Compressed Matrix Package (CMPACK)
Accelerating Virtual Prototyping and Digital Engineering
ARA’s Compressed Matrix Package (CMPACK) fast solver software library accelerates simulations of complex radio frequency, acoustic, and other wave physics applications. CMPACK provides fast, accurate simulation results for virtual prototyping and digital engineering that save time and money.
Modern C++
Library Features
Highly Scalable
Significantly Faster
than Competing Fast Solvers
Significantly Faster
than Competing Fast Solvers
Highly Scalable
Modern C++
Library Features
CMPACK Features
- Modern C++ library featuring automated testing, modularity, and composability
- Significantly faster than competing fast-solve libraries
- Well-documented API with examples to facilitate integration
- Hybrid threaded and optional Message Passing Interface parallelization which scales from workstations to supercomputers
- Compressed matrix build, factor, solve, and multiply
- Support for both symmetric and general matrices
- Fast-update solver for when only a portion of the simulated geometry is updated
CMPACK Features
- Modern C++ library featuring automated testing, modularity, and composability
- Significantly faster than competing fast-solve libraries
- Well-documented API with examples to facilitate integration
- Hybrid threaded and optional Message Passing Interface parallelization which scales from workstations to supercomputers
- Compressed matrix build, factor, solve, and multiply
- Support for both symmetric and general matrices
- Fast-update solver for when only a portion of the simulated geometry is updated
Predicting Radio Frequency Physics with Speed and Accuracy
Virtual prototyping can accelerate the design process for everything from fighter jets to commercially-available autonomous vehicles. Faster simulation times and reduced memory requirements allow customers to explore millions of potential designs virtually using modeling and simulation before a physical prototype is ever created.
ARA’s CMPACK helps solve complicated matrix systems, which can grow to tens of millions of unknowns at higher frequencies, in order to accurately predict how radio waves will interact with equipment. Competitor simulations can take days or weeks to produce high-fidelity results. ARA’s CMPACK cuts this time down to just hours while maintaining a high level of accuracy.
CMPACK’s Scalability
CMPACK’s scalability allows the fast solver to factor large matrices in record times on individual workstations up to supercomputers. Using a ten million unknown matrix as an example and the industry-standard fast solving method, Adaptive Cross Approximation (ACA), CMPACK was able to factor the matrix in 12.4 hours. CMPACK’s unique Domain Decomposition Method (DDM) factored the same ten million unknown matrix in just 14.8 minutes.
z-1 (LU Factor) Runtime: CMPACK-DDM on 1,664 cores and 16TB of Memory
