NAG Library for the Xeon Phi™ Processor and Coprocessor
The NAG Library has been developed in conjunction with Intel to harness the performance of the Intel Xeon Phi™ family. For the coprocessor many of the algorithms in the Library are tuned to run significantly faster both in offload or native modes. For the processor thhe Library has been ported to use the full Xeon Phi processor functionality.
The NAG Library has proven scalability, speed and correctness making it easy to exploit the power of the Xeon Phi family. It contains all the powerful, robust and flexible algorithms in the NAG Library but is specifically optimized for the Xeon Phi systems. By using the NAG Library you are future proofing your numerical software investment as your hardware grows.
Specialised parallel functionality
- Fast Fourier Transforms (FFTs)
- Random Number Generators
- Partial Differential Equations
- Curve and Surface Fitting
- Correlation and Regression Analysis
- Multivariate Methods
- Time Series Analysis
- Financial Option Pricing
- Dense Linear Algebra (LAPACK)
- Sparse Iterative Solvers
- Sparse Direct Solvers (based on SuperLU)
- Sparse Iterative Eigensolvers (based on ARPACK)
- Ordinary Differential Equations (ODEs)
- Multivariate Statistics
The documentation for the NAG C Library version for Intel® Xeon Phi™ Processor (Mark 26) can be viewed here.
The documentation for the NAG Library for Xeon Phi™ Coprocessor (Mark 23) can be viewed on-line in the following format:
- HTML: The full manual available as HTML (using HTML5 features such as MathML and SVG in suitable browsers)
The Library is organised into Chapters - each being documented with its own Introduction and Contents list followed by a comprehensive document for each function detailing its purpose, description, list of parameters and possible error exits. Example programs and results are also supplied. All examples are available on-line to facilitate their use as templates for the users' calling programs.
Full NAG Library runs directly on the Xeon Phi processor.
The NAG Library can run in native mode, can run in a programmer controlled offload mode; and, can also be set to automatically manage offloads to help optimize total system performance depending on the size of calculations.
When working in native mode all of the library functions and routines are executed on the coprocessor; whilst working in the two offload modes when certain routines are called from the host they may execute regions of parallel work on the coprocessor. In automated offload the NAG Performance Parameters module controls which routines are run on the coprocessor, based on the nature of the routine and the scale of the problem together with the configuration of the host processor system. In the programmer controlled mode the automation is overridden by user-controlled switches that can target a specific coprocessor when required.
The table shows current implementations.
|Xeon Phi Processor||C version|
|Intel C||64 bit||CSLM626DD|
|Xeon Phi Coprocessor||Fortran version|
|Intel Fortran||64 bit||FSLM623DC|
Services and Support
NAG’s Technical Support Service is provided by a team of specialists in numerical and statistical software development, in fact the NAG Library and Compiler development team share responsibility for the support of our software. We strongly believe that in order to effectively support complex software the technicians must be both experienced and understand the intricacies of the computational techniques. This conviction is reflected in the composition of the team most of whom are qualified to PhD level and have combined experience of software support in excess of 50 years.
NAG provides accurate, documented, numerical software and services to help you make sure that your results are accurate. The validity of each NAG routine is tested for each platform that it is enabled for. Only when an implementation satisfies our stringent accuracy standards is it released. As a result, you can rely on the proven accuracy and reliability of NAG to give you the right answers. NAG is an ISO 9001 certified organization.
The numerical codes that underpin the results from your software are not future proof. While the mathematics does not change, the codes have a limited lifespan because of new hardware structures, mathematical innovation and changes in application needs. NAG Numerical Services help you and your organization find and implement the optimum numerical computation solutions. NAG works with your team to impart skills and techniques that will help solve your numerical software problems.
Your users, developers and managers can all benefit from NAG's highly regarded training courses. All of the training courses shown below have been delivered successfully either from NAG offices or at client premises. Training courses can be tailored to suit your particular requirements and be targeted to novice, intermediate or experienced levels. Specialized mentoring and development programs are also available for HPC managers.
NAG was founded on collaboration as an inter-University collaborative venture combining the talents of mathematicians and computer scientists. NAG has continued to collaborate with individuals and organizations over the past four decades and today longstanding and new partners are delivering tangible benefits to users and students all over the world.