Conditions for Code Replacement of MATLAB Functions with HVX

R2026b

Each MATLAB^® function that can be used with the Embedded Coder^® Support Package for Qualcomm^® Hexagon^® Processors requires specific conditions to allow code replacement by using the Qualcomm Hexagon Vector eXtensions (HVX) library. You use this code replacement library to generate C/C++ code from a Simulink^® model or from MATLAB code.

When you satisfy the specified conditions, the generated C/C++ code uses HVX support to achieve optimized execution on Qualcomm Hexagon processors.

HVX supports these MATLAB functions for floating-point data only when you configure the required properties, as specified in this table.

Basic Math Operations

Function	Input Requirements	Property Requirements	Equivalent Functions
sqrt	Input must be a vector or matrix Real value	Length must be ≥ 4	qhmath_hvx_sqrt_af
rsqrt (Fixed-Point Designer)	Input must be a vector or matrix Real value	-	qhmath_hvx_rsqrt_af
log	Input must be a vector or matrix Real value	-	qhmath_hvx_log_af
log2	Input must be a vector or matrix Real value	-	qhmath_hvx_log2_af
log10	Input must be a vector or matrix Real value	-	qhmath_hvx_log10_af
exp	Input must be a vector or matrix Real value	-	qhmath_hvx_exp_af
power	Input must be a vector or matrix Real value	-	qhmath_hvx_pow_af

Trigonometric Operations

Function	Input Requirements	Property Requirements	Equivalent Functions
sin	Input must be a vector or matrix Real value	-	qhmath_hvx_sin_af
cos	Input must be a vector or matrix Real value	-	qhmath_hvx_cos_af
asin	For MATLAB and row-major of Simulink block Input must be a vector or matrix Real value	-	qhmath_hvx_asin_af
acos	For MATLAB and row-major of Simulink block Input must be a vector or matrix Real value	-	qhmath_hvx_acos_af
atan	Input must be a vector or matrix Real value	-	qhmath_hvx_atan_af
sinh	Input must be a vector or matrix Real value	-	qhmath_hvx_sinh_af
cosh	Input must be a vector or matrix Real value	-	qhmath_hvx_cosh_af
tanh	Input must be a vector or matrix Real value	-	qhmath_hvx_tanh_af
tan	Input must be a vector or matrix Real value	-	qhmath_hvx_tan_af
hypot	Input must be a vector or matrix Real value	Length must be ≥ 25	qhmath_hvx_hypot_af

Fourier Transform Operations

Function	Input Requirements	Output Requirements	Input Alignment	Output Alignment	Property Requirements	Equivalent Functions
fft	Input must be real Supports input sizes that are powers of 2, starting from 64.	Output must be complex	128	128	Note To enable code replacement, use the syntax fft(x). The code generator replaces only this form with a HVX implementation. If you specify additional inputs such as n or dim (for example, fft(X,n) or fft(X,n,dim)), the code generator generates the default plain C code.	mw_qhdsp_hvx_r1dfft_af_builtin
fft	Input must be complex Supports input sizes that are powers of 2, starting from 64.	Output must be complex	128	128		mw_qhdsp_hvx_c1dfft_af_opt_builtin
ifft	Input must be complex Supports input sizes that are powers of 2, starting from 32.	Output must be complex	128	128	Note To enable code replacement, use the syntax ifft(x). The code generator replaces only this form with a HVX implementation. If you specify additional inputs such as n or dim (for example, ifft(X,n) or ifft(X,n,dim)), the code generator generates the default plain C code.	mw_qhdsp_hvx_c1difft_af_builtin