Functionality Locality, Mixture & Control = Logic = Memory
Xiangjun Peng
TL;DR
The paper expands the traditional notion of locality by introducing Functionality Locality, where the order of access alone can switch the functional role of data without any spatial movement. It then defines Mixture as a length-dependent bit sequence whose capabilities (compute, query, move) depend on layout (bit-parallel vs bit-serial) and introduces an Indexes ≈ Values principle that enables full functionality without layout changes. A central unifying idea, Control = Logic = Memory, reframes architecture by blending control signals, computation, and memory across time, revisiting Von Neumann and Harvard models and arguing for a memory-centric computational paradigm. The work further discusses broad implications, including mass-energy correlations, layout-directed functionalities tied to orbital analogies, and fractal-inspired infinite vs finite encodings, offering a high-level analytic framework to guide memory-centric architectures and future research directions.
Abstract
This work provides new insights and constructs to the field of computer architecture and systems, and these insights are expected to be useful for the broad software stack. First, this work introduces Functionality Locality: this form of Functionality Locality shows that functionalities can be changed with a single piece of information, by solely changing the access order. This broadens the scope of ``principle of locality", which originally includes spatial and temporal locality. Second, this work coins the term Mixture, by incorporating the layout-directed functionalities with the original quantifiers such as scalar and vector. The implications of Mixture significantly expands new understanding of quantifiers, and this work identifies several important ones (from the author perspective). Third, with Functionality and Mixture, this work identifies the principle ``Control = Logic = Memory", and provides a revisit to Von Neumann architectures and Harvard architectures. This centers the focus on the memory, and brings further guidelines on memory-centric architectures with a new analytic framework. Fourth, this work discusses several important implications from this work in a variety of aspects.