ipu Namespace Reference

IPU intrinsic functions. More...

Functions
int	andc (int src0, int src1)
	Targets the andc instruction. More...
unsigned	andc (unsigned src0, unsigned src1)
	Targets the andc instruction. More...
float	andc (float src0, float src1)
	Targets the andc instruction. More...
float2	andc (float2 src0, float2 src1)
	Targets the andc64 instruction. More...
int	bitrev8 (int src)
	Targets the bitrev8 instruction. More...
unsigned	cms (int src)
	Targets the cms instruction. More...
float2	roll32 (float2 src0, float2 src1)
	Targets the roll32 instruction. More...
int	roll8l (int src0, int src1)
	Targets the roll8l instruction. More...
int	roll8r (int src0, int src1)
	Targets the roll8r instruction. More...
int	shuf8x8hi (int src0, int src1)
	Targets the shuf8x8hi instruction. More...
int	shuf8x8lo (int src0, int src1)
	Targets the shuf8x8lo instruction. More...
float2	sort4x32hi (float2 src0, float2 src1)
	Targets the sort4x32hi instruction. More...
float2	sort4x32lo (float2 src0, float2 src1)
	Targets the sort4x32lo instruction. More...
int	sort8 (int src)
	Targets the sort8 instruction. More...
int	sort8x8hi (int src0, int src1)
	Targets the sort8x8hi instruction. More...
int	sort8x8lo (int src0, int src1)
	Targets the sort8x8lo instruction. More...
int	swap8 (int src)
	Targets the sort8 instruction. More...
half2	absadd (half2 src0, half2 src1)
	Targets the f16v2absadd instruction. More...
half4	absadd (half4 src0, half4 src1)
	Targets the f16v4absadd instruction. More...
float2	absadd (float2 src0, float2 src1)
	Targets the f32v2absadd instruction. More...
float	absadd (float src0, float src1)
	Targets the f32absadd instruction. More...
half2	absmax (half2 src0, half2 src1)
	Targets the f16v2absmax instruction. More...
half4	absmax (half4 src0, half4 src1)
	Targets the f16v4absmax instruction. More...
float2	absmax (float2 src0, float2 src1)
	Targets the f32v2absmax instruction. More...
float	absmax (float src0, float src1)
	Targets the f32absmax instruction. More...
half2	max (half2 src0, half2 src1)
	Targets the f16v2max instruction. More...
half4	max (half4 src0, half4 src1)
	Targets the f16v4max instruction. More...
float2	max (float2 src0, float2 src1)
	Targets the f32v2max instruction. More...
float	max (float src0, float src1)
	Targets the f32max instruction. More...
half2	maxc (half4 src)
	Targets the f16v4maxc instruction. More...
half2	min (half2 src0, half2 src1)
	Targets the f16v2min instruction. More...
half4	min (half4 src0, half4 src1)
	Targets the f16v4min instruction. More...
float2	min (float2 src0, float2 src1)
	Targets the f32v2min instruction. More...
float	min (float src0, float src1)
	Targets the f32min instruction. More...
half2	clamp (half2 src0, half2 src1)
	Targets the f16v2clamp instruction. More...
half4	clamp (half4 src0, half2 src1)
	Targets the f16v4clamp instruction. More...
float2	clamp (float2 src0, float2 src1)
	Targets the f32v2clamp instruction. More...
float	clamp (float src0, float2 src1)
	Targets the f32clamp instruction. More...
void	cmac (half2 src0, half2 src1)
	Targets the f16v2cmac instruction. More...
void	cmac (half4 src0, half4 src1)
	Targets the f16v4cmac instruction. More...
half2	exp (half2 src)
	Targets the f16v2exp instruction. More...
float	exp (float src)
	Targets the f32exp instruction. More...
half2	exp2 (half2 src)
	Targets the f16v2exp instruction. More...
float	exp2 (float src)
	Targets the f32exp instruction. More...
half2	log2 (half2 src)
	Targets the f16v2log2 instruction. More...
float	log2 (float src)
	Targets the f32ln instruction. More...
half2	tanh (half2 src)
	Targets the f16v2tanh instruction. More...
float	tanh (float src)
	Targets the f32tanh instruction. More...
half2	ln (half2 src)
	Targets the f16v2ln instruction. More...
float	ln (float src)
	Targets the f32ln instruction. More...
float2	axpy (float2 src0, float2 src1)
	Targets the f32v2axpy instruction. More...
half2	f16v2grand ()
	Targets the f16v2grand instruction. More...
float2	f32v2grand ()
	Targets the f32v2grand instruction. More...
half4	rmask (half4 src0, float src1)
	Targets the f16v4rmask instruction. More...
float2	rmask (float2 src0, float src1)
	Targets the f32v2rmask instruction. More...
half2	sigm (half2 src)
	Targets the f16v2sigm instruction. More...
float	sigm (float src)
	Targets the f32sigm instruction. More...
float	sum (half2 src)
	Targets the f16v2sum instruction. More...
float2	sum (half4 src)
	Targets the f16v4sum instruction. More...
half2	cmpeq (half2 src0, half2 src1)
	Targets the f16v2cmpeq instruction. More...
half4	cmpeq (half4 src0, half4 src1)
	Targets the f16v4cmpeq instruction. More...
float2	cmpeq (float2 src0, float2 src1)
	Targets the f32v2cmpeq instruction. More...
float	cmpeq (float src0, float src1)
	Targets the f32cmpeq instruction. More...
half2	cmpge (half2 src0, half2 src1)
	Targets the f16v2cmpge instruction. More...
half4	cmpge (half4 src0, half4 src1)
	Targets the f16v4cmpge instruction. More...
float2	cmpge (float2 src0, float2 src1)
	Targets the f32v2cmpge instruction. More...
float	cmpge (float src0, float src1)
	Targets the f32cmpge instruction. More...
half2	cmpgt (half2 src0, half2 src1)
	Targets the f16v2cmpgt instruction. More...
half4	cmpgt (half4 src0, half4 src1)
	Targets the f16v4cmpgt instruction. More...
float2	cmpgt (float2 src0, float2 src1)
	Targets the f32v2cmpgt instruction. More...
float	cmpgt (float src0, float src1)
	Targets the f32cmpgt instruction. More...
half2	cmple (half2 src0, half2 src1)
	Targets the f16v2cmple instruction. More...
half4	cmple (half4 src0, half4 src1)
	Targets the f16v4cmple instruction. More...
float2	cmple (float2 src0, float2 src1)
	Targets the f32v2cmple instruction. More...
float	cmple (float src0, float src1)
	Targets the f32cmple instruction. More...
half2	cmplt (half2 src0, half2 src1)
	Targets the f16v2cmplt instruction. More...
half4	cmplt (half4 src0, half4 src1)
	Targets the f16v4cmplt instruction. More...
float2	cmplt (float2 src0, float2 src1)
	Targets the f32v2cmplt instruction. More...
float	cmplt (float src0, float src1)
	Targets the f32cmplt instruction. More...
half2	cmpne (half2 src0, half2 src1)
	Targets the f16v2cmpne instruction. More...
half4	cmpne (half4 src0, half4 src1)
	Targets the f16v4cmpne instruction. More...
float2	cmpne (float2 src0, float2 src1)
	Targets the f32v2cmpne instruction. More...
float	cmpne (float src0, float src1)
	Targets the f32cmpne instruction. More...
unsigned	clz (int src)
	Targets the clz instruction. More...
unsigned	popc (int src)
	Targets the popc instruction. More...
short2	roll16 (short2 src0, short2 src1)
	Targets the roll16 instruction. More...
ushort2	roll16 (ushort2 src0, ushort2 src1)
	Targets the roll16 instruction. More...
half2	roll16 (half2 src0, half2 src1)
	Targets the roll16 instruction. More...
short2	sort4x16hi (short2 src0, short2 src1)
	Targets the sort4x16hi instruction. More...
ushort2	sort4x16hi (ushort2 src0, ushort2 src1)
	Targets the sort4x16hi instruction. More...
half2	sort4x16hi (half2 src0, half2 src1)
	Targets the sort4x16hi instruction. More...
short2	sort4x16lo (short2 src0, short2 src1)
	Targets the sort4x16lo instruction. More...
ushort2	sort4x16lo (ushort2 src0, ushort2 src1)
	Targets the sort4x16lo instruction. More...
half2	sort4x16lo (half2 src0, half2 src1)
	Targets the sort4x16lo instruction. More...
half	load_postinc (const half **a, int i)
	Post-incrementing load, targeting the ldb16step instruction. More...
void	store_postinc (half2 **a, half2 v, int i)
	Post-incrementing store, targeting the st32step instruction. More...
half2	load_postinc (const half2 **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (half4 **a, half4 v, int i)
	Post-incrementing store, targeting the st64step instruction. More...
half4	load_postinc (const half4 **a, int i)
	Post-incrementing load, targeting the ld64step instruction. More...
void	store_postinc (float **a, float v, int i)
	Post-incrementing store, targeting the st32step instruction. More...
float	load_postinc (const float **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (float2 **a, float2 v, int i)
	Post-incrementing store, targeting the st64step instruction. More...
float2	load_postinc (const float2 **a, int i)
	Post-incrementing load, targeting the ld64step instruction. More...
void	store_postinc (int **a, int v, int i)
	Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise. More...
int	load_postinc (const int **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (unsigned **a, unsigned v, int i)
	Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise. More...
unsigned	load_postinc (const unsigned **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (int2 **a, int2 v, int i)
	Post-incrementing store. More...
int2	load_postinc (const int2 **a, int i)
	Post-incrementing load. More...
void	store_postinc (uint2 **a, uint2 v, int i)
	Post-incrementing store. More...
uint2	load_postinc (const uint2 **a, int i)
	Post-incrementing load. More...
void	store_postinc (short **a, short v, int i)
	Post-incrementing store. More...
short	load_postinc (const short **a, int i)
	Post-incrementing load, targeting the lds16step instruction. More...
void	store_postinc (unsigned short **a, unsigned short v, int i)
	Post-incrementing store. More...
unsigned short	load_postinc (const unsigned short **a, int i)
	Post-incrementing load, targeting the ldz16step instruction. More...
void	store_postinc (short2 **a, short2 v, int i)
	Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise. More...
short2	load_postinc (const short2 **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (ushort2 **a, ushort2 v, int i)
	Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise. More...
ushort2	load_postinc (const ushort2 **a, int i)
	Post-incrementing load, targeting the ld32step instruction. More...
void	store_postinc (short4 **a, short4 v, int i)
	Post-incrementing store. More...
short4	load_postinc (const short4 **a, int i)
	Post-incrementing load. More...
void	store_postinc (ushort4 **a, ushort4 v, int i)
	Post-incrementing store. More...
ushort4	load_postinc (const ushort4 **a, int i)
	Post-incrementing load. More...
void	store_postinc (char **a, char v, int i)
	Post-incrementing store. More...
char	load_postinc (const char **a, int i)
	Post-incrementing load, targeting the lds8step instruction. More...
void	store_postinc (unsigned char **a, unsigned char v, int i)
	Post-incrementing store. More...
unsigned char	load_postinc (const unsigned char **a, int i)
	Post-incrementing load, targeting the ldz8step instruction. More...
float	acos (float x)
	The arccos function, the inverse of cosine. More...
float2	acos (float2 x)
	The arccos function, the inverse of cosine. More...
half	acos (half x)
	The arccos function, the inverse of cosine. More...
half2	acos (half2 x)
	The arccos function, the inverse of cosine. More...
half4	acos (half4 x)
	The arccos function, the inverse of cosine. More...
float	acosh (float x)
	The arccosh function, the inverse of the hyperbolic cosine. More...
float2	acosh (float2 x)
	The arccosh function, the inverse of the hyperbolic cosine. More...
half	acosh (half x)
	The arccosh function, the inverse of the hyperbolic cosine. More...
half2	acosh (half2 x)
	The arccosh function, the inverse of the hyperbolic cosine. More...
half4	acosh (half4 x)
	The arccosh function, the inverse of the hyperbolic cosine. More...
float	asin (float x)
	The arcsin function, the inverse of sine. More...
float2	asin (float2 x)
	The arcsin function, the inverse of sine. More...
half	asin (half x)
	The arcsin function, the inverse of sine. More...
half2	asin (half2 x)
	The arcsin function, the inverse of sine. More...
half4	asin (half4 x)
	The arcsin function, the inverse of sine. More...
float	asinh (float x)
	The arcsinh function, the inverse of the hyperbolic sine. More...
float2	asinh (float2 x)
	The arcsinh function, the inverse of the hyperbolic sine. More...
half	asinh (half x)
	The arcsinh function, the inverse of the hyperbolic sine. More...
half2	asinh (half2 x)
	The arcsinh function, the inverse of the hyperbolic sine. More...
half4	asinh (half4 x)
	The arcsinh function, the inverse of the hyperbolic sine. More...
float	atan (float x)
	The arctan function, the inverse of tangent. More...
float2	atan (float2 x)
	The arctan function, the inverse of tangent. More...
half	atan (half x)
	The arctan function, the inverse of tangent. More...
half2	atan (half2 x)
	The arctan function, the inverse of tangent. More...
half4	atan (half4 x)
	The arctan function, the inverse of tangent. More...
float	atanh (float x)
	The arctanh function, the inverse of the hyperbolic tangent. More...
float2	atanh (float2 x)
	The arctanh function, the inverse of the hyperbolic tangent. More...
half	atanh (half x)
	The arctanh function, the inverse of the hyperbolic tangent. More...
half2	atanh (half2 x)
	The arctanh function, the inverse of the hyperbolic tangent. More...
half4	atanh (half4 x)
	The arctanh function, the inverse of the hyperbolic tangent. More...
float	cbrt (float x)
	The cubic root function. More...
float2	cbrt (float2 x)
	The cubic root function. More...
half	cbrt (half x)
	The cubic root function. More...
half2	cbrt (half2 x)
	The cubic root function. More...
half4	cbrt (half4 x)
	The cubic root function. More...
float	ceil (float x)
	Rounds up input to the closest integral value. More...
float2	ceil (float2 x)
	Rounds up input to the closest integral value. More...
half	ceil (half x)
	Rounds up input to the closest integral value. More...
half2	ceil (half2 x)
	Rounds up input to the closest integral value. More...
half4	ceil (half4 x)
	Rounds up input to the closest integral value. More...
float	cos (float x)
	The trigonometric cosine function. More...
float2	cos (float2 x)
	The trigonometric cosine function. More...
half	cos (half x)
	The trigonometric cosine function. More...
half2	cos (half2 x)
	The trigonometric cosine function. More...
half4	cos (half4 x)
	The trigonometric cosine function. More...
float	cosh (float x)
	The hyperbolic cosine function. More...
float2	cosh (float2 x)
	The hyperbolic cosine function. More...
half	cosh (half x)
	The hyperbolic cosine function. More...
half2	cosh (half2 x)
	The hyperbolic cosine function. More...
half4	cosh (half4 x)
	The hyperbolic cosine function. More...
float	erf (float x)
	The error function. More...
float2	erf (float2 x)
	The error function. More...
half	erf (half x)
	The error function. More...
half2	erf (half2 x)
	The error function. More...
half4	erf (half4 x)
	The error function. More...
float	erfc (float x)
	The complementary error function. More...
float2	erfc (float2 x)
	The complementary error function. More...
half	erfc (half x)
	The complementary error function. More...
half2	erfc (half2 x)
	The complementary error function. More...
half4	erfc (half4 x)
	The complementary error function. More...
float2	exp (float2 x)
	The base-e exponential function. More...
half	exp (half x)
	The base-e exponential function. More...
half4	exp (half4 x)
	The base-e exponential function. More...
float2	exp2 (float2 x)
	The base-2 exponential function. More...
half	exp2 (half x)
	The base-2 exponential function. More...
half4	exp2 (half4 x)
	The base-2 exponential function. More...
float	expm1 (float x)
	The base-e exponential function, minus one: exp(x) - 1. More...
float2	expm1 (float2 x)
	The base-e exponential function, minus one: exp(x) - 1. More...
half	expm1 (half x)
	The base-e exponential function, minus one: exp(x) - 1. More...
half2	expm1 (half2 x)
	The base-e exponential function, minus one: exp(x) - 1. More...
half4	expm1 (half4 x)
	The base-e exponential function, minus one: exp(x) - 1. More...
float	fabs (float x)
	Computes the absolute value of the input. More...
float2	fabs (float2 x)
	Computes the absolute value of the input. More...
half	fabs (half x)
	Computes the absolute value of the input. More...
half2	fabs (half2 x)
	Computes the absolute value of the input. More...
half4	fabs (half4 x)
	Computes the absolute value of the input. More...
float	floor (float x)
	Rounds down input to the closest integral value. More...
float2	floor (float2 x)
	Rounds down input to the closest integral value. More...
half	floor (half x)
	Rounds down input to the closest integral value. More...
half2	floor (half2 x)
	Rounds down input to the closest integral value. More...
half4	floor (half4 x)
	Rounds down input to the closest integral value. More...
float	log (float x)
	The natural logarithm. More...
float2	log (float2 x)
	The natural logarithm. More...
half	log (half x)
	The natural logarithm. More...
half2	log (half2 x)
	The natural logarithm. More...
half4	log (half4 x)
	The natural logarithm. More...
float	log10 (float x)
	The base-10 logarithm. More...
float2	log10 (float2 x)
	The base-10 logarithm. More...
half	log10 (half x)
	The base-10 logarithm. More...
half2	log10 (half2 x)
	The base-10 logarithm. More...
half4	log10 (half4 x)
	The base-10 logarithm. More...
float	log1p (float x)
	The natural logarithm of 1 + x. More...
float2	log1p (float2 x)
	The natural logarithm of 1 + x. More...
half	log1p (half x)
	The natural logarithm of 1 + x. More...
half2	log1p (half2 x)
	The natural logarithm of 1 + x. More...
half4	log1p (half4 x)
	The natural logarithm of 1 + x. More...
float2	log2 (float2 x)
	The base-2 logarithm. More...
half	log2 (half x)
	The base-2 logarithm. More...
half4	log2 (half4 x)
	The base-2 logarithm. More...
float	nearbyint (float x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
float2	nearbyint (float2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half	nearbyint (half x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half2	nearbyint (half2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half4	nearbyint (half4 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
float	rint (float x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
float2	rint (float2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half	rint (half x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half2	rint (half2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
half4	rint (half4 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
float	round (float x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
float2	round (float2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
half	round (half x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
half2	round (half2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
half4	round (half4 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
float	sin (float x)
	The trigonometric sine function. More...
float2	sin (float2 x)
	The trigonometric sine function. More...
half	sin (half x)
	The trigonometric sine function. More...
half2	sin (half2 x)
	The trigonometric sine function. More...
half4	sin (half4 x)
	The trigonometric sine function. More...
float	sinh (float x)
	The hyperbolic sine function. More...
float2	sinh (float2 x)
	The hyperbolic sine function. More...
half	sinh (half x)
	The hyperbolic sine function. More...
half2	sinh (half2 x)
	The hyperbolic sine function. More...
half4	sinh (half4 x)
	The hyperbolic sine function. More...
float	sqrt (float x)
	The square root function. More...
float2	sqrt (float2 x)
	The square root function. More...
half	sqrt (half x)
	The square root function. More...
half2	sqrt (half2 x)
	The square root function. More...
half4	sqrt (half4 x)
	The square root function. More...
float	rsqrt (float x)
	The reciprocal square root function. More...
float2	rsqrt (float2 x)
	The reciprocal square root function. More...
half	rsqrt (half x)
	The reciprocal square root function. More...
half2	rsqrt (half2 x)
	The reciprocal square root function. More...
half4	rsqrt (half4 x)
	The reciprocal square root function. More...
float	tan (float x)
	The trigonometric tangent function. More...
float2	tan (float2 x)
	The trigonometric tangent function. More...
half	tan (half x)
	The trigonometric tangent function. More...
half2	tan (half2 x)
	The trigonometric tangent function. More...
half4	tan (half4 x)
	The trigonometric tangent function. More...
float2	tanh (float2 x)
	The hyperbolic tangent function. More...
half	tanh (half x)
	The hyperbolic tangent function. More...
half4	tanh (half4 x)
	The hyperbolic tangent function. More...
float	tgamma (float x)
	The gamma function. More...
float2	tgamma (float2 x)
	The gamma function. More...
half	tgamma (half x)
	The gamma function. More...
half2	tgamma (half2 x)
	The gamma function. More...
half4	tgamma (half4 x)
	The gamma function. More...
float	trunc (float x)
	Rounds input towards zero to the nearest integral value that is not larger in magnitude than x. More...
float2	trunc (float2 x)
	Rounds input towards zero to the nearest integral value that is not larger in magnitude than x. More...
half	trunc (half x)
	Rounds input towards zero to the nearest integral value that is not larger in magnitude than x. More...
half2	trunc (half2 x)
	Rounds input towards zero to the nearest integral value that is not larger in magnitude than x. More...
half4	trunc (half4 x)
	Rounds input towards zero to the nearest integral value that is not larger in magnitude than x. More...
float	sigmoid (float x)
	The sigmoid function, ie 1/(1 + exp(- x )). More...
float2	sigmoid (float2 x)
	The sigmoid function, ie 1/(1 + exp(- x )). More...
half	sigmoid (half x)
	The sigmoid function, ie 1/(1 + exp(- x )). More...
half2	sigmoid (half2 x)
	The sigmoid function, ie 1/(1 + exp(- x )). More...
half4	sigmoid (half4 x)
	The sigmoid function, ie 1/(1 + exp(- x )). More...
float	atan2 (float x, float y)
	The arctangent of ( y )/( x ), in radians. More...
float2	atan2 (float2 x, float2 y)
	The arctangent of ( y )/( x ), in radians. More...
half	atan2 (half x, half y)
	The arctangent of ( y )/( x ), in radians. More...
half2	atan2 (half2 x, half2 y)
	The arctangent of ( y )/( x ), in radians. More...
half4	atan2 (half4 x, half4 y)
	The arctangent of ( y )/( x ), in radians. More...
float	copysign (float x, float y)
	Composes a value of magnitude x with the sign of y. More...
float2	copysign (float2 x, float2 y)
	Composes a value of magnitude x with the sign of y. More...
half	copysign (half x, half y)
	Composes a value of magnitude x with the sign of y. More...
half2	copysign (half2 x, half2 y)
	Composes a value of magnitude x with the sign of y. More...
half4	copysign (half4 x, half4 y)
	Composes a value of magnitude x with the sign of y. More...
float	fdim (float x, float y)
	Calculates the absolute difference between the two inputs. More...
float2	fdim (float2 x, float2 y)
	Calculates the absolute difference between the two inputs. More...
half	fdim (half x, half y)
	Calculates the absolute difference between the two inputs. More...
half2	fdim (half2 x, half2 y)
	Calculates the absolute difference between the two inputs. More...
half4	fdim (half4 x, half4 y)
	Calculates the absolute difference between the two inputs. More...
float	fmax (float x, float y)
	Calculates the maximum of the two inputs. More...
float2	fmax (float2 x, float2 y)
	Calculates the maximum of the two inputs. More...
half	fmax (half x, half y)
	Calculates the maximum of the two inputs. More...
half2	fmax (half2 x, half2 y)
	Calculates the maximum of the two inputs. More...
half4	fmax (half4 x, half4 y)
	Calculates the maximum of the two inputs. More...
float	fmin (float x, float y)
	Calculates the minimum of the two inputs. More...
float2	fmin (float2 x, float2 y)
	Calculates the minimum of the two inputs. More...
half	fmin (half x, half y)
	Calculates the minimum of the two inputs. More...
half2	fmin (half2 x, half2 y)
	Calculates the minimum of the two inputs. More...
half4	fmin (half4 x, half4 y)
	Calculates the minimum of the two inputs. More...
float	fmod (float x, float y)
	Calculates the remainder of the division x / y rounded towards zero. More...
float2	fmod (float2 x, float2 y)
	Calculates the remainder of the division x / y rounded towards zero. More...
half	fmod (half x, half y)
	Calculates the remainder of the division x / y rounded towards zero. More...
half2	fmod (half2 x, half2 y)
	Calculates the remainder of the division x / y rounded towards zero. More...
half4	fmod (half4 x, half4 y)
	Calculates the remainder of the division x / y rounded towards zero. More...
float	hypot (float x, float y)
	Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs. More...
float2	hypot (float2 x, float2 y)
	Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs. More...
half	hypot (half x, half y)
	Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs. More...
half2	hypot (half2 x, half2 y)
	Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs. More...
half4	hypot (half4 x, half4 y)
	Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs. More...
float	pow (float x, float y)
	Calculates x to the power of y. More...
float2	pow (float2 x, float2 y)
	Calculates x to the power of y. More...
half	pow (half x, half y)
	Calculates x to the power of y. More...
half2	pow (half2 x, half2 y)
	Calculates x to the power of y. More...
half4	pow (half4 x, half4 y)
	Calculates x to the power of y. More...
float	remainder (float x, float y)
	Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number. More...
float2	remainder (float2 x, float2 y)
	Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number. More...
half	remainder (half x, half y)
	Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number. More...
half2	remainder (half2 x, half2 y)
	Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number. More...
half4	remainder (half4 x, half4 y)
	Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number. More...
float	fma (float x, float y, float z)
	Computes ( x * y ) + z. More...
float2	fma (float2 x, float2 y, float2 z)
	Computes ( x * y ) + z. More...
half	fma (half x, half y, half z)
	Computes ( x * y ) + z. More...
half2	fma (half2 x, half2 y, half2 z)
	Computes ( x * y ) + z. More...
half4	fma (half4 x, half4 y, half4 z)
	Computes ( x * y ) + z. More...
long long	llrint (float x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
longlong2	llrint (float2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long long	llrint (half x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
longlong2	llrint (half2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
longlong4	llrint (half4 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long long	llround (float x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
longlong2	llround (float2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long long	llround (half x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
longlong2	llround (half2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
longlong4	llround (half4 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long	lrint (float x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long2	lrint (float2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long	lrint (half x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long2	lrint (half2 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long4	lrint (half4 x)
	Rounds input to a nearby integral value, using the current rounding mode. More...
long	lround (float x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long2	lround (float2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long	lround (half x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long2	lround (half2 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...
long4	lround (half4 x)
	Rounds input to nearest integral value, with halfway cases rounded away from zero. More...

Detailed Description

IPU intrinsic functions.

Function Documentation

absadd() [1/4]

float ipu::absadd ( float  src0, float  src1  )

inline

Targets the f32absadd instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

The result of a scalar addition of absolute values src0 and src1.

absadd() [2/4]

float2 ipu::absadd ( float2  src0, float2  src1  )

inline

Targets the f32v2absadd instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The result of an element-wise addition of absolute values in src0 and src1.

absadd() [3/4]

half2 ipu::absadd ( half2  src0, half2  src1  )

inline

Targets the f16v2absadd instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The result of an element-wise addition of absolute values in src0 and src1.

absadd() [4/4]

half4 ipu::absadd ( half4  src0, half4  src1  )

inline

Targets the f16v4absadd instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

The result of an element-wise addition of absolute values in src0 and src1.

absmax() [1/4]

float ipu::absmax ( float  src0, float  src1  )

inline

Targets the f32absmax instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

The maximum of absolute values src0 and src1.

absmax() [2/4]

float2 ipu::absmax ( float2  src0, float2  src1  )

inline

Targets the f32v2absmax instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The element-wise maximum of absolute values in src0 and src1.

absmax() [3/4]

half2 ipu::absmax ( half2  src0, half2  src1  )

inline

Targets the f16v2absmax instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The element-wise maximum of absolute values in src0 and src1.

absmax() [4/4]

half4 ipu::absmax ( half4  src0, half4  src1  )

inline

Targets the f16v4absmax instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

The element-wise maximum of absolute values in src0 and src1.

acos() [1/5]

float ipu::acos ( float  x )

inline

The arccos function, the inverse of cosine.

Parameters

x	A value of type float.

Returns

The result of acos of x.

acos() [2/5]

float2 ipu::acos ( float2  x )

inline

The arccos function, the inverse of cosine.

Parameters

x	A value of type float2.

Returns

The element-wise result of acos of x.

acos() [3/5]

half ipu::acos ( half  x )

inline

The arccos function, the inverse of cosine.

Parameters

x	A value of type half.

Returns

The result of acos of x.

acos() [4/5]

half2 ipu::acos ( half2  x )

inline

The arccos function, the inverse of cosine.

Parameters

x	A value of type half2.

Returns

The element-wise result of acos of x.

acos() [5/5]

half4 ipu::acos ( half4  x )

inline

The arccos function, the inverse of cosine.

Parameters

x	A value of type half4.

Returns

The element-wise result of acos of x.

acosh() [1/5]

float ipu::acosh ( float  x )

inline

The arccosh function, the inverse of the hyperbolic cosine.

Parameters

x	A value of type float.

Returns

The result of acosh of x.

acosh() [2/5]

float2 ipu::acosh ( float2  x )

inline

The arccosh function, the inverse of the hyperbolic cosine.

Parameters

x	A value of type float2.

Returns

The element-wise result of acosh of x.

acosh() [3/5]

half ipu::acosh ( half  x )

inline

The arccosh function, the inverse of the hyperbolic cosine.

Parameters

x	A value of type half.

Returns

The result of acosh of x.

acosh() [4/5]

half2 ipu::acosh ( half2  x )

inline

The arccosh function, the inverse of the hyperbolic cosine.

Parameters

x	A value of type half2.

Returns

The element-wise result of acosh of x.

acosh() [5/5]

half4 ipu::acosh ( half4  x )

inline

The arccosh function, the inverse of the hyperbolic cosine.

Parameters

x	A value of type half4.

Returns

The element-wise result of acosh of x.

andc() [1/4]

float ipu::andc ( float  src0, float  src1  )

inline

Targets the andc instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

The bitwise logical and of src0 and the negated value of src1 of type float.

andc() [2/4]

float2 ipu::andc ( float2  src0, float2  src1  )

inline

Targets the andc64 instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The bitwise logical and of src0 and the negated value of src1 of type float2.

andc() [3/4]

int ipu::andc ( int  src0, int  src1  )

inline

Targets the andc instruction.

Parameters

src0	A value of type int.
src1	A value of type int, can be a 12-bit constant.

Returns

The bitwise logical and of src0 and the negated value of src1 of type int.

andc() [4/4]

unsigned ipu::andc ( unsigned  src0, unsigned  src1  )

inline

Targets the andc instruction.

Parameters

src0	A value of type unsigned.
src1	A value of type unsigned, can be a 12-bit constant.

Returns

The bitwise logical and of src0 and the negated value of src1 of type unsigned.

asin() [1/5]

float ipu::asin ( float  x )

inline

The arcsin function, the inverse of sine.

Parameters

x	A value of type float.

Returns

The result of asin of x.

asin() [2/5]

float2 ipu::asin ( float2  x )

inline

The arcsin function, the inverse of sine.

Parameters

x	A value of type float2.

Returns

The element-wise result of asin of x.

asin() [3/5]

half ipu::asin ( half  x )

inline

The arcsin function, the inverse of sine.

Parameters

x	A value of type half.

Returns

The result of asin of x.

asin() [4/5]

half2 ipu::asin ( half2  x )

inline

The arcsin function, the inverse of sine.

Parameters

x	A value of type half2.

Returns

The element-wise result of asin of x.

asin() [5/5]

half4 ipu::asin ( half4  x )

inline

The arcsin function, the inverse of sine.

Parameters

x	A value of type half4.

Returns

The element-wise result of asin of x.

asinh() [1/5]

float ipu::asinh ( float  x )

inline

The arcsinh function, the inverse of the hyperbolic sine.

Parameters

x	A value of type float.

Returns

The result of asinh of x.

asinh() [2/5]

float2 ipu::asinh ( float2  x )

inline

The arcsinh function, the inverse of the hyperbolic sine.

Parameters

x	A value of type float2.

Returns

The element-wise result of asinh of x.

asinh() [3/5]

half ipu::asinh ( half  x )

inline

The arcsinh function, the inverse of the hyperbolic sine.

Parameters

x	A value of type half.

Returns

The result of asinh of x.

asinh() [4/5]

half2 ipu::asinh ( half2  x )

inline

The arcsinh function, the inverse of the hyperbolic sine.

Parameters

x	A value of type half2.

Returns

The element-wise result of asinh of x.

asinh() [5/5]

half4 ipu::asinh ( half4  x )

inline

The arcsinh function, the inverse of the hyperbolic sine.

Parameters

x	A value of type half4.

Returns

The element-wise result of asinh of x.

atan() [1/5]

float ipu::atan ( float  x )

inline

The arctan function, the inverse of tangent.

Parameters

x	A value of type float.

Returns

The result of atan of x.

atan() [2/5]

float2 ipu::atan ( float2  x )

inline

The arctan function, the inverse of tangent.

Parameters

x	A value of type float2.

Returns

The element-wise result of atan of x.

atan() [3/5]

half ipu::atan ( half  x )

inline

The arctan function, the inverse of tangent.

Parameters

x	A value of type half.

Returns

The result of atan of x.

atan() [4/5]

half2 ipu::atan ( half2  x )

inline

The arctan function, the inverse of tangent.

Parameters

x	A value of type half2.

Returns

The element-wise result of atan of x.

atan() [5/5]

half4 ipu::atan ( half4  x )

inline

The arctan function, the inverse of tangent.

Parameters

x	A value of type half4.

Returns

The element-wise result of atan of x.

atan2() [1/5]

float ipu::atan2 ( float  x, float  y  )

inline

The arctangent of ( y )/( x ), in radians.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of atan2 of x and y.

atan2() [2/5]

float2 ipu::atan2 ( float2  x, float2  y  )

inline

The arctangent of ( y )/( x ), in radians.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of atan2 of x and y.

atan2() [3/5]

half ipu::atan2 ( half  x, half  y  )

inline

The arctangent of ( y )/( x ), in radians.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of atan2 of x and y.

atan2() [4/5]

half2 ipu::atan2 ( half2  x, half2  y  )

inline

The arctangent of ( y )/( x ), in radians.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of atan2 of x and y.

atan2() [5/5]

half4 ipu::atan2 ( half4  x, half4  y  )

inline

The arctangent of ( y )/( x ), in radians.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of atan2 of x and y.

atanh() [1/5]

float ipu::atanh ( float  x )

inline

The arctanh function, the inverse of the hyperbolic tangent.

Parameters

x	A value of type float.

Returns

The result of atanh of x.

atanh() [2/5]

float2 ipu::atanh ( float2  x )

inline

The arctanh function, the inverse of the hyperbolic tangent.

Parameters

x	A value of type float2.

Returns

The element-wise result of atanh of x.

atanh() [3/5]

half ipu::atanh ( half  x )

inline

The arctanh function, the inverse of the hyperbolic tangent.

Parameters

x	A value of type half.

Returns

The result of atanh of x.

atanh() [4/5]

half2 ipu::atanh ( half2  x )

inline

The arctanh function, the inverse of the hyperbolic tangent.

Parameters

x	A value of type half2.

Returns

The element-wise result of atanh of x.

atanh() [5/5]

half4 ipu::atanh ( half4  x )

inline

The arctanh function, the inverse of the hyperbolic tangent.

Parameters

x	A value of type half4.

Returns

The element-wise result of atanh of x.

axpy()

float2 ipu::axpy ( float2  src0, float2  src1  )

inline

Targets the f32v2axpy instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The single precision two-element vector res = a*src0 + src1. The scalar multiplicand a is provided by the internal state element $TAS.

bitrev8()

int ipu::bitrev8 ( int  src )

inline

Targets the bitrev8 instruction.

Parameters

src	A value of type int.

Returns

A result of type int that is equivalent to the value of src with the bit order of each byte reversed.

cbrt() [1/5]

float ipu::cbrt ( float  x )

inline

The cubic root function.

Parameters

x	A value of type float.

Returns

The result of cbrt of x.

cbrt() [2/5]

float2 ipu::cbrt ( float2  x )

inline

The cubic root function.

Parameters

x	A value of type float2.

Returns

The element-wise result of cbrt of x.

cbrt() [3/5]

half ipu::cbrt ( half  x )

inline

The cubic root function.

Parameters

x	A value of type half.

Returns

The result of cbrt of x.

cbrt() [4/5]

half2 ipu::cbrt ( half2  x )

inline

The cubic root function.

Parameters

x	A value of type half2.

Returns

The element-wise result of cbrt of x.

cbrt() [5/5]

half4 ipu::cbrt ( half4  x )

inline

The cubic root function.

Parameters

x	A value of type half4.

Returns

The element-wise result of cbrt of x.

ceil() [1/5]

float ipu::ceil ( float  x )

inline

Rounds up input to the closest integral value.

Parameters

x	A value of type float.

Returns

The smallest integral value that is not less than x.

ceil() [2/5]

float2 ipu::ceil ( float2  x )

inline

Rounds up input to the closest integral value.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the smallest integral value that is not less than x[i].

ceil() [3/5]

half ipu::ceil ( half  x )

inline

Rounds up input to the closest integral value.

Parameters

x	A value of type half.

Returns

The smallest integral value that is not less than x.

ceil() [4/5]

half2 ipu::ceil ( half2  x )

inline

Rounds up input to the closest integral value.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the smallest integral value that is not less than x[i].

ceil() [5/5]

half4 ipu::ceil ( half4  x )

inline

Rounds up input to the closest integral value.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the smallest integral value that is not less than x[i].

clamp() [1/4]

float ipu::clamp ( float  src0, float2  src1  )

inline

Targets the f32clamp instruction.

Parameters

src0	A value of type float.
src1	A value of type float2.

Returns

The median of src0 and the two elements in src1.

clamp() [2/4]

float2 ipu::clamp ( float2  src0, float2  src1  )

inline

Targets the f32v2clamp instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The min-of-maximum result of src0 and src1, of type float2. The first element is the median of the first element of src0 and the two elements in src1. The second element is the median of the second element of src0 and the two elements in src1.

clamp() [3/4]

half2 ipu::clamp ( half2  src0, half2  src1  )

inline

Targets the f16v2clamp instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The min-of-maximum result of src0 and src1, of type half2. The first element is the median value of the first element of src0 and the two elements in src1. The second element is the median of the second element of src0 and the two elements in src1.

clamp() [4/4]

half4 ipu::clamp ( half4  src0, half2  src1  )

inline

Targets the f16v4clamp instruction.

Parameters

src0	A value of type half4.
src1	A value of type half2.

Returns

The min-of-maximum result of src0 and src1, of type half4. Each element is the median of the element in src0 at the same index, and the two values in src1.

clz()

unsigned ipu::clz ( int  src )

inline

Targets the clz instruction.

Parameters

src	A value of type int.

Returns

The number of higher bits in src that are zero.

cmac() [1/2]

void ipu::cmac ( half2  src0, half2  src1  )

inline

Targets the f16v2cmac instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

cmac() [2/2]

void ipu::cmac ( half4  src0, half4  src1  )

inline

Targets the f16v4cmac instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

cmpeq() [1/4]

float ipu::cmpeq ( float  src0, float  src1  )

inline

Targets the f32cmpeq instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Equality test of src0 and src1. If src0 == src1 the result will be 0xffff, and 0x0000 otherwise.

cmpeq() [2/4]

float2 ipu::cmpeq ( float2  src0, float2  src1  )

inline

Targets the f32v2cmpeq instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise equality test of src0 and src1. If src0[i] == src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpeq() [3/4]

half2 ipu::cmpeq ( half2  src0, half2  src1  )

inline

Targets the f16v2cmpeq instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise equality test of src0 and src1. If src0[i] == src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpeq() [4/4]

half4 ipu::cmpeq ( half4  src0, half4  src1  )

inline

Targets the f16v4cmpeq instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise equality test of src0 and src1. If src0[i] == src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpge() [1/4]

float ipu::cmpge ( float  src0, float  src1  )

inline

Targets the f32cmpge instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Greater-than-or-equal-to test of src0 and src1. If src0 >= src1 the result will be 0xffff, and 0x0000 otherwise.

cmpge() [2/4]

float2 ipu::cmpge ( float2  src0, float2  src1  )

inline

Targets the f32v2cmpge instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise greater-than-or-equal-to test of src0 and src1. If src0[i] >= src1[i] the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpge() [3/4]

half2 ipu::cmpge ( half2  src0, half2  src1  )

inline

Targets the f16v2cmpge instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise greater-than-or-equal-to test of src0 and src1. If src0[i] >= src1[i] the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpge() [4/4]

half4 ipu::cmpge ( half4  src0, half4  src1  )

inline

Targets the f16v4cmpge instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise greater-than-or-equal-to test of src0 and src1. If src0[i] >= src1[i] the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpgt() [1/4]

float ipu::cmpgt ( float  src0, float  src1  )

inline

Targets the f32cmpgt instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Greater-than test of src0 and src1. If src0 > src1 the result will be 0xffff, and 0x0000 otherwise.

cmpgt() [2/4]

float2 ipu::cmpgt ( float2  src0, float2  src1  )

inline

Targets the f32v2cmpgt instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise greater-than test of src0 and src1. If src0 > src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpgt() [3/4]

half2 ipu::cmpgt ( half2  src0, half2  src1  )

inline

Targets the f16v2cmpgt instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise greater-than test of src0 and src1. If src0 > src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpgt() [4/4]

half4 ipu::cmpgt ( half4  src0, half4  src1  )

inline

Targets the f16v4cmpgt instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise greater-than test of src0 and src1. If src0 > src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmple() [1/4]

float ipu::cmple ( float  src0, float  src1  )

inline

Targets the f32cmple instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Less-than-or-equal-to test of src0 and src1. If src0 <= src1 the result will be 0xffff, and 0x0000 otherwise.

cmple() [2/4]

float2 ipu::cmple ( float2  src0, float2  src1  )

inline

Targets the f32v2cmple instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise less-than-or-equal-to test of src0 and src1. If src0 <= src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmple() [3/4]

half2 ipu::cmple ( half2  src0, half2  src1  )

inline

Targets the f16v2cmple instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise less-than-or-equal-to test of src0 and src1. If src0 <= src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmple() [4/4]

half4 ipu::cmple ( half4  src0, half4  src1  )

inline

Targets the f16v4cmple instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise less-than-or-equal-to test of src0 and src1. If src0 <= src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmplt() [1/4]

float ipu::cmplt ( float  src0, float  src1  )

inline

Targets the f32cmplt instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Less-than test of src0 and src1. If src0 < src1 the result will be 0xffff, and 0x0000 otherwise.

cmplt() [2/4]

float2 ipu::cmplt ( float2  src0, float2  src1  )

inline

Targets the f32v2cmplt instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise less-than test of src0 and src1. If src0 < src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmplt() [3/4]

half2 ipu::cmplt ( half2  src0, half2  src1  )

inline

Targets the f16v2cmplt instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise less-than test of src0 and src1. If src0 < src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmplt() [4/4]

half4 ipu::cmplt ( half4  src0, half4  src1  )

inline

Targets the f16v4cmplt instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise less-than test of src0 and src1. If src0 < src1 the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpne() [1/4]

float ipu::cmpne ( float  src0, float  src1  )

inline

Targets the f32cmpne instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

Inequality test of src0 and src1. If src0 != src1 the result will be 0xffff, and 0x0000 otherwise.

cmpne() [2/4]

float2 ipu::cmpne ( float2  src0, float2  src1  )

inline

Targets the f32v2cmpne instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

Element-wise inequality test of src0 and src1. If src0[i] != src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpne() [3/4]

half2 ipu::cmpne ( half2  src0, half2  src1  )

inline

Targets the f16v2cmpne instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

Element-wise inequality test of src0 and src1. If src0[i] != src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cmpne() [4/4]

half4 ipu::cmpne ( half4  src0, half4  src1  )

inline

Targets the f16v4cmpne instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

Element-wise inequality test of src0 and src1. If src0[i] != src1[i], the result vector element at index i will be 0xffff, and 0x0000 otherwise.

cms()

unsigned ipu::cms ( int  src )

inline

Targets the cms instruction.

Parameters

src	A value of type int.

Returns

The number of higher order bits in src that match the sign bit (bit 31), as an unsigned.

copysign() [1/5]

float ipu::copysign ( float  x, float  y  )

inline

Composes a value of magnitude x with the sign of y.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of copysign of x and y.

copysign() [2/5]

float2 ipu::copysign ( float2  x, float2  y  )

inline

Composes a value of magnitude x with the sign of y.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of copysign of x and y.

copysign() [3/5]

half ipu::copysign ( half  x, half  y  )

inline

Composes a value of magnitude x with the sign of y.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of copysign of x and y.

copysign() [4/5]

half2 ipu::copysign ( half2  x, half2  y  )

inline

Composes a value of magnitude x with the sign of y.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of copysign of x and y.

copysign() [5/5]

half4 ipu::copysign ( half4  x, half4  y  )

inline

Composes a value of magnitude x with the sign of y.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of copysign of x and y.

cos() [1/5]

float ipu::cos ( float  x )

inline

The trigonometric cosine function.

Parameters

x	A value of type float.

Returns

The result of cos of x.

cos() [2/5]

float2 ipu::cos ( float2  x )

inline

The trigonometric cosine function.

Parameters

x	A value of type float2.

Returns

The element-wise result of cos of x.

cos() [3/5]

half ipu::cos ( half  x )

inline

The trigonometric cosine function.

Parameters

x	A value of type half.

Returns

The result of cos of x.

cos() [4/5]

half2 ipu::cos ( half2  x )

inline

The trigonometric cosine function.

Parameters

x	A value of type half2.

Returns

The element-wise result of cos of x.

cos() [5/5]

half4 ipu::cos ( half4  x )

inline

The trigonometric cosine function.

Parameters

x	A value of type half4.

Returns

The element-wise result of cos of x.

cosh() [1/5]

float ipu::cosh ( float  x )

inline

The hyperbolic cosine function.

Parameters

x	A value of type float.

Returns

The result of cosh of x.

cosh() [2/5]

float2 ipu::cosh ( float2  x )

inline

The hyperbolic cosine function.

Parameters

x	A value of type float2.

Returns

The element-wise result of cosh of x.

cosh() [3/5]

half ipu::cosh ( half  x )

inline

The hyperbolic cosine function.

Parameters

x	A value of type half.

Returns

The result of cosh of x.

cosh() [4/5]

half2 ipu::cosh ( half2  x )

inline

The hyperbolic cosine function.

Parameters

x	A value of type half2.

Returns

The element-wise result of cosh of x.

cosh() [5/5]

half4 ipu::cosh ( half4  x )

inline

The hyperbolic cosine function.

Parameters

x	A value of type half4.

Returns

The element-wise result of cosh of x.

erf() [1/5]

float ipu::erf ( float  x )

inline

The error function.

Parameters

x	A value of type float.

Returns

The error function value for x.

erf() [2/5]

float2 ipu::erf ( float2  x )

inline

The error function.

Parameters

x	A value of type float2.

Returns

The error function value for x.

erf() [3/5]

half ipu::erf ( half  x )

inline

The error function.

Parameters

x	A value of type half.

Returns

The error function value for x.

erf() [4/5]

half2 ipu::erf ( half2  x )

inline

The error function.

Parameters

x	A value of type half2.

Returns

The error function value for x.

erf() [5/5]

half4 ipu::erf ( half4  x )

inline

The error function.

Parameters

x	A value of type half4.

Returns

The error function value for x.

erfc() [1/5]

float ipu::erfc ( float  x )

inline

The complementary error function.

Parameters

x	A value of type float.

Returns

The complementary error function value for x.

erfc() [2/5]

float2 ipu::erfc ( float2  x )

inline

The complementary error function.

Parameters

x	A value of type float2.

Returns

The complementary error function value for x.

erfc() [3/5]

half ipu::erfc ( half  x )

inline

The complementary error function.

Parameters

x	A value of type half.

Returns

The complementary error function value for x.

erfc() [4/5]

half2 ipu::erfc ( half2  x )

inline

The complementary error function.

Parameters

x	A value of type half2.

Returns

The complementary error function value for x.

erfc() [5/5]

half4 ipu::erfc ( half4  x )

inline

The complementary error function.

Parameters

x	A value of type half4.

Returns

The complementary error function value for x.

exp() [1/5]

float ipu::exp ( float  x )

inline

Targets the f32exp instruction.

The base-e exponential function.

Parameters

src	A value of type float.

Returns

The result of e^{src}.

Parameters

x	A value of type float.

Returns

The result of exp of x.

exp() [2/5]

float2 ipu::exp ( float2  x )

inline

The base-e exponential function.

Parameters

x	A value of type float2.

Returns

The element-wise result of exp of x.

exp() [3/5]

half ipu::exp ( half  x )

inline

The base-e exponential function.

Parameters

x	A value of type half.

Returns

The result of exp of x.

exp() [4/5]

half2 ipu::exp ( half2  x )

inline

Targets the f16v2exp instruction.

The base-e exponential function.

Parameters

src	A value of type half2.

Returns

A vector of the results of e^X of the two elements in src.

Parameters

x	A value of type half2.

Returns

The element-wise result of exp of x.

exp() [5/5]

half4 ipu::exp ( half4  x )

inline

The base-e exponential function.

Parameters

x	A value of type half4.

Returns

The element-wise result of exp of x.

exp2() [1/5]

float ipu::exp2 ( float  x )

inline

Targets the f32exp instruction.

The base-2 exponential function.

Parameters

src	A value of type float.

Returns

The result of 2^{src}.

Parameters

x	A value of type float.

Returns

The result of exp2 of x.

exp2() [2/5]

float2 ipu::exp2 ( float2  x )

inline

The base-2 exponential function.

Parameters

x	A value of type float2.

Returns

The element-wise result of exp2 of x.

exp2() [3/5]

half ipu::exp2 ( half  x )

inline

The base-2 exponential function.

Parameters

x	A value of type half.

Returns

The result of exp2 of x.

exp2() [4/5]

half2 ipu::exp2 ( half2  x )

inline

Targets the f16v2exp instruction.

The base-2 exponential function.

Parameters

src	A value of type half2.

Returns

A vector of the results of 2^X of the two elements in src.

Parameters

x	A value of type half2.

Returns

The element-wise result of exp2 of x.

exp2() [5/5]

half4 ipu::exp2 ( half4  x )

inline

The base-2 exponential function.

Parameters

x	A value of type half4.

Returns

The element-wise result of exp2 of x.

expm1() [1/5]

float ipu::expm1 ( float  x )

inline

The base-e exponential function, minus one: exp(x) - 1.

Parameters

x	A value of type float.

Returns

The result of expm1 of x.

expm1() [2/5]

float2 ipu::expm1 ( float2  x )

inline

The base-e exponential function, minus one: exp(x) - 1.

Parameters

x	A value of type float2.

Returns

The element-wise result of expm1 of x.

expm1() [3/5]

half ipu::expm1 ( half  x )

inline

The base-e exponential function, minus one: exp(x) - 1.

Parameters

x	A value of type half.

Returns

The result of expm1 of x.

expm1() [4/5]

half2 ipu::expm1 ( half2  x )

inline

The base-e exponential function, minus one: exp(x) - 1.

Parameters

x	A value of type half2.

Returns

The element-wise result of expm1 of x.

expm1() [5/5]

half4 ipu::expm1 ( half4  x )

inline

The base-e exponential function, minus one: exp(x) - 1.

Parameters

x	A value of type half4.

Returns

The element-wise result of expm1 of x.

f16v2grand()

half2 ipu::f16v2grand ( )

inline

Targets the f16v2grand instruction.

Returns

Gaussian distribution, two-element half-precision random vector.

f32v2grand()

float2 ipu::f32v2grand ( )

inline

Targets the f32v2grand instruction.

Returns

Gaussian distribution, two-element single-precision random vector.

fabs() [1/5]

float ipu::fabs ( float  x )

inline

Computes the absolute value of the input.

Parameters

x	A value of type float.

Returns

The absolute value of x.

fabs() [2/5]

float2 ipu::fabs ( float2  x )

inline

Computes the absolute value of the input.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the absolute value of x[i].

fabs() [3/5]

half ipu::fabs ( half  x )

inline

Computes the absolute value of the input.

Parameters

x	A value of type half.

Returns

The absolute value of x.

fabs() [4/5]

half2 ipu::fabs ( half2  x )

inline

Computes the absolute value of the input.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the absolute value of x[i].

fabs() [5/5]

half4 ipu::fabs ( half4  x )

inline

Computes the absolute value of the input.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the absolute value of x[i].

fdim() [1/5]

float ipu::fdim ( float  x, float  y  )

inline

Calculates the absolute difference between the two inputs.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of fdim of x and y.

fdim() [2/5]

float2 ipu::fdim ( float2  x, float2  y  )

inline

Calculates the absolute difference between the two inputs.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of fdim of x and y.

fdim() [3/5]

half ipu::fdim ( half  x, half  y  )

inline

Calculates the absolute difference between the two inputs.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of fdim of x and y.

fdim() [4/5]

half2 ipu::fdim ( half2  x, half2  y  )

inline

Calculates the absolute difference between the two inputs.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of fdim of x and y.

fdim() [5/5]

half4 ipu::fdim ( half4  x, half4  y  )

inline

Calculates the absolute difference between the two inputs.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of fdim of x and y.

floor() [1/5]

float ipu::floor ( float  x )

inline

Rounds down input to the closest integral value.

Parameters

x	A value of type float.

Returns

The largest integral value that is not greater than x.

floor() [2/5]

float2 ipu::floor ( float2  x )

inline

Rounds down input to the closest integral value.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the largest integral value that is not greater than x[i].

floor() [3/5]

half ipu::floor ( half  x )

inline

Rounds down input to the closest integral value.

Parameters

x	A value of type half.

Returns

The largest integral value that is not greater than x.

floor() [4/5]

half2 ipu::floor ( half2  x )

inline

Rounds down input to the closest integral value.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the largest integral value that is not greater than x[i].

floor() [5/5]

half4 ipu::floor ( half4  x )

inline

Rounds down input to the closest integral value.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the largest integral value that is not greater than x[i].

fma() [1/5]

float ipu::fma ( float  x, float  y, float  z  )

inline

Computes ( x * y ) + z.

Parameters

x	A value of type float.
y	A value of type float.
z	A value of type float.

Returns

The result of ( x * y ) + z.

fma() [2/5]

float2 ipu::fma ( float2  x, float2  y, float2  z  )

inline

Computes ( x * y ) + z.

Parameters

x	A value of type float2.
y	A value of type float2.
z	A value of type float2.

Returns

A vector where every ith element is the result of ( x[i] * y[i] ) + z[i].

fma() [3/5]

half ipu::fma ( half  x, half  y, half  z  )

inline

Computes ( x * y ) + z.

Parameters

x	A value of type half.
y	A value of type half.
z	A value of type half.

Returns

The result of ( x * y ) + z.

fma() [4/5]

half2 ipu::fma ( half2  x, half2  y, half2  z  )

inline

Computes ( x * y ) + z.

Parameters

x	A value of type half2.
y	A value of type half2.
z	A value of type half2.

Returns

A vector where every ith element is the result of ( x[i] * y[i] ) + z[i].

fma() [5/5]

half4 ipu::fma ( half4  x, half4  y, half4  z  )

inline

Computes ( x * y ) + z.

Parameters

x	A value of type half4.
y	A value of type half4.
z	A value of type half4.

Returns

A vector where every ith element is the result of ( x[i] * y[i] ) + z[i].

fmax() [1/5]

float ipu::fmax ( float  x, float  y  )

inline

Calculates the maximum of the two inputs.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The larger value of x and y. If one of them is a NaN, returns the other.

fmax() [2/5]

float2 ipu::fmax ( float2  x, float2  y  )

inline

Calculates the maximum of the two inputs.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

A vector where every ith element is the larger of x[i] and y[i]. If either of them is a NaN, the other is set as the maximum.

fmax() [3/5]

half ipu::fmax ( half  x, half  y  )

inline

Calculates the maximum of the two inputs.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The larger value of x and y. If one of them is a NaN, returns the other.

fmax() [4/5]

half2 ipu::fmax ( half2  x, half2  y  )

inline

Calculates the maximum of the two inputs.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

A vector where every ith element is the larger of x[i] and y[i]. If either of them is a NaN, the other is set as the maximum.

fmax() [5/5]

half4 ipu::fmax ( half4  x, half4  y  )

inline

Calculates the maximum of the two inputs.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

A vector where every ith element is the larger of x[i] and y[i]. If either of them is a NaN, the other is set as the maximum.

fmin() [1/5]

float ipu::fmin ( float  x, float  y  )

inline

Calculates the minimum of the two inputs.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The smaller value of x and y. If one of them is a NaN, returns the other.

fmin() [2/5]

float2 ipu::fmin ( float2  x, float2  y  )

inline

Calculates the minimum of the two inputs.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

A vector where every ith element is the smaller of x[i] and y[i]. If either of them is a NaN, the other is set as the minimum.

fmin() [3/5]

half ipu::fmin ( half  x, half  y  )

inline

Calculates the minimum of the two inputs.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The smaller value of x and y. If one of them is a NaN, returns the other.

fmin() [4/5]

half2 ipu::fmin ( half2  x, half2  y  )

inline

Calculates the minimum of the two inputs.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

A vector where every ith element is the smaller of x[i] and y[i]. If either of them is a NaN, the other is set as the minimum.

fmin() [5/5]

half4 ipu::fmin ( half4  x, half4  y  )

inline

Calculates the minimum of the two inputs.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

A vector where every ith element is the smaller of x[i] and y[i]. If either of them is a NaN, the other is set as the minimum.

fmod() [1/5]

float ipu::fmod ( float  x, float  y  )

inline

Calculates the remainder of the division x / y rounded towards zero.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of fmod of x and y.

fmod() [2/5]

float2 ipu::fmod ( float2  x, float2  y  )

inline

Calculates the remainder of the division x / y rounded towards zero.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of fmod of x and y.

fmod() [3/5]

half ipu::fmod ( half  x, half  y  )

inline

Calculates the remainder of the division x / y rounded towards zero.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of fmod of x and y.

fmod() [4/5]

half2 ipu::fmod ( half2  x, half2  y  )

inline

Calculates the remainder of the division x / y rounded towards zero.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of fmod of x and y.

fmod() [5/5]

half4 ipu::fmod ( half4  x, half4  y  )

inline

Calculates the remainder of the division x / y rounded towards zero.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of fmod of x and y.

hypot() [1/5]

float ipu::hypot ( float  x, float  y  )

inline

Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of hypot of x and y.

hypot() [2/5]

float2 ipu::hypot ( float2  x, float2  y  )

inline

Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of hypot of x and y.

hypot() [3/5]

half ipu::hypot ( half  x, half  y  )

inline

Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of hypot of x and y.

hypot() [4/5]

half2 ipu::hypot ( half2  x, half2  y  )

inline

Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of hypot of x and y.

hypot() [5/5]

half4 ipu::hypot ( half4  x, half4  y  )

inline

Calculates the hypotenuse of the right-angled triangle whose two shorter sides are of lengths given by the two inputs.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of hypot of x and y.

llrint() [1/5]

long long ipu::llrint ( float  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float.

Returns

The value of x rounded to a nearby integral of type long long.

llrint() [2/5]

longlong2 ipu::llrint ( float2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long long.

llrint() [3/5]

long long ipu::llrint ( half  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half.

Returns

The value of x rounded to a nearby integral of type long long.

llrint() [4/5]

longlong2 ipu::llrint ( half2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long long.

llrint() [5/5]

longlong4 ipu::llrint ( half4  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long long.

llround() [1/5]

long long ipu::llround ( float  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float.

Returns

The nearest integral value to x as a long long, with halfway cases rounded away from zero.

llround() [2/5]

longlong2 ipu::llround ( float2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long long, with halfway cases rounded away from zero.

llround() [3/5]

long long ipu::llround ( half  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half.

Returns

The nearest integral value to x as a long long, with halfway cases rounded away from zero.

llround() [4/5]

longlong2 ipu::llround ( half2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long long, with halfway cases rounded away from zero.

llround() [5/5]

longlong4 ipu::llround ( half4  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long long, with halfway cases rounded away from zero.

ln() [1/2]

float ipu::ln ( float  src )

inline

Targets the f32ln instruction.

Parameters

src	A value of type half2.

Returns

The result of the natural log of src.

ln() [2/2]

half2 ipu::ln ( half2  src )

inline

Targets the f16v2ln instruction.

Parameters

src	A value of type half2.

Returns

A vector of the results of the natural log of the two elements in src.

load_postinc() [1/17]

char ipu::load_postinc ( const char **  a, int  i  )

inline

Post-incrementing load, targeting the lds8step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a char.

load_postinc() [2/17]

float ipu::load_postinc ( const float **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a float.

load_postinc() [3/17]

float2 ipu::load_postinc ( const float2 **  a, int  i  )

inline

Post-incrementing load, targeting the ld64step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a float2.

load_postinc() [4/17]

half ipu::load_postinc ( const half **  a, int  i  )

inline

Post-incrementing load, targeting the ldb16step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a half.

load_postinc() [5/17]

half2 ipu::load_postinc ( const half2 **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a half2.

load_postinc() [6/17]

half4 ipu::load_postinc ( const half4 **  a, int  i  )

inline

Post-incrementing load, targeting the ld64step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a half4.

load_postinc() [7/17]

int ipu::load_postinc ( const int **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as an int.

load_postinc() [8/17]

int2 ipu::load_postinc ( const int2 **  a, int  i  )

inline

Post-incrementing load.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as an int2.

load_postinc() [9/17]

short ipu::load_postinc ( const short **  a, int  i  )

inline

Post-incrementing load, targeting the lds16step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a short.

load_postinc() [10/17]

short2 ipu::load_postinc ( const short2 **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a short2.

load_postinc() [11/17]

short4 ipu::load_postinc ( const short4 **  a, int  i  )

inline

Post-incrementing load.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a short4.

load_postinc() [12/17]

uint2 ipu::load_postinc ( const uint2 **  a, int  i  )

inline

Post-incrementing load.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a uint2.

load_postinc() [13/17]

unsigned ipu::load_postinc ( const unsigned **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as an unsigned.

load_postinc() [14/17]

unsigned char ipu::load_postinc ( const unsigned char **  a, int  i  )

inline

Post-incrementing load, targeting the ldz8step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as an unsigned char.

load_postinc() [15/17]

unsigned short ipu::load_postinc ( const unsigned short **  a, int  i  )

inline

Post-incrementing load, targeting the ldz16step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as an unsigned short.

load_postinc() [16/17]

ushort2 ipu::load_postinc ( const ushort2 **  a, int  i  )

inline

Post-incrementing load, targeting the ld32step instruction.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a ushort2.

load_postinc() [17/17]

ushort4 ipu::load_postinc ( const ushort4 **  a, int  i  )

inline

Post-incrementing load.

Parameters

a	Address of the variable holding the address to load from. Gets incremented by i after the load.
i	Value by which to increment a after load.

Returns

The value in memory whose address is given by a, as a ushort4.

log() [1/5]

float ipu::log ( float  x )

inline

The natural logarithm.

Parameters

x	A value of type float.

Returns

The result of log of x.

log() [2/5]

float2 ipu::log ( float2  x )

inline

The natural logarithm.

Parameters

x	A value of type float2.

Returns

The element-wise result of log of x.

log() [3/5]

half ipu::log ( half  x )

inline

The natural logarithm.

Parameters

x	A value of type half.

Returns

The result of log of x.

log() [4/5]

half2 ipu::log ( half2  x )

inline

The natural logarithm.

Parameters

x	A value of type half2.

Returns

The element-wise result of log of x.

log() [5/5]

half4 ipu::log ( half4  x )

inline

The natural logarithm.

Parameters

x	A value of type half4.

Returns

The element-wise result of log of x.

log10() [1/5]

float ipu::log10 ( float  x )

inline

The base-10 logarithm.

Parameters

x	A value of type float.

Returns

The result of log10 of x.

log10() [2/5]

float2 ipu::log10 ( float2  x )

inline

The base-10 logarithm.

Parameters

x	A value of type float2.

Returns

The element-wise result of log10 of x.

log10() [3/5]

half ipu::log10 ( half  x )

inline

The base-10 logarithm.

Parameters

x	A value of type half.

Returns

The result of log10 of x.

log10() [4/5]

half2 ipu::log10 ( half2  x )

inline

The base-10 logarithm.

Parameters

x	A value of type half2.

Returns

The element-wise result of log10 of x.

log10() [5/5]

half4 ipu::log10 ( half4  x )

inline

The base-10 logarithm.

Parameters

x	A value of type half4.

Returns

The element-wise result of log10 of x.

log1p() [1/5]

float ipu::log1p ( float  x )

inline

The natural logarithm of 1 + x.

Parameters

x	A value of type float.

Returns

The result of log1p of x.

log1p() [2/5]

float2 ipu::log1p ( float2  x )

inline

The natural logarithm of 1 + x.

Parameters

x	A value of type float2.

Returns

The element-wise result of log1p of x.

log1p() [3/5]

half ipu::log1p ( half  x )

inline

The natural logarithm of 1 + x.

Parameters

x	A value of type half.

Returns

The result of log1p of x.

log1p() [4/5]

half2 ipu::log1p ( half2  x )

inline

The natural logarithm of 1 + x.

Parameters

x	A value of type half2.

Returns

The element-wise result of log1p of x.

log1p() [5/5]

half4 ipu::log1p ( half4  x )

inline

The natural logarithm of 1 + x.

Parameters

x	A value of type half4.

Returns

The element-wise result of log1p of x.

log2() [1/5]

float ipu::log2 ( float  x )

inline

Targets the f32ln instruction.

The base-2 logarithm.

Parameters

src	A value of type half2.

Returns

The result of the log (base 2) of src.

Parameters

x	A value of type float.

Returns

The result of log2 of x.

log2() [2/5]

float2 ipu::log2 ( float2  x )

inline

The base-2 logarithm.

Parameters

x	A value of type float2.

Returns

The element-wise result of log2 of x.

log2() [3/5]

half ipu::log2 ( half  x )

inline

The base-2 logarithm.

Parameters

x	A value of type half.

Returns

The result of log2 of x.

log2() [4/5]

half2 ipu::log2 ( half2  x )

inline

Targets the f16v2log2 instruction.

The base-2 logarithm.

Parameters

src	A value of type half2.

Returns

A vector of the results of the log (base 2) of the two elements in src.

Parameters

x	A value of type half2.

Returns

The element-wise result of log2 of x.

log2() [5/5]

half4 ipu::log2 ( half4  x )

inline

The base-2 logarithm.

Parameters

x	A value of type half4.

Returns

The element-wise result of log2 of x.

lrint() [1/5]

long ipu::lrint ( float  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float.

Returns

The value of x rounded to a nearby integral of type long.

lrint() [2/5]

long2 ipu::lrint ( float2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long.

lrint() [3/5]

long ipu::lrint ( half  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half.

Returns

The value of x rounded to a nearby integral of type long.

lrint() [4/5]

long2 ipu::lrint ( half2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long.

lrint() [5/5]

long4 ipu::lrint ( half4  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is x[i] rounded to a nearby integral, of type long.

lround() [1/5]

long ipu::lround ( float  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float.

Returns

The nearest integral value to x as a long, with halfway cases rounded away from zero.

lround() [2/5]

long2 ipu::lround ( float2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long, with halfway cases rounded away from zero.

lround() [3/5]

long ipu::lround ( half  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half.

Returns

The nearest integral value to x as a long, with halfway cases rounded away from zero.

lround() [4/5]

long2 ipu::lround ( half2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long, with halfway cases rounded away from zero.

lround() [5/5]

long4 ipu::lround ( half4  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the nearest integral value to x[i] as a long, with halfway cases rounded away from zero.

max() [1/4]

float ipu::max ( float  src0, float  src1  )

inline

Targets the f32max instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

The maximum of src0 and src1.

max() [2/4]

float2 ipu::max ( float2  src0, float2  src1  )

inline

Targets the f32v2max instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The element-wise maximum of src0 and src1.

max() [3/4]

half2 ipu::max ( half2  src0, half2  src1  )

inline

Targets the f16v2max instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The element-wise maximum of src0 and src1.

max() [4/4]

half4 ipu::max ( half4  src0, half4  src1  )

inline

Targets the f16v4max instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

The element-wise maximum of src0 and src1.

maxc()

half2 ipu::maxc ( half4  src )

inline

Targets the f16v4maxc instruction.

Parameters

src	A value of type half4.

Returns

The 2x2 lateral maximum of src. The 0th element in the result vector is the maximum of src[0] and src[1], and the 1st element is the maximum of src[2] and src[3].

min() [1/4]

float ipu::min ( float  src0, float  src1  )

inline

Targets the f32min instruction.

Parameters

src0	A value of type float.
src1	A value of type float.

Returns

The minimum of src0 and src1.

min() [2/4]

float2 ipu::min ( float2  src0, float2  src1  )

inline

Targets the f32v2min instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The element-wise minimum of src0 and src1.

min() [3/4]

half2 ipu::min ( half2  src0, half2  src1  )

inline

Targets the f16v2min instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The element-wise minimum of src0 and src1.

min() [4/4]

half4 ipu::min ( half4  src0, half4  src1  )

inline

Targets the f16v4min instruction.

Parameters

src0	A value of type half4.
src1	A value of type half4.

Returns

The element-wise minimum of src0 and src1.

nearbyint() [1/5]

float ipu::nearbyint ( float  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float.

Returns

The value of x rounded to a nearby integral.

nearbyint() [2/5]

float2 ipu::nearbyint ( float2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

nearbyint() [3/5]

half ipu::nearbyint ( half  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half.

Returns

The value of x rounded to a nearby integral.

nearbyint() [4/5]

half2 ipu::nearbyint ( half2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

nearbyint() [5/5]

half4 ipu::nearbyint ( half4  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

popc()

unsigned ipu::popc ( int  src )

inline

Targets the popc instruction.

Parameters

src	A value of type int.

Returns

The number of set bits in src.

pow() [1/5]

float ipu::pow ( float  x, float  y  )

inline

Calculates x to the power of y.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of ( x )^( y )

pow() [2/5]

float2 ipu::pow ( float2  x, float2  y  )

inline

Calculates x to the power of y.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

A vector where every ith element is x[i] raised to the power of y[i].

pow() [3/5]

half ipu::pow ( half  x, half  y  )

inline

Calculates x to the power of y.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of ( x )^( y )

pow() [4/5]

half2 ipu::pow ( half2  x, half2  y  )

inline

Calculates x to the power of y.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

A vector where every ith element is x[i] raised to the power of y[i].

pow() [5/5]

half4 ipu::pow ( half4  x, half4  y  )

inline

Calculates x to the power of y.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

A vector where every ith element is x[i] raised to the power of y[i].

remainder() [1/5]

float ipu::remainder ( float  x, float  y  )

inline

Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number.

Parameters

x	A value of type float.
y	A value of type float.

Returns

The result of remainder of x and y.

remainder() [2/5]

float2 ipu::remainder ( float2  x, float2  y  )

inline

Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number.

Parameters

x	A value of type float2.
y	A value of type float2.

Returns

The element-wise result of remainder of x and y.

remainder() [3/5]

half ipu::remainder ( half  x, half  y  )

inline

Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number.

Parameters

x	A value of type half.
y	A value of type half.

Returns

The result of remainder of x and y.

remainder() [4/5]

half2 ipu::remainder ( half2  x, half2  y  )

inline

Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number.

Parameters

x	A value of type half2.
y	A value of type half2.

Returns

The element-wise result of remainder of x and y.

remainder() [5/5]

half4 ipu::remainder ( half4  x, half4  y  )

inline

Calculates the remainder of the division x / y, rounded to the nearest integral value, with halfway cases rounded to the even number.

Parameters

x	A value of type half4.
y	A value of type half4.

Returns

The element-wise result of remainder of x and y.

rint() [1/5]

float ipu::rint ( float  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float.

Returns

The value of x rounded to a nearby integral.

rint() [2/5]

float2 ipu::rint ( float2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

rint() [3/5]

half ipu::rint ( half  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half.

Returns

The value of x rounded to a nearby integral.

rint() [4/5]

half2 ipu::rint ( half2  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

rint() [5/5]

half4 ipu::rint ( half4  x )

inline

Rounds input to a nearby integral value, using the current rounding mode.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is x[i] rounded to a nearby integral.

rmask() [1/2]

float2 ipu::rmask ( float2  src0, float  src1  )

inline

Targets the f32v2rmask instruction.

Parameters

src0	A value of type float2.
src1	A value of type float.

Returns

The result is a masked version of src0, with each element of the input being individually masked with the probability specified by the bottom 17-bits of src1:

• if src1[16] == 1, no masking is applied;
• if src1[16:0] == 0, the result is a zero vector;
• otherwise each element is individually unmasked with probability src1[15:0] / 65536. PRNG is used by this instruction to generate 2 x 16-bit random values from the discrete uniform distribution.

rmask() [2/2]

half4 ipu::rmask ( half4  src0, float  src1  )

inline

Targets the f16v4rmask instruction.

Parameters

src0	A value of type half4.
src1	A value of type float.

Returns

The result is a masked version of src0, with each element of the input being individually masked with the probability specified by the bottom 17-bits of src1:

• if src1[16] == 1, no masking is applied;
• if src1[16:0] == 0, the result is a zero vector;
• otherwise each element is individually unmasked with probability src1[15:0] / 65536. PRNG is used by this instruction to generate 4 x 16-bit random values from the discrete uniform distribution.

roll16() [1/3]

half2 ipu::roll16 ( half2  src0, half2  src1  )

inline

Targets the roll16 instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The result of a SIMD roll permutation on the 4 16-bit values across src0 and src1, as a half2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 1 |

roll16() [2/3]

short2 ipu::roll16 ( short2  src0, short2  src1  )

inline

Targets the roll16 instruction.

Parameters

src0	A value of type short2.
src1	A value of type short2.

Returns

The result of a SIMD roll permutation on the 4 16-bit values across src0 and src1, as a short2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 1 |

roll16() [3/3]

ushort2 ipu::roll16 ( ushort2  src0, ushort2  src1  )

inline

Targets the roll16 instruction.

Parameters

src0	A value of type ushort2.
src1	A value of type ushort2.

Returns

The result of a SIMD roll permutation on the 4 16-bit values across src0 and src1, as a ushort2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 1 |

roll32()

float2 ipu::roll32 ( float2  src0, float2  src1  )

inline

Targets the roll32 instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The result of a SIMD roll permutation on the 4 32-bit float values across src0 and src1, as a float2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 1 |

roll8l()

int ipu::roll8l ( int  src0, int  src1  )

inline

Targets the roll8l instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The result of a SIMD roll-left permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 6 | 5 | 4 | 3 |

roll8r()

int ipu::roll8r ( int  src0, int  src1  )

inline

Targets the roll8r instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The result of a SIMD roll-right permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 4 | 3 | 2 | 1 |

round() [1/5]

float ipu::round ( float  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float.

Returns

The result of round of x.

round() [2/5]

float2 ipu::round ( float2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type float2.

Returns

The element-wise result of round of x.

round() [3/5]

half ipu::round ( half  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half.

Returns

The result of round of x.

round() [4/5]

half2 ipu::round ( half2  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half2.

Returns

The element-wise result of round of x.

round() [5/5]

half4 ipu::round ( half4  x )

inline

Rounds input to nearest integral value, with halfway cases rounded away from zero.

Parameters

x	A value of type half4.

Returns

The element-wise result of round of x.

rsqrt() [1/5]

float ipu::rsqrt ( float  x )

inline

The reciprocal square root function.

Parameters

x	A value of type float.

Returns

The reciprocal of the square root of x.

rsqrt() [2/5]

float2 ipu::rsqrt ( float2  x )

inline

The reciprocal square root function.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the reciprocal of the square root of x[i].

rsqrt() [3/5]

half ipu::rsqrt ( half  x )

inline

The reciprocal square root function.

Parameters

x	A value of type half.

Returns

The reciprocal of the square root of x.

rsqrt() [4/5]

half2 ipu::rsqrt ( half2  x )

inline

The reciprocal square root function.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the reciprocal of the square root of x[i].

rsqrt() [5/5]

half4 ipu::rsqrt ( half4  x )

inline

The reciprocal square root function.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the reciprocal of the square root of x[i].

shuf8x8hi()

int ipu::shuf8x8hi ( int  src0, int  src1  )

inline

Targets the shuf8x8hi instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The upper word of a SIMD shuffle permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 7 | 3 | 6 | 2 |

shuf8x8lo()

int ipu::shuf8x8lo ( int  src0, int  src1  )

inline

Targets the shuf8x8lo instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The lower word of a SIMD shuffle permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 5 | 1 | 4 | 0 |

sigm() [1/2]

float ipu::sigm ( float  src )

inline

Targets the f32sigm instruction.

Parameters

src	A value of type float.

Returns

The result of an element-wise application of the sigmoid function on src.

sigm() [2/2]

half2 ipu::sigm ( half2  src )

inline

Targets the f16v2sigm instruction.

Parameters

src	A value of type half2.

Returns

The result of an element-wise application of the sigmoid function on src.

sigmoid() [1/5]

float ipu::sigmoid ( float  x )

inline

The sigmoid function, ie 1/(1 + exp(- x )).

Parameters

x	A value of type float.

Returns

The result of sigmoid of x.

sigmoid() [2/5]

float2 ipu::sigmoid ( float2  x )

inline

The sigmoid function, ie 1/(1 + exp(- x )).

Parameters

x	A value of type float2.

Returns

The element-wise result of sigmoid of x.

sigmoid() [3/5]

half ipu::sigmoid ( half  x )

inline

The sigmoid function, ie 1/(1 + exp(- x )).

Parameters

x	A value of type half.

Returns

The result of sigmoid of x.

sigmoid() [4/5]

half2 ipu::sigmoid ( half2  x )

inline

The sigmoid function, ie 1/(1 + exp(- x )).

Parameters

x	A value of type half2.

Returns

The element-wise result of sigmoid of x.

sigmoid() [5/5]

half4 ipu::sigmoid ( half4  x )

inline

The sigmoid function, ie 1/(1 + exp(- x )).

Parameters

x	A value of type half4.

Returns

The element-wise result of sigmoid of x.

sin() [1/5]

float ipu::sin ( float  x )

inline

The trigonometric sine function.

Parameters

x	A value of type float.

Returns

The result of sin of x.

sin() [2/5]

float2 ipu::sin ( float2  x )

inline

The trigonometric sine function.

Parameters

x	A value of type float2.

Returns

The element-wise result of sin of x.

sin() [3/5]

half ipu::sin ( half  x )

inline

The trigonometric sine function.

Parameters

x	A value of type half.

Returns

The result of sin of x.

sin() [4/5]

half2 ipu::sin ( half2  x )

inline

The trigonometric sine function.

Parameters

x	A value of type half2.

Returns

The element-wise result of sin of x.

sin() [5/5]

half4 ipu::sin ( half4  x )

inline

The trigonometric sine function.

Parameters

x	A value of type half4.

Returns

The element-wise result of sin of x.

sinh() [1/5]

float ipu::sinh ( float  x )

inline

The hyperbolic sine function.

Parameters

x	A value of type float.

Returns

The result of sinh of x.

sinh() [2/5]

float2 ipu::sinh ( float2  x )

inline

The hyperbolic sine function.

Parameters

x	A value of type float2.

Returns

The element-wise result of sinh of x.

sinh() [3/5]

half ipu::sinh ( half  x )

inline

The hyperbolic sine function.

Parameters

x	A value of type half.

Returns

The result of sinh of x.

sinh() [4/5]

half2 ipu::sinh ( half2  x )

inline

The hyperbolic sine function.

Parameters

x	A value of type half2.

Returns

The element-wise result of sinh of x.

sinh() [5/5]

half4 ipu::sinh ( half4  x )

inline

The hyperbolic sine function.

Parameters

x	A value of type half4.

Returns

The element-wise result of sinh of x.

sort4x16hi() [1/3]

half2 ipu::sort4x16hi ( half2  src0, half2  src1  )

inline

Targets the sort4x16hi instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a half2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 3 | 1 |

sort4x16hi() [2/3]

short2 ipu::sort4x16hi ( short2  src0, short2  src1  )

inline

Targets the sort4x16hi instruction.

Parameters

src0	A value of type short2.
src1	A value of type short2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a short2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 3 | 1 |

sort4x16hi() [3/3]

ushort2 ipu::sort4x16hi ( ushort2  src0, ushort2  src1  )

inline

Targets the sort4x16hi instruction.

Parameters

src0	A value of type ushort2.
src1	A value of type ushort2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a ushort2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 3 | 1 |

sort4x16lo() [1/3]

half2 ipu::sort4x16lo ( half2  src0, half2  src1  )

inline

Targets the sort4x16lo instruction.

Parameters

src0	A value of type half2.
src1	A value of type half2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a half2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 0 |

sort4x16lo() [2/3]

short2 ipu::sort4x16lo ( short2  src0, short2  src1  )

inline

Targets the sort4x16lo instruction.

Parameters

src0	A value of type short2.
src1	A value of type short2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a short2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 0 |

sort4x16lo() [3/3]

ushort2 ipu::sort4x16lo ( ushort2  src0, ushort2  src1  )

inline

Targets the sort4x16lo instruction.

Parameters

src0	A value of type ushort2.
src1	A value of type ushort2.

Returns

The result of a SIMD sort permutation on the 4 16-bit values across src0 and src1, as a ushort2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 0 |

sort4x32hi()

float2 ipu::sort4x32hi ( float2  src0, float2  src1  )

inline

Targets the sort4x32hi instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The result of a SIMD sort permutation on the 4 32-bit float values across src0 and src1, as a float2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 3 | 1 |

sort4x32lo()

float2 ipu::sort4x32lo ( float2  src0, float2  src1  )

inline

Targets the sort4x32lo instruction.

Parameters

src0	A value of type float2.
src1	A value of type float2.

Returns

The result of a SIMD sort permutation on the 4 32-bit float values across src0 and src1, as a float2. src0 src1 -> Result | 3 | 2 | | 1 | 0 | | 2 | 0 |

sort8()

int ipu::sort8 ( int  src )

inline

Targets the sort8 instruction.

Parameters

src	A value of type int.

Returns

The result of a SIMD sort8 permutation on the 4 8-bit values in src, as an int. src -> Result | 3 | 2 | 1 | 0 | | 3 | 1 | 2 | 0 |

sort8x8hi()

int ipu::sort8x8hi ( int  src0, int  src1  )

inline

Targets the sort8x8hi instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The upper word of the result of a SIMD sort8 permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 7 | 5 | 3 | 1 |

sort8x8lo()

int ipu::sort8x8lo ( int  src0, int  src1  )

inline

Targets the sort8x8lo instruction.

Parameters

src0	A value of type int.
src1	A value of type int.

Returns

The lower word of the result of a SIMD sort8 permutation on the 8 8-bit values across src0 and src1, as an int. src0 src1 -> Result | 7 | 6 | 5 | 4 | | 3 | 2 | 1 | 0 | | 6 | 4 | 2 | 0 |

sqrt() [1/5]

float ipu::sqrt ( float  x )

inline

The square root function.

Parameters

x	A value of type float.

Returns

The result of sqrt of x.

sqrt() [2/5]

float2 ipu::sqrt ( float2  x )

inline

The square root function.

Parameters

x	A value of type float2.

Returns

The element-wise result of sqrt of x.

sqrt() [3/5]

half ipu::sqrt ( half  x )

inline

The square root function.

Parameters

x	A value of type half.

Returns

The result of sqrt of x.

sqrt() [4/5]

half2 ipu::sqrt ( half2  x )

inline

The square root function.

Parameters

x	A value of type half2.

Returns

The element-wise result of sqrt of x.

sqrt() [5/5]

half4 ipu::sqrt ( half4  x )

inline

The square root function.

Parameters

x	A value of type half4.

Returns

The element-wise result of sqrt of x.

store_postinc() [1/16]

void ipu::store_postinc ( char **  a, char  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type char.
i	Value by which to increment a after store.

store_postinc() [2/16]

void ipu::store_postinc ( float **  a, float  v, int  i  )

inline

Post-incrementing store, targeting the st32step instruction.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type float.
i	Value by which to increment a after store.

store_postinc() [3/16]

void ipu::store_postinc ( float2 **  a, float2  v, int  i  )

inline

Post-incrementing store, targeting the st64step instruction.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type float2.
i	Value by which to increment a after store.

store_postinc() [4/16]

void ipu::store_postinc ( half2 **  a, half2  v, int  i  )

inline

Post-incrementing store, targeting the st32step instruction.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type half2.
i	Value by which to increment a after store.

store_postinc() [5/16]

void ipu::store_postinc ( half4 **  a, half4  v, int  i  )

inline

Post-incrementing store, targeting the st64step instruction.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type half4.
i	Value by which to increment a after store.

store_postinc() [6/16]

void ipu::store_postinc ( int **  a, int  v, int  i  )

inline

Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type int.
i	Value by which to increment a after store.

store_postinc() [7/16]

void ipu::store_postinc ( int2 **  a, int2  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type int2.
i	Value by which to increment a after store.

store_postinc() [8/16]

void ipu::store_postinc ( short **  a, short  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type short.
i	Value by which to increment a after store.

store_postinc() [9/16]

void ipu::store_postinc ( short2 **  a, short2  v, int  i  )

inline

Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type short2.
i	Value by which to increment a after store.

store_postinc() [10/16]

void ipu::store_postinc ( short4 **  a, short4  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type short4.
i	Value by which to increment a after store.

store_postinc() [11/16]

void ipu::store_postinc ( uint2 **  a, uint2  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type uint2.
i	Value by which to increment a after store.

store_postinc() [12/16]

void ipu::store_postinc ( unsigned **  a, unsigned  v, int  i  )

inline

Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type unsigned.
i	Value by which to increment a after store.

store_postinc() [13/16]

void ipu::store_postinc ( unsigned char **  a, unsigned char  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type unsigned char.
i	Value by which to increment a after store.

store_postinc() [14/16]

void ipu::store_postinc ( unsigned short **  a, unsigned short  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type unsigned short.
i	Value by which to increment a after store.

store_postinc() [15/16]

void ipu::store_postinc ( ushort2 **  a, ushort2  v, int  i  )

inline

Post-incrementing store, targeting the stm32step instruction if i is a variable stride, and st32step otherwise.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type ushort2.
i	Value by which to increment a after store.

store_postinc() [16/16]

void ipu::store_postinc ( ushort4 **  a, ushort4  v, int  i  )

inline

Post-incrementing store.

Parameters

a	Address of the variable holding the address to store to. Gets incremented by i after the store.
v	Value to store, of type ushort4.
i	Value by which to increment a after store.

sum() [1/2]

float ipu::sum ( half2  src )

inline

Targets the f16v2sum instruction.

Parameters

src	A value of type half2.

Returns

The sum of the two elements in src as a float.

sum() [2/2]

float2 ipu::sum ( half4  src )

inline

Targets the f16v4sum instruction.

Parameters

src	A value of type half2.

Returns

The 2x2 lateral summation of the elements in src as a float2. The first element is the sum of src[0] and src[1], the second element is the sum of src[2] and src[3].

swap8()

int ipu::swap8 ( int  src )

inline

Targets the sort8 instruction.

Parameters

src	A value of type int.

Returns

The result of a SIMD swap permutation on the 4 8-bit values in src, as an int. src -> Result | 3 | 2 | 1 | 0 | | 2 | 3 | 0 | 1 |

tan() [1/5]

float ipu::tan ( float  x )

inline

The trigonometric tangent function.

Parameters

x	A value of type float.

Returns

The result of tan of x.

tan() [2/5]

float2 ipu::tan ( float2  x )

inline

The trigonometric tangent function.

Parameters

x	A value of type float2.

Returns

The element-wise result of tan of x.

tan() [3/5]

half ipu::tan ( half  x )

inline

The trigonometric tangent function.

Parameters

x	A value of type half.

Returns

The result of tan of x.

tan() [4/5]

half2 ipu::tan ( half2  x )

inline

The trigonometric tangent function.

Parameters

x	A value of type half2.

Returns

The element-wise result of tan of x.

tan() [5/5]

half4 ipu::tan ( half4  x )

inline

The trigonometric tangent function.

Parameters

x	A value of type half4.

Returns

The element-wise result of tan of x.

tanh() [1/5]

float ipu::tanh ( float  x )

inline

Targets the f32tanh instruction.

The hyperbolic tangent function.

Parameters

src	A value of type float.

Returns

The result of tanh(src)`.

Parameters

x	A value of type float.

Returns

The result of tanh of x.

tanh() [2/5]

float2 ipu::tanh ( float2  x )

inline

The hyperbolic tangent function.

Parameters

x	A value of type float2.

Returns

The element-wise result of tanh of x.

tanh() [3/5]

half ipu::tanh ( half  x )

inline

The hyperbolic tangent function.

Parameters

x	A value of type half.

Returns

The result of tanh of x.

tanh() [4/5]

half2 ipu::tanh ( half2  x )

inline

Targets the f16v2tanh instruction.

The hyperbolic tangent function.

Parameters

src	A value of type half2.

Returns

The result of tanh(src)`.

Parameters

x	A value of type half2.

Returns

The element-wise result of tanh of x.

tanh() [5/5]

half4 ipu::tanh ( half4  x )

inline

The hyperbolic tangent function.

Parameters

x	A value of type half4.

Returns

The element-wise result of tanh of x.

tgamma() [1/5]

float ipu::tgamma ( float  x )

inline

The gamma function.

Parameters

x	A value of type float.

Returns

The result of tgamma of x.

tgamma() [2/5]

float2 ipu::tgamma ( float2  x )

inline

The gamma function.

Parameters

x	A value of type float2.

Returns

The element-wise result of tgamma of x.

tgamma() [3/5]

half ipu::tgamma ( half  x )

inline

The gamma function.

Parameters

x	A value of type half.

Returns

The result of tgamma of x.

tgamma() [4/5]

half2 ipu::tgamma ( half2  x )

inline

The gamma function.

Parameters

x	A value of type half2.

Returns

The element-wise result of tgamma of x.

tgamma() [5/5]

half4 ipu::tgamma ( half4  x )

inline

The gamma function.

Parameters

x	A value of type half4.

Returns

The element-wise result of tgamma of x.

trunc() [1/5]

float ipu::trunc ( float  x )

inline

Rounds input towards zero to the nearest integral value that is not larger in magnitude than x.

Parameters

x	A value of type float.

Returns

The nearest integral value that is not larger in magnitude than x.

trunc() [2/5]

float2 ipu::trunc ( float2  x )

inline

Rounds input towards zero to the nearest integral value that is not larger in magnitude than x.

Parameters

x	A value of type float2.

Returns

A vector where every ith element is the nearest integral value to x[i] that is not larger in magnitude.

trunc() [3/5]

half ipu::trunc ( half  x )

inline

Rounds input towards zero to the nearest integral value that is not larger in magnitude than x.

Parameters

x	A value of type half.

Returns

The nearest integral value that is not larger in magnitude than x.

trunc() [4/5]

half2 ipu::trunc ( half2  x )

inline

Rounds input towards zero to the nearest integral value that is not larger in magnitude than x.

Parameters

x	A value of type half2.

Returns

A vector where every ith element is the nearest integral value to x[i] that is not larger in magnitude.

trunc() [5/5]

half4 ipu::trunc ( half4  x )

inline

Rounds input towards zero to the nearest integral value that is not larger in magnitude than x.

Parameters

x	A value of type half4.

Returns

A vector where every ith element is the nearest integral value to x[i] that is not larger in magnitude.