Introduction

Compiler automatic vectorization converts the operation of processing one pair of scalars at a time into the operation of parallel processing of multiple pairs of vector operations, which can significantly accelerate some program operations with loops. In addition to the preceding interfaces, KML_SVML also provides vectorized interfaces for the compiler. For details, see Table 1. During compilation, you can add different compilation options for different compilers so that the compiler can automatically call the corresponding vectorized interfaces.

**Table 1** Compiler vectorization interfaces
SVML Interface	MATH Interface	Description
_ZGVnN2v_sin _ZGVnN4v_sinf	sin sinf	Sine function
_ZGVnN2v_cos _ZGVnN4v_cosf	cos cosf	Cosine function
_ZGVnN2v_tan _ZGVnN4v_tanf	tan tanf	Tangent function
_ZGVnN2v_atan _ZGVnN4v_atanf	atan atanf	Arc tangent function
_ZGVnN2vv_atan2 _ZGVnN4vv_atan2f	atan2 atan2f	Arc tangent function
_ZGVnN2v_sinh _ZGVnN4v_sinhf	sinh sinhf	Hyperbolic sine function
_ZGVnN2v_cosh _ZGVnN4v_coshf	cosh coshf	Hyperbolic cosine function
_ZGVnN2v_tanh _ZGVnN4v_tanhf	tanh tanhf	Hyperbolic tangent function
_ZGVnN2v_exp _ZGVnN4v_expf	exp expf	Exponential function (base e)
_ZGVnN2v_exp2 _ZGVnN4v_exp2f	exp2 exp2f	Exponential function (base e)
_ZGVnN2v_expm1 _ZGVnN4v_expm1f	expm1 expm1f	Exponential function (base e)
_ZGVnN2v_log _ZGVnN4v_logf	log logf	Logarithmic function (base e)
_ZGVnN2v_log10 _ZGVnN4v_log10f	log10 log10f	Logarithmic function (base 10)
_ZGVnN2v_log1p _ZGVnN4v_log1pf	log1p log1pf	Logarithmic function (base e)
_ZGVnN2vv_pow _ZGVnN4vv_powf	pow powf	Power function

The preceding interfaces are provided for the compiler. You are not advised to call them directly.

Parent topic: Automatic Vectorization