Diagnosis technology

Although there exists several commercial tools for analysis of electromagnetic effects, there is currently no tool to diagnose the critical regions in design contributing to undesired electromagnetic performance.

Using Simyog technology a mesh-based sensitivity profile can be generated to highlight regions of the layout contributing most to an objective parameter. This precludes the necessity for the designer to parameterize the design layout, run multiple solves or rely on expert intervention to identify regions for re-design.

Learning-based fast re-simulation

In spite of the advances in compression and preconditioning techniques, today’s electromagnetic solvers, both commercial and academic, still consume large amounts of compute time for the solution of large scale problems. This limits the usage of such tools to the verification stage while preventing its wide-spread use in the design stage, which requires repetitive runs on design variants. Further, designers often seek ways to parameterize or optimize models based on selected parameters. However, these methods often take a large number of repetitive 3D full-wave electromagnetic solver runs on design variant geometries.

Currently all 3D full-wave solvers follow the conventional model-mesh- solve workflow to analyze any electromagnetic structure. Therefore, they treat each model independently regardless of any similarity with a previously solved modeland therefore sacrifices the possibility of accelerating a model solution from information harnessed from a prior solution of a similar model. This is a missed opportunity particularly in the solution of design variants which involve multiple models with near–identical geometrical features.

Using Simyog technology, we can drastically reduce the solution time, using information harnessed from previously completed solution of a design variant.

Hardware acceleration and matrix compression

Simyog’s linear scaling compression algorithm results in a favorable O(NlogN) time and memory complexity.

Further, the algorithms are chosen and tailored for minimizing serial content in accordance to Amdahl’s law.

Using Simyog Technology parallel solutions have been implemented on multiple hardware acceleration frameworks. Recently on many-core Intel Knights Landing platform a speed-up of 60x was demonstrated on 68 cores.