COMP 2101 · Making the Impossible Run in Real Time

Led by Carmackian Engineering Simulacrum

How do you make the impossible run in real time? The engineering of exploitable structure, from Commander Keen to Quake.

1. Module 1

   Adaptive Tile Refresh: The First Trick

   Led by Carmackian Engineering Simulacrum

   The question

   Smooth animation on a PC in 1990 was impossible. Smooth scrolling in a tile-based side-scroller was possible. The specific structural property that made the difference: only the changed tiles needed redrawing. What does this tell you about the word "impossible"?

   Outcome

   The student can state the general/specific principle and apply it to new cases.

   Sub-units

   1. ○ 1.1 The General/Specific Distinction Open this unit → ✓ ↻
   2. ○ 1.2 Find Three More Cases Open this unit → ✓ ↻
   Begin the course → Print certificate of completion
2. Module 2

   BSP Trees and Doom

   Led by Carmackian Engineering Simulacrum

   The question

   Rendering visibility naively is O(n²) per frame. BSP trees precompute the spatial subdivision once, delivering O(n) per frame. The price: static geometry only. Doom's levels were static. How do you recognise when an algorithm's assumptions perfectly match your problem?

   Outcome

   The student can explain BSP trees and identify their scope conditions.

   Sub-units

   1. ○ 2.1 BSP Construction and Traversal Open this unit → ✓ ↻
   2. ○ 2.2 Essay: When Does BSP Fail? Open this unit → ✓ ↻
   Open this module → Print certificate of completion
3. Module 3

   Quake and Full 3D

   Led by Carmackian Engineering Simulacrum

   The question

   The fast inverse square root used bit manipulation on floating-point representations to approximate 1/sqrt(x). It was a masterpiece of low-level optimisation. Three years later, hardware made it irrelevant. When is hand-optimisation worth the investment?

   Outcome

   The student can explain the fast inverse square root and evaluate the hardware timing problem.

   Sub-units

   1. ○ 3.1 The Fast Inverse Square Root Open this unit → ✓ ↻
   Open this module → Print certificate of completion
4. Module 4

   Profiling-Driven Optimisation

   Led by Carmackian Engineering Simulacrum

   The question

   You think you know where the bottleneck is. You are almost always wrong. Amdahl's Law: even if you optimise 80% of the program to run infinitely fast, the remaining 20% limits your total speedup to 5x. Where does the algorithm end and the hardware begin?

   Outcome

   The student can apply Amdahl's Law and describe a profiling-driven optimisation workflow.

   Sub-units

   1. ○ 4.1 Apply Amdahl's Law Open this unit → ✓ ↻
   Open this module → Print certificate of completion
5. Module 5

   The Impossible Deadline

   Led by Carmackian Engineering Simulacrum

   The question

   The 60fps deadline. What structural properties of the specific problem can you exploit? What can you precompute, approximate, render less often? The trick is almost always there — finding it is the engineering creativity.

   Outcome

   The student can apply the impossible-deadline framework and take a defended position on engineering creativity.

   Sub-units

   1. ○ 5.1 Final Essay: Engineering Creativity Open this unit → ✓ ↻
   Open this module → Print certificate of completion