Code Optimization Techniques

Several common techniques that I've learned in my Optimization class:

• Instancing: static (pack geometries into the same vertex buffer), dynamic (similar to static, but transformed per frame), Shader (use vertex constants and GPU transforms the data), Hardware (supported by GPU).
  
• Reduce draw calls: Since the process of transferring data to the GPU includes overhead, the concept of instancing is to pack more data into the same buffer to improve the performance.


• Data Dependencies: How to write the code in special ways for the CPU to trade memory for computational speed.


• Threading: Avoid deadlocks and other abnormalities.


• Memory allocation: Avoid "new" memory at all cost.


• Benchmark, bottleneck, and performance tuning.


• SIMD: Single Instruction Multiple Data.