Advancing Open-Source Statistical Frameworks
Fujitsu is developing a high-performance, energy-efficient 2nm ARM-based CPU, FUJITSU-MONAKA, for next-gen data centers, focusing on optimizing performance, usability, and open-source software (OSS) development. Collaborating with the OSS community, we contribute to AI-accelerated software stacks for data-intensive workloads. This talk will highlight Fujitsu’s recent contributions to the open software ecosystem for statistical computing:
- Implemented statistical kernel routines in the Vector Statistical Library (VSL) to fill gaps where open-source options were insufficient, enhancing workflows with reference backend support.
- Integrated OpenRNG, an ARM open-source RNG library, to maintain bitwise reproducibility between ARM and x86 systems and consistency in RNG functions across platforms.
Tested on ARM-based Graviton3 using GCC and OpenBLAS, these contributions enhance oneDAL’s versatility, supporting diverse architectures and unifying the compute ecosystem around open standards.
Speaker(s)
Rakshith GB - Senior Engineer, Fujitsu Research of India Pvt. Ltd.
Rakshith G B is a Senior Engineer at FRIPL, leading R&D efforts in tuning oneDAL for ARM platforms. With extensive experience in high-performance computing, he specializes in optimizing software for computer vision, video compression, and machine learning applications. His work has significantly advanced the efficiency and scalability of data analytics libraries on ARM architectures.
Dhanus M Lal - Software Engineer, Fujitsu Research of India Pvt. Ltd.
Dhanus joined Fujitsu Research India as a Software Developer in 2023. Prior to this role, he earned his BSc (Research) and MSc in Mathematics from the Indian Institute of Science, Bangalore. His areas of interest include theoretical computer science, computational mathematics, software performance analysis, and related fields. Dhanus has also contributed to several open-source projects.
