Multi-objective memory allocation-based code optimizations for real-time systems

The thesis proposes compiler-level multi-objective memory allocation-based optimizations to simultaneously optimize the program’s worst-case execution time, energy consumption, and code size. It introduces multi-objective static memory allocation that optimizes multiple objectives. The thesis tackles compilation time bottlenecks caused while solving the multi-objective static allocation problem by introducing approximation models that approximate optimization objectives and search space reduction techniques. The thesis proposes a novel multi-objective dynamic memory allocation technique to circumvent the small scratchpad memory size constraint on the static allocation problem. The thesis introduces Direct Memory Access (DMA)- and compression-awareness to the multi-objective dynamic allocation problem, which tackles the high dynamic allocation-based worst-case execution time, energy, and code size costs.

Subjects

Multi-Objective Optimization

Compiler

Static and dynamic SPM allocation

Metaheuristic algorithms

Real-time systems

Approximation models and search space reduction

DDC Class

004: Computer Sciences

Lizenz

https://creativecommons.org/licenses/by/4.0/

Name

Dissertation_ShashankJadhav.pdf

Size

24.69 MB

Format

Adobe PDF

Options

Multi-objective memory allocation-based code optimizations for real-time systems