VXLIB_multiply_i8u_i16s_o16s

Functions
VXLIB_STATUS	VXLIB_multiply_i8u_i16s_o16s (const uint8_t src0[restrict], const VXLIB_bufParams2D_t *src0_addr, const int16_t src1[restrict], const VXLIB_bufParams2D_t *src1_addr, int16_t dst[restrict], const VXLIB_bufParams2D_t *dst_addr, uint16_t overflow_policy, VXLIB_F32 scale)

Detailed Description

Function Documentation

VXLIB_multiply_i8u_i16s_o16s()

VXLIB_STATUS VXLIB_multiply_i8u_i16s_o16s	(	const uint8_t	src0[restrict],
		const VXLIB_bufParams2D_t *	src0_addr,
		const int16_t	src1[restrict],
		const VXLIB_bufParams2D_t *	src1_addr,
		int16_t	dst[restrict],
		const VXLIB_bufParams2D_t *	dst_addr,
		uint16_t	overflow_policy,
		VXLIB_F32	scale
	)

Description:

Performs element-wise multiplication between one unsigned 8-bit image, one signed 16-bit image and a scalar value and results in a signed 16-bit image. The overflow handling is controlled by an overflow-policy parameter. Use VXLIB_CONVERT_POLICY_WRAP as the parameter for wrap and VXLIB_CONVERT_POLICY_SATURATE for saturate.

Method:

The multiplication is computed using the following equation:

 dst(x,y) = src0(x,y) * src1(x,y) * scale

Parameters

[in]	src0[]	Pointer to array containing first input image (UQ8.0)
[in]	src0_addr[]	Pointer to structure containing dimensional information of src0
[in]	src1[]	Pointer to array containing second input image (SQ15.0)
[in]	src1_addr[]	Pointer to structure containing dimensional information of src1
[out]	dst[]	Pointer to array containing output image (SQ15.0)
[in]	dst_addr[]	Pointer to structure containing dimensional information of dst
[in]	overflow_policy	Parameter indicating overflow policy
[in]	scale	Parameter indicating scale

Assumptions:

• I/O buffer pointers are assumed to be not aliased.
• Negative scale values are not supported.

Performance Considerations:

• For best performance, the following parameter settings are recommended:
  • Set widths equal to strides
  • Align all pointers to 8 byte boundaries
  • Set all stride values to a multiple of 8
  • Set all width values to a multiple of 8
  • Scale values of positive integer values or 1/2^n values are optimized for speed.