According to Tachyum, both companies will look to jointly explore and evaluate the benefits of incorporating support for its Prodigy processors.
MatLogica gives quantitative and scientific developers tools to quickly build computational analytics that scales with modern hardware. The company’s primary product is AADC, a just-in-time (JIT) compiler tailored for complex repetitive calculations and Automatic Adjoint Differentiation (AAD), which can achieve up to 100x speed-ups while also calculating risks automatically and precisely using AAD.
MatLogica’s compilation technology will utilise Tachyum Prodigy large vector registers and vector instructions for complex risk calculations. A Prodigy-enabled MatLogica AADC deployment will be able to help customers to migrate their software to Tachyum technology without changes in their software stack.
“At MatLogica, we are committed to delivering tools for developers wishing to retain the advantages of object-oriented languages whilst driving performance with data-oriented design. We look forward to extending our toolkit for Tachyum Prodigy, to enable native support across CPU, GPU and TPU and unlocking the full potential of Tachyum’s technology,” said Dmitri Goloubentsev, CTO of MatLogica. “By using MatLogica’s AADC on Tachyum, users should see unprecedented levels of performance, simplified software infrastructure, lower ownership costs, as well as reduced carbon footprint for compute-intensive tasks.”
Tachyum’s Prodigy integrates 128 high-performance custom designed 64-bit compute cores with the functionality of a CPU, a GPU, and a TPU in a single device within a homogeneous architecture. This allows it to deliver performance that is up to 4x faster than that of the highest performing x86 processors (for cloud workloads) and up to 3x that of the highest performing GPU for HPC and 6x for AI applications.
According to Tachyum, Prodigy’s architecture delivers industry-leading performance in both data centre and AI workloads. During off peak hours, for example, Prodigy-powered data centre servers can be seamlessly and dynamically switched to AI workloads, eliminating the need for expensive dedicated AI hardware, and dramatically helps to increase server utilisation.
“AI is an important space for Tachyum and, with Prodigy, we will continue to democratise AI through cost-effective availability for everybody,” said Dr. Radoslav Danilak, founder and CEO of Tachyum. “By signing an MoU with MatLogica, we will be better able to serve the needs of developers looking to seamlessly calculate automatic analytical sensitivities. This combination of MatLogica’s AADC and our Prodigy will provide unprecedented levels of performance for computationally intensive programmes and enable a new approach to the development of Machine Learning applications.”
