26 lines
1.3 KiB
TeX
26 lines
1.3 KiB
TeX
%% SCRAP: archive/legacy/OPTIMIZATION_OPPORTUNITIES
|
|
%% SOURCE: docs/working/archive/legacy/OPTIMIZATION_OPPORTUNITIES.md
|
|
%% STATUS: SUPERSEDED
|
|
%% FITS: none
|
|
%% EDITORIAL: lifted — prose rewritten to press voice
|
|
|
|
\section*{Optimization Opportunity Analysis (2025-11-19)}
|
|
|
|
A bang-for-buck analysis conducted on 2025-11-19 ranked ten candidate
|
|
optimizations for the StarForth physics engine by impact-to-effort ratio.
|
|
The two highest-ranked items---decay slope inference (Opportunity~1) and
|
|
variance-based window width tuning (Opportunity~2)---were subsequently
|
|
implemented and validated; they now constitute Feedback Loops~6 and~5
|
|
respectively in the production architecture.
|
|
|
|
The remaining eight opportunities (decay rate design-of-experiments, rolling
|
|
window sizing, hotwords cache threshold tuning, binary-search validation, ML
|
|
prefetch model, NUMA placement, dictionary reorganization, and SIMD
|
|
acceleration) were assessed at lower priority. Of these, opportunities related
|
|
to the hotwords cache and rolling window sizing were addressed implicitly
|
|
through the DoE experimental program reported in the \textsc{ssrn} paper.
|
|
The ML prefetch, NUMA, and SIMD items were explicitly deferred.
|
|
|
|
%% TODO(bob): Confirm which of OPP 3--10 were formally closed vs.\ remain open
|
|
%% as future work. This document predates the final OPP report series.
|