Erik McClure

So I have this dll that’s doing a bunch of horrible WinAPI calls for me. It’s designed to abstract away all the pain and unholy functions feeding on innocent blood. Little did I know that trying to cage WinAPI into a more manageable corner would be my undoing.

The window is created and assigned a WndProc callback function inside this DLL, as per the various absurd requirements involving how you create windows in Windows. Everything works just fine and dandy until the application window suddenly closes, an error “ding” is heard, and the program silently crashes without any sort of error message whatsoever.

What just happened?

Well, I’m afraid you just triggered an assertion. But instead of doing what an assertion is supposed to do, which is immediately crash the program so you know what the fuck just happened, it instead somehow manages to create an infinite callback loop caused by the operating system actually trying to assign the assertion dialog to the window. You know, the same window that was part of an application that’s supposed to be in the middle of CRASHING?! This causes another message to be sent, thus causing another assertion to be violated, et cetera until the stack overflows, except it doesn’t actually tell you the stack overflows, it just crashes with absolutely no explanation. Even while in a debugger, which is truly amazing.

Specifically, the assertion triggers a dialog box to be displayed, but this dialog box forcibly sends another WM_CANCELMODE message to the application, apparently completely bypassing the entire point of having a message queue, because the program never gets to go back to it. It’s WndProc is simply called with WM_CANCELMODE because fuck you, and so if your assertion fails when you’re processing WM_CANCELMODE, it will simply do this over and over and over until the entire window tree collapses in on itself from the intense gravitational pull of astronomical amounts of stupidity.

The resulting black hole sucks in all nearby sources of sanity, which include some sort of error message about the stack overflowing, or perhaps the actual fucking assertion error, and then explodes, leaving you without a goddamn clue about what just happened. This is usually followed by copious amounts of screaming, then disbelief, then finally assuming the fetal position and praying for forgiveness for attempting to touch the windows API, as the Microsoft gods laugh and drag another developer’s soul into their black circle of hell.

Somehow, every time I touch the windows API, it always ends with me curled into a ball, moaning “what the fuck” over and over again while crying.

Technology tends to serve one of two purposes - to make us more efficient at some task, or to entertain us in our resulting free time. However, when we fixate on productivity to the exclusion of everything else, we often forget about the big picture. Perhaps the best example of this are people insisting that real coders need to use Vim to be productive due to it’s unmatched text editing powers.

This is totally absurd. I spend maybe 10% of my time actually writing code, 30% debugging that code, and the remaining 60% trying to solve a problem. Even if I could write all my code instantly, I have only improved my productivity by 10%. I’ve found that changing my code patterns to let me catch bugs faster and earlier has had a much more significant impact on my coding speed, because taking a chunk out of 30% has a much greater effect on my overall productivity.

But what about the 60%? I’m sure I could make some of that go away with more powerful visualization tools and intense mental training, but when I hit a problem that’s simply really hard to solve, nothing short of a cybernetic implant that makes my brain bigger is going to make a dent in how much time I spend thinking about something unless I want to make a stupid mistake and regret it later.

The issue that’s arising from our hyperproductive tools is that our productivity is beginning to outstrip our ability to think. We are so productive we can’t think fast enough to utilize it. Vim may be the most amazing text editor ever, but it doesn’t matter because I spend more time thinking than I do actually editing text. We’re so focused on making everyone super productive we seem to forget that we are beginning to receive diminishing returns from some of our efforts.

One consequence of this is that, obviously, we should be focusing on tools to help us think faster. We need to do profiling of people’s lives to find the chokepoints and focus on those instead of doing the equivalent of micro-optimizations in C code that’s only called every 5 minutes. That, however, does not concern me. What does concern me is the repeated mistakes futurists make when attempting to predict the future of technology.

This Microsoft video is a superb example of good technology predictions implemented in the worst way possible. The entire video treats human beings as machinery that needs to complete various tasks in the quickest way possible, instead of recognizing that human beings are, in fact, people. Many of the human interactions feel fake because all the interactions are treated simply as tasks that must be completed efficiently, regardless of how beneficial the time saved actually is. Productivity is not important, the way it feels is important.

Futuristic architecture often makes this same mistake, creating cold, bland environments that, while they do feel futuristic, are not things anyone would want to live or work in. We need environments that feel warm, inviting, and natural. When building futuristic environments, we must be extremely careful about where the future is peeking in, because there is such a thing as too much future in one room.

We make things look like wood simply because we like how wood looks, not because we need to build anything out of wood anymore. We like trees growing around our houses. We make our lights look like the sun instead of actually being white. We are constantly making arbitrary choices that have nothing to do with productivity and everything to do with making us feel comfortable. Until designers recognize this and stop sucking the life out of everything in an effort to make it more “productive”, they will inevitably be shunned in favor of slightly less efficient, but more inviting designs.

Design is an optimization problem that must maximize both productivity and feel, not one or the other. Some people actually like color in their IDEs and webpages that consist of more than flat text, faint lines and whitespace.

[ 1 · 2 · 3 · 4 · 5 · 6 · 7 ]

If you are familiar with C#, you should be familiar with the difference between C#’s struct and class declarations. Namely, a struct is a value type and a class is a reference type, meaning that if you pass a struct to a function, its default behavior is for the entire struct to be copied into the function’s parameter, so any modifications made to it won’t affect whatever was passed in. On the flip side, a class is a reference value, so a reference is passed into the function, and any changes made to that reference will be reflected in the object that was originally passed into the function.

// Takes an integer, or a basic value type
public static int add(int v)
{
  v+=3;
  return 4+v;
}

public struct Oppa
{
  public string gangnam;
}

// Takes a struct, or a complex value type
public static Oppa style(Oppa g)
{
  g.gangnam="notstyle";
  return g;
}

public class Psy
{
  public int style;
}

// Takes a class, or a reference type
public static void change(Psy psy)
{
  psy.style=5;
}

// Takes an integer, but forces it to be passed by reference instead of by value.
public static int addref(ref int v)
{
  v+=3;
  return 4+v;
}

int a = 0;
int b = add(a);
// a is still 0
// b is now 7

int c = addref(a);
// a is now 3, because it was passed by reference
// c is now 7

Oppa s1;
s1.gangnam="style";
Oppa s2 = style(s1);
//s1.gangnam is still "style"
//s2.gangnam is now "notstyle"

Psy psy = new Psy();
psy.style=0;
change(psy);
// psy.style is now 5, because it was passed by reference

// Integer passed by value
int add(int v)
{
  v+=3;
  return 4+v;
}

class Psy
{
public:
  int style;
};

// Class passed by value
Psy change(Psy psy)
{
  psy.style=5;
  return psy;
}

// Integer passed by reference
int addref(int& v)
{
  v+=3;
  return 4+v;
}

// Class passed by reference
Psy changeref(Psy& psy)
{
  psy.style=5;
  return psy;
}

int horse = 2;
int korea = add(horse);
// horse is still 2
// korea is now 9

int horse2 = 2;
int korea2 = addref(horse2);
// horse2 is now 5
// korea2 is now 9

Psy psy;
psy.style = 0;
Psy ysp = change(psy);
// psy.style is still 0
// ysp.style is now 5

Psy psy2;
psy2.style = 0;
Psy ysp2 = changeref(psy2);
// psy2.style is now 5
// ysp2.style is also 5

class myString
{
public:
  // The copy constructor, which copies the string over instead of copying the pointer
  myString(const myString& copy)
  {
    size_t len = strlen(copy._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,copy._str,sizeof(char)*len);
  }
  // Normal constructor
  myString(const char* str)
  {
    size_t len = strlen(str);
    _str=new char[len];
    memcpy(_str,str,sizeof(char)*len);
  }
  // Destructor that deallocates our string
  ~myString()
  {
    delete [] _str;
  }

private:
  char* _str;
};

int a = 3;
int b = 2;
int& ra = a;
int* pa = &a;

b = a; //b is now 3
a = 0; //b is still 3, but a is now 0
b = ra; // b is now 0
a = 5; // b is still 0 but now a is 5
b = *pa; // b is now 5
b = 8; // b is now 8 but a is still 5

ra = b; //a is now 8! This assigns b's values to ra, it does NOT change the reference!
ra = 9; //a is now 9, and b is still 8! ra STILL refers to a, and NOTHING can change that.

pa = &b; // Now pa points to to b
a = *pa; // a is now 8, because pointers CAN be changed.
*pa = 7; // Now b is 7, but a is still 8

int*& rpa = pa; //Now we have a reference to a pointer (C++11)
//rpa = 5; // COMPILER ERROR, rpa is a reference to a POINTER
int** ppa = &pa;
//rpa = ppa; // COMPILER ERROR, rpa is a REFERENCE to a pointer, not a pointer to a pointer!
rpa = &a; //now pa points to a again. This does NOT change the reference!
b = *pa; // Now b is 8, the same as a.

class someClass
{
someClass operator =(anything b); // me=other
someClass operator +(anything b); // me+other
someClass operator -(anything b); // me-other
someClass operator +(); // +me
someClass operator -(); // -me (negation)
someClass operator *(anything b); // me*other
someClass operator /(anything b); // me/other
someClass operator %(anything b); // me%other
someClass& operator ++(); // ++me
someClass& operator ++(int); // me++
someClass& operator --(); // --me
someClass& operator --(int); // me--
// All operators can TECHNICALLY return any value whatsoever, but for many of them only certain values make sense.
bool operator ==(anything b); 
bool operator !=(anything b);
bool operator >(anything b);
bool operator <(anything b);
bool operator >=(anything b);
bool operator <=(anything b);
bool operator !(); // !me
// These operators do not usually return someClass, but rather a type specific to what the class does.
anything operator &&(anything b); 
anything operator ||(anything b);

anything operator ~();
anything operator &(anything b);
anything operator |(anything b);
anything operator ^(anything b);
anything operator <<(anything b);
anything operator >>(anything b);
someClass& operator +=(anything b); // Should always return *this;
someClass& operator -=(anything b);
someClass& operator *=(anything b);
someClass& operator /=(anything b);
someClass& operator %=(anything b);
someClass& operator &=(anything b);
someClass& operator |=(anything b);
someClass& operator ^=(anything b);
someClass& operator <<=(anything b);
someClass& operator >>=(anything b);
anything operator [](anything b); // This will almost always return a reference to some internal array type, like myElement&
anything operator *();
anything operator &();
anything* operator ->(); // This has to return a pointer or some other type that has the -> operator defined.

anything operator ->*(anything a);
anything operator ()(anything a1, U a2, ...);
anything operator ,(anything b);
operator otherThing(); // Allows this class to have an implicit conversion to type otherThing
void* operator new(size_t x); // These are called when you write new someClass()
void* operator new[](size_tx); // new someClass[num]
void operator delete(void*x); // delete pointer_to_someClass
void operator delete[](void*x); // delete [] pointer_to_someClass

};

// These are global operators that behave more like C# operators, but must be defined outside of classes, and a few operators do not have global overloads, which is why they are missing from this list. Again, operators can technically take or return any value, but normally you only override these so you can handle some other type being on the left side.
someClass operator +(anything a, someClass b);
someClass operator -(anything a, someClass b);
someClass operator +(someClass a);
someClass operator -(someClass a);
someClass operator *(anything a, someClass b);
someClass operator /(anything a, someClass b);
someClass operator %(anything a, someClass b);
someClass operator ++(someClass a);
someClass operator ++(someClass a, int); // Note the unnamed dummy-parameter int - this differentiates between prefix and suffix increment operators.
someClass operator --(someClass a);
someClass operator --(someClass a, int); // Note the unnamed dummy-parameter int - this differentiates between prefix and suffix decrement operators.

bool operator ==(anything a, someClass b);
bool operator !=(anything a, someClass b);
bool operator >(anything a, someClass b);
bool operator <(anything a, someClass b);
bool operator >=(anything a, someClass b);
bool operator <=(anything a, someClass b);
bool operator !(someClass a);
bool operator &&(anything a, someClass b);
bool operator ||(anything a, someClass b);

someClass operator ~(someClass a);
someClass operator &(anything a, someClass b);
someClass operator |(anything a, someClass b);
someClass operator ^(anything a, someClass b);
someClass operator <<(anything a, someClass b);
someClass operator >>(anything a, someClass b);
someClass operator +=(anything a, someClass b);
someClass operator -=(anything a, someClass b);
someClass operator *=(anything a, someClass b);
someClass operator /=(anything a, someClass b);
someClass operator %=(anything a, someClass b);
someClass operator &=(anything a, someClass b);
someClass operator |=(anything a, someClass b);
someClass operator ^=(anything a, someClass b);
someClass operator <<=(anything a, someClass b);
someClass operator >>=(anything a, someClass b);
someClass operator *(someClass a);
someClass operator &(someClass a);

someClass operator ->*(anything a, someClass b);
someClass operator ,(anything a, someClass b);
void* operator new(size_t x);
void* operator new[](size_t x);
void operator delete(void* x);
void operator delete[](void*x);

class myString
{
public:
  // The copy constructor, which copies the string over instead of copying the pointer
  myString(const myString& copy)
  {
    size_t len = strlen(copy._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,copy._str,sizeof(char)*len);
  }
  // Normal constructor
  myString(const char* str)
  {
    size_t len = strlen(str);
    _str=new char[len];
    memcpy(_str,str,sizeof(char)*len);
  }
  // Destructor that deallocates our string
  ~myString()
  {
    delete [] _str;
  }

  // Assignment operator, does the same thing the copy constructor does, but also mimics the destructor by deleting _str. NOTE: It is considered bad practice to call the destructor directly. Use a Clear() method or something equivalent instead.
  myString& operator=(const myString& right)
  {
    delete [] _str;
    size_t len = strlen(right._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,right._str,sizeof(char)*len);
  }

private:
  char* _str;
};

Move semantics are designed to solve the following problem. If we have a series of dynamic string objects being concatenated, with normal copy constructors we run into a serious problem:

std::string result = std::string("Oppa") + std::string(" Gangnam") + std::string(" Style") + std::string(" by") + std::string(" Psy");
// This is evaluated by first creating a new string object with its own memory allocation, then deallocating both " by" and " Psy" after copying their contents into the new one
//std::string result = std::string("Oppa") + std::string(" Gangnam") + std::string(" Style") + std::string(" by Psy");
// Then another new object is made and " by Psy" and " Style" are deallocated
//std::string result = std::string("Oppa") + std::string(" Gangnam") + std::string(" Style by Psy");
// And so on and so forth
//std::string result = std::string("Oppa") + std::string(" Gangnam Style by Psy");
//std::string result = std::string("Oppa Gangnam Style by Psy");
// So just to add 5 strings together, we've had to allocate room for 5 additional strings in the middle of it, 4 of which are then simply deallocated!

class myString
{
public:
  // The copy constructor, which copies the string over instead of copying the pointer
  myString(const myString& copy)
  {
    size_t len = strlen(copy._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,copy._str,sizeof(char)*len);
  }
  // Move Constructor
  myString(myString&& mov)
  {
    _str = mov._str;
    mov._str=NULL;
  }
  // Normal constructor
  myString(const char* str)
  {
    size_t len = strlen(str);
    _str=new char[len];
    memcpy(_str,str,sizeof(char)*len);
  }
  // Destructor that deallocates our string
  ~myString()
  {
    if(_str!=NULL) // Make sure we only delete _str if it isn't NULL!
      delete [] _str;
  }

  // Assignment operator, does the same thing the copy constructor does, but also mimics the destructor by deleting _str. NOTE: It is considered bad practice to call the destructor directly. Use a Clear() method or something equivalent instead.
  myString& operator=(const myString& right)
  {
    delete [] _str;
    size_t len = strlen(right._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,right._str,sizeof(char)*len);
    return *this;
  }

private:
  char* _str;
};

The idea behind a move constructor is that, instead of copying the values into our object, we move them into our object, setting the source to some NULL value. Notice that this can only work for pointers, or objects containing pointers. Integers, floats, and other similar types can’t really be “moved”, so instead their values are simply copied over. Consequently, move semantics is only beneficial for types like strings that involve dynamic memory allocation. However, because we must set the source pointers to 0, that means we can’t use const myString&&, because then we wouldn’t be able to modify the source pointers! This is why a move constructor is declared without a const modifier, which makes sense, since we intend to modify the object.

But wait, just like a copy constructor has an assignment copy operator, a move constructor has an equivalent assignment move operator. Just like the copy assignment, the move operator behaves exactly like the move constructor, but must destroy the existing object beforehand. The assignment move operator is declared like this:

myString& operator=(myString&& right)
  {
    delete [] _str;
    _str=right._str;
    right._str=0;
    return *this;
  }

Move semantics can be used for some interesting things, like unique pointers, that only have move semantics - by disabling the copy constructor, you can create an object that is impossible to copy, and can therefore only be moved, which guarantees that there will only be one copy of its contents in existence. std::unique_ptr is an implementation of this provided in C++0x. Note that if a data structure requires copy semantics, std::unique_ptr will throw a compiler error, instead of simply mysteriously failing like the deprecated std::autoptr.

There is an important detail when you are using inheritance or objects with move semantics:

class Substring : myString
{
  Substring(Substring&& mov) : myString(std::move(mov))
  {
    _sub = std::move(mov._sub);
  }

  Substring& operator=(Substring&& right)
  {
    myString::operator=(std::move(right));
    _sub = std::move(mov._sub);
    return *this;
  }

  myString _sub;
};

There’s some other weird things you can do with move semantics. This most interesting part is the strange behavior of && when it is appended to existing references.

• A& & becomes A&
• A& && becomes A&
• A&& & becomes A&
• A&& && becomes A&&

By taking advantage of the second and fourth lines, we can perform perfect forwarding. Perfect forwarding allows us to pass an argument as either a normal reference (A&) or an rvalue (A&&) and then forward it into another function, preserving its status as an rvalue or a normal reference, including whether or not it’s const A& or const A&&. Perfect forwarding can be implemented like so:

template<typename U>
void Set(U && other)
{
  _str=std::forward<U>(other);
}

class myString
{
public:
  // The copy constructor, which copies the string over instead of copying the pointer
  myString(const myString& copy)
  {
    size_t len = strlen(copy._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,copy._str,sizeof(char)*len);
  }
  // Move Constructor
  myString(myString&& mov)
  {
    _str = mov._str;
    mov._str=NULL;
  }
  // Normal constructor
  myString(const char* str)
  {
    size_t len = strlen(str);
    _str=new char[len];
    memcpy(_str,str,sizeof(char)*len);
  }
  // Destructor that deallocates our string
  ~myString()
  {
    if(_str!=NULL) // Make sure we only delete _str if it isn't NULL!
      delete [] _str;
  }
  void Set(myString&& str)
  {
    _set<myString&&>(std::move(str));
  }
  void Set(const myString& str)
  {
    _set<const myString&>(str);
  }


  // Assignment operator, does the same thing the copy constructor does, but also mimics the destructor by deleting _str. NOTE: It is considered bad practice to call the destructor directly. Use a Clear() method or something equivalent instead.
  myString& operator=(const myString& right)
  {
    delete [] _str;
    size_t len = strlen(right._str)+1; //+1 for null terminator
    _str=new char[len];
    memcpy(_str,right._str,sizeof(char)*len);
    return *this;
  }

private:
  template<typename U>
  void _set(U && other)
  {
    _str=std::forward<U>(other);
  }

  char* _str;
};

I see a lot of people get excited about extreme concurrency in modern hardware bringing us closer to the magical holy grail of raytracing. It seems that everyone thinks that once we have raytracing, we can fully simulate entire digital worlds, everything will be photorealistic, and graphics will become a “solved problem”. This simply isn’t true, and in fact highlights several fundamental misconceptions about the problems faced by modern games and other interactive media.

For those unfamiliar with the term, raytracing is the process of rendering a 3D scene by tracing the path of a beam of light after it is emitted from a light source, calculating its properties as it bounces off various objects in the world until it finally hits the virtual camera. At least, you hope it hits the camera. You see, to be perfectly accurate, you have to cast a bajillion rays of light out from the light sources and then see which ones end up hitting the camera at some point. This is obviously a problem, because most of the rays don’t actually hit the camera, and are simply wasted. Because this brute force method is so incredibly inefficient, many complex algorithms (such as photon-mapping and Metropolis light transport) have been developed to yield approximations that are thousands of times more efficient. These techniques are almost always focused on attempting to find paths from the light source to the camera, so rays can be cast in the reverse direction. Some early approximations actually cast rays out from the camera until they hit an object, then calculated the lighting information from the distance and angle, disregarding other objects in the scene. While highly efficient, this method produced extremely inaccurate results.

It is with a certain irony that raytracing is touted as being a precise, super-accurate rendering method when all raytracing is actually done via approximations in the first place. Pixar uses photon-mapping for its movies. Most raytracers operate on stochastic sampling approximations. We can already do raytracing in realtime, if we get approximate enough, it just looks boring and is extremely limited. Graphics development doesn’t just stop when someone develops realtime raytracing, because there will always be room for a better approximation.

1. Photorealism

The meaning of photorealism is difficult to pin down, in part because the term is inherently subjective. If you define photorealism as being able to render a virtual scene such that it precisely matches a photo, then it is almost impossible to achieve in any sort of natural environment where the slightest wind can push a tree branch out of alignment.

This quickly gives rise to defining photorealism as rendering a virtual scene such that it is indistinguishable from a photograph of a similar scene, even if they aren’t exactly the same. This, however, raises the issue of just how indistinguishable it needs to be. This seems like a bizarre concept, but there are different degrees of “indistinguishable” due to the differences between people’s observational capacities. Many people will never notice a slightly misaligned shadow or a reflection that’s a tad too bright. For others, they will stand out like sore thumbs and completely destroy their suspension of disbelief.

We have yet another problem in that the entire concept of “photorealism” has nothing to do with how humans see the world in the first place. Photos are inherently linear, while human experience a much more dynamic, log-based lighting scale. This gives rise to HDR photography, which actually has almost nothing to do with the HDR implemented in games. Games simply change the brightness of the entire scene, instead of combining the brightness of multiple exposures to brighten some areas and darken others in the same photo. If all photos are not created equal, then exactly which photo are we talking about when we say “photorealistic”?

2. Complexity

Raytracing is often cited as allowing an order of magnitude more detail in models by being able to efficiently process many more polygons. This is only sort of true in that raytracing is not subject to the same computational constraints that rasterization is. Rasterization must render every single triangle in the scene, whereas raytracing is only interested in whether or not a ray hits a triangle. Unfortunately, it still has to navigate through the scene representation. Even if a raytracer could handle a scene with a billion polygons efficiently, this raises completely unrelated problems involving RAM access times and cache pollution that suddenly become actual performance bottlenecks instead of micro-optimizations.

In addition, raytracing approximation algorithms almost always take advantage of rays that degrade quickly, such that they can only bounce 10-15 times before becoming irrelevant. This is fine and dandy for walking around in a city or a forest, but what about a kitchen? Even though raytracing is much better at handling reflections accurately, highly reflective materials cripple the raytracer, because now rays are bouncing hundreds of times off a myriad of surfaces instead of just 10. If not handled properly, it can absolutely devastate performance, which is catastrophic for game engines that must maintain constant render times.

3. Scale

How do you raytrace stars? Do you simply wrap a sphere around the sky and give it a “star” material? Do you make them all point sources infinitely far away? How does this work in a space game, where half the stars you see can actually be visited, and the other half are entire galaxies? How do you accurately simulate an entire solar system down to the surface of a planet, as the Kerbal Space Program developers had to? Trying to figure out how to represent that kind of information in a meaningful form with only 64 bits of precision, if you are lucky, is a problem completely separate from raytracing, yet of increasingly relevant concern as games continue to expand their horizons more and more. How do we simulate an entire galaxy? How can we maintain meaningful precision when faced with astronomical scales, and how does this factor in to our rendering pipeline? These are problems that arise in any rendering pipeline, regardless of what techniques it uses, due to fundamental limitations in our representations of numbers.

4. Materials

Do you know what methane clouds look like? What about writing an aerogel shader? Raytracing, by itself, doesn’t simply figure out how a given material works, you have to tell it how each material behaves, and its accuracy is wholly dependent on how accurate your description of the material is. This isn’t easy, either, it requires advanced mathematical models and heaps of data collection. In many places we’re actually still trying to figure out how to build physically correct material equations in the first place. Did you know that Dreamworks had to rewrite part of their cloud shader¹ for How To Train Your Dragon? It turns out that getting clouds to look good when your character is flying directly beneath them with a hand raised is really hard.

This is just for common lighting phenomena! How are you going to write shaders for things like pools of magic water and birefringent calcite crystals? How about trying to accurately simulate circular polarizers when most raytracers don’t even know what polarization is? Does being photorealistic require you to simulate the Tyndall Effect for caustics in crystals and particulate matter? There are so many tiny little details all around us that affect everything from the color of our iris to the creation of rainbows. Just how much does our raytracer need to simulate in order to be photorealistic?

5. Physics

What if we ignored the first four problems and simply assumed we had managed to make a perfect, magical photorealistic raytracer. Congratulations, you’ve managed to co-opt the entirety of your CPU for the task of rendering a static 3D scene, leaving nothing left for the physics. All we’ve managed to accomplish is taking the “interactive” out of “interactive media”. Being able to influence the world around us is a key ingredient to immersion in games, and this requires more and more accurate physics, which are arguably just as difficult to calculate as raytracing is. The most advanced real-time physics engine to-date is the Lagoa Multiphysics, and it can only just barely simulate a tiny scene in a well-controlled environment before it completely decimates a modern CPU. This is without any complex rendering at all. Now try doing that for a scene with a radius of several miles. Oh, and remember our issue with scaling? This applies to physics too! Except with physics, its an order of magnitude even more difficult.

6. Content

As many developers have been discovering, procedural generation is not magic pixie dust you can sprinkle on problems to make them go away. Yet, without advances in content generation, we are forced to hire armies of artists to create the absurd amounts of detail required by modern games. Raytracing doesn’t solve this problem, it makes it worse. In any given square mile of a human settlement, there are billions of individual objects, ranging from pine cones, to rocks, to TV sets, to crumbs, all of which technically have physics, and must be kept track of, and rendered, and even more importantly, modeled.

Despite multiple attempts at leveraging procedural generation, the content problem has simply refused to go away. Until we can effectively harness the power of procedural generation, augmented artistic tools, and automatic design morphing, the advent of fully photorealistic raytracing will be useless. The best graphics engine in the world is nothing without art.

7. AI

I know that raytracing is exciting, sometimes simply as a demonstration of raw computational power. But it always disheartens me when people fantasize about playing amazingly detailed games indistinguishable from real life when that simply isn’t going to happen, even with the inevitable development² of realtime raytracing. By the time it becomes commercially viable, it will simply be yet another incremental step in our eternal quest for infinite realism. It is an important step, and one we should strive for, but it alone is not sufficient to spark a revolution.