13.07.2015 Views

Intel(R) - Computational and Systems Biology at MIT

Intel(R) - Computational and Systems Biology at MIT

Intel(R) - Computational and Systems Biology at MIT

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Linking Your Applic<strong>at</strong>ion with <strong>Intel</strong>® M<strong>at</strong>h Kernel Library 5st<strong>at</strong>ic linking of user code myprog.f, sequential version of sparse solver, <strong>and</strong>sequential <strong>Intel</strong> MKL supporting LP64 interface.For other linking examples, see the <strong>Intel</strong> MKL support website <strong>at</strong>http://www.intel.com/support/performancetools/libraries/mkl/.Linking with Interface LibrariesLinking with the Absoft compilersYou can use <strong>Intel</strong> MKL with the Absoft compilers on systems based on <strong>Intel</strong>® 64 or IA-32architecture. Table 5-2 explains which Interface layer library must be included in the linkline to link with the Absoft compilers.Table 5-2Interface layer library for linking with the Absoft compilersArchiectureProgrammingInterface St<strong>at</strong>ic Linking Dynamic LinkingIA-32 Does not m<strong>at</strong>ter libmkl_intel.a libmkl_intel.so<strong>Intel</strong>® 64 ILP64 libmkl_gf_ilp64.a libmkl_gf_ilp64.so<strong>Intel</strong>® 64 LP64 libmkl_gf_lp64.a libmkl_gf_lp64.soLinking with Threading LibrariesIn the past, only few compilers other than <strong>Intel</strong>® ones supported threading of the user'sapplic<strong>at</strong>ion. Starting with <strong>Intel</strong> MKL 10.0 timeframe, additional compilers will be offeringOpenMP* threading. If an applic<strong>at</strong>ion compiled with such a threading compiler usedOpenMP threading <strong>and</strong> called threaded parts of <strong>Intel</strong> MKL versions lower than 10.0, theremight be difficulties. They may arise because MKL is threaded using the <strong>Intel</strong>® compilers,<strong>and</strong> threading libraries from different compilers are not comp<strong>at</strong>ible. This can lead toperformance issues, <strong>and</strong> perhaps even failures when incomp<strong>at</strong>ible threading is used withinthe same applic<strong>at</strong>ion. Starting with <strong>Intel</strong> MKL 10.0, several solutions are available incertain cases. Those solutions are provided both from the Threading Layer <strong>and</strong> the suppliedrun-time libraries found in the Compiler Support RTL Layer.The Solution in Layers. With this release of <strong>Intel</strong> MKL, the library is structured as layers.One of those layers is a Threading Layer. Because of the internal structure of the library, allof the threading represents a small amount of code. This code is compiled by differentcompilers (such as a gnu compiler on Linux*) <strong>and</strong> the appropri<strong>at</strong>e layer linked in with thethreaded applic<strong>at</strong>ion.5-9

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!