Benchmarks

[faiface]

Benchmarks are incredibly useful when optimizing performance. They are also useful when we want to tell people how much we improved the performance of the library (for example: "The performance of drawing primitive shapes was improved by 80%", sounds good).

Technology

• We'll use the standard "testing"
• Each benchmark should be located in the same folder as the functionality it's benchmarking
• Related benchmarks should be located together in a single appropriately named file
• Benchmark files should have a different package than the functionality they're benchmarking (pixe -> pixel_test)

What to benchmark

Here's the list so far. Will most likely expand.

• Pixel
  • ToRGBA
  • All Matrix operations
  • Sprite drawing to Batch
• IMDraw
  • All IMDraw drawing operations (circle, line, etc.) separately for different number of points (i.e. don't use b.N as the number of points, but as the number of repeated drawings)
• Text
  • Creating Atlas from several different rune sets (small, large)
  • Writing text (by small chunks, by large chunks, b.N should be the number of chunks, not their size)