Denis Barthou

Software & Tools

I have directly contributed to MAQAO and AFF3CT software, that have now active communities of users, in both industry and academia.

MAQAO

Modular Assembly Quality Analyzed and Optimizer is a parallel performance analysis and optimization framework that I initiated in 2004. Designed for developers and performance experts, it provides advanced capabilities for analyzing and optimizing low-level code. MAQAO has stood the test of time: it is actively maintained and extended at the University of Versailles, supported by a long-lasting community of users and contributors. It has become part of the VI-HPS Institute, is listed among the EU Innovation Radar technologies, and has been featured in an AWS blog highlighting its potential. GitLab Web page

MAQAO allows to explore scalability at the loop level and to quickly identify which parts of the code are the limiting factors when scaling. By computing the difference between the theoretical maximum speed up and the measured speed up MAQAO provides an efficiency ratio indicating the loop efficiency for each configuration. Figure from AWS blog.

BER/FER display using AFF3CT simulation, here on polar codes.

AFF3CT

A Fast Forward Error Correction Toolbox is a high-performance simulation framework dedicated to Forward Error Correction (FEC, or channel coding). It was initiated during A. Cassagne’s PhD in 2017 and supports a wide range of codes—from well-established Turbo codes to modern Polar and LDPC codes. AFF3CT has grown into a vibrant, active and industrial community maintained and developped at Sorbonne University, IMS Lab/Bordeaux University and Inria. Users and developers gather annually during the AFF3CT Day, illustrating the maturity and engagement around the project. Thanks to its performance and robustness, AFF3CT is now used in both academic research and industrial settings. GitHub Web page

I have also contributed to other software tools through PhD supervision and research projects (e.g., MIPP, PARCOACH), focusing on SIMD abstraction, MPI correctness checking. These projects are actively maintained by their respective communities

• Optimal scheduling algorithms for software-defined radio pipelined and replicated task chains on multicore architectures, Diane Orhan, Laércio Lima Pilla, Denis Barthou, Adrien Cassagne, Olivier Aumage, Romain Tajan, Christophe Jégo, Camille Leroux. In Journal of Parallel and Distributed Computing, vol 204, 2025 [ DOI | http ]

  This paper presents an optimal and greedy scheduling method for streams on chains of tasks, using both replication (for parallelism) and pipeline, with limited number of cores. It optimizes the throughput of the stream.

• PolyTOPS: Reconfigurable and Flexible Polyhedral Scheduler. Gianpietro Consolaro, Zhen Zhang, Harenome Razanajato, Nelson Lossing, Nassim Tchoulak, Adilla Susungi, Artur Cesar Araujo Alves, Renwei Zhang, Denis Barthou, Corinne Ancourt, and Cédric Bastoul. IEEE/ACM International Symposium on Code Generation and Optimization (CGO), 2024. [ DOI | http ]

  PolyTOPS is a configurable, tunable polyhedral scheduler, applied to both compute intensive and AI codes. The configurability allows auto-tuning.

• Exploring scheduling algorithms for parallel task graphs: a modern game engine case study. Mustapha Regragui, Baptiste Coye, Laércio Lima Pilla, Raymond Namyst, and Denis Barthou. In International European Conference on Parallel and Distributed Computing (Euro-Par), 2022. [DOI | .pdf ]

  This paper describes how task scheduling methods can impact game engine performance. This was evaluated on Ubisoft game engines.

• AFF3CT: A Fast Forward Error Correction Toolbox! Adrien Cassagne, Olivier Hartmann, Mathieu Leonardon, Kun He, Camille Leroux, Romain Tajan, Olivier Aumage, Denis Barthou, Thibaud Tonnellier, Vincent Pignoly, Bertrand Le Gal, and Christophe Jego. SoftwareX, 2019. [ DOI | .pdf ]

  This is the reference paper for AFF3CT, proposing an efficient framework for the writing of error correcting codes in software defined radios.

• Adaptive Partitioning for Iterated Sequences of Irregular OpenCL Kernels. Pierre Huchant, Denis Barthou, and Marie-Christine Counilh. In SBAC-PAD - 30th International Symposium on Computer Architecture and High Performance Computing, 2018. [ DOI | .pdf ]

  The method proposed in this paper is to automatically split and distribute a kernel on multiple GPUs. Provided the kernel is repeatedly executed, a dynamic and efficient load balancing is performed, leading to optimal performance for instance on physic simulation such as n-body simulation.

• Rewriting System for Profile-Guided Data Layout Transformations on Binaries. Olivier Aumage, Christopher Haine, and Denis Barthou. In Int. European Conference on Parallel and Distributed Computing, 2017. [ bib | DOI | .pdf ]

  This paper proposes a method based on MAQAO that identifies the access patterns to the data (SoA, AoS, ...) and allows to evaluate the impact of changing this layout on performance. The transformation is directly performed on the assembly code and allows the programmer to quickly have an estimation before an expensive rewrite of the data layout.

• PARCOACH: Combining static and dynamic validation of MPI collective communications. Emmanuelle Saillard, Patrick Carribault, and Denis Barthou. International Journal of High Performance Computing Applications, 2014. [ DOI | .pdf ]

  Parcoach analyzes statically MPI collectives and adds code to prevent deadlocks whenever the situation may arise, identifying the root cause of the deadlock.

• Hydra: Automatic algorithm exploration from linear algebra equations. Alexandre Duchâteau, David A. Padua, and Denis Barthou. In IEEE/ACM International Symposium on Code Generation and Optimization (CGO), 2013. [ DOI ]

  Hydra takes a mathematical equation on matrices and decomposes it recursively in order to create parallel tasks. It enables the automatic production of efficient solvers requiring very little or no cod- ing at all and delivering performance approximating that of the highly tuned library routines such as Intel’s MKL

• Performance Tuning of x86 OpenMP Codes with MAQAO. Denis Barthou, Andres Charif Rubial, William Jalby, Souad Koliai, and Cédric Valensi. In Tools for High Performance Computing, 2009. [ DOI | .pdf ]

  The journal article is describing how, but analyzing the binary code, MAQAO can help for performance engineering and debugging on multicore architecture.

• Fuzzy Array Dataflow Analysis. Jean-Francois Collard, Denis Barthou, and Paul Feautrier. In ACM SIGPLAN Symposium on Principles & Practice of Parallel Programming (PPOPP), , 1995. [ DOI ] [ DOI ]

  This paper is on a new method to extend the polydral model to non-affine constraints, in particular for dependence analysis. That is the core of my PhD.