Romain Dolbeau
Fellow, HPC Software at SiPearl
Greater Rennes Metropolitan Area France
Summary
🤩
Rockstar🎓
Top SchoolRomain Dolbeau is a Fellow in HPC Software with 24 years of experience in high-performance systems, specializing in low-level optimization, parallel programming (MPI/OpenMP), and hardware acceleration (CUDA, SIMD). He blends deep academic roots in computer architecture with industrial leadership at SiPearl and Bull, advising on performance tuning and hardware/software co-design. A hands-on contributor to prominent open-source projects like FFTW and OpenZFS, he has implemented SIMD kernels for AVX-512, NEON and even KCVI, and added architecture support including RISC-V. Beyond coding, he brings systems expertise from Linux kernel contributions to production-grade ZFS optimizations and embedded SoC work in Litex, reflecting a rare full-stack performance mindset from silicon to application. Known for teaching and running workshops, he translates complex microarchitectural behavior into practical guidance that boosts real-world application throughput.
24 years of coding experience
DEA, Computer Science, DEA, Computer Science at Université de Rennes IDEUG, licence, maitrise, Computer Science, DEUG, licence, maitrise, Computer Science at Paris-Sud University (Paris XI)Ecole normale supérieure de Cachan
French, English
Github Skills (27)
filesystem10
assembly10
avx10
risc-v10
c1110
hardware10
x8610
c1710
device-tree10
dts10
performance-optimization10
zfs10
x86-6410
assemble10
sys10
Programming languages (11)
Github contributions (5)
FFTW/fftw3
Sep 2013 - May 2015
DO NOT CHECK OUT THESE FILES FROM GITHUB UNLESS YOU KNOW WHAT YOU ARE DOING. (See below.)
Role in this project:
Backend Developer
Backend DeveloperContributions:2 reviews, 19 commits, 6 PRs in 1 year 8 months
Contributions summary:Romain's contributions primarily involve adding and optimizing SIMD (Single Instruction, Multiple Data) support for the FFTW3 library, specifically for Intel's Knights Corner (KCVI) and AVX-512 architectures, as well as ARM's NEON and generic SIMD implementations. Their work includes implementing low-level SIMD instructions to improve double-precision performance, implementing and fixing various typos, and addressing stack alignment issues to ensure efficient execution. The user's work demonstrates a strong focus on performance optimization at a low level.
abaqusc-plus-plus
openzfs/zfs
Oct 2016 - Apr 2021
OpenZFS on Linux and FreeBSD
Role in this project:Back-end Developer & Systems Engineer
Back-end Developer & Systems EngineerContributions:17 commits, 15 PRs, 122 comments in 4 years 7 months
Contributions summary:Romain primarily contributed to low-level system code optimization and the implementation of specific CPU architecture-specific code paths within the ZFS file system. Their work focused on leveraging SIMD instructions (NEON, AVX-512, AltiVec) for performance improvements, including parity generation, byteswapping, and checksum calculations. They also added support for new CPU architectures like RISC-V and made adjustments to existing code to accommodate changes in compiler behavior, ensuring compatibility and optimized performance across various hardware platforms.
openzfstarballlinuxfreebsdsystem-software
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.