Conditions for Code Replacement of Math Operators with HVX

Each math operators that can be used with the Support Package for Qualcomm Hexagon Processors requires specific conditions to allow code replacement by using the Qualcomm Hexagon Library (QHL). You use this code replacement when generating C/C++ code from a model or from MATLAB^® code.

If you do not meet the specific requirements, then the generated C/C++ code runs on the Qualcomm Hexagon Processors. However, this generated code does not use library support.

The library supports these math operators only when you set specific properties, as indicated in this table:

	Operators	Input Requirements	Parameter Requirements	Equivalent Functions
1	`add`	Input can be vector/matrix Real and Complex Values	Length should be >=1024 and a multiple of 32	Vector and matrix real inputs `qhblas_hvx_vector_add_af` Vector and matrix complex inputs `mw_qhblas_hvx_vector_add_acf`
2	`sub` (Fixed-Point Designer)	Input can be vector/matrix Real and Complex Values	Length should be >=1024 and a multiple of 32	Vector and matrix real inputs `qhblas_vector_sub_af` Vector and matrix complex inputs `mw_qhblas_vector_sub_acf`
3	Mul - elementwise	Input can be vector/matrix Real Values	Length should be >=1024 and a multiple of 32	Vector and matrix real inputs `qhblas_hvx_vector_hadamard_af`
4	`Scaling`	Input can be real/complex/vector/matrix Scale factor has to be a real scalar		Vector and matrix real inputs `qhblas_hvx_vector_scaling_af` Vector and matrix complex inputs `mw_qhblas_hvx_vector_scaling_acf`
5	`Matrix Multiplication`	Real Values Input can be a vector/matrix	Matrix1 dimensions = mn Matrix2 dimensions = nk if RowMajor & k = 1 or ColMajor & m = 1 (2n) - 1 >= 127 & mk >= 4 else (2n) - 1>= 15 & mk >= 64	For Vector and Matrix real inputs `mw_qhblas_hvx_matrix_matrix_mpy_af`

Conditions for Code Replacement of Math Operators with HVX

See Also

Related Topics