Composable and Modular Code Generation in MLIR: A Structured and Retargetable Approach to Tensor Com...
Composable and Modular Code Generation in MLIR: A Structured and Retargetable Approach to Tensor Compiler Construction
About this item
Full title
Author / Creator
Publisher
Ithaca: Cornell University Library, arXiv.org
Journal title
Language
English
Formats
Publication information
Publisher
Ithaca: Cornell University Library, arXiv.org
Subjects
More information
Scope and Contents
Contents
Despite significant investment in software infrastructure, machine learning systems, runtimes and compilers do not compose properly. We propose a new design aiming at providing unprecedented degrees of modularity, composability and genericity. This paper discusses a structured approach to the construction of domain-specific code generators for tens...
Alternative Titles
Full title
Composable and Modular Code Generation in MLIR: A Structured and Retargetable Approach to Tensor Compiler Construction
Authors, Artists and Contributors
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_2626562805
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_2626562805
Other Identifiers
E-ISSN
2331-8422
