Benchmark MethodologyThese README numbers are measured, hardware-specific samples generated from the repo's explicit benchmark tasks.

Benchmark Methodology

These README numbers are measured, hardware-specific samples generated from the repo's explicit benchmark tasks.

  • Scenes: Triangle, Embedding Galaxy, and Force Graph
  • Platforms: JVM, Web, and Android emulator
  • Resolution: 1920x1080
  • Warmup: 120 frames
  • Measured window: 600 frames
  • Repeats: 3 per scene/platform
  • Published metrics:
    • boot_to_first_frame_ms
    • avg_fps
    • p95_frame_ms
    • peak_heap_delta_mb
    • workload_label
    • environment_label

Notes:

  • boot_to_first_frame_ms measures benchmark-mode startup through the first rendered frame for that platform runner.
  • avg_fps and p95_frame_ms are derived from wall-clock frame intervals observed by the benchmark loop, not GPU timestamp queries.
  • peak_heap_delta_mb is reported as a delta from the baseline heap captured before scene boot.
  • Memory reporting is platform-specific by design:
    • JVM: JVM heap delta
    • Web: JS heap delta from performance.memory when Chrome exposes it
    • Android: app heap plus native heap estimate delta from the benchmark process
  • Android rows are emulator results and must be treated as emulator / host-GPU-assisted measurements, not physical-device claims.
  • The synthetic performance tests in the repo remain useful for validation guardrails, but they are not the source of the README benchmark table.
Architected in Kotlin. Rendered with Materia. Powered by Aether.
© 2026 Yousef.?