Kazem Shekofteh

Kazem Shekofteh

Kazem Shekofteh, Dr.

Kazem Shekofteh is a postdoctoral researcher at the Institute of Computer Engineering (ZITI), Heidelberg University, where he has been teaching since 2022. His research focuses on GPU computing, performance analysis of parallel programs, and high-performance computing for bioinformatics, with growing interest in emerging accelerators such as Tenstorrent and Graphcore IPUs. He received his PhD (2019) and MSc from Ferdowsi University of Mashhad, Iran, and previously held an assistant professorship at Shandiz Institute of Higher Education. He was a visiting scholar at Heidelberg University in late 2016, and his work has appeared in leading journals such as IEEE Transactions on Parallel and Distributed Systems. At Heidelberg he lectures on GPU Computing, Consistency and Coherence, and the Advanced Seminar. In 2026 he received a heiSKILLS Fellowship “KI in der Lehre (AI in Teaching)” for an AI-powered personalized assignment system.

Research Interests

  • Performance — characterization and analysis of parallel programs on GPUs and accelerators
  • Emerging hardware — algorithm design for Tenstorrent and Graphcore IPU
  • Parallel computing challenges — irregularity and sparsity
  • Applications — bioinformatics on accelerators

Awards & Fellowships

Recent Service (last 4 years)

Co-Organizer

  • 2026 — Workshop on IoT, Edge, and Mobile for Embedded Machine Learning (ITEM), Naples, Italy, September 11.

Co-Chair

Program Committee Member

Poster Committee Member

Student Volunteer Co-Chair

Reviewer

Journals

Conferences

Publications

  1. Philipp Hematty, S.-Kazem Shekofteh and Holger Fröning
    Parameter Savings Are Not Memory Savings: Butterfly Attention on a Tile Processor
    24th International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms (HeteroPar 2026) , collocated with Euro-PAR (accepted for publication), 2026
    link | bib 
    @inproceedings{HeteroPar2026PhilippKazem,
  title = {Parameter Savings Are Not Memory Savings: Butterfly Attention on a Tile Processor},
  author = {Hematty, Philipp and Shekofteh, S.-Kazem and Fr{{\"o}}ning, Holger},
  booktitle = {24th International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms (HeteroPar 2026) , collocated with Euro-PAR (accepted for publication)},
  year = {2026},
  url = {https://eventi.unibo.it/heteropar2026}
}
  2. S.-Kazem Shekofteh, Nils Kochendörfer and Holger Fröning
    Characterizing Needleman-Wunsch Sequence Alignment on the Graphcore IPU: Memory Trade-offs and Optimization Strategies
    32nd International European Conference on Parallel and Distributed Computing (Euro-PAR) (accepted for publication) (Euro-PAR), 2026
    link | bib 
    @inproceedings{EuroPAR2026ipu,
  title = {Characterizing Needleman-Wunsch Sequence Alignment on the Graphcore IPU: Memory Trade-offs and Optimization Strategies},
  author = {Shekofteh, S.-Kazem and Kochendörfer, Nils and Fr{{\"o}}ning, Holger},
  booktitle = {32nd International European Conference on Parallel and Distributed Computing (Euro-PAR) (accepted for publication)},
  series = {Euro-PAR},
  year = {2026},
  url = {https://2026.euro-par.org/program/main-conference}
}
  3. S.-Kazem Shekofteh, Nils Kochendörfer and Holger Fröning
    Are IPUs Viable for Genomic Sequence Alignment? Performance and Memory Analysis
    Workshop Biosignale 2026, Innsbruck, Austria, 2026
    link | arXiv | bib 
    @inproceedings{Biosignale2026Kazem,
  title = {Are IPUs Viable for Genomic Sequence Alignment? Performance and Memory Analysis},
  author = {Shekofteh, S.-Kazem and Kochendörfer, Nils and Fr{{\"o}}ning, Holger},
  booktitle = {Workshop Biosignale 2026, Innsbruck, Austria},
  year = {2026},
  url = {https://doi.org/10.15203/99106-194-6-21},
}
  4. Arjan Siddhpura, S.-Kazem Shekofteh and Holger Fröning
    Comparative Benchmark of Scaled-YOLOv4 on Graphcore IPU and NVIDIA GPU
    19th Workshop on Design and Architectures for Signal and Image Processing (DASIP 2026), collocated with HiPEAC Conference, 2026
    link | bib 
    @inproceedings{DASIP26ArjanKazem,
  title = {Comparative Benchmark of Scaled-YOLOv4 on Graphcore IPU and NVIDIA GPU},
  author = {Siddhpura, Arjan and Shekofteh, S.-Kazem and Fr{{\"o}}ning, Holger},
  booktitle = {19th Workshop on Design and Architectures for Signal and Image Processing (DASIP 2026), collocated with HiPEAC Conference},
  year = {2026},
  url = {https://link.springer.com/chapter/10.1007/978-3-032-23871-9_9}
}
  5. S.-Kazem Shekofteh, Daniel Bogacz, Christian Alles and Holger Fröning
    Butterfly factorization for vision transformers on multi-IPU systems
    Parallel Computing, 127, 103165, 2026
    link | bib 
    @article{SHEKOFTEH2026103165,
  title = {Butterfly factorization for vision transformers on multi-IPU systems},
  journal = {Parallel Computing},
  volume = {127},
  pages = {103165},
  year = {2026},
  issn = {0167-8191},
  doi = {https://doi.org/10.1016/j.parco.2025.103165},
  url = {https://www.sciencedirect.com/science/article/pii/S0167819125000419},
  author = {Shekofteh, S.-Kazem and Bogacz, Daniel and Alles, Christian and Fr{{\"o}}ning, Holger},
  keywords = {Intelligence processing units, Butterfly factorization, Vision transformer, Machine learning, Model compression}
}
  6. S.-Kazem Shekofteh, Christian Alles and Holger Fröning
    Reducing Memory Requirements for the IPU using Butterfly Factorizations
    SC ’23 Workshops of The International Conference on High Performance Computing, Network, Storage, and Analysis, SC-W 2023, Denver, CO, USA, November 12-17, 2023, 1255–1263, ACM, 2023
    link | arXiv | bib 
    @inproceedings{DBLP:conf/sc/ShekoftehAF23,
  author = {Shekofteh, S.-Kazem and Alles, Christian and Fr{\"{o}}ning, Holger},
  title = {Reducing Memory Requirements for the {IPU} using Butterfly Factorizations},
  booktitle = {{SC} '23 Workshops of The International Conference
                    on High Performance Computing, Network, Storage, and Analysis, {SC-W}
                    2023, Denver, CO, USA, November 12-17, 2023},
  pages = {1255--1263},
  publisher = {{ACM}},
  year = {2023},
  url = {https://doi.org/10.1145/3624062.3624196},
  doi = {10.1145/3624062.3624196},
  timestamp = {Tue, 28 Nov 2023 00:00:00 +0100},
}
  7. S.-Kazem Shekofteh, Christian Alles, Nils Kochendörfer and Holger Fröning
    On Performance Analysis of Graphcore IPUs: Analyzing Squared and Skewed Matrix Multiplication
    CoRR, abs/2310.00256, 2023
    link | bib 
    @article{DBLP:journals/corr/abs-2310-00256,
  author = {Shekofteh, S.-Kazem and Alles, Christian and Kochend{\"{o}}rfer, Nils and Fr{\"{o}}ning, Holger},
  title = {On Performance Analysis of Graphcore IPUs: Analyzing Squared and Skewed
                    Matrix Multiplication},
  journal = {CoRR},
  volume = {abs/2310.00256},
  year = {2023},
  url = {https://arxiv.org/abs/2310.00256},
  doi = {10.48550/ARXIV.2310.00256},
  eprinttype = {arXiv},
  eprint = {2310.00256},
  timestamp = {Wed, 18 Oct 2023 01:00:00 +0200},
}
  8. S.-Kazem Shekofteh, Hamid Noori, Mahmoud Naghibzadeh, Holger Fröning and Hadi Sadoghi Yazdi
    cCUDA: Effective Co-Scheduling of Concurrent Kernels on GPUs
    IEEE Trans. Parallel Distributed Syst., 31(4), 766–778, 2020
    link | bib 
    @article{DBLP:journals/tpds/ShekoftehNNFY20,
  author = {Shekofteh, S.-Kazem and Noori, Hamid and Naghibzadeh, Mahmoud and Fr{\"{o}}ning, Holger and Yazdi, Hadi Sadoghi},
  title = {cCUDA: Effective Co-Scheduling of Concurrent Kernels on GPUs},
  journal = {{IEEE} Trans. Parallel Distributed Syst.},
  volume = {31},
  number = {4},
  pages = {766--778},
  year = {2020},
  url = {https://doi.org/10.1109/TPDS.2019.2944602},
  doi = {10.1109/TPDS.2019.2944602},
  timestamp = {Fri, 02 Oct 2020 01:00:00 +0200},
}
  9. S.-Kazem Shekofteh, Hamid Noori, Mahmoud Naghibzadeh, Hadi Sadoghi Yazdi and Holger Fröning
    Metric Selection for GPU Kernel Classification
    ACM Trans. Archit. Code Optim., 15(4), 68:1–68:27, 2019
    link | bib 
    @article{DBLP:journals/taco/ShekoftehNNYF19,
  author = {Shekofteh, S.-Kazem and Noori, Hamid and Naghibzadeh, Mahmoud and Yazdi, Hadi Sadoghi and Fr{\"{o}}ning, Holger},
  title = {Metric Selection for {GPU} Kernel Classification},
  journal = {{ACM} Trans. Archit. Code Optim.},
  volume = {15},
  number = {4},
  pages = {68:1--68:27},
  year = {2019},
  url = {https://doi.org/10.1145/3295690},
  doi = {10.1145/3295690},
  timestamp = {Sat, 08 Jan 2022 00:00:00 +0100},
}
  1. Mohamad Beheshti Roui, S.-Kazem Shekofteh, Hamid Noori and Ahad Harati
    Efficient scheduling of streams on GPGPUs
    The Journal of Supercomputing, 76(11), 9270–9302, 2020
    link | bib 
    @article{roui2020efficient,
  author = {Beheshti Roui, Mohamad and Shekofteh, S.-Kazem and Noori, Hamid and Harati, Ahad},
  date = {2020/11/01},
  date-added = {2024-04-04 16:55:55 +0200},
  date-modified = {2024-04-04 16:56:58 +0200},
  doi = {10.1007/s11227-020-03209-x},
  id = {Beheshti Roui2020},
  isbn = {1573-0484},
  journal = {The Journal of Supercomputing},
  number = {11},
  pages = {9270--9302},
  title = {Efficient scheduling of streams on GPGPUs},
  url = {https://doi.org/10.1007/s11227-020-03209-x},
  volume = {76},
  year = {2020},
  bdsk-url-1 = {https://doi.org/10.1007/s11227-020-03209-x}
}
  2. F. Khorshahiyan, S.-Kazem Shekofteh and H. Noori
    Predicting Execution Time of CUDA Kernels with Unified Memory Capability
    2019 9th International Conference on Computer and Knowledge Engineering (ICCKE), 2019 9th International Conference on Computer and Knowledge Engineering (ICCKE), 437–443, 2019
    link | bib 
    @proceedings{khorshahiyan2019predicting,
  author = {Khorshahiyan, F. and Shekofteh, S.-Kazem and Noori, H.},
  booktitle = {2019 9th International Conference on Computer and Knowledge Engineering (ICCKE)},
  date-added = {2024-04-04 16:57:38 +0200},
  date-modified = {2024-04-04 16:58:19 +0200},
  doi = {10.1109/ICCKE48569.2019.8964952},
  isbn = {2643-279X},
  journal = {2019 9th International Conference on Computer and Knowledge Engineering (ICCKE)},
  journal1 = {2019 9th International Conference on Computer and Knowledge Engineering (ICCKE)},
  pages = {437--443},
  title = {Predicting Execution Time of CUDA Kernels with Unified Memory Capability},
  year = {2019},
  year1 = {24-25 Oct. 2019},
  url = {https://doi.org/10.1109/ICCKE48569.2019.8964952}
}
  3. Ahmadreza Montazerolghaem, S.-Kazem Shekofteh, M. H. Yaghmaee and Mahmoud Naghibzadeh
    A load scheduler for SIP proxy servers: design, implementation and evaluation of a history weighted window approach
    International Journal of Communication Systems, 30(3), e2980, John Wiley & Sons, Ltd, 2017
    link | bib 
    @article{montazerolghaem2017load,
  author = {Montazerolghaem, Ahmadreza and Shekofteh, S.-Kazem and Yaghmaee, M. H. and Naghibzadeh, Mahmoud},
  date = {2017/02/01},
  date-added = {2024-04-04 17:00:11 +0200},
  date-modified = {2024-04-04 17:00:43 +0200},
  doi = {https://doi.org/10.1002/dac.2980},
  isbn = {1074-5351},
  journal = {International Journal of Communication Systems},
  journal1 = {International Journal of Communication Systems},
  journal2 = {International Journal of Communication Systems},
  journal3 = {Int J Commun Syst},
  keywords = {load balancer; scheduler; session initiation protocol; asterisk; overload},
  month = {2024/04/04},
  n2 = {Summary The widespread use of Session Initiation Protocol as a signalling protocol has created various challenges. An important one is that its throughput can be severely degraded when an overload happens in the proxy server because of several retransmissions from the user agent. One common approach to overcome this problem is ?load balancing?. A balancer needs to know the status of proxy servers, which are continuously gathered implicitly or explicitly. Implicit methods have averagely less overhead than explicit ones. This paper attempts to prevent throughput reduction by balancing the loads among available proxy servers properly using an implicit mechanism called History Weighted Average Response time. The proposed algorithm is robust because it incurs no extra processing to proxy servers. The novelty of the mechanism is making use of ?response time history? to estimate the load being currently processed on servers. By implementing in a real testbed, throughput and scalability are improved compared with an important state-of-the-art similar algorithm. This improvement stems from no need for modification in SIP protocol, easy implementation and application, simple computations for making decision and no need for extra feedback between servers and load balancer. Copyright ? 2015 John Wiley \& Sons, Ltd.},
  number = {3},
  pages = {e2980},
  publisher = {John Wiley \& Sons, Ltd},
  title = {A load scheduler for SIP proxy servers: design, implementation and evaluation of a history weighted window approach},
  url = {https://doi.org/10.1002/dac.2980},
  volume = {30},
  year = {2017},
  year1 = {2017},
}
  4. A. Montazerolghaem, S.-Kazem Shekofteh, G. Khojaste, M. Naghibzadeh and M. -H. Yaghmaee-M
    A novel load scheduling for session initiation protocol networks
    2014 4th International Conference on Computer and Knowledge Engineering (ICCKE), 2014 4th International Conference on Computer and Knowledge Engineering (ICCKE), 509–514, 2014
    link | bib 
    @proceedings{montazerolghaem2014novel,
  author = {Montazerolghaem, A. and Shekofteh, S.-Kazem and Khojaste, G. and Naghibzadeh, M. and Yaghmaee-M, M. -H.},
  booktitle = {2014 4th International Conference on Computer and Knowledge Engineering (ICCKE)},
  date-added = {2024-04-04 17:01:42 +0200},
  date-modified = {2024-04-04 17:01:53 +0200},
  doi = {10.1109/ICCKE.2014.6993376},
  journal = {2014 4th International Conference on Computer and Knowledge Engineering (ICCKE)},
  journal1 = {2014 4th International Conference on Computer and Knowledge Engineering (ICCKE)},
  pages = {509--514},
  title = {A novel load scheduling for session initiation protocol networks},
  year = {2014},
  year1 = {29-30 Oct. 2014},
  url = {https://doi.org/10.1109/ICCKE.2014.6993376}
}
  5. Javad Mohebbi Najm Abad, S.-Kazem Shekofteh, Hamid Tabatabaee and Maryam Mehrnejad
    CoreIIScheduler: Scheduling Tasks in a Multi-core-Based Grid Using NSGA-II Technique
    Intelligent Informatics, 507–518, Springer Berlin Heidelberg, 2013
    link | bib 
    @proceedings{najmabad2013corell,
  address = {Berlin, Heidelberg},
  author = {Najm Abad, Javad Mohebbi and Shekofteh, S.-Kazem and Tabatabaee, Hamid and Mehrnejad, Maryam},
  booktitle = {Intelligent Informatics},
  date = {2013//},
  date-added = {2024-04-04 17:02:50 +0200},
  date-modified = {2024-04-04 17:03:11 +0200},
  editor = {Abraham, Ajith and Thampi, Sabu M},
  id = {10.1007/978-3-642-32063-7{\_}54},
  isbn = {978-3-642-32063-7},
  pages = {507--518},
  publisher = {Springer Berlin Heidelberg},
  title = {CoreIIScheduler: Scheduling Tasks in a Multi-core-Based Grid Using NSGA-II Technique},
  year = {2013},
  url = {https://doi.org/10.1007/978-3-642-32063-7_54}
}
  6. S.-Kazem Shekofteh, Hossein Deldari and Maryam Baradaran Khalkhali
    Reducing cache contention in a multi-core processor via a scheduler
    2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE), 6, V6-555-V6-558, 2010
    link | bib 
    @inproceedings{shekofteh2010reducing,
  author = {Shekofteh, S.-Kazem and Deldari, Hossein and Khalkhali, Maryam Baradaran},
  booktitle = {2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE)},
  date-added = {2024-04-04 17:06:21 +0200},
  date-modified = {2024-04-04 17:06:30 +0200},
  doi = {10.1109/ICACTE.2010.5579213},
  keywords = {component;multi-core architecture;resource contention;shared cache;thread scheduling},
  pages = {V6-555-V6-558},
  title = {Reducing cache contention in a multi-core processor via a scheduler},
  volume = {6},
  year = {2010},
  url = {https://doi.org/10.1109/ICACTE.2010.5579213}
}
  7. H. Salami, H. Saadatfar, Farhad Rahmani Fard, S.-Kazem Shekofteh and H. Deldari
    Improving cluster computing performance based on job futurity prediction
    2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE), 2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE), 6, V6-303–V6-307, 2010
    link | bib 
    @proceedings{salami2010improving,
  author = {Salami, H. and Saadatfar, H. and Fard, Farhad Rahmani and Shekofteh, S.-Kazem and Deldari, H.},
  booktitle = {2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE)},
  date-added = {2024-04-04 17:05:56 +0200},
  date-modified = {2024-04-04 17:06:09 +0200},
  doi = {10.1109/ICACTE.2010.5579820},
  isbn = {2154-7505},
  journal = {2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE)},
  journal1 = {2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE)},
  pages = {V6-303--V6-307},
  title = {Improving cluster computing performance based on job futurity prediction},
  vo = {6},
  volume = {6},
  year = {2010},
  year1 = {20-22 Aug. 2010},
  url = {https://doi.org/10.1109/ICACTE.2010.5579820}
}
  8. M. Baradaran -Khalkhali, S.-Kazem Shekofteh, S. Toosizadeh and M. . -R. Akbarzadeh -T
    Exploiting fuzzy approximator to head pose estimation
    Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2010, Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2010, 68–72, 2010
    | bib 
    @proceedings{khalkhali2010exploiting,
  author = {-Khalkhali, M. Baradaran and Shekofteh, S.-Kazem and Toosizadeh, S. and -T, M. . -R. Akbarzadeh},
  booktitle = {Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2010},
  date-added = {2024-04-04 17:04:59 +0200},
  date-modified = {2024-04-04 17:05:17 +0200},
  isbn = {2326-0319},
  journal = {Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2010},
  journal1 = {Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2010},
  pages = {68--72},
  title = {Exploiting fuzzy approximator to head pose estimation},
  year = {2010},
  year1 = {23-25 Sept. 2010}
}
  9. S.-Kazem Shekofteh, M. Baradaran-K, S. Toosizadeh, M. -R. Akbarzadeh-T and M. Hashemi
    Head pose estimation using fuzzy approximator augmented by redundant membership functions
    2010 2nd International Conference on Software Technology and Engineering, 2010 2nd International Conference on Software Technology and Engineering, 2, V2-306–V2-310, 2010
    link | bib 
    @proceedings{shekofteh2010head,
  author = {Shekofteh, S.-Kazem and Baradaran-K, M. and Toosizadeh, S. and Akbarzadeh-T, M. -R. and Hashemi, M.},
  booktitle = {2010 2nd International Conference on Software Technology and Engineering},
  date-added = {2024-04-04 17:03:53 +0200},
  date-modified = {2024-04-04 17:04:03 +0200},
  doi = {10.1109/ICSTE.2010.5608799},
  journal = {2010 2nd International Conference on Software Technology and Engineering},
  journal1 = {2010 2nd International Conference on Software Technology and Engineering},
  pages = {V2-306--V2-310},
  title = {Head pose estimation using fuzzy approximator augmented by redundant membership functions},
  vo = {2},
  volume = {2},
  year = {2010},
  year1 = {3-5 Oct. 2010},
  url = {https://doi.org/10.1109/ICSTE.2010.5608799}
}
  10. M. Baradaran-K, S.-Kazem Shekofteh, S. Toosizadeh and M. -R. Akbarzadeh-T
    A fuzzy approximator with Gaussian membership functions to estimate a human’s head pose
    2010 10th International Conference on Intelligent Systems Design and Applications, 2010 10th International Conference on Intelligent Systems Design and Applications, 1154–1158, 2010
    link | bib 
    @proceedings{baradaran2010fuzzy,
  author = {Baradaran-K, M. and Shekofteh, S.-Kazem and Toosizadeh, S. and Akbarzadeh-T, M. -R.},
  booktitle = {2010 10th International Conference on Intelligent Systems Design and Applications},
  date-added = {2024-04-04 17:03:24 +0200},
  date-modified = {2024-04-04 17:03:34 +0200},
  doi = {10.1109/ISDA.2010.5687029},
  isbn = {2164-7151},
  journal = {2010 10th International Conference on Intelligent Systems Design and Applications},
  journal1 = {2010 10th International Conference on Intelligent Systems Design and Applications},
  pages = {1154--1158},
  title = {A fuzzy approximator with Gaussian membership functions to estimate a human's head pose},
  year = {2010},
  year1 = {29 Nov.-1 Dec. 2010},
  url = {https://doi.org/10.1109/ISDA.2010.5687029}
}

Connctions and External Links