Intel® C++ Compiler 16.0 User and Reference Guide
This topic only applies to Intel® Many Integrated Core Architecture (Intel® MIC Architecture).
The intrinsic functions supporting Intel® Initial Many Core Instructions (Intel® IMCI) Short Vector Math Library (SVML) or math intrinsics are vector variants of corresponding scalar math operations. They take vector arguments and perform scalar math operation on each element of the source vectors. The result of the operation is returned in a vector.
The masked variants of the SVML intrinsics have two additional parameters: v1_old and k1, where v1_old is the old value of the destination vector and k1 is the mask parameter.
Masked variants operate only on those elements of the source vectors for which the corresponding bit in the mask parameter is set to '1'. Elements in the result vector for which corresponding bit in mask parameter is '0' take their values from the v1_old parameter.
The SVML intrinsics do not have any corresponding instructions. The prototypes for SVML intrinsics are available in the immintrin.h file.
Many routines in the svml library are more highly optimized for Intel® microprocessors than for non-Intel microprocessors.