Access Analytics Int'l
Advanced Analysis Capabilities and BCSLIB-EXT


HOME



ABOUT US



PRODUCTS



SERVICES



PARTNERS



CONTACT US


 
Products

Access Analytics & BCSLIB-EXT
Access Analytics acquired BCSLIB-EXT in 2007. As owner of the intellectual property, we provide licenses to the tool suite in binary or source code, and continued support for existing customers and clients.

Contact us for more information about incorporating BCSLIB-EXT into your products.

BCSLIB-EXT Capabilities and Opportunities

BCSLIB-EXT is an advanced analysis engine. It contains powerful linear algebra methods that enable a broad range of analysis capabilities. It has unique capabilities for solving large sparse matrix problems, such as eigenanalysis, linear system solves, least squares solutions, null space basis computations, as well as rank detection, and variable dependency identification.

It is built for engineering applications and other demanding programming environments. It uses robust, reliable implementations of innovative and state-of-the-art algorithms. The coding paradigm and structure makes it efficient and portable across a wide range of computers and compilers.

As an advanced sparse tool suite it also exploits problem sparsity to produce computationally fast and data efficient solutions. In addition, for those extremely large problems, BCSLIB-EXT includes an internally managed out-of-core capability to insure efficiency and maximal use of system resources for finding the solution to the most challenging problems. For example, a major vendor of CAD/FEM software uses BCSLIB-EXT to solve structural statics problems five times faster than with their old technology while increasing their problem size capability by an order of magnitude.

BCSLIB-EXT is in use by many major commercial software companies for empowering their analysis suites. It is broadly used in the CAE industry as the analysis engine for finite element model analysis.

In addition BCSLIB-EXT is used by commercial entities around the world to enable their own unique analysis abilities, from nonlinear simulations and multidisciplinary design optimizations, to product effective and model analysis. A U.S. auto manufacturer builds quieter cars through Nastran analysis of acoustic vibration problems with millions of degrees of freedom.

BCSLIB-EXT has been used in a broad range of demanding solutions including problems in circuit analysis, chemical process control, machine tool path definition, constrained data fitting, finite element analysis, large scale optimization, determination of optimal low-thrust trajectories for satellite orbits and many other applications.

BCSLIB-EXT also provides a software environment for the continued development of advanced analysis capabilities, especially those which are large and sparse and require efficient and effective use of computing environments, exploiting multicore and parallel environments to having access to high performance computing environments.


BCSLIB-GPU Capabilities and Flexibility

BCSLIB-GPU is a CUDA implementation of BCSLIB-EXT. It is designed as "plug-and-play" with only one additional subroutine to be called. That new subroutine allows the user to specify the GPGPU number to be used as well as other thresholds that can tuune the performance. BCSLIB-GPU has all computational aspects of BCSLIB-EXT implemented to expoit the GPGPU, from frontal matrix assembly to in-core and out-of-core frontal factorization and forward and back solves. The block shift inverted Lanczos method is also is CUDA enabled over all aspects of the computations that can take advantage of the GPGPU.

BCSLLIB-GPU has demonstrated speedups of 2 times as fast and up to 7 times as fast as the multi-threaded BCSLIB-EXT. Performance varies depending on problem size and problem formulation.

Because of BCSLIB-GPU's plug and play design, all functionality that you had with BCSLIB-EXT is available with BCSLIB-GPU, together with the performance enhancement of CUDA enabled GPGPUs. Furthermore, it is available immediately without any major code changes, simply add in the BCSLIB-GPU library and invoke one new subroutine call to specify the GPGPU number, and you're up and running on your GPGPU enabled platform.

BCSLIB-GPU is in use in commercial software codes in the CAE simulation market today, and performing at better rates than their hardware vendor tuned alternatives.


2014 Access Analytics International, LLC