of more concern to me is the quality of OpenGL implementations - and for what it's worth, I have not found cases where a missing API is standing between me and the hardware.  A number of developers have posted concern that OpenGL drivers are made too complex (and thus too unreliable or slow or expensive to maintain) because the OpenGL spec has too many old features.  I have to leave that to the driver writers themselves to decide!  But when we profile X-Plane, we either see a driver that's very fast, or a driver that's being slow on a modern code path, in a way that is simply buggy.

Finally, I may be biased by the particular application I work on, but new APIs that replace the old ones don't do me a lot of good unless they get me better performance.  X-Plane runs on a wide range of hardware; we can't drop cards that don't support the latest and greatest APIs.  So let's imagine that OpenGL 3.0 contained some of the features whose absence generated such fury.  Now if I want to take advantage of these features, I need to code that part of the rendering engine twice: once with the new implementation and once with the old implementation.  If that doesn't get me better speed, I don't want the extra code and complexity and wider matrix of cases to debug and test.

In particular, the dilemma for anyone designing a renderer on top of modern OpenGL cards is: how to create an implementation that is efficient on hardware whose capabilities is so different.  I'll comment on that more in my next post.  But for the purposes of OpenGL 3.0: I'm not in a position to drop support for old implementations of the GL, so it doesn't bother me at all that the spec doesn't drop support either.

The real test for OpenGL is not when a major revision is published; it is when the next generation of hardware comes out.

MeshTool vs. Draped Polygons

An author asked me some questions that I think are so important that I'll blog the answers:

• The new texture paging system (LOAD_CENTER) works for both terrain textures (.ter files) and draped polygons (.pol files). You do not have to use draped polygons to get texture paging - you can use paging in a base mesh!
• Orthophoto terrain via a (.ter) file is by far the preferred method for orthophoto sceneries - it is a vastly better option than draped polygons. Draped polygons are horribly wasteful of hardware resources, and should really only be used for tiny areas, e.g. airport surface areas. If you are using even a moderate amount of orthophotos, make a base mesh!
• MeshTool is the future of photo scenery, and will continue to be the way to make high performance orthophoto meshes for X-Plane.

The future of MeshTool is bug fixes, a richer syntax, and some day maybe a UI front-end.