File header from my first programming assignment in 4 years

What can I say... I'm a bit of a nerd, and I guess I missed doing this stuff.

/********************
 * render_point.cpp *
 ********************

 (1) Christopher Casey
 (2) CS180, Spring 2010
 (3) Assignment #1
 (4) Due Wednesday, January 27

SUMMARY: An x86 FPU Assembly implementation of Mandelbrot Set
calculations.

OBJECTIVE: Use inline assembler to provide a speed increase
over the following reference code:

  //jsh 1/08
  double a = 0.0, b = 0.0, norm2 = 0.0;
  for (n = 0; norm2 < 4.0 && n < N; ++n) {
    double c = a*a - b*b + x;
    b = 2.0*a*b + y;
    a = c;
    norm2 = a*a + b*b; 
  }

SOLUTION NOTES:
I initially used FPU assembler to convert the body of the
loop. This saved on a few calls to the `fld` instruction as
items were copied from system memory to the FPU stack. On my
second iteration of optimization, I wrote the loop condition
in asm as well. In addition to saving a few more `fld` calls,
this allowed me to re-use the `a*a` and `b*b` values
calculated for norm2.

I've also added code that tests to see if the point is in the
cardioid or the period-2 bulb. I found a write-up about these
optimizations and decided to try implementing them. They are
available via pre-processor directives so they can be turned
off, as they only increase performance when in the default
view. These optimizations were written in C for legibility
(and because they aren't the inner-most loop). The speed
increase realized in the default view is enormous -- the
scene is rendered in about 1/5 the time.

Optimization can still be made to the loop: like the reference
code, it iterates once more than necessary when reaching max
iterations. However, since we default to 5120 iterations this
would provide a hardly noticable increase. One could also
optimize the return value by providing a look-up table rather
than performing integer-to-double-to-integer conversions to
convert `n` to a COLORREF type. Doing this would require
explicit knowledge of the value of `N`.

KNOWN ISSUES:
My solution returns a slightly different set of values for the
points than the reference code. The discrepancy is due to two
different issues. First, I believe that a minor bug exists in
the reference code. The code tests for `norm2 < 4.0`, but
typical Mandelbrot Set implementations test for `norm2 <= 4.0`.
My code corrects this.

The second, and more broad issue is that because my floating
point operations occur in a different order than the ones
generated by the reference code (and I add a+a instead of
multiplying a*2), the floating point error produces slightly
different results.

The error is slight, and I don't feel either issue requires
further investigation at this time.

BENCHMARK:
Tested on an Intel i7 920 Windows 7 64-bit, mandelbrot.cpp as driver, max 5120 iterations

Default View (lower_left_x = -1.5f; lower_left_y = -1.0f; scale = 2.0f/length;):
Reference code [/Od]:         9.38 seconds
Reference code [/O2]:         6.77 seconds
  (40% speed increase over /Od)
Reference code [/Ox]:         6.53 seconds
  (40% speed increase over /Od)
__asm [/Od]:                  4.60 seconds
  (100% speed increase over reference + /Od,
   or 45% increase over /O2)
__asm (+optimizations) [/Od]: 0.80 seconds
  (1000% speed increase over reference + /Od)

Alternate View (lower_left_x = -0.25; lower_left_y = 0.775; scale = 0.25/length;):
Reference code [/Od]:         7.03 seconds
Reference code [/O2]:         5.71 seconds
  (20% speed increase over /Od)
Reference code [/Ox]:         5.70 seconds
  (20% speed increase over /Od)
__asm [/Od]:                  2.40 seconds
 (190% speed increase over reference + /Od,
  or 130% increase over /O2)
__asm (+optimizations) [/Od]: 2.40 seconds
 (190% speed increase over reference + /Od)

REFERENCES:
Assignment details:
https://faculty.digipen.edu/~jhanson/protected/cs180assignment1.pdf
https://faculty.digipen.edu/~jhanson/protected/render_point.h
https://faculty.digipen.edu/~jhanson/protected/mandelbrot.cpp
https://faculty.digipen.edu/~jhanson/protected/render_point0.cpp

Optimization notes:
http://www.nitorijournal.org/?p=517
  (Cardioid and period-2 Bulb detection)

Additional notes: (researched, but not used in this project)
http://users.softlab.ece.ntua.gr/~ttsiod/mandelSSE.html
 (CUDA and SSE-based mandelbrots)
http://www.ozone3d.net/tutorials/mandelbrot_set_p4.php
 (General Purpose GPU Stream processing mandelbrots)
*/

A letter to my professor

Professor ["C"],

In today's lecture we discussed frame rate, and they eye's ability to detect motion. I was a little concerned that some misconceptions were being fueled about the human eye's ability to detect motion. To say that the eye is "only capable" of seeing 12, 24, 30, or 60 "frames" is not correct. When I brought this up in class, you mentioned the example of a spinning wheel appearing to move backwards. I feel this example is flawed. Generally, this illusion applies to the wheel being captured on film at 24 hz, where we obviously have a stroboscopic effect causing the wheel to appear to turn backwards. It can also be observed in indoor lighting, particularly with flourescents, as the lighting itself operates at 60 hz.

There have actually been some real-world, sunlit experiments where people would perceive a wheel as rotating backwards, but I believe they are largely considered to be the result of eye fatigue, and not of some explicit limit in the brain's ability to interpret the data from the retina. I wonder if this may also be in part a result of our brains becoming so accustomed to strobic lights in modern life.

I found an interesting article on the subject:

"Illusory motion reversal is caused by rivalry, not by

perceptual snapshots of the visual field."

Keith Kline, Alex O. Holcombe, David M. Eagleman

http://wexler.free.fr/library/files/kline%20(2004)%20illusory%20motion%20reversal%20is%20caused%20by%20rivalry,%20not%20by%20perceptual%20snapshots%20of%20the%20visual%20field.pdf

Much of the article is technical neuroscience that I don't understand, but their conclusion is clear: motion reversal illusion cannot be explained by snapshots of the visual field.

Also, you posed the question in class today, "what does 'real-time' really mean?" I think it's a term that's become a little bit bastardized to mean "fast, snappy response time," but the technical definition is that a real-time system has a fixed amount of time allotted to complete a task.

Someone brought up the idea that a turn-based game like chess is not "real-time." Actually, in a tournament chess match with a hard timer, chess is a real-time game. Because the operation (selecting and executing a move) must be completed by a discrete moment, a computer opponent can be considered a real-time system. If no deadline is given, the operation can run as a simple batch process with no concern to the time it takes to complete.

Rigorously speaking, most game engines don't run in true real-time any more; it's just very important that we get really close, and handle things properly when we don't make our 16ms deadline :)

Arkham Asylum's crazy copy protection

The new game Batman: Arkham Asylum uses a fancy new copy protection scheme. In addition to the usual copy protection fare, this game hooks in additional check that occurs during gameplay. If the game has been cracked (had the copy protection removed), it will disable Batman's glide ability, effectively halting the player's progress through the game. What's curious about this is that the game never actually announces to the player that they've been caught trying to pirate the software -- it just appears that Batman can't proceed for some unknown reason.

One illegitimate user has already been caught by outting himself on the official support forums: http://forums.eidosgames.com/showthread.php?t=95030.

I'm left a bit puzzled about the efficacy of this copy protection strategy. Certainly, it will reduce the number of players who are able to enjoy the game without paying for it. It also wastes the time and bandwidth of the would-be pirates who download the game illegally, which is at least a moral victory. However, this system does very little to turn pirates into customers. In fact, I believe it may have an opposite effect.

Because the software doesn't explicitly notify the user that their copy is believed to be illegititmate, players are likely to blame the problem on a bug in the software. Because most stores don't accept returns on opened software items, it is unlikely that a user would risk purchasing the software in order to fix the problem. In fact, I believe he's more likely to tell his friends that the game is buggy and discourage them from purchasing it as well. If they continue to have this experience with software, they'll become less and less likely to buy products in stores. In the long run, this distrust can be hurtful to the entire industry.

While I'm definitely in favor of having a layered copy protection scheme, I think that covert copy protection is a sad, sad mistake. Instead, why not pop up a message to let the player know they won't be able to proceed until they legitimize their copy? You still get to play a gotcha on all the crackers -- they'll have to play through the entire game to make sure it's clean -- and you get to clearly tell your early adopters that you welcome them as customers more than you despise them as pirates.

Making a killing in the WoW auction house

I've been playing World of Warcraft quite a bit lately, and I thought I'd share a little about my recent experience on the auction house. I just leveled a new character and funded much of its gear with money from my main, so I was feeling a little poor all around. The alt had taken Inscription and Enchanting as professions, and I decided to see how they fared in terms of profitability... Wow.

Inscription allows the character to turn herbs into specialty pigments and inks, and then further convert the inks into glyphs. Two key points set it apart from other professions:

First, glyphs transcend character levels. A level 80 Paladin may desire many of the same glyphs as a level 20 Paladin would. This means that nearly every glyph that a scribe learns throughout leveling his profession is a valuable commodity.

Second, every glyph in the game can be normalized to the same underlying cost to produce. This is thanks to the fact that a trader in the high-level city of Dalaran will gladly trade you any ink in the game for your Ink of the Sea, the ink produced by the highest level of herbs currently in the game.

With these two factors in mind, I was able to produce a modest set of 100 glyphs to list on the auction house as an experiment. The accepted markup on glyphs is astounding! My cost to produce and sell a glyph is 3 gold plus about 5 seconds worth of time crafting and listing the item. I have been selling glyphs on the auction house for anywhere from 10-30 gold. That's up to 1000% return on my investment.

With the glyph business and a few side projects (dipping a toe into enchanting and Darkmoon Card sales), my net income is in the 1000s of gold every week -- for context, this character leveled decadently over the course of two months with less than a 10,000 gold stipend from my main character, an amount I'm confident I can soon produce weekly after some more refinement of my business.

An interesting article on IT turnover

I ran across this article the other day:

http://thedailywtf.com/Articles/Up-or-Out-Solving-the-IT-Turnover-Crisis.aspx

(go ahead and give it a read, I've got time.)

I'd be really interested in finding a way to implement this in the games industry. I think that because projects generally fit the 2-4 year span described here, it should be an easy match. Game sequels seem to benefit most when they are given fresh ideas for gameplay, storytelling techniques, and technical breakthroughs. It seems much easier to facilitate this if you keep only your 'very best' and bring in other new talent to rebuild your product.

As a current employee of a worldwide publisher, I suspect that once my value reaches apex at my current studio I should be able to find somewhere else to go under the same umbrella.

CONFIRMED features for SOCOM 4?

I know someone who secretly works at Sony, and he is making "SOCOM 4?"

Here is the CONFIRMED list of features:
- Cube Maps
- A Fortress
- A Bunker
- Not Voiced By Jason Statham
- No Aliens
- Maybe Aliens

More information as it develops!

P.S.

My garbage can smells like something died in it. I think I'm going to set it on fire.