LEMAITRE Florian
Supervision : Lionel LACASSAGNE
Co-supervision : Benjamen Couturier (CERN)
Tracking haute fréquence pour architectures SIMD : optimisation de la reconstruction LHCb
During this thesis, we studied linear algebra systems with small matrices (typically from 2x2 to 5x5) used within the LHCb experiment (and also in other domains like computer vision). Linear algebra libraries like Eigen, Magma or the MKL are not optimized for such small matrices.
We used and combined many well-known transforms helping SIMD and some unusual transforms like the fast reciprocal square root computation. We wrote a code generator in order to simplify the use of such transforms and to have a portable code.
We tested these optimizations and analyzed their impact on the speed of simple algorithm. Batch processing in SoA is crucial to process fast these small matrices. We also analyzed how the accuracy of the results depends on the precision of the data.
We implemented these transforms in order to speed-up the Cholesky factorization of small matrices (up to 12x12). The processing speed is capped if the fast reciprocal square root computation is not used. We got a speed up between x10 and x33 using F32.
Finally, we studied and sped up the Kalman filter in its general form. Our 4x4 F32 implementation is x90 faster. The Kalman filter used within LHCb has been sped up by x2.2 compared to the current SIMD version and by at least x2.3 compared to filters used other high energy physics experiments.
Defence : 02/13/2019
Jury members :
Albert COHEN Google [Rapporteur]
Daniel MENARD IETR (Université de Rennes) [Rapporteur]
Lionel LACASSAGNE LIP6 (Sorbonne Université)
Emmanuel CHAILLOUX LIP6 (Sorbonne Université)
Michèle GOUIFFÈS LIMSI (Université Paris-Sud)
Bertrand LE GAL IMS (Université de Bordeaux)
2016-2023 Publications
-
2023
- Th. Romera, A. Petreto, F. Lemaitre, M. Bouyer, Quentin L. Meunier, L. Lacassagne, D. Etiemble : “Optical flow algorithms optimized for speed, energy and accuracy on embedded GPUs”, Journal of Real-Time Image Processing, vol. 20 (2), pp. 32, (Springer Verlag) (2023)
-
2022
- N. Maurice, F. Lemaitre, J. Sopena, L. Lacassagne : “LSL3D : Etiquetage en Composantes Connexe par segments pour volumes 3D”, COMPAS 2022 - ConfĂ©rence francophone d'informatique en ParallĂ©lisme, Architecture et Système, Amiens, France (2022)
- F. Lemaitre, N. Maurice, L. Lacassagne : “An efficient run-based Connected Component Labeling algorithm for processing holes”, Binary is the new Black and White workshop @ IEEE ICIAP 2022, Lecce, Italy (2022)
- N. Maurice, F. Lemaitre, J. Sopena, L. Lacassagne : “LSL3D: a run-based Connected Component Labeling algorithm for 3D volumes”, Binary is the new Black and White workshop @ IEEE ICIAP 2022, Lecce, Italy (2022)
- N. Blin, E. Carlinet, F. Lemaitre, L. Lacassagne, Th. GĂ©raud : “Max-tree Computation on GPUs”, (2022)
- M. Millet, N. Rambaux, A. Petreto, F. Lemaitre, L. Lacassagne : “Meteorix - A new processing chain for real-time detection and tracking of meteors from space”, WGN, Journal of the International Meteor Organization, vol. 49 (6), (International Meteor Organization) (2022)
-
2021
- M. Millet, N. Rambaux, A. Petreto, F. Lemaitre, L. Lacassagne : “Meteorix: a new processing chain for detection and tracking of meteors from space”, IMC 2021 - International Meteor Conference, confĂ©rence virtuelle, France (2021)
- M. Millet, N. Rambaux, A. Petreto, F. Lemaitre, L. Lacassagne : “DĂ©tection temps rĂ©el de mĂ©tĂ©ores Ă bord d’un nanosatellite, application au projet Meteorix”, ORASIS 2021, Saint FerrĂ©ol, France (2021)
- Th. Romera, A. Petreto, F. Lemaitre, M. Bouyer, Q. Meunier, L. Lacassagne : “Implementations Impact on Iterative Image Processing for Embedded GPU”, European Signal Processing Conference (EUSIPCO), Dublin, Ireland (2021)
- F. Lemaitre, A. Hennequin, L. Lacassagne : “Taming Voting Algorithms on Gpus for an Efficient Connected Component Analysis Algorithm”, ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, Canada, pp. 7903-7907, (IEEE) (2021)
- F. Lemaitre, A. Hennequin, L. Lacassagne : “Taming Voting Algorithms on GPUs for an Efficient Connected Component Analysis Algorithm”, GPU Technical Conference, San Jose, United States (2021)
- F. Lemaitre, L. Lacassagne : “A new run-based Connected Component Labeling for efficiently analyzing and processing holes”, (2021)
-
2020
- F. Lemaitre, A. Hennequin, L. Lacassagne : “How to speed Connected Component Labeling up with SIMD RLE algorithms”, Workshop on Programming Models for SIMD/Vector Processing (WPMVP@PPoPP), San Diego, Californie, United States (2020)
- A. Petreto, Th. Romera, F. Lemaitre, M. Bouyer, B. Gaillard, P. Menard, Q. Meunier, L. Lacassagne : “Real-time embedded video denoiser prototype”, 9th International Symposium - Optronics in Defense and Security (Optro), Paris, France (2020)
-
2019
- F. Lemaitre : “Optimisation et transformations d’algorithmes pour l’expĂ©rience LHCb”, thesis, phd defence 02/13/2019, supervision Lacassagne, Lionel, co-supervision : Benjamen, Couturier (CERN) (2019)
- A. Petreto, Th. Romera, F. Lemaitre, I. Masliah, B. Gaillard, M. Bouyer, Q. Meunier, L. Lacassagne : “DĂ©bruitage temps rĂ©el embarquĂ© pour vidĂ©os fortement bruitĂ©es”, COMPAS 2019, Anglet, France (2019)
- A. Petreto, Th. Romera, I. Masliah, B. Gaillard, M. Bouyer, Q. Meunier, L. Lacassagne, F. Lemaitre : “A New Real-Time Embedded Video Denoising Algorithm”, DASIP 2019 - The Conference on Design and Architectures for Signal and Image Processing, MontrĂ©al, Canada (2019)
-
2018
- F. Lemaitre, B. Couturier, L. Lacassagne : “Small SIMD Matrices for CERN High Throughput Computing”, WPMVP 2018 Workshop on Programming Models for SIMD/Vector Processing, Vienna, Austria, (ACM Press) (2018)
-
2017
- F. Lemaitre, B. Couturier, L. Lacassagne : “Cholesky Factorization on SIMD multi-core architectures”, Journal of Systems Architecture, (Elsevier) (2017)
-
2016
- F. Lemaitre, L. Lacassagne : “Batched Cholesky Factorization for tiny matrices”, Design and Architectures for Signal and Image Processing (DASIP), Rennes, France, pp. 1-8 (2016)