Intel Software Adrenaline

Programming CodeBook

Issue link:

Contents of this Issue


Page 8 of 16

Math Processing 9 MATH PROCESSING Intel® Math Kernel Library (Intel® MKL) is the industry's leading math library—with support for Intel® Xeon® processors and Intel® Xeon Phi™ coprocessors. Intel Xeon Phi coprocessor support has been extended, including: • Heterogeneous computing • Takes advantage of both multicore host and manycore coprocessors. • Optimized for wider (512-bit) SIMD instructions and threaded for many cores • Highly optimized functions Resources for Manycore Developers » Faster Math Performance with Intel® Math Kernel Library (video) » Automatic Offload with Intel® Math Kernel Library (video) » Sample Usage Models on Intel® Xeon Phi™ Coprocessors •Offloading is automatic and transparent •Takes advantage of multiple coprocessors Automatic Offload (AO) Explore •By default, Intel® MKL determines when to offload; work division between host and coprocessors Intel® Math Kernel Library forum » Intel® Math Kernel Library Link Options Wizard » User's Guide » Performance charts » •Automatic host and target parallelism •Users can still specify work division between host and target (for BLAS only) •Offloading is explicitly controlled by compiler pragmas or directives Compiler Assisted Offload (CAO) •Exploit the full potential of compiler's offloading facility •More flexibility in data transfer and remote execution management •Advantage of data persistence: reusing transferred data for multiple operations •Use the coprocessor as an independent compute node •Programs can be built to run only on the coprocessor by using the –mmic build option Native Execution •Intel® MKL function calls inside an offloaded code region execute natively •Get better performance when input data are already available on the coprocessor, and output is not immediately needed on the host side For in-depth coverage, view the webinar: Get Ready for Intel® Math Kernel Library on Intel® Xeon Phi™ Coprocessor » For more information regarding performance and optimization choices in Intel® software products, visit: Privacy | Terms of Use | Learn More

Articles in this issue

Links on this page

view archives of Intel Software Adrenaline - Programming CodeBook