nrf_axon_compiler_sig_tanh_dequant_consts()

void nrf_axon_compiler_sig_tanh_dequant_consts	(	int32_t *	multiplier,
		uint8_t *	shift )

adjusts dequantization multiplier and scaler to produce a q3.12

The input dequantization need to be multplied by 2^12. This function will massage the original values to perform the additional multiply by some combination of increasing the multiplier or decreasing the shift.

Parameters

[in]

out] multiplier input values are 1st multplied by this.

Parameters

[in]

out] shift input values are then rounded by this many bits.