AHS v25.0 – Adaptive Hybrid Sort (Bit Allocation)

A high-performance evolution of the Adaptive Hybrid Sort, version 25.0 introduces a sophisticated System Calibration phase. This version moves beyond static thresholds, dynamically probing your machine’s unique hardware profile to determine the exact crossover point where 11-bit cache locality outperforms 16-bit raw throughput.

Core Architecture Breakdown

ComponentFunctionality
System Calibration On startup, the engine executes a hardware-specific benchmark across varying data sizes. It identifies the "Crossover Point" (where the 11-bit engine's cache efficiency overcomes the 16-bit engine's lower pass count) and locks this threshold for the current session.
Bit Allocation Control Provides an Engine Selection manual override, allowing users to force either the 11-Bit (L1 Cache Optimized) or 16-Bit (Fast Pass) logic, or remain in "Auto-Tune" mode to utilize the calibrated hardware profile.
Fused Transformation To maximize speed, the engine utilizes a Fused Scatter/Decode process. During the final radix pass, the data is decoded from its sortable integer format back into its original floating-point value simultaneously with the memory scatter, eliminating an entire O(N) pass.
WebGL2 Pipeline A robust rendering system that uploads data as an R32F texture. It features a Dual-Path Fetch system in the fragment shader: a direct 1D fetch for standard sizes and a tiled 2D fetch for datasets exceeding the hardware's MAX_TEXTURE_SIZE.
Dynamic Telemetry A real-time status dashboard that monitors SAB (SharedArrayBuffer) status, provides a countdown of sorting passes, and displays high-precision execution timing for both visual and benchmarked runs.
Pattern Library A refreshed UI for selecting from 30 data distributions. It uses a <dialog> modal with high-definition mini-canvas previews that adapt to the selected data range (Both, Positive, or Negative).

Leave a comment

Log in with itch.io to leave a comment.