Intel(R) - Computational and Systems Biology at MIT

6 Intel® Math Kernel Library User’s GuideOperating on DenormalsIf an Intel MKL function operates on denormals, that is, non-zero numbers that are smallerthan the smallest possible non-zero number supported by a given floating-point format, orproduces denormals during the computation (for instance, if the incoming data is too closeto the underflow threshold), you may experience considerable performance drop. The CPUstate may be set so that floating-point operations on denormals invoke the exceptionhandler that slows down the application.To resolve the issue, before compiling the main program, turn on the -ftz option, if youare using the Intel® compiler or any other compiler that can control this feature. In thiscase, denormals are treated as zeros at processor level and the exception handler is notinvoked. Note, however, that setting this option slightly impacts the accuracy.Another way to bring the performance back to norm is proper scaling of the input data toavoid numbers near the underflow threshold.FFT Optimized RadicesYou can gain performance of Intel MKL FFT if length of the data vector permits factorizationinto powers of optimized radices.In Intel MKL, the list of optimized radices depends upon the architecture:• 2, 3, 4, 5 for IA-32 architecture• 2, 3, 4, 5 for Intel® 64 architecture• 2, 3, 4, 5, 7, 11 for IA-64 architecture.Using Intel® MKL Memory ManagementIntel MKL has the memory management software that controls memory buffers for use bythe library functions. New buffers that the library allocates when certain functions (Level 3BLAS or FFT) are called are not deallocated until the program ends. To get the amount ofmemory allocated by the memory management software, call the MKL_MemStat()function. If at some point your program needs to free memory, it may do so with a call toMKL_FreeBuffers(). If another call is made to a library function that needs a memorybuffer, then the memory manager will again allocate the buffers and they will again remainallocated until either the program ends or the program deallocates the memory.This behavior facilitates better performance. However, some tools may report the behavioras a memory leak. You can release memory in your program through the use of a functionmade available in Intel MKL or you can force memory releasing after each call by setting anenvironment variable.6-16