55 lines
2.0 KiB
TeX
55 lines
2.0 KiB
TeX
%% SCRAP: archive/research/README
|
|
%% SOURCE: docs/working/archive/research/README.md
|
|
%% STATUS: HISTORICAL
|
|
%% FITS: vol3-research/app if useful
|
|
%% EDITORIAL: lifted — prose rewritten to press voice
|
|
|
|
\section{Research Directory Overview}
|
|
|
|
This directory index was the navigation hub for the \texttt{docs/06-research/}
|
|
subtree before the 2026-06-16 documentation reorganisation. It is preserved
|
|
as a historical reference; canonical research content has been promoted into
|
|
the formal volume.
|
|
|
|
\subsection{Key Research Contributions}
|
|
|
|
Three primary contributions are identified.
|
|
|
|
\textbf{Physics-grounded self-adaptive runtime.}
|
|
StarForth demonstrates a novel approach to VM optimisation: an execution
|
|
heat model grounded in thermodynamic metaphor, a rolling window of truth
|
|
for deterministic metric seeding, and an inference engine for adaptive
|
|
parameter tuning.
|
|
|
|
\textbf{Formally proven deterministic behaviour.}
|
|
Zero percent algorithmic variance across ninety experimental runs.
|
|
Determinism is established independently for heat decay (Loop~\#3),
|
|
window-width inference (Loop~\#5), and pipelining metrics (Loop~\#4),
|
|
and reproduces across platforms.
|
|
|
|
\textbf{Design of Experiments methodology.}
|
|
A rigorous factorial DoE framework drives all experimental work, including
|
|
ANOVA and Levene's-test statistical validation, heartbeat-driven data
|
|
collection, and reproducible protocols.
|
|
|
|
\subsection{Publication Materials}
|
|
|
|
A comprehensive peer-review submission package was assembled under
|
|
\texttt{archive/phase-1/Reference/physics\_experiment/PEER\_REVIEW\_SUBMISSION/},
|
|
including a main paper draft, formal verification interpretation, variance
|
|
analysis summary, and supplementary materials.
|
|
|
|
A DARPA SSM proposal and a provisional patent application were filed during
|
|
this period.
|
|
|
|
\subsection{Experimental Reproducibility}
|
|
|
|
All experiments are reproducible via:
|
|
|
|
\begin{lstlisting}[language=bash]
|
|
make fastest
|
|
./build/amd64/fastest/starforth --doe
|
|
\end{lstlisting}
|
|
|
|
Results should demonstrate 0\% algorithmic variance.
|