The Compiler as a Decomposed Optimization System

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Bilar, Daniyel Yaacov
Format: Recurso digital
Sprache:Englisch
Veröffentlicht: Zenodo 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866901794049753088
author Bilar, Daniyel Yaacov
author_facet Bilar, Daniyel Yaacov
contents <p>This paper revisits the central claim of Bilar (2007) that the structure of executables is the product of an<br>engineered optimization process. We introduce the Network Utility Maximization (NUM) framework and<br>demonstrate that modern compiler architectures can be formally mapped onto distinct decomposition strategies<br>from optimization theory. Specifically, we posit that ahead-of-time (AOT) pass pipelines implement primal<br>decomposition, just-in-time (JIT) compilers with profile-guided optimization implement dual decomposition,<br>speculative optimization implements penalty methods, and link-time optimization implements the alternating<br>direction method of multipliers (ADMM). This reframes the compiler not as a monolithic optimizer, but as a<br>system of interlocked, resource-allocating subsystems. The formalism generates testable predictions, which we<br>validate empirically in a series of companion papers (<a href="https://zenodo.org/records/18828679">Empirical Validation of Shadow-Price-Guided Inlining in MIR</a>) <br><br>This work provides a mathematical foundation for the paper’s core insight and suggests new directions for compiler design.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18715390
institution Zenodo
language eng
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle The Compiler as a Decomposed Optimization System
Bilar, Daniyel Yaacov
compiler optimization
network utility maximization
distributed optimization
decomposition methods
JIT compilation
profile-guided optimization
<p>This paper revisits the central claim of Bilar (2007) that the structure of executables is the product of an<br>engineered optimization process. We introduce the Network Utility Maximization (NUM) framework and<br>demonstrate that modern compiler architectures can be formally mapped onto distinct decomposition strategies<br>from optimization theory. Specifically, we posit that ahead-of-time (AOT) pass pipelines implement primal<br>decomposition, just-in-time (JIT) compilers with profile-guided optimization implement dual decomposition,<br>speculative optimization implements penalty methods, and link-time optimization implements the alternating<br>direction method of multipliers (ADMM). This reframes the compiler not as a monolithic optimizer, but as a<br>system of interlocked, resource-allocating subsystems. The formalism generates testable predictions, which we<br>validate empirically in a series of companion papers (<a href="https://zenodo.org/records/18828679">Empirical Validation of Shadow-Price-Guided Inlining in MIR</a>) <br><br>This work provides a mathematical foundation for the paper’s core insight and suggests new directions for compiler design.</p>
title The Compiler as a Decomposed Optimization System
topic compiler optimization
network utility maximization
distributed optimization
decomposition methods
JIT compilation
profile-guided optimization
url https://doi.org/10.5281/zenodo.18715390