# An Efficient Data Structure for Dynamic Two-Dimensional Reconfiguration

## TL;DR

This paper introduces a novel quadtree-based data structure for efficient dynamic defragmentation and reallocation on partially reconfigurable FPGAs, exploiting their two-dimensional geometry.

## Contribution

It proposes a new quadtree-inspired method for dynamic segmentation and communication infrastructure adjustment, improving upon existing approaches.

## Key findings

- Simulations suggest the worst-case bounds are overly pessimistic.
- The method effectively manages dynamic reconfiguration with reduced fragmentation.
- The approach exploits 2D chip geometry for better reconfiguration efficiency.

## Abstract

In the presence of dynamic insertions and deletions into a partially reconfigurable FPGA, fragmentation is unavoidable. This poses the challenge of developing efficient approaches to dynamic defragmentation and reallocation. One key aspect is to develop efficient algorithms and data structures that exploit the two-dimensional geometry of a chip, instead of just one. We propose a new method for this task, based on the fractal structure of a quadtree, which allows dynamic segmentation of the chip area, along with dynamically adjusting the necessary communication infrastructure. We describe a number of algorithmic aspects, and present different solutions. We also provide a number of basic simulations that indicate that the theoretical worst-case bound may be pessimistic.