Intel(R) - Computational and Systems Biology at MIT

More documents

Recommendations

Info

3 Intel® Math Kernel Library User’s GuideInterface Layer. The layer essentially provides matching between the compiled code of yourapplication and the threading and/or computational parts of the library. This layer mayallow for matching like these:• Provides LP64 interface to Intel MKL ILP64 software(see Support for ILP64 Programming for details)• Provides means to deal with the way different compilers return function values• For those software vendors that use Cray-style names, provides mapping ofsingle-precision names to double-precision ones in applications that employ ILP64.Threading Layer. This layer provides a way for threaded Intel MKL to share supportedcompiler threading. The layer also provides for a sequential version of the library. Whatwas internal to the library previously, now is essentially exposed in the threading layer bybeing compiled for different environments (threaded or sequential) and compilers (Intel,gnu, and so on).Computational Layer. It is the heart of Intel MKL and has only one variant for anyprocessor/operating system family, such as 32-bit Intel® processors on a 32-bit operatingsystem. The computational layer can accommodate multiple architectures throughidentification of the architecture or architectural feature and choose the appropriate binarycode at execution. Intel MKL may be thought of as the large computational layer that isunaffected by different computational environments. Then, as it has no RTL requirements,RTLs refer not to the computational layer but to one of the layers above it, that is, theinterface layer or the threading layer. The most likely case is matching the threading layerwith the RTL layer.Compiler Support RTL Layer. This layer has run-time library support functions. For example,libguide and libiomp are RTLs providing threading support for the OpenMP* threadingin Intel MKL.See also the “Linking Examples” section in chapter 5.Sequential Version of the LibraryStarting with release 9.1, the Intel MKL package provides support for sequential, that is,non-threaded, version of the library. It requires no Compiler Support RTL layer, that is, noLegacy OpenMP* or Compatibility OpenMP* run-time libraries, and does not respond to theenvironment variable OMP_NUM_THREADS (see the Using Intel® MKL Parallelism section inchapter 6 for details). This version of Intel MKL runs unthreaded code. However, it isthread-safe, which means that you can use it in a parallel region from your own OpenMPcode. You should use sequential version only if you have a particular reason not to use IntelMKL threading. The sequential version (layer) may be helpful when using Intel MKL withprograms threaded with non-Intel compilers or in other situations where you may, forvarious reasons, need a non-threaded version of the library. For more information, seesection Avoiding Conflicts in the Execution Environment in chapter 6.3-4
Intel® Math Kernel Library Structure 3To obtain sequential version of Intel MKL, in the Threading layer, choose the*sequential.* library to link (see Directory Structure in Detail).Note that the sequential library depends on the POSIX threads library (pthread), which isused to make Intel MKL software thread-safe and should be included in the link line (seeLinking Examples in chapter 5).Support for ILP64 ProgrammingThe terms "LP64" and "ILP64" are used for certain historical reasons and due to theprogramming models philosophy described here:http ://www .unix.org/version2/whatsnew/lp64_wp.html .Intel MKL ILP64 libraries do not completely follow the programming models philosophy.However, the general idea is the same: use 64-bit integer type for indexing huge arrays,that is, arrays with more than 2 31 -1 elements.It’s up to you to choose which interface to use. You should definitely choose LP64 interfacefor compatibility with the previous Intel MKL versions, as "LP64" is just a new name for theonly interface that the Intel MKL versions lower than 9.1 provided. You should definitelychoose the ILP64 interface if your application uses Intel MKL for calculations with hugedata arrays (of more than 2 31 -1 elements) or the library may be used so in future.The LP64 and ILP64 interfaces are supported in the Interface layer. Once the appropriatelibrary in the Interface layer is selected (see Directory Structure in Detail), all librariesbelow the Interface layer are compiled using the chosen interface.As the differences between the ILP64 and LP64 interfaces are out of scope of the Intel MKLReference Manual, you are encouraged to browse the include files, examples, and tests forthe ILP64 interface details. To do this, see the following directories, respectively:/include/examples/testsThis section shows• How the ILP64 concept is implemented specifically for Intel MKL• How to compile your code for the ILP64 interface• How to code for the ILP64 interface• How to browse the Intel MKL include files for the ILP64 interfaceThis section also explains limitations of the ILP64 support.3-5
Page 3 and 4: INFORMATION IN THIS DOCUMENT IS PRO
Page 7 and 8: Intel® Math Kernel Library User’
Page 9 and 10: Overview 1Intel® Math Kernel Libra
Page 11 and 12: Overview 1Chapter 3Chapter 4Chapter
Page 13 and 14: Getting Started 2This chapter helps
Page 16 and 17: 2 Intel® Math Kernel Library User
Page 38 and 39: Configuring YourDevelopment Environ
Page 40 and 41: Configuring Your Development Enviro
Page 42 and 43: Configuring Your Development Enviro
Page 70 and 71:
6 Intel® Math Kernel Library User
Page 72 and 73:
Page 74 and 75:
Page 76 and 77:
Page 78 and 79:
Page 80 and 81:
Page 82 and 83:
Page 84 and 85:
Page 86 and 87:
Page 88 and 89:
Page 90 and 91:
Page 92 and 93:
Page 94 and 95:
LINPACK and MP LINPACKBenchmarks 10
Page 96 and 97:
LINPACK and MP LINPACK Benchmarks 1
Page 98 and 99:
Page 100 and 101:
Page 102 and 103:
Page 104 and 105:
Page 106 and 107:
A Intel® Math Kernel Library User
Page 108 and 109:
IndexAAbsoft compiler, linking with
Page 110:
Intel MKL choice, particular cases,
show all

Intel(R) - Computational and Systems Biology at MIT

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

Delete template?

Save as template?