Prism: Symbolic Superoptimization of Tensor Programs

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wu, Mengdi, Jiang, Xiaoyu, Padon, Oded, Jia, Zhihao
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917414303694848
author Wu, Mengdi
Jiang, Xiaoyu
Padon, Oded
Jia, Zhihao
author_facet Wu, Mengdi
Jiang, Xiaoyu
Padon, Oded
Jia, Zhihao
contents This paper presents Prism, the first symbolic superoptimizer for tensor programs. The key idea is sGraph, a symbolic, hierarchical representation that compactly encodes large classes of tensor programs by symbolically representing some execution parameters. Prism organizes optimization as a two-level search: it constructs symbolic graphs that represent families of programs, and then instantiates them into concrete implementations. This formulation enables structured pruning of provably suboptimal regions of the search space using symbolic reasoning over operator semantics, algebraic identities, and hardware constraints. We develop techniques for efficient symbolic graph generation, equivalence verification via e-graph rewriting, and parameter instantiation through auto-tuning. Together, these components allow Prism to bridge the rigor of exhaustive search with the scalability required for modern ML workloads. Evaluation on five commonly used LLM workloads shows that Prism achieves up to $2.2\times$ speedup over best superoptimizers and $4.9\times$ over best compiler-based approaches, while reducing end-to-end optimization time by up to $3.4\times$.
format Preprint
id arxiv_https___arxiv_org_abs_2604_15272
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Prism: Symbolic Superoptimization of Tensor Programs
Wu, Mengdi
Jiang, Xiaoyu
Padon, Oded
Jia, Zhihao
Programming Languages
Artificial Intelligence
Machine Learning
This paper presents Prism, the first symbolic superoptimizer for tensor programs. The key idea is sGraph, a symbolic, hierarchical representation that compactly encodes large classes of tensor programs by symbolically representing some execution parameters. Prism organizes optimization as a two-level search: it constructs symbolic graphs that represent families of programs, and then instantiates them into concrete implementations. This formulation enables structured pruning of provably suboptimal regions of the search space using symbolic reasoning over operator semantics, algebraic identities, and hardware constraints. We develop techniques for efficient symbolic graph generation, equivalence verification via e-graph rewriting, and parameter instantiation through auto-tuning. Together, these components allow Prism to bridge the rigor of exhaustive search with the scalability required for modern ML workloads. Evaluation on five commonly used LLM workloads shows that Prism achieves up to $2.2\times$ speedup over best superoptimizers and $4.9\times$ over best compiler-based approaches, while reducing end-to-end optimization time by up to $3.4\times$.
title Prism: Symbolic Superoptimization of Tensor Programs
topic Programming Languages
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2604.15272