Overview

Vector MAC combined with vector comparisons with 16 bit real by 16 bit real

Functions
v16acc48	mac16_abs (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v16acc48	mac16_abs (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_max (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer .

v16acc48	mac16_max (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_max (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_maxdiff (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v16acc48	mac16_maxdiff (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_maxdiff (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_min (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer .

v16acc48	mac16_min (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mac16_min (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_abs (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v8acc48	mac8_abs (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_max (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer .

v8acc48	mac8_max (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_max (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_maxdiff (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v8acc48	mac8_maxdiff (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_maxdiff (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_min (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer .

v8acc48	mac8_min (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mac8_min (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_abs (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v16acc48	msc16_abs (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_max (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer .

v16acc48	msc16_max (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_max (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_maxdiff (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v16acc48	msc16_maxdiff (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_maxdiff (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_min (v16acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer .

v16acc48	msc16_min (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	msc16_min (v16acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_abs (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v8acc48	msc8_abs (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_max (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer .

v8acc48	msc8_max (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_max (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_maxdiff (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v8acc48	msc8_maxdiff (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_maxdiff (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_min (v8acc48 acc, v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer .

v8acc48	msc8_min (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	msc8_min (v8acc48 acc, v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_abs (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v16acc48	mul16_abs (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_max (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer .

v16acc48	mul16_max (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_max (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_maxdiff (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v16acc48	mul16_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_min (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer .

v16acc48	mul16_min (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	mul16_min (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_abs (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v8acc48	mul8_abs (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_max (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer .

v8acc48	mul8_max (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_max (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_maxdiff (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v8acc48	mul8_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_min (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer .

v8acc48	mul8_min (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	mul8_min (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_abs (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v16acc48	negmul16_abs (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_max (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer .

v16acc48	negmul16_max (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_max (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_maxdiff (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v16acc48	negmul16_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_min (v64int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer .

v16acc48	negmul16_min (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v16acc48	negmul16_min (v32int16 xbuff, int xstart, unsigned int xoffsets, unsigned int xoffsets_hi, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, unsigned int zoffsets_hi, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_abs (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

v8acc48	negmul8_abs (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_max (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer .

v8acc48	negmul8_max (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_max (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_maxdiff (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

v8acc48	negmul8_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_maxdiff (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_min (v64int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer .

v8acc48	negmul8_min (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

v8acc48	negmul8_min (v32int16 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int16 ybuff, int ystart, unsigned int ysquare, v16int16 zbuff, int zstart, unsigned int zoffsets, int zstep)
	Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

Function Documentation

v16acc48 mac16_abs	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 += z00*abs(x00) + z01*abs(x01)
acc1 += z10*abs(x10) + z11*abs(x11)
acc2 += z20*abs(x20) + z21*abs(x21)
acc3 += z30*abs(x30) + z31*abs(x31)
acc4 += z40*abs(x40) + z41*abs(x41)
acc5 += z50*abs(x50) + z51*abs(x51)
acc6 += z60*abs(x60) + z61*abs(x61)
acc7 += z70*abs(x70) + z71*abs(x71)
acc8 += z80*abs(x80) + z81*abs(x81)
acc9 += z90*abs(x90) + z91*abs(x91)
acc10 += z100*abs(x100) + z101*abs(x101)
acc11 += z110*abs(x110) + z111*abs(x111)
acc12 += z120*abs(x120) + z121*abs(x121)
acc13 += z130*abs(x130) + z131*abs(x131)
acc14 += z140*abs(x140) + z141*abs(x141)
acc15 += z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_abs	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 += z00*abs(x00) + z01*abs(x01)
acc1 += z10*abs(x10) + z11*abs(x11)
acc2 += z20*abs(x20) + z21*abs(x21)
acc3 += z30*abs(x30) + z31*abs(x31)
acc4 += z40*abs(x40) + z41*abs(x41)
acc5 += z50*abs(x50) + z51*abs(x51)
acc6 += z60*abs(x60) + z61*abs(x61)
acc7 += z70*abs(x70) + z71*abs(x71)
acc8 += z80*abs(x80) + z81*abs(x81)
acc9 += z90*abs(x90) + z91*abs(x91)
acc10 += z100*abs(x100) + z101*abs(x101)
acc11 += z110*abs(x110) + z111*abs(x111)
acc12 += z120*abs(x120) + z121*abs(x121)
acc13 += z130*abs(x130) + z131*abs(x131)
acc14 += z140*abs(x140) + z141*abs(x141)
acc15 += z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_max	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 += z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 += z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 += z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 += z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 += z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 += z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 += z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 += z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_max	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 += z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 += z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 += z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 += z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 += z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 += z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 += z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 += z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_max	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 += z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 += z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 += z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 += z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 += z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 += z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 += z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 += z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_maxdiff	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 += z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 += z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 += z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 += z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 += z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 += z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 += z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 += z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_maxdiff	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 += z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 += z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 += z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 += z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 += z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 += z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 += z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 += z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_maxdiff	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 += z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 += z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 += z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 += z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 += z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 += z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 += z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 += z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_min	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 += z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 += z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 += z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 += z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 += z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 += z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 += z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 += z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_min	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 += z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 += z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 += z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 += z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 += z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 += z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 += z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 += z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mac16_min	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 += z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 += z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 += z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 += z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 += z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 += z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 += z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 += z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_abs	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 += z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 += z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 += z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 += z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 += z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 += z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 += z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 += z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_abs	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 += z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 += z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 += z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 += z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 += z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 += z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 += z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 += z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_max	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_max	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_max	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 += z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 += z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 += z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 += z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 += z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 += z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 += z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_maxdiff	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_maxdiff	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_maxdiff	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 += z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 += z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 += z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 += z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 += z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 += z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 += z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_min	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_min	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mac8_min	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-accumulate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 += z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 += z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 += z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 += z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 += z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 += z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 += z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 += z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_abs	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 -= z00*abs(x00) + z01*abs(x01)
acc1 -= z10*abs(x10) + z11*abs(x11)
acc2 -= z20*abs(x20) + z21*abs(x21)
acc3 -= z30*abs(x30) + z31*abs(x31)
acc4 -= z40*abs(x40) + z41*abs(x41)
acc5 -= z50*abs(x50) + z51*abs(x51)
acc6 -= z60*abs(x60) + z61*abs(x61)
acc7 -= z70*abs(x70) + z71*abs(x71)
acc8 -= z80*abs(x80) + z81*abs(x81)
acc9 -= z90*abs(x90) + z91*abs(x91)
acc10 -= z100*abs(x100) + z101*abs(x101)
acc11 -= z110*abs(x110) + z111*abs(x111)
acc12 -= z120*abs(x120) + z121*abs(x121)
acc13 -= z130*abs(x130) + z131*abs(x131)
acc14 -= z140*abs(x140) + z141*abs(x141)
acc15 -= z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_abs	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*abs(x00) + z01*abs(x01)
acc1 -= z10*abs(x10) + z11*abs(x11)
acc2 -= z20*abs(x20) + z21*abs(x21)
acc3 -= z30*abs(x30) + z31*abs(x31)
acc4 -= z40*abs(x40) + z41*abs(x41)
acc5 -= z50*abs(x50) + z51*abs(x51)
acc6 -= z60*abs(x60) + z61*abs(x61)
acc7 -= z70*abs(x70) + z71*abs(x71)
acc8 -= z80*abs(x80) + z81*abs(x81)
acc9 -= z90*abs(x90) + z91*abs(x91)
acc10 -= z100*abs(x100) + z101*abs(x101)
acc11 -= z110*abs(x110) + z111*abs(x111)
acc12 -= z120*abs(x120) + z121*abs(x121)
acc13 -= z130*abs(x130) + z131*abs(x131)
acc14 -= z140*abs(x140) + z141*abs(x141)
acc15 -= z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_max	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 -= z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 -= z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 -= z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 -= z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 -= z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 -= z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 -= z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 -= z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_max	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 -= z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 -= z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 -= z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 -= z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 -= z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 -= z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 -= z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 -= z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_max	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 -= z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 -= z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 -= z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 -= z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 -= z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 -= z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 -= z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 -= z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_maxdiff	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 -= z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 -= z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 -= z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 -= z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 -= z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 -= z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 -= z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 -= z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_maxdiff	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 -= z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 -= z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 -= z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 -= z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 -= z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 -= z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 -= z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 -= z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_maxdiff	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 -= z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 -= z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 -= z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 -= z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 -= z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 -= z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 -= z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 -= z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_min	(	v16acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 -= z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 -= z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 -= z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 -= z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 -= z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 -= z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 -= z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 -= z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_min	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 -= z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 -= z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 -= z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 -= z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 -= z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 -= z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 -= z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 -= z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 msc16_min	(	v16acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 -= z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 -= z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 -= z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 -= z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 -= z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 -= z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 -= z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 -= z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

acc	Incoming accumulation vector (16 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_abs	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 -= z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 -= z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 -= z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 -= z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 -= z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 -= z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 -= z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 -= z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_abs	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 -= z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 -= z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 -= z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 -= z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 -= z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 -= z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 -= z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_max	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_max	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_max	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 -= z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 -= z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 -= z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 -= z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 -= z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 -= z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 -= z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_maxdiff	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_maxdiff	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_maxdiff	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 -= z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 -= z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 -= z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 -= z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 -= z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 -= z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 -= z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_min	(	v8acc48	acc,
		v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_min	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 msc8_min	(	v8acc48	acc,
		v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-subtract intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 -= z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 -= z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 -= z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 -= z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 -= z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 -= z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 -= z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 -= z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

acc	Incoming accumulation vector (8 x int48 lanes)
xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_abs	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 = z00*abs(x00) + z01*abs(x01)
acc1 = z10*abs(x10) + z11*abs(x11)
acc2 = z20*abs(x20) + z21*abs(x21)
acc3 = z30*abs(x30) + z31*abs(x31)
acc4 = z40*abs(x40) + z41*abs(x41)
acc5 = z50*abs(x50) + z51*abs(x51)
acc6 = z60*abs(x60) + z61*abs(x61)
acc7 = z70*abs(x70) + z71*abs(x71)
acc8 = z80*abs(x80) + z81*abs(x81)
acc9 = z90*abs(x90) + z91*abs(x91)
acc10 = z100*abs(x100) + z101*abs(x101)
acc11 = z110*abs(x110) + z111*abs(x111)
acc12 = z120*abs(x120) + z121*abs(x121)
acc13 = z130*abs(x130) + z131*abs(x131)
acc14 = z140*abs(x140) + z141*abs(x141)
acc15 = z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_abs	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 = z00*abs(x00) + z01*abs(x01)
acc1 = z10*abs(x10) + z11*abs(x11)
acc2 = z20*abs(x20) + z21*abs(x21)
acc3 = z30*abs(x30) + z31*abs(x31)
acc4 = z40*abs(x40) + z41*abs(x41)
acc5 = z50*abs(x50) + z51*abs(x51)
acc6 = z60*abs(x60) + z61*abs(x61)
acc7 = z70*abs(x70) + z71*abs(x71)
acc8 = z80*abs(x80) + z81*abs(x81)
acc9 = z90*abs(x90) + z91*abs(x91)
acc10 = z100*abs(x100) + z101*abs(x101)
acc11 = z110*abs(x110) + z111*abs(x111)
acc12 = z120*abs(x120) + z121*abs(x121)
acc13 = z130*abs(x130) + z131*abs(x131)
acc14 = z140*abs(x140) + z141*abs(x141)
acc15 = z150*abs(x150) + z151*abs(x151)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_max	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 = z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 = z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 = z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 = z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 = z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 = z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 = z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 = z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 = z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 = z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 = z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 = z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 = z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 = z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 = z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 = z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71))
acc8 = z80*(max(x80 , y80)) + z81*(max(x81 , y81))
acc9 = z90*(max(x90 , y90)) + z91*(max(x91 , y91))
acc10 = z100*(max(x100 , y100)) + z101*(max(x101 , y101))
acc11 = z110*(max(x110 , y110)) + z111*(max(x111 , y111))
acc12 = z120*(max(x120 , y120)) + z121*(max(x121 , y121))
acc13 = z130*(max(x130 , y130)) + z131*(max(x131 , y131))
acc14 = z140*(max(x140 , y140)) + z141*(max(x141 , y141))
acc15 = z150*(max(x150 , y150)) + z151*(max(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_maxdiff	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 = z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 = z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 = z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 = z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 = z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 = z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 = z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 = z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 = z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 = z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 = z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 = z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 = z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 = z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 = z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 = z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0)
acc8 = z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0)
acc9 = z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0)
acc10 = z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0)
acc11 = z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0)
acc12 = z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0)
acc13 = z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0)
acc14 = z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0)
acc15 = z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0)

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_min	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 = z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 = z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 = z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 = z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 = z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 = z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 = z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 = z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 = z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 = z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 = z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 = z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 = z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 = z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 = z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 = z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 mul16_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71))
acc8 = z80*(min(x80 , y80)) + z81*(min(x81 , y81))
acc9 = z90*(min(x90 , y90)) + z91*(min(x91 , y91))
acc10 = z100*(min(x100 , y100)) + z101*(min(x101 , y101))
acc11 = z110*(min(x110 , y110)) + z111*(min(x111 , y111))
acc12 = z120*(min(x120 , y120)) + z121*(min(x121 , y121))
acc13 = z130*(min(x130 , y130)) + z131*(min(x131 , y131))
acc14 = z140*(min(x140 , y140)) + z141*(min(x141 , y141))
acc15 = z150*(min(x150 , y150)) + z151*(min(x151 , y151))

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_abs	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 = z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 = z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 = z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 = z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 = z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 = z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 = z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 = z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_abs	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 = z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03)
acc1 = z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13)
acc2 = z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23)
acc3 = z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33)
acc4 = z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43)
acc5 = z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53)
acc6 = z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63)
acc7 = z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_max	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03))
acc1 = z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13))
acc2 = z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23))
acc3 = z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33))
acc4 = z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43))
acc5 = z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53))
acc6 = z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63))
acc7 = z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_maxdiff	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0)
acc1 = z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0)
acc2 = z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0)
acc3 = z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0)
acc4 = z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0)
acc5 = z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0)
acc6 = z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0)
acc7 = z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0)

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_min	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 mul8_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03))
acc1 = z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13))
acc2 = z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23))
acc3 = z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33))
acc4 = z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43))
acc5 = z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53))
acc6 = z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63))
acc7 = z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73))

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_abs	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 = -( z00*abs(x00) + z01*abs(x01) )
acc1 = -( z10*abs(x10) + z11*abs(x11) )
acc2 = -( z20*abs(x20) + z21*abs(x21) )
acc3 = -( z30*abs(x30) + z31*abs(x31) )
acc4 = -( z40*abs(x40) + z41*abs(x41) )
acc5 = -( z50*abs(x50) + z51*abs(x51) )
acc6 = -( z60*abs(x60) + z61*abs(x61) )
acc7 = -( z70*abs(x70) + z71*abs(x71) )
acc8 = -( z80*abs(x80) + z81*abs(x81) )
acc9 = -( z90*abs(x90) + z91*abs(x91) )
acc10 = -( z100*abs(x100) + z101*abs(x101) )
acc11 = -( z110*abs(x110) + z111*abs(x111) )
acc12 = -( z120*abs(x120) + z121*abs(x121) )
acc13 = -( z130*abs(x130) + z131*abs(x131) )
acc14 = -( z140*abs(x140) + z141*abs(x141) )
acc15 = -( z150*abs(x150) + z151*abs(x151) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_abs	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*abs(x00) + z01*abs(x01) )
acc1 = -( z10*abs(x10) + z11*abs(x11) )
acc2 = -( z20*abs(x20) + z21*abs(x21) )
acc3 = -( z30*abs(x30) + z31*abs(x31) )
acc4 = -( z40*abs(x40) + z41*abs(x41) )
acc5 = -( z50*abs(x50) + z51*abs(x51) )
acc6 = -( z60*abs(x60) + z61*abs(x61) )
acc7 = -( z70*abs(x70) + z71*abs(x71) )
acc8 = -( z80*abs(x80) + z81*abs(x81) )
acc9 = -( z90*abs(x90) + z91*abs(x91) )
acc10 = -( z100*abs(x100) + z101*abs(x101) )
acc11 = -( z110*abs(x110) + z111*abs(x111) )
acc12 = -( z120*abs(x120) + z121*abs(x121) )
acc13 = -( z130*abs(x130) + z131*abs(x131) )
acc14 = -( z140*abs(x140) + z141*abs(x141) )
acc15 = -( z150*abs(x150) + z151*abs(x151) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_max	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) )
acc8 = -( z80*(max(x80 , y80)) + z81*(max(x81 , y81)) )
acc9 = -( z90*(max(x90 , y90)) + z91*(max(x91 , y91)) )
acc10 = -( z100*(max(x100 , y100)) + z101*(max(x101 , y101)) )
acc11 = -( z110*(max(x110 , y110)) + z111*(max(x111 , y111)) )
acc12 = -( z120*(max(x120 , y120)) + z121*(max(x121 , y121)) )
acc13 = -( z130*(max(x130 , y130)) + z131*(max(x131 , y131)) )
acc14 = -( z140*(max(x140 , y140)) + z141*(max(x141 , y141)) )
acc15 = -( z150*(max(x150 , y150)) + z151*(max(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) )
acc8 = -( z80*(max(x80 , y80)) + z81*(max(x81 , y81)) )
acc9 = -( z90*(max(x90 , y90)) + z91*(max(x91 , y91)) )
acc10 = -( z100*(max(x100 , y100)) + z101*(max(x101 , y101)) )
acc11 = -( z110*(max(x110 , y110)) + z111*(max(x111 , y111)) )
acc12 = -( z120*(max(x120 , y120)) + z121*(max(x121 , y121)) )
acc13 = -( z130*(max(x130 , y130)) + z131*(max(x131 , y131)) )
acc14 = -( z140*(max(x140 , y140)) + z141*(max(x141 , y141)) )
acc15 = -( z150*(max(x150 , y150)) + z151*(max(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) )
acc8 = -( z80*(max(x80 , y80)) + z81*(max(x81 , y81)) )
acc9 = -( z90*(max(x90 , y90)) + z91*(max(x91 , y91)) )
acc10 = -( z100*(max(x100 , y100)) + z101*(max(x101 , y101)) )
acc11 = -( z110*(max(x110 , y110)) + z111*(max(x111 , y111)) )
acc12 = -( z120*(max(x120 , y120)) + z121*(max(x121 , y121)) )
acc13 = -( z130*(max(x130 , y130)) + z131*(max(x131 , y131)) )
acc14 = -( z140*(max(x140 , y140)) + z141*(max(x141 , y141)) )
acc15 = -( z150*(max(x150 , y150)) + z151*(max(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_maxdiff	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) )
acc8 = -( z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0) )
acc9 = -( z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0) )
acc10 = -( z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0) )
acc11 = -( z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0) )
acc12 = -( z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0) )
acc13 = -( z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0) )
acc14 = -( z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0) )
acc15 = -( z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) )
acc8 = -( z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0) )
acc9 = -( z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0) )
acc10 = -( z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0) )
acc11 = -( z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0) )
acc12 = -( z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0) )
acc13 = -( z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0) )
acc14 = -( z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0) )
acc15 = -( z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) )
acc8 = -( z80*((x80 - y80 > 0) ? (x80 - y80):0) + z81*((x81 - y81 > 0) ? (x81 - y81):0) )
acc9 = -( z90*((x90 - y90 > 0) ? (x90 - y90):0) + z91*((x91 - y91 > 0) ? (x91 - y91):0) )
acc10 = -( z100*((x100 - y100 > 0) ? (x100 - y100):0) + z101*((x101 - y101 > 0) ? (x101 - y101):0) )
acc11 = -( z110*((x110 - y110 > 0) ? (x110 - y110):0) + z111*((x111 - y111 > 0) ? (x111 - y111):0) )
acc12 = -( z120*((x120 - y120 > 0) ? (x120 - y120):0) + z121*((x121 - y121 > 0) ? (x121 - y121):0) )
acc13 = -( z130*((x130 - y130 > 0) ? (x130 - y130):0) + z131*((x131 - y131 > 0) ? (x131 - y131):0) )
acc14 = -( z140*((x140 - y140 > 0) ? (x140 - y140):0) + z141*((x141 - y141 > 0) ? (x141 - y141):0) )
acc15 = -( z150*((x150 - y150 > 0) ? (x150 - y150):0) + z151*((x151 - y151 > 0) ? (x151 - y151):0) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_min	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) )
acc8 = -( z80*(min(x80 , y80)) + z81*(min(x81 , y81)) )
acc9 = -( z90*(min(x90 , y90)) + z91*(min(x91 , y91)) )
acc10 = -( z100*(min(x100 , y100)) + z101*(min(x101 , y101)) )
acc11 = -( z110*(min(x110 , y110)) + z111*(min(x111 , y111)) )
acc12 = -( z120*(min(x120 , y120)) + z121*(min(x121 , y121)) )
acc13 = -( z130*(min(x130 , y130)) + z131*(min(x131 , y131)) )
acc14 = -( z140*(min(x140 , y140)) + z141*(min(x141 , y141)) )
acc15 = -( z150*(min(x150 , y150)) + z151*(min(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) )
acc8 = -( z80*(min(x80 , y80)) + z81*(min(x81 , y81)) )
acc9 = -( z90*(min(x90 , y90)) + z91*(min(x91 , y91)) )
acc10 = -( z100*(min(x100 , y100)) + z101*(min(x101 , y101)) )
acc11 = -( z110*(min(x110 , y110)) + z111*(min(x111 , y111)) )
acc12 = -( z120*(min(x120 , y120)) + z121*(min(x121 , y121)) )
acc13 = -( z130*(min(x130 , y130)) + z131*(min(x131 , y131)) )
acc14 = -( z140*(min(x140 , y140)) + z141*(min(x141 , y141)) )
acc15 = -( z150*(min(x150 , y150)) + z151*(min(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		unsigned int	xoffsets_hi,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		unsigned int	zoffsets_hi,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) )
acc8 = -( z80*(min(x80 , y80)) + z81*(min(x81 , y81)) )
acc9 = -( z90*(min(x90 , y90)) + z91*(min(x91 , y91)) )
acc10 = -( z100*(min(x100 , y100)) + z101*(min(x101 , y101)) )
acc11 = -( z110*(min(x110 , y110)) + z111*(min(x111 , y111)) )
acc12 = -( z120*(min(x120 , y120)) + z121*(min(x121 , y121)) )
acc13 = -( z130*(min(x130 , y130)) + z131*(min(x131 , y131)) )
acc14 = -( z140*(min(x140 , y140)) + z141*(min(x141 , y141)) )
acc15 = -( z150*(min(x150 , y150)) + z151*(min(x151 , y151)) )

Returns: Returned accumulation vector (16 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
xoffsets_hi	4b offset for each lane, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to 8th lane
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zoffsets_hi	4b offset for each lane, applied to input from Z buffer. LSB apply to 8th lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_abs	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer .

acc0 = -( z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03) )
acc1 = -( z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13) )
acc2 = -( z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23) )
acc3 = -( z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33) )
acc4 = -( z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43) )
acc5 = -( z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53) )
acc6 = -( z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63) )
acc7 = -( z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_abs	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the absolute value in the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*abs(x00) + z01*abs(x01) + z02*abs(x02) + z03*abs(x03) )
acc1 = -( z10*abs(x10) + z11*abs(x11) + z12*abs(x12) + z13*abs(x13) )
acc2 = -( z20*abs(x20) + z21*abs(x21) + z22*abs(x22) + z23*abs(x23) )
acc3 = -( z30*abs(x30) + z31*abs(x31) + z32*abs(x32) + z33*abs(x33) )
acc4 = -( z40*abs(x40) + z41*abs(x41) + z42*abs(x42) + z43*abs(x43) )
acc5 = -( z50*abs(x50) + z51*abs(x51) + z52*abs(x52) + z53*abs(x53) )
acc6 = -( z60*abs(x60) + z61*abs(x61) + z62*abs(x62) + z63*abs(x63) )
acc7 = -( z70*abs(x70) + z71*abs(x71) + z72*abs(x72) + z73*abs(x73) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to the x buffer. LSB apply to first lane
xstep	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_max	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer .

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_max	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(max(x00 , y00)) + z01*(max(x01 , y01)) + z02*(max(x02 , y02)) + z03*(max(x03 , y03)) )
acc1 = -( z10*(max(x10 , y10)) + z11*(max(x11 , y11)) + z12*(max(x12 , y12)) + z13*(max(x13 , y13)) )
acc2 = -( z20*(max(x20 , y20)) + z21*(max(x21 , y21)) + z22*(max(x22 , y22)) + z23*(max(x23 , y23)) )
acc3 = -( z30*(max(x30 , y30)) + z31*(max(x31 , y31)) + z32*(max(x32 , y32)) + z33*(max(x33 , y33)) )
acc4 = -( z40*(max(x40 , y40)) + z41*(max(x41 , y41)) + z42*(max(x42 , y42)) + z43*(max(x43 , y43)) )
acc5 = -( z50*(max(x50 , y50)) + z51*(max(x51 , y51)) + z52*(max(x52 , y52)) + z53*(max(x53 , y53)) )
acc6 = -( z60*(max(x60 , y60)) + z61*(max(x61 , y61)) + z62*(max(x62 , y62)) + z63*(max(x63 , y63)) )
acc7 = -( z70*(max(x70 , y70)) + z71*(max(x71 , y71)) + z72*(max(x72 , y72)) + z73*(max(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_maxdiff	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer .

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_maxdiff	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the maximum difference between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*((x00 - y00 > 0) ? (x00 - y00):0) + z01*((x01 - y01 > 0) ? (x01 - y01):0) + z02*((x02 - y02 > 0) ? (x02 - y02):0) + z03*((x03 - y03 > 0) ? (x03 - y03):0) )
acc1 = -( z10*((x10 - y10 > 0) ? (x10 - y10):0) + z11*((x11 - y11 > 0) ? (x11 - y11):0) + z12*((x12 - y12 > 0) ? (x12 - y12):0) + z13*((x13 - y13 > 0) ? (x13 - y13):0) )
acc2 = -( z20*((x20 - y20 > 0) ? (x20 - y20):0) + z21*((x21 - y21 > 0) ? (x21 - y21):0) + z22*((x22 - y22 > 0) ? (x22 - y22):0) + z23*((x23 - y23 > 0) ? (x23 - y23):0) )
acc3 = -( z30*((x30 - y30 > 0) ? (x30 - y30):0) + z31*((x31 - y31 > 0) ? (x31 - y31):0) + z32*((x32 - y32 > 0) ? (x32 - y32):0) + z33*((x33 - y33 > 0) ? (x33 - y33):0) )
acc4 = -( z40*((x40 - y40 > 0) ? (x40 - y40):0) + z41*((x41 - y41 > 0) ? (x41 - y41):0) + z42*((x42 - y42 > 0) ? (x42 - y42):0) + z43*((x43 - y43 > 0) ? (x43 - y43):0) )
acc5 = -( z50*((x50 - y50 > 0) ? (x50 - y50):0) + z51*((x51 - y51 > 0) ? (x51 - y51):0) + z52*((x52 - y52 > 0) ? (x52 - y52):0) + z53*((x53 - y53 > 0) ? (x53 - y53):0) )
acc6 = -( z60*((x60 - y60 > 0) ? (x60 - y60):0) + z61*((x61 - y61 > 0) ? (x61 - y61):0) + z62*((x62 - y62 > 0) ? (x62 - y62):0) + z63*((x63 - y63 > 0) ? (x63 - y63):0) )
acc7 = -( z70*((x70 - y70 > 0) ? (x70 - y70):0) + z71*((x71 - y71 > 0) ? (x71 - y71):0) + z72*((x72 - y72 > 0) ? (x72 - y72):0) + z73*((x73 - y73 > 0) ? (x73 - y73):0) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_min	(	v64int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer .

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 64 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ystart	Starting position offset applied to all lanes for input from Y buffer.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8_min	(	v32int16	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int16	ybuff,
		int	ystart,
		unsigned int	ysquare,
		v16int16	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep
	)

Multiply-negate intrinsic function after computing the minimum between the selected lanes from the input buffer using small X input buffer.

acc0 = -( z00*(min(x00 , y00)) + z01*(min(x01 , y01)) + z02*(min(x02 , y02)) + z03*(min(x03 , y03)) )
acc1 = -( z10*(min(x10 , y10)) + z11*(min(x11 , y11)) + z12*(min(x12 , y12)) + z13*(min(x13 , y13)) )
acc2 = -( z20*(min(x20 , y20)) + z21*(min(x21 , y21)) + z22*(min(x22 , y22)) + z23*(min(x23 , y23)) )
acc3 = -( z30*(min(x30 , y30)) + z31*(min(x31 , y31)) + z32*(min(x32 , y32)) + z33*(min(x33 , y33)) )
acc4 = -( z40*(min(x40 , y40)) + z41*(min(x41 , y41)) + z42*(min(x42 , y42)) + z43*(min(x43 , y43)) )
acc5 = -( z50*(min(x50 , y50)) + z51*(min(x51 , y51)) + z52*(min(x52 , y52)) + z53*(min(x53 , y53)) )
acc6 = -( z60*(min(x60 , y60)) + z61*(min(x61 , y61)) + z62*(min(x62 , y62)) + z63*(min(x63 , y63)) )
acc7 = -( z70*(min(x70 , y70)) + z71*(min(x71 , y71)) + z72*(min(x72 , y72)) + z73*(min(x73 , y73)) )

Returns: Returned accumulation vector (8 x int48 lanes)

Parameters

xbuff	Input buffer of 32 elements of type int16
xstart	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xoffsets	4b offset for each lane, while each second lane is an offset to the lane before + 1, applied to both x and y buffers. LSB apply to first lane
xstep	Step between each column for selection in the x and y buffers. Ystep is symmetric to xstep (ystep advances by -xstep). xstep is restricted to multiples of 2 as granularity for xbuff is 32-bit.
xsquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
ybuff	Right input buffer of 32 elements of type int16
ystart	Starting position offset applied to all lanes for input from Y buffer. ystart is restricted to multiples of 2 as granularity for ybuff is 32-bit.
ysquare	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	Input buffer of 16 elements of type int16
zstart	Starting position offset applied to all lanes for input from Z buffer. This must be a compile time constant. Only the 4 LSB of the argument are used.
zoffsets	4b offset for each lane, applied to input from Z buffer. LSB apply to first lane
zstep	Step between each column for selection in the zbuffer.

Note

This intrinsic uses the 'square' parameter, to have more information on how to use this please go here
For more information on how data selection works from the buffers go here. For this intrinsic, the data buffer uses the 16bx16b scheme and the coefficient buffer uses the general scheme.
Parameter 'zstart' must be a compile time constant.

Overview

Functions

Function Documentation