The configuration parameter DFTI_PRECISION denotes the floating-point precision in which the transform is to be carried out. A setting of DFTI_SINGLE stands for single precision, and a setting of DFTI_DOUBLE stands for double precision. The data is meant to be presented in this precision; the computation will be carried out in this precision; and the result will be delivered in this precision. This is one of the four settable configuration parameters that do not have default values. The user must set them explicitly, most conveniently at the call to descriptor creation function DftiCreateDescriptor.
Fortran module MKL_DFTI also defines named constants DFTI_SINGLE_R and DFTI_DOUBLE_R, with the same semantics as DFTI_SINGLE and DFTI_DOUBLE, respectively. These constants are not recommended for direct use. They only facilitate access to precision-specific functions for the generic name DftiCreateDescriptor through the modification of the "USE MKL_DFTI" statement, as described in section CreateDescriptor.
Copyright © 1994 - 2010, Intel Corporation. All rights reserved.