The Aldwin division of ANEO, a French consulting agency based in Boulogne-Billancourt, and its Advanced Computing Technologies team (HPC|IA-Cloud|Big Data) is now working together with NEC to leverage the leading-edge vector technology of SX-Aurora TSUBASA for porting and optimizing Aldwin's in-house software for American option pricing. The code is implemented in C++17 and designed to capitalize on the full capabilities of both scalar and vector architectures. It uses OpenMP to multithread the application, while the vectorization is implemented using the Eigen library.
As porting the code proved to be straightforward, the first results could be achieved very quickly. The execution time on an 8-core SX-Aurora TSUBASA is approximately five times shorter than on a 40-core CPU based on the x86 architecture, which is largely due to the very high memory bandwidth with SX-Aurora TSUBASA of about 1200 Gbyte/s.
"SX-Aurora TSUBASA is a vector supercomputer embedded on a PCIe card, and the performance shows it. We have benchmarked the platform with an application tailored to take advantage of vectorization on the CPU, and we have been very impressed with the performance characteristics of NEC's latest generation of vector technology", stated Wilfried Kirschenmann CTO of Aldwin by ANEO.
"We are very happy to enter this highly promising technology partnership with ANEO. We are impressed with their technical expertise and deep understanding of supercomputing technology and their ability to leverage real-world applications for finance organisations", stated Bruno Lecointe, Operation Director, HPC Western Europe, NEC Deutschland.
More information on ANEO's software technology approach for SX-Aurora TSUBASA can be found in a whitepaper available online .