Kazem Cheshmi

I am an Assistant Professor in the Department of Electrical and Computer Engineering at McMaster University. My recent research focuses on compiler optimization techniques for accelerating scientific computing and machine learning applications on parallel architectures.

I lead the SwiftWare Lab.

SONAD'25 submission/registration is open.

Future Students

I am always looking for motivated and hard-working students. Please apply to the ECE department at McMaster if you are interested in building compilers for scientific and machine learning applications. Please also email me your application and possibly (link to) your coding project samples. I strongly encourage applications from students that are a member of underrepresented groups to increase diversity in STEM.

Teaching

  • COMP ENG 4SP4. High-Performance Programming. (Fall 2024, 2025)
  • COMP ENG 3DY4. Computer Systems Integration Project. (Winter 2023, 2024)
  • ECE 718: Special Topics in Computation, Compiler Design for High Performance Computing. (Fall 2023 - 2025)
  • CSC367: Parallel Programming. (Winter 2021)

Service

  • SONDA'25 Organizer, Program Committee at PPoPP'25, Session Chair at Supercomputing'23, Program Committee member at IPDPS'23, Program Committee member at Supercomputing'23, External Review Committee at SPAA'22, Web Chair at PPoPP'22, Program Committee member at ICPP'22, Session Chair at ICPP'22, Reviewer at Journal of Supercomputing, ACM Transactions on Architecture and Code Optimization, and Journal of Parallel and Distributed Computing.

Major Awards

Publications

  • Loop Fusion in Matrix Multiplications with Sparse Dependence
    Mohammad Mahdi Salehi and Kazem Cheshmi
    International Conference on Supercomputing, ICS'25 (Accepted)

  • Runtime Composition of Iterations for Fusing Loop-carried Sparse Dependence
    Kazem Cheshmi, Michelle Mills Strout, and Maryam Mehri Dehnavi
    International Conference for High Performance Computing, Networking, Storage and Analysis, SC'23

  • Register Tiling for Unstructured Sparsity in Neural Network Inference
    Lucas Wilkinson*, Kazem Cheshmi*, and Maryam Mehri Dehnavi (*Equal contributions)
    ACM SIGPLAN Conference on Programming Language Design and Implementation, PLDI'23 (Selected for MIT PL Review)

  • Vectorizing Sparse Matrix Computations with Partially-Strided Codelets
    Kazem Cheshmi*, Zachary Cetinic*, and Maryam Mehri Dehnavi (*Equal contributions)
    International Conference for High Performance Computing, Networking, Storage and Analysis, SC'22

  • HDagg: Hybrid Aggregation of Loop-carried Dependence Iterations in Sparse Matrix Computations
    Behrooz Zarebavani, Kazem Cheshmi, Bangtian Liu, Michelle Mills Strout, and Maryam Mehri Dehnavi
    IEEE International Parallel and Distributed Processing Symposium , IPDPS'22

  • Optimizing Sparse Computations Jointly
    Kazem Cheshmi, Michelle Mills Strout, and Maryam Mehri Dehnavi
    Symposium on Principles and Practice of Parallel Programming , PPoPP'22

  • NASOQ: Numerically Accurate Sparsity-Oriented QP Solver
    Kazem Cheshmi, Danny M. Kaufman, Shoaib Kamil, and Maryam Mehri Dehnavi
    ACM Transactions on Graphics , SIGGRAPH'20

  • MatRox: Modular approach for improving data locality in Hierarchical (Mat)rix App(Rox)imation
    Bangtian Liu*, Kazem Cheshmi*, Saeed Soori, Michelle Mills Strout, Maryam Mehri Dehnavi (*Equal contributions)
    Symposium on Principles and Practice of Parallel Programming , PPoPP'20

  • Sparse computation data dependence simplification for efficient compiler-generated inspectors
    Mahdi Soltan Mohammadi, Tomofumi Yuki, Kazem Cheshmi, Eddie C Davis, Mary Hall, Maryam Mehri Dehnavi, Payal Nandy, Catherine Olschanowsky, Anand Venkat, Michelle Mills Strout
    ACM SIGPLAN Conference on Programming Language Design and Implementation, PLDI'19


  • ParSy: inspection and transformation of sparse matrix computations for parallelism
    Kazem Cheshmi, S. Kamil, Michelle Mills Strout, and Maryam Mehri Dehnavi
    International Conference for High Performance Computing, Networking, Storage and Analysis, SC'18.

  • Extending index-array properties for data dependence analysis
    Mahdi Soltan Mohammadi, Kazem Cheshmi, Maryam Mehri Dehnavi, Anand Venkat, TomofumiYuki, and Michelle M. Strout
    International Workshop on Languages and Compilers for Parallel Computing, LCPC'18.

  • Sparsity-aware storage format selection
    Kazem Cheshmi, Leila Cheshmi, and Maryam Mehri Dehnavi.
    International Conference on High Performance Computing & Simulation, HPCS'18 (Best paper poster award).

  • Sympiler: Transforming Sparse Matrix Codes by Decoupling Symbolic Analysis
    Kazem Cheshmi, Shoaib Kamil, Michelle Mills Strout, and Maryam Mehri Dehnavi
    International Conference for High Performance Computing, Networking, Storage and Analysis, SC'17

  • A clustered GALS Network-on-Chip architecture with communication-aware mapping
    Kazem Cheshmi, Siamak Mohammadi, Daniel Versick, Djamshid Tavangarian, and Jelena Trajkovic
    Euromicro International Conference on Parallel, Distributed and Network-Based Processing, PDP'15.

  • Chameleon: Channel efficient optical network-on-chip
    Sebastien Le Beux, Hui Li, Ian O’Connor, Kazem Cheshmi, Xuchen Liu, Jelena Trajkovic, and Gabriela Nicolescu
    Design, Automation & Test in Europe Conference & Exhibition, DATE'14.

  • Quota setting router architecture for quality of service in GALS network-on-chip
    Kazem Cheshmi, Jelena Trajkovic, Mohammadreza Soltaniyeh, and Siamak Mohammadi
    International Symposium on Rapid System Prototyping, RSP'13.

Open-Source

  • Sympiler : is a domain-specific code-generator and library for sparse computation. It containts a set of techniques for loop tiling, splitting, fusion, and vectorization for sparse codes, such as sparse factorization and matrix multiplication.

  • NASOQ : is a scalable and efficient Quadratic Programming solver that obtains solutions for requested accuracies.




“Capable is he who is wise - Happiness from wisdom will arise.” - Ferdowsi