Xplicit Computing Objective Massively-Parallel Unified Technical Environment
In 2014 we began work on a revolutionary new computing platform -- to meet 21st Century challenges
Numerical Super-Code: C++ with OpenCL
A giant leap in heterogeneous computing technology; applies modern software and hardware paradigms to yield a real-time framework everyone can apply to their projects.
"I think if I were rating 'what’s the most advanced at Space Access' I’d put [xcompute] as number 1."- Dr David Livingston The Space Show - Transcript
"As someone who is coming from the other side of the process (helping humans build mental models with which they can solve problems computationally!), I think of your effort as a crucial one to bring the full power of computing's tools and capabilities to a place humans can productively leverage." - Zachary Dodds Harvey Mudd College, Dept of Computer Science
Facilitating teams through concept, preliminary, critical design, and manufacture stages.
1. Organize complex systems using generic building-blocks.
2. Assign geometries and models (comprised of algorithms).
3. Connect and couple results across teams and disciplines.
4. Refine and improve systems fidelity as needed.
systems can contain sub-systems, visualized as a tree
Modern analyses require systems interact over wide space-time scales and with increasing topological complexity. State-of-art software abstractions provide a way to divide each problem and integrate as part of the whole. Algorithms then call upon varying geometry types and runtime behavior is redirected to take advantage of key characteristic optimizations by type, yielding up to 50x performance benefit.
explicit topology and implicit signed-distance fields
Continuous Galerkin Finite Element Method (FEM-CG) accurately and quickly solves steady-state and time-varying analyses, supporting structural, thermal, and continuum multi-physics. Ground-up algorithms utilize ViennaCL solvers that run on any device - including GPU!
local stresses on complex structures and components
Sophisticated compressible and incompressible CFD capabilities are based on Finite Volume (FVM), Finite Difference (FDM), and Lattice-Boltzmann methods (LBM). A modular approach to algorithms enables natural inclusion of viscid, inviscid, heat transfer, and turbulence modeling on any device!
real-time conservation-based transport and state equations