GETTING THE MOST OUT OF MANYCORE
The world’s biggest supercomputing facilities continue to push boundaries, adopting new technologies as they push forward to the challenging goal of exa-scale computing. These innovations inevitably impact the work of everyday engineers and scientists, and over time, updates in workstation-scale technology track the architectural evolution of large-scale supercomputing systems. At present workstations are shifting to a manycore layout, where massively parallel GPU co-processors share the computing workload with multiple CPU core. For a more details view on this topic, please register for our upcoming webinar.
The obvious move for computational fluid dynamics simulations is to start adopting a manycore mindset that reduces O&M cost on computing equipment and makes high-performance simulations more accessible to the user base. Envenio’s general purpose CFD software tool, EXN/Aero is a manycore-optimized solution for engineers looking to scale up CFD simulation capability without a corresponding increase in license & compute resource budgets.
As CFD analysis demands greater fidelity, speed and complexity from its simulation tools, software designers need to optimize and parallelize to keep up. Our manycore mindset locks us into a steep scalability trajectory that non-optimized solvers will find harder and harder to match over time.
CELL & INTERFACE DESIGN
The patented design of the EXN/Aero solver subdivides simulation data and associated compute tasks into two type of objects: cells and interfaces. Cell objects contain CFD mesh (structured or unstructured mesh blocks) and these represent the bulk of the computational load. Interface objects on the other hand, are multi-function data transformation utilities with many uses including boundary condition input, communications between cells, reduction of output data and interpretation of sensor inputs.
CELL-BASED MAPPING MODULE (CBMM)
Not every computing technology is equally good at every computing task. Key to manycore computing is providing this flexible load balancing mechanism that lets the various architectures play to their strengths. The CBMM is our answer to this problem. It detects the available compute resources, and then manages assignment of the cell and interface compute tasks on-line, according to their attributes and data type. This approach maximizes parallelism for all solver operations and removes bottlenecks.
FLEXIBLE PARALLELIZATION SCHEMES
The cell and interface design, allows for greater creativity and flexibility in the way a CFD run is parallelized. Bulky, expensive cell tasks are computed on GPU devices much of the time while interface tasks are distributed based on their proximity to cell objects. This minimizes the communication overhead for interface information from interface to cell, and from host memory to co-processor memory .
EXN/Aero’s CBMM also allows multiple simulation instances to run concurrently on the same shared resource, ultimately speeding up the engineering workflow. Provided the total available host & device memory are not exceeded, then additional runs do not drastically affect each other’s performance.
The ability of manycore systems to run parallel instances makes it possible to access another dimension of parallelism. EXN/Aero’s prototype multi-propogator parallel-in-time (para-real) solution technique planned for release in 2017 will enable users to process multiple time slabs simultaneously, greatly reducing the time required for steady-state runs and decreasing start-up times for unsteady runs.
INVESTING IN NEW TECHNIQUES
Envenio’s investment into forward-looking technologies continues, taking advantage of established trends to deliver low-cost powerful manycore supercomputing at all enterprise scales. An example of this is the development of space-time (para-real) parallelization, which helps overcome some of the new latency barriers that appear in a manycore environment.
EXN/Aero is built from the ground up to maximize performance in hybrid parallel computing environments. Today this means a mix of CPU and GPU, but we are ready to adapt as new technologies continue to trickle down from the exa-scale installations.