exp_tile

template<bool approx = false, bool fast_and_approx = true, uint32_t scale = 0x3F800000>
void ckernel::exp_tile_init()

Please refer to documentation for any_init.

Template scale parameter is used when aprox and fast_and_approx are true and exp_tile is called with scale_en set to true.

template<bool approx = false, bool fast_and_approx = true, bool scale_en = false, bool skip_positive_check = false, int iterations = 8>
void ckernel::exp_tile(uint32_t idst, int vector_mode = (int)VectorMode::RC, uint16_t scale = 0x3F80)

Performs element-wise computation of exponential on each element of a tile in the DST register. The DST register buffer must be in an acquired state via an acquire_dst call. This call is blocking and is only available on the compute engine.

Return value: None

Template Parameter	Description	Type	Valid Range	Default
approx	Enable approximate mode.	bool	true, false	false
fast_and_approx	If approx is true, enable fast approximation.	bool	true, false	true
scale_en	Enable input scaling by a constant factor in approximate or non-approximate mode	bool	true, false	false
skip_positive_check	Skip large-positive input check	bool	true, false	false
iterations	Number of iterations over 32-SFPU lanes to run	int	Positive integer	8

Argument	Description	Type	Valid Range	Required
idst	The index of the tile in DST register buffer to perform the computation on	uint32_t	Must be less than the size of the DST register buffer	True
vector_mode	Specifies the vector mode for computation (default: VectorMode::RC)	int	Subject to specific hardware/kernel limits	False
scale	Scale factor to apply in approximate or non-approximate mode if scale_en is true (default: 0x3F80, 1.0f in FP16b)	uint16_t	Valid FP16b representation	False