Software Development

One night, I was listening to Coast to Coast AM and they had Neal Adams as the guest. Neal Adams is best known for his work on Batman and Green Arrow. He was asked how come movies have such a hard time with their heroes, but comic books really have done a good job? What Neal said really surprised me. He said back in the early days, every single idea no matter how good or bad was tried. EVERYTHING!!! Not one single idea was left unturned. If an artist had an idea, it was tried. So it did not take long that new and original ideas were hard to come by until it became very difficult for new artists to get something published. He also said that none of the heroes we know today exist as they did originally. So basically, everything stems from really bad ideas.

How can this be? How could every single comic book hero come from bad ideas? Neal discussed more of the process. They would have limited trial runs. Many of these became limited editions collectors items if the comic was later picked up for general publication. If they did not do well, they were scrapped. Those that did well, the heroes evolved over time depending on the readers' response to them. He mentioned so many examples that I can hardly remember any of them. One example that everyone knows about is that Superman could originally only jump over tall buildings. Today, he can fly and even turn back time.

Read more...

In a previous article, I showed how you can draw from front to back using OpenGL. Here is the final set of equations that we used for blending. S is the source and D is the destination.

D[B] += S[B] * D[A]
D[G] += S[G] * D[A]
D[R] += S[R] * D[A]
D[A] *= S[A]

Source and destination are in precomputed format. A precomputed format is where you apply the following transformation.

B *= A
G *= A
R *= A
A = 1-A

A on the right hand side is the original Alpha value. All values are relative to 255. So 128 for alpha would actually be 128/255 in all the above equations. D3D takes care of all this for you automatically.

You do not need to convert all your images beforehand to precomputed format. Instead, you can use the fixed function pipeline to precompute your image on the fly.

  // Set texture for texture stage 0
  pD3DDevice->SetTexture(0,pMyTexture);

  pD3DDevice->SetTextureStageState(0,D3DTSS_COLOROP,D3DTOP_MODULATE);
  pD3DDevice->SetTextureStageState(0,D3DTSS_COLORARG1,D3DTA_TEXTURE|D3DTA_ALPHAREPLICATE);
  pD3DDevice->SetTextureStageState(0,D3DTSS_COLORARG2,D3DTA_TEXTURE);
  pD3DDevice->SetTextureStageState(0,D3DTSS_ALPHAOP,D3DTOP_SELECTARG1);
  pD3DDevice->SetTextureStageState(0,D3DTSS_ALPHAARG1,D3DTA_TEXTURE|D3DTA_COMPLEMENT);

  pD3DDevice->SetTextureStageState(1,D3DTSS_COLOROP,D3DTOP_DISABLE);
  pD3DDevice->SetTextureStageState(1,D3DTSS_ALPHAOP,D3DTOP_DISABLE);

This should work on ALL video cards no matter how old. Then we would apply the alpha blending equations.

Read more...

Ok, so if you're on a Windows XP box and you're developing something like a compiler or debugger, you're gonna need some tools, right? Tools like an API for loading and unloading modules. Tools that let you inspect symbols and line information stored in the debug section of your object and exe files. Tools that let you query what module, what source file, what line number a certain function that caused a breakpoint or bad memory allocation occurred. Tools that let you walk the stack and display a stack trace. These are things that compiler and debugger writers need to deal with. Even people who just want self inspecting code will need this stuff.

So I'm going along doing all of this. And as a good little boy, I didn't make my own tools though I nearly ended up doing so trying to work around the problems I was getting. I'm using the Borland C++ Builder compiler because it has properties and closures and I don't feel like writing extra code to enable a feature I've been using for ages. So BCB it was. Only problem is that it uses a different name mangling scheme to MS. Not only that, but Borland doesn't output CodeView debug format that Windows debugging API wants. So I wrote some code that converts Borland debug format to CodeView format. I got some of this code online, but it worked badly and didn't work for the newer versions of Borland. I also output line numbers which wasn't available in the original online code.

So I'm good to go, right? I can add this to my code to automatically check for memory leaks and see where each and every allocation and deallocation is made. All this isn't so much for the final product of Project V as it is a standard set of tools I've used for ages on larger projects for personal use to make life easier. It's basically like having a debugger within your application that tells you right away if anything is wrong. And for those who dislike debuggers, this is useful to find out WHERE the error is. Not to fix it.

Read more...

What is computability? Computability is the ability to obtain a result from a set of inputs for a specific objective. Simple operations are additions, multiplication, etc. Sorting is another example. So what is needed to be able to compute everything that is computable?

Alan Turing created a mathematical model of his Turing machine which is described as being able to compute anything that is computable. What it does not say is what is required to satisfy this requirement. This means there are many mathematical models that are equivalent. Before getting into that, let's look at what operations are required by a Turing machine to be truly universal.

• Ability to hold a state.
  
• Ability to select where next data item comes from.
  
• Ability to select where output is stored.
  
• Boolean algebra (enables AND, OR, NOT, ADD, SUB, MUL, DIV, REM, etc.)
  

Read more...

Parallel programming will come in many forms. Right now, most people think of multicore systems as the main type of parallel programming. However, I don't think this will be the way of the future. In the past, I said things will get smaller and faster and more numerous. Before we get there, there is an alternative and I've somewhat discussed it on a prior date. It's to have massive amount of data being transformed by the exact same piece of code. That's why I'm really fascinated by 3D hardware. In every image operation, the same set of operations is done on each and every pixel. Before that, each and every vertex is operated upon by its own, but identical, set of operations.

Why is this important? Anyone that's taken a parallel computing course knows that there are four types of computing.

1. SISD
2. SIMD
3. MISD
4. MIMD

Read more...