Enabling Efficient Sparse Computations Using Linear Algebra Aware Compilers
quality 1/10 · low quality
0 net
AI Summary
LAPIS is a compiler framework built on MLIR that optimizes sparse linear algebra operations across diverse architectures using Kokkos dialect for performance portability and a partition dialect for distributed memory execution. The framework demonstrates MLIR's capability to enable linear algebra-level optimizations for both sparse and dense kernels on GPUs, with applications to graph algorithms, relational databases, and scientific machine learning.
Tags
Entities
LAPIS
MLIR
Kokkos
GraphBLAS
TenSQL
Sandia National Laboratories
Rajamanickam, Sivasankaran
Kelley, Brian Michael
Sadayappan, Ponnuswamy
Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers (Technical Report) | OSTI.GOV Skip to main content Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers Technical Report 路 Mon Sep 01 00:00:00 EDT 2025 DOI: https://doi.org/10.2172/3013883 路 OSTI ID: 3013883 Rajamanickam, Sivasankaran Search OSTI.GOV for author "Rajamanickam, Sivasankaran" Search OSTI.GOV for ORCID "0000-0002-5854-409X" View ORCID profile [1] ; Kelley, Brian Michael Search OSTI.GOV for author "Kelley, Brian Michael" Search OSTI.GOV for ORCID "0000-0003-3607-360X" View ORCID profile [1] ; Sadayappan, Ponnuswamy [2] ; Rountev, Atanas [3] ; Roose, Jonathan [1] ; Eydenberg, Michael Shannon Search OSTI.GOV for author "Eydenberg, Michael Shannon" Search OSTI.GOV for ORCID "0000-0002-5400-8089" View ORCID profile [1] ; Alvey-Blanco, Addison Jordan Search OSTI.GOV for author "Alvey-Blanco, Addison Jordan" Search OSTI.GOV for ORCID "0009-0002-7528-7301" View ORCID profile [1] ; Vaidya, Miheer [2] ; Singh, Shreya [2] ; Mantri, Devanshu [2] Sandia National Lab. (SNL-NM), Albuquerque, NM (United States) Univ. of Utah, Salt Lake City, UT (United States) The Ohio State Univ., Columbus, OH (United States) + Show Author Affiliations This project developed the LAPIS compiler framework, built on the Multilevel Intermediate Representation (MLIR), to optimize sparse linear algebra operations and support performance portability across diverse architectures. The main innovation of LAPIS is the Kokkos dialect, which allows for lowering codes from a high productivity language to different architectures in an elegant way. The dialect also allows the conversion of lower-level MLIR code to C++ Kokkos code, facilitating the integration of scientific machine learning (SciML) models into applications. To extend LAPIS for distributed memory architectures, a new partition dialect was created to manage the distribution of sparse tensors and express communication patterns for sparse linear algebra operations. This dialect also supports the distributed execution of operators and includes algorithmic optimizations to minimize communication to improve performance. The project also demonstrates that MLIR can enable effective linear algebra-level optimizations, improving performance on different GPUs for both sparse and dense linear algebra kernels. Key applications of LAPIS include sparse linear algebra and graph kernels, TenSQL, a relational database management solution built on GraphBLAS, and the development of subgraph isomorphism and monomorphism kernels, showcasing performance portability. In summary, the LAPIS framework supports productivity, performance, portability, and distributed memory execution, while also enabling linear algebra-level optimizations that are challenging in traditional programming languages, with successful applications ranging from simple sparse linear algebra to complex graph kernels. View Technical Report Cite Citation Formats MLA APA Chicago BibTeX Rajamanickam, Sivasankaran, et al. "Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers." , Sep. 2025. https://doi.org/10.2172/3013883 馃棊 Copy to clipboard Rajamanickam, Sivasankaran, Kelley, Brian Michael, Sadayappan, Ponnuswamy, Rountev, Atanas, Roose, Jonathan, Eydenberg, Michael Shannon, Alvey-Blanco, Addison Jordan, Vaidya, Miheer, Singh, Shreya, & Mantri, Devanshu (2025). Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers. https://doi.org/10.2172/3013883 馃棊 Copy to clipboard Rajamanickam, Sivasankaran, Kelley, Brian Michael, Sadayappan, Ponnuswamy, et al., "Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers," (2025), https://doi.org/10.2172/3013883 馃棊 Copy to clipboard @techreport{osti_3013883, author = {Rajamanickam, Sivasankaran and Kelley, Brian Michael and Sadayappan, Ponnuswamy and Rountev, Atanas and Roose, Jonathan and Eydenberg, Michael Shannon and Alvey-Blanco, Addison Jordan and Vaidya, Miheer and Singh, Shreya and Mantri, Devanshu}, title = {Enabling Efficient Sparse Computations using Linear Algebra Aware Compilers}, institution = {Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States)}, annote = {This project developed the LAPIS compiler framework, built on the Multilevel Intermediate Representation (MLIR), to optimize sparse linear algebra operations and support performance portability across diverse architectures. The main innovation of LAPIS is the Kokkos dialect, which allows for lowering codes from a high productivity language to different architectures in an elegant way. The dialect also allows the conversion of lower-level MLIR code to C++ Kokkos code, facilitating the integration of scientific machine learning (SciML) models into applications. To extend LAPIS for distributed memory architectures, a new partition dialect was created to manage the distribution of sparse tensors and express communication patterns for sparse linear algebra operations. This dialect also supports the distributed execution of operators and includes algorithmic optimizations to minimize communication to improve performance. The project also demonstrates that MLIR can enable effective linear algebra-level optimizations, improving performance on different GPUs for both sparse and dense linear algebra kernels. Key applications of LAPIS include sparse linear algebra and graph kernels, TenSQL, a relational database management solution built on GraphBLAS, and the development of subgraph isomorphism and monomorphism kernels, showcasing performance portability. In summary, the LAPIS framework supports productivity, performance, portability, and distributed memory execution, while also enabling linear algebra-level optimizations that are challenging in traditional programming languages, with successful applications ranging from simple sparse linear algebra to complex graph kernels.}, doi = {10.2172/3013883}, url = {https://www.osti.gov/biblio/3013883}, place = {United States}, year = {2025}, month = {09}} 馃棊 Copy to clipboard Export Endnote RIS CSV/Excel XML JSON Share Facebook Twitter / X LinkedIn Email Save You must Sign In or Create an Account in order to save documents to your library. Print Details Similar Records / Subjects Research Organization: Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States) Sponsoring Organization: USDOE National Nuclear Security Administration (NNSA); USDOE Laboratory Directed Research and Development (LDRD) Program DOE Contract Number: NA0003525 OSTI ID: 3013883 Report Number(s): SAND--2025-11870R; 1789932 Country of Publication: United States Language: English Similar Records A High Performance Sparse Tensor Algebra Compiler in MLIR Conference 路 Sun Dec 19 23:00:00 EST 2021 路 OSTI ID: 1855960 Automatic Code Generation for High-Performance Graph Algorithms Conference 路 Tue Dec 26 23:00:00 EST 2023 路 OSTI ID: 2376153 An implementation of SISAL for distributed-memory architectures Thesis/Dissertation 路 Thu Jun 01 00:00:00 EDT 1995 路 OSTI ID: 176572 Related Subjects 97 MATHEMATICS AND COMPUTING