8 bit Real x 8 bit Real

Overview

Vector MAC intrinsics with 8 bit real by 8 bit real

Functions
v16acc48	mac16 (v16acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v16acc48	mac16 (v16acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v16acc48	mac16 (v16acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v16acc48	mac16 (v16acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v16acc48	mac16 (v16acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v16acc48	mac16 (v16acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v16acc48	mac16 (v16acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v16acc48	mac16 (v16acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v8acc48	mac8 (v8acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v8acc48	mac8 (v8acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v8acc48	mac8 (v8acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v8acc48	mac8 (v8acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v8acc48	mac8 (v8acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v8acc48	mac8 (v8acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v8acc48	mac8 (v8acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function .
v8acc48	mac8 (v8acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-accumulate intrinsic function using small X input buffer.
v16acc48	msc16 (v16acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v16acc48	msc16 (v16acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v16acc48	msc16 (v16acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v16acc48	msc16 (v16acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v16acc48	msc16 (v16acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v16acc48	msc16 (v16acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v16acc48	msc16 (v16acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v16acc48	msc16 (v16acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v8acc48	msc8 (v8acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v8acc48	msc8 (v8acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v8acc48	msc8 (v8acc48 acc, v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v8acc48	msc8 (v8acc48 acc, v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v8acc48	msc8 (v8acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v8acc48	msc8 (v8acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v8acc48	msc8 (v8acc48 acc, v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function .
v8acc48	msc8 (v8acc48 acc, v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-subtract intrinsic function using small X input buffer.
v16acc48	mul16 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v16acc48	mul16 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v16acc48	mul16 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v16acc48	mul16 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v16acc48	mul16 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v16acc48	mul16 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v16acc48	mul16 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v16acc48	mul16 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v8acc48	mul8 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v8acc48	mul8 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v8acc48	mul8 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v8acc48	mul8 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v8acc48	mul8 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v8acc48	mul8 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v8acc48	mul8 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function .
v8acc48	mul8 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply intrinsic function using small X input buffer.
v16acc48	negmul16 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v16acc48	negmul16 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v16acc48	negmul16 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v16acc48	negmul16 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v16acc48	negmul16 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v16acc48	negmul16 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v16acc48	negmul16 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v16acc48	negmul16 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v8acc48	negmul8 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v8acc48	negmul8 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v8acc48	negmul8 (v128int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v8acc48	negmul8 (v64int8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v8acc48	negmul8 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v8acc48	negmul8 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32int8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.
v8acc48	negmul8 (v128uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function .
v8acc48	negmul8 (v64uint8 xbuff, int xstart, unsigned int xoffsets, int xstep, unsigned int xsquare, v32uint8 zbuff, int zstart, unsigned int zoffsets, int zstep, unsigned int zsquare)
	Multiply-negate intrinsic function using small X input buffer.

Function Documentation

v16acc48 mac16	(	v16acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mac16	(	v16acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 += z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 += z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 += z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 += z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 += z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 += z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 += z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 += z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function .

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mac8	(	v8acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-accumulate intrinsic function using small X input buffer.

acc0 += z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 += z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 += z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 += z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 += z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 += z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 += z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 += z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 msc16	(	v16acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 -= z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 -= z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 -= z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 -= z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 -= z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 -= z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 -= z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 -= z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
acc	v16acc48	Incoming accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function .

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 msc8	(	v8acc48	acc,
		v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-subtract intrinsic function using small X input buffer.

acc0 -= z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 -= z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 -= z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 -= z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 -= z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 -= z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 -= z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 -= z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
acc	v8acc48	Incoming accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 mul16	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77
acc8 = z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87
acc9 = z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97
acc10 = z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107
acc11 = z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117
acc12 = z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127
acc13 = z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137
acc14 = z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147
acc15 = z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function .

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 mul8	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply intrinsic function using small X input buffer.

acc0 = z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015
acc1 = z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115
acc2 = z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215
acc3 = z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315
acc4 = z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415
acc5 = z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515
acc6 = z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615
acc7 = z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v16acc48 negmul16	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 )
acc8 = -( z80*x80 + z81*x81 + z82*x82 + z83*x83 + z84*x84 + z85*x85 + z86*x86 + z87*x87 )
acc9 = -( z90*x90 + z91*x91 + z92*x92 + z93*x93 + z94*x94 + z95*x95 + z96*x96 + z97*x97 )
acc10 = -( z100*x100 + z101*x101 + z102*x102 + z103*x103 + z104*x104 + z105*x105 + z106*x106 + z107*x107 )
acc11 = -( z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 + z116*x116 + z117*x117 )
acc12 = -( z120*x120 + z121*x121 + z122*x122 + z123*x123 + z124*x124 + z125*x125 + z126*x126 + z127*x127 )
acc13 = -( z130*x130 + z131*x131 + z132*x132 + z133*x133 + z134*x134 + z135*x135 + z136*x136 + z137*x137 )
acc14 = -( z140*x140 + z141*x141 + z142*x142 + z143*x143 + z144*x144 + z145*x145 + z146*x146 + z147*x147 )
acc15 = -( z150*x150 + z151*x151 + z152*x152 + z153*x153 + z154*x154 + z155*x155 + z156*x156 + z157*x157 )

Parameters

Input/Output	Type	Comments
return	v16acc48	Returned accumulation vector (16 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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
xstep	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane, applied to input from Z buffer, corresponds to 2x the lane number and each second lane is an offset to the lane before + 1. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v128int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128int8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v64int8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64int8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32int8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32int8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v128uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function .

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v128uint8	Input buffer of 128 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.

v8acc48 negmul8	(	v64uint8	xbuff,
		int	xstart,
		unsigned int	xoffsets,
		int	xstep,
		unsigned int	xsquare,
		v32uint8	zbuff,
		int	zstart,
		unsigned int	zoffsets,
		int	zstep,
		unsigned int	zsquare
	)

Multiply-negate intrinsic function using small X input buffer.

acc0 = -( z00*x00 + z01*x01 + z02*x02 + z03*x03 + z04*x04 + z05*x05 + z06*x06 + z07*x07 + z08*x08 + z09*x09 + z010*x010 + z011*x011 + z012*x012 + z013*x013 + z014*x014 + z015*x015 )
acc1 = -( z10*x10 + z11*x11 + z12*x12 + z13*x13 + z14*x14 + z15*x15 + z16*x16 + z17*x17 + z18*x18 + z19*x19 + z110*x110 + z111*x111 + z112*x112 + z113*x113 + z114*x114 + z115*x115 )
acc2 = -( z20*x20 + z21*x21 + z22*x22 + z23*x23 + z24*x24 + z25*x25 + z26*x26 + z27*x27 + z28*x28 + z29*x29 + z210*x210 + z211*x211 + z212*x212 + z213*x213 + z214*x214 + z215*x215 )
acc3 = -( z30*x30 + z31*x31 + z32*x32 + z33*x33 + z34*x34 + z35*x35 + z36*x36 + z37*x37 + z38*x38 + z39*x39 + z310*x310 + z311*x311 + z312*x312 + z313*x313 + z314*x314 + z315*x315 )
acc4 = -( z40*x40 + z41*x41 + z42*x42 + z43*x43 + z44*x44 + z45*x45 + z46*x46 + z47*x47 + z48*x48 + z49*x49 + z410*x410 + z411*x411 + z412*x412 + z413*x413 + z414*x414 + z415*x415 )
acc5 = -( z50*x50 + z51*x51 + z52*x52 + z53*x53 + z54*x54 + z55*x55 + z56*x56 + z57*x57 + z58*x58 + z59*x59 + z510*x510 + z511*x511 + z512*x512 + z513*x513 + z514*x514 + z515*x515 )
acc6 = -( z60*x60 + z61*x61 + z62*x62 + z63*x63 + z64*x64 + z65*x65 + z66*x66 + z67*x67 + z68*x68 + z69*x69 + z610*x610 + z611*x611 + z612*x612 + z613*x613 + z614*x614 + z615*x615 )
acc7 = -( z70*x70 + z71*x71 + z72*x72 + z73*x73 + z74*x74 + z75*x75 + z76*x76 + z77*x77 + z78*x78 + z79*x79 + z710*x710 + z711*x711 + z712*x712 + z713*x713 + z714*x714 + z715*x715 )

Parameters

Input/Output	Type	Comments
return	v8acc48	Returned accumulation vector (8 x int48 lanes)
xbuff	v64uint8	Input buffer of 64 elements of type int8
xstart	int	Starting position offset applied to all lanes of input from X buffer. xstart is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xoffsets	unsigned int	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	unsigned int	Step between each column for selection in the xbuffer. xstep is restricted to multiples of 4 as granularity for xbuff is 32-bit.
xsquare	unsigned int	Select order of the mini-permute square (default=0x3210). LSB apply to first element
zbuff	v32uint8	Input buffer of 32 elements of type int8
zstart	int	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. zstart is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zoffsets	unsigned int	4b offset for each lane while each second lane is an offset to the lane before + 1, applied to input from Z buffer. LSB apply to first lane
zstep	int	Step between each column for selection in the zbuffer. zstep is restricted to multiples of 2 as granularity for zbuff is 16-bit.
zsquare	unsigned int	4b offset for each lane, Select order of the mini-permute square (default=0x3210). LSB apply to first element

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 8bx8b scheme and the coefficient buffer uses the 8bx8b scheme.
• Parameter 'zstart' must be a compile time constant.