Photo from Moahim (CC BY-SA 3.0 pl)

IMPACT 2026

16th International Workshop on Polyhedral Compilation Techniques

January 28, 2026 | Kraków, Poland In conjunction with HiPEAC 2026, January 26-28, 2026

Polyhedral compilation techniques have been at the forefront of research and innovation, playing a central role in a wide range of domain-specific languages and optimizing compilers. Thanks to a unified formalism for parallelization and memory optimization, they provide high impact in competitive areas such as machine learning, scientific computing, and media processing. Like a well-aged bottle of wine, these techniques have evolved and branched into new domains over time, with successful applications spanning across various products.

IMPACT is a unique event focusing on polyhedral compilation. The workshop brings together researchers and practitioners for a high-quality one-day event including a keynote, paper presentations, and work-in-progress discussions. The workshop has a well established tradition of high quality, extensive peer reviews, with a PC balancing core polyhedral expertise with applications and broader computing systems research.

Program

ICE Kraków - Room: S11 (L3)

HiPEAC session site

• Opening
  [Slides]
• Keynote: Retire Linear Algebra Libraries - 4 years later
  Albert Cohen
  [Abstract]
  Four years ago at the HiPEAC Industry Session, and later again in a couple of invited presentations, I argued for retiring expert-designed linear algebra libraries. More broadly, the motivation for automating the production of highly tuned performance software has only grown more critical since then. I also advocated for compiler technology reinventing itself in this context. First, make it an effective goal of (domain specific) compilers: summit the peak, don’t just speed up the valley walk to the base of the mountain! Then, in doing so, don’t sacrifice all the learnings of reusable compiler construction: your compiler can be more powerful than you think! But progress has been slow. We have even witnessed regression to less automated processes in specific scenarios, driven by urgency and fast-paced hardware innovation. I will reflect on this trend, on progress made, on hardships met along the way, on the positioning of polyhedral compilation technology in this race, and on the supporting and competing roles of generative AI.
  [Bio]
  Albert Cohen is passionate about programming languages and compilers, and leveraging ML to automate the most complex problems in code generation and HW/SW codesign. Graduating from ENS Lyon and Université Paris Saclay, he joined INRIA and École Polytechnique (part-time associate professor). He has been a visiting scholar at the University of Illinois, a visiting professor at Philips Research and Facebook Artificial Intelligence Research, before joining Google in 2018. His work on programming languages and high-performance systems resulted in 250 peer-reviewed publications together with 30 PhD students and international collaborators, with technology transfer to production compilers. Albert played a pioneering role in the design and adoption of MLIR for machine learning and domain-specific computing.
  [Slides]

• Z-polyhedra and LBLs in PolyLib
  Vincent Loechner, Dhimiter Riza
  [Abstract]

  Z-polyhedra and linearly bounded lattices (LBLs) provide expressive representations for sets of integer points beyond classical rational or integer polyhedra. Although Z-polyhedra were introduced in PolyLib more than two decades ago, their implementation was limited and did not support robust manipulation of general unions of Z-polyhedra or LBLs.

  In this paper, we present a sound, complete and unified implementation of Z-polyhedra, LBLs, and their finite unions in PolyLib. Our approach relies on the representation of an LBL as the affine lattice function image of a coordinate polyhedron. A normalization scheme enables union, intersection, difference, image, preimage, inclusion testing, and conversion of a union of LBLs to a union of Z-domains. Our implementation is using efficient algorithms whenever possible but is also able to compute the most complex by nature cases, carefully handling lattice normalization, equality elimination and holes. It has been validated on a wide range of test cases.

  This work significantly extends PolyLib’s capabilities and makes advanced lattice-based representations practically usable in polyhedral compilation and related applications.

  [Paper] [Slides] [BibTeX]

              @inproceedings{loechner26-paper1,
                title     = {{Z}-polyhedra and {LBL}s in {PolyLib}},
                author    = {Loechner, Vincent and Riza, Dhimiter},
                booktitle = {16th International Workshop on Polyhedral Compilation Techniques
                             (IMPACT 2026, in conjunction with HiPEAC 2026)},
                year      = 2026,
                url       = {https://impact-workshop.org/impact2026/#loechner26-paper1},
              }
            

• Towards Optimising Programs with Sketch-Guided Polyhedral Compilation
  Valeran Maytié, Reuben Carolan, Christophe Alias, Cédric Bastoul, Thomas Kœhler
  [Abstract]

  When programmers use semi-automatic compilers, they typically write optimisation scripts, to guide the compiler towards key optimisations. Instead of writing scripts, it may be preferable to write sketches that focus on the desired structure of the optimised code, without worrying about individual transformations.

  In this article, we present a new semi-automatic, sketch-guided compilation approach. We introduce a sketch language that enables expressing the result of imperative loop transformations and a new polyhedral algorithm capable of generating code constrained by both a sketch and a computation specification.

  [Paper] [Slides] [BibTeX]

              @inproceedings{maytie26-paper2,
                title     = {Towards Optimising Programs with Sketch-Guided Polyhedral Compilation},
                author    = {Maytié, Valeran and Carolan, Reuben and Alias, Christophe and Bastoul, Cédric and Kœhler, Thomas},
                booktitle = {16th International Workshop on Polyhedral Compilation Techniques
                             (IMPACT 2026, in conjunction with HiPEAC 2026)},
                year      = 2026,
                url       = {https://impact-workshop.org/impact2026/#maytie26-paper2},
              }
            

• Guiding Polyhedral Scheduling for Vectorization through Constraints Generated from an SLP Algorithm
  Tom Hammer, Stéphane Genaud, Vincent Loechner
  [Abstract]

  Polyhedral schedulers present well established techniques to extract parallelism, improve data locality, and generate tiled code for statically analyzable loops. However, as the polyhedral model abstracts programs in a mathematical representation detached from language, architectural, and hardware specific constraints, encoding vectorization in an affine form can prove challenging.

  In this paper, we present an approach to integrate information on vectorization decisions made by an SLP algorithm (Autovesk) into a polyhedral compiler (Pluto) through the addition of constraints to the schedule. We execute the SLP vectorization algorithm preserving annotated statement instance information. From its output, we create a set of constraints aiming to enforce vectorization. These optional constraints are injected during the scheduling process of the polyhedral compiler. We evaluate the performance and make use of hardware counters to check the relevancy of our method on the Polybench/C suite.

  [Paper] [BibTeX]

              @inproceedings{hammer26-paper3,
                title     = {Guiding Polyhedral Scheduling for Vectorization through Constraints Generated from an {SLP} Algorithm},
                author    = {Hammer, Tom and Genaud, Stéphane and Loechner, Vincent},
                booktitle = {16th International Workshop on Polyhedral Compilation Techniques
                             (IMPACT 2026, in conjunction with HiPEAC 2026)},
                year      = 2026,
                url       = {https://impact-workshop.org/impact2026/#hammer26-paper3},
              }
            

• New Insights on Scalar Promotion with the Polyhedral Model
  Alec Sadler, Nathan Chandanson, Hugo Thievenaz, Christophe Alias
  [Abstract]

  Memory accesses are a well known bottleneck whose impact might be mitigated by using properly the memory hierarchy until registers. In this paper, we address scalar promotion, a technique to turn temporary arrays into a collection of scalar variables to be allocated to registers. We revisit array scalarization in the light of the recent advances of the polyhedral model. We propose a general algorithm for array scalarization and we show a scalarization of stencil computations thanks to a preliminary preprocessing. Our scalarization algorithm operates on the polyhedral intermediate representation and could be plugged in a polyhedral compiler among other passes. In particular, our scalarization algorithm is parametrized by the program schedule, possibly computed by a previous compilation pass. We present a preliminary experimental validation with promising results.

  [Paper] [Slides] [BibTeX]

              @inproceedings{sadler26-paper4,
                title     = {New Insights on Scalar Promotion with the Polyhedral Model},
                author    = {Sadler, Alec and Chandanson, Nathan and Thievenaz, Hugo and Alias, Christophe},
                booktitle = {16th International Workshop on Polyhedral Compilation Techniques
                             (IMPACT 2026, in conjunction with HiPEAC 2026)},
                year      = 2026,
                url       = {https://impact-workshop.org/impact2026/#sadler26-paper4},
              }
            

Ask questions and interact with the polyhedral community.

Feel free to come with an unscheduled set of slides and wild ideas to discuss spontaneously.

• [Slides Wild Idea]

Call For Papers

We welcome both theoretical and experimental papers and presentations on all aspects of polyhedral compilation. We also welcome submissions describing preliminary results, crazy new ideas, position papers, experience reports, education material, and available tools, with an aim to stimulate discussions, collaborations, and advances in the field. The following illustrates potential IMPACT papers:

• Thorough theoretical discussion of a preliminary idea with an attempt to place it in context but no experimental results.
• Experimental results comparing two or more existing ideas, followed by a detailed analysis.
• Presentation of an existing idea in a different way, including illustrations of how the idea applies to new use cases, code, architectures, etc. Attribution should as clear as possible.

Topics of interest include, but are not limited to:

• program optimization (automatic parallelization, tiling, etc.);
• code generation;
• data/communication management on GPUs, accelerators and distributed systems;
• hardware/high-level synthesis for affine programs;
• static analysis;
• program verification;
• model checking;
• theoretical foundations of the polyhedral model;
• extensions of the polyhedral model;
• scalability and robustness of polyhedral compilation techniques.

Important Dates

Submission

Submissions

Paper submissions should not exceed 10 pages (recommended 8 pages), excluding references, formatted as per ACM SIGPLAN proceedings format. Short paper submissions, even with only 2 pages, and presentation-only submissions are welcome as well.

Please use version 1.54 or above of the following templates to prepare your manuscript: https://www.acm.org/publications/proceedings-template

Make sure to use the "sigplan" subformat.

Visit http://sigplan.org/Resources/Author/ for further information on SIGPLAN manuscript formatting.

NOTE: older versions of the article template use smaller fonts for submission, please double-check that you are using the recent style file (in particular, various LaTeX distributions ship older versions of the acmart style file, please download the most recent one from ACM).

Submissions should use PDF format and be printable on US Letter or A4 paper.

Please submit your manuscripts through EasyChair: https://easychair.org/conferences/?conf=impact2026

Proceedings

Proceedings will be posted online. If the final version of an accepted paper does not sufficiently address the comments of the reviewers, then it may be accompanied by a note from the program committee. Publication at IMPACT will not prevent later publication in conferences or journals of the presented work. However, simultaneous submission to IMPACT and other workshop, conference, or journal is often prohibited by the policy of other venues. For instance, a manuscript overlapping significantly with IMPACT submission cannot be submitted to PLDI 2026 or any other overlapping SIGPLAN event.

Presentations

The presentation should take no longer than 25 minutes to present. Please make sure that at least one of the authors can attend the workshop if your work is accepted. Registration are processed through the HiPEAC website.

Workshop Chair

Organizing Committee

Contact Us

For any inquiries or questions you may have regarding the IMPACT 2026 workshop, please feel free to reach out to Guillaume Iooss and Vincent Loechner at guillaume.iooss [-a-] inria.fr and loechner [-a-] unistra.fr.