simd.h File Reference

Classes for SIMD processing. More...

#include <algorithm>
#include <cstring>
#include <OpenImageIO/Imath.h>
#include <OpenImageIO/dassert.h>
#include <OpenImageIO/platform.h>

Include dependency graph for simd.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	simd::simd_raw_t< T, N >
struct	simd::simd_raw_t< T, N >::type
struct	simd::simd_bool_t< N >
struct	simd::simd_bool_t< N >::type
struct	simd::simd_bool_t< 16 >
struct	simd::VecType< T, elements >
struct	simd::VecType< int, 1 >
struct	simd::VecType< float, 1 >
struct	simd::VecType< int, 4 >
struct	simd::VecType< float, 4 >
struct	simd::VecType< float, 3 >
struct	simd::VecType< bool, 4 >
struct	simd::VecType< int, 8 >
struct	simd::VecType< float, 8 >
struct	simd::VecType< bool, 8 >
struct	simd::VecType< int, 16 >
struct	simd::VecType< float, 16 >
struct	simd::VecType< bool, 16 >
struct	simd::SimdSize< T >
struct	simd::SimdSize< vint4 >
struct	simd::SimdSize< vfloat4 >
struct	simd::SimdSize< vfloat3 >
struct	simd::SimdSize< vbool4 >
struct	simd::SimdSize< vint8 >
struct	simd::SimdSize< vfloat8 >
struct	simd::SimdSize< vbool8 >
struct	simd::SimdSize< vint16 >
struct	simd::SimdSize< vfloat16 >
struct	simd::SimdSize< vbool16 >
struct	simd::SimdElements< T >
struct	simd::SimdElements< vfloat3 >
struct	simd::SimdTypeName< T >
	Template giving a printable name for each type. More...
struct	simd::SimdTypeName< vfloat4 >
struct	simd::SimdTypeName< vint4 >
struct	simd::SimdTypeName< vbool4 >
struct	simd::SimdTypeName< vfloat8 >
struct	simd::SimdTypeName< vint8 >
struct	simd::SimdTypeName< vbool8 >
struct	simd::SimdTypeName< vfloat16 >
struct	simd::SimdTypeName< vint16 >
struct	simd::SimdTypeName< vbool16 >
class	simd::vbool4
class	simd::vbool8
class	simd::vbool16
class	simd::vint4
	Integer 4-vector, accelerated by SIMD instructions when available. More...
class	simd::vint8
	Integer 8-vector, accelerated by SIMD instructions when available. More...
class	simd::vint16
	Integer 16-vector, accelerated by SIMD instructions when available. More...
class	simd::vfloat4
class	simd::vfloat3
class	simd::matrix44
class	simd::vfloat8
class	simd::vfloat16

Namespaces
	simd

Macros
#define	OIIO_SIMD_SSE   0
#define	OIIO_SIMD_AVX   0
#define	OIIO_AVX512VL_ENABLED   0
#define	OIIO_AVX512DQ_ENABLED   0
#define	OIIO_AVX512PF_ENABLED   0
#define	OIIO_AVX512ER_ENABLED   0
#define	OIIO_AVX512CD_ENABLED   0
#define	OIIO_AVX512BW_ENABLED   0
#define	OIIO_FMA_ENABLED   0
#define	OIIO_AVX512IFMA_ENABLED   0
#define	OIIO_F16C_ENABLED   0
#define	OIIO_SIMD_NEON   0
#define	OIIO_SIMD   0
#define	OIIO_SIMD4_ALIGN
#define	OIIO_SIMD_MAX_SIZE_BYTES   16
#define	OIIO_SIMD8_ALIGN   OIIO_SIMD4_ALIGN
#define	OIIO_SIMD16_ALIGN   OIIO_SIMD8_ALIGN
#define	OIIO_SIMD_HAS_MATRIX4   1 /* matrix44 defined */
#define	OIIO_SIMD_HAS_FLOAT8   1 /* DEPRECATED(1.8) */
#define	OIIO_SIMD_HAS_SIMD8   1 /* vfloat8, vint8, vbool8 defined */
#define	OIIO_SIMD_HAS_SIMD16   1 /* vfloat16, vint16, vbool16 defined */
#define	OIIO_SIMD_FLOAT4_CONST(name, val)   static const OIIO_SIMD4_ALIGN float name[4] = { (val), (val), (val), (val) }
#define	OIIO_SIMD_FLOAT4_CONST4(name, v0, v1, v2, v3)   static const OIIO_SIMD4_ALIGN float name[4] = { (v0), (v1), (v2), (v3) }
#define	OIIO_SIMD_INT4_CONST(name, val)   static const OIIO_SIMD4_ALIGN int name[4] = { (val), (val), (val), (val) }
#define	OIIO_SIMD_INT4_CONST4(name, v0, v1, v2, v3)   static const OIIO_SIMD4_ALIGN int name[4] = { (v0), (v1), (v2), (v3) }
#define	OIIO_SIMD_UINT4_CONST(name, val)   static const OIIO_SIMD4_ALIGN uint32_t name[4] = { (val), (val), (val), (val) }
#define	OIIO_SIMD_UINT4_CONST4(name, v0, v1, v2, v3)   static const OIIO_SIMD4_ALIGN uint32_t name[4] = { (v0), (v1), (v2), (v3) }
#define	OIIO_SIMD_FLOAT8_CONST(name, val)
#define	OIIO_SIMD_FLOAT8_CONST8(name, v0, v1, v2, v3, v4, v5, v6, v7)
#define	OIIO_SIMD_INT8_CONST(name, val)
#define	OIIO_SIMD_INT8_CONST8(name, v0, v1, v2, v3, v4, v5, v6, v7)
#define	OIIO_SIMD_UINT8_CONST(name, val)
#define	OIIO_SIMD_UINT8_CONST8(name, v0, v1, v2, v3, v4, v5, v6, v7)
#define	OIIO_SIMD_VFLOAT16_CONST(name, val)
#define	OIIO_SIMD_VFLOAT16_CONST16(name, v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15)
#define	OIIO_SIMD_INT16_CONST(name, val)
#define	OIIO_SIMD_INT16_CONST16(name, v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15)
#define	OIIO_SIMD_UINT16_CONST(name, val)
#define	OIIO_SIMD_UINT16_CONST16(name, v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15)
#define	SIMD_DO(x)   for (int i = 0; i < elements; ++i) x
#define	SIMD_CONSTRUCT(x)   for (int i = 0; i < elements; ++i) m_val[i] = (x)
#define	SIMD_CONSTRUCT_PAD(x)
#define	SIMD_RETURN(T, x)   T r; for (int i = 0; i < r.elements; ++i) r[i] = (x); return r
#define	SIMD_RETURN_REDUCE(T, init, op)   T r = init; for (int i = 0; i < v.elements; ++i) op; return r

Typedefs
typedef vbool4	simd::mask4
typedef vbool4	simd::bool4
typedef vbool8	simd::bool8
typedef vint4	simd::int4
typedef vint8	simd::int8
typedef vfloat3	simd::float3
typedef vfloat4	simd::float4
typedef vfloat8	simd::float8

Functions
template<int i0, int i1, int i2, int i3>
OIIO_FORCEINLINE vbool4	simd::shuffle (const vbool4 &a)
template<int i>
OIIO_FORCEINLINE vbool4	simd::shuffle (const vbool4 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE bool	simd::extract (const vbool4 &a)
template<int i>
OIIO_FORCEINLINE vbool4	simd::insert (const vbool4 &a, bool val)
	Helper: substitute val for a[i]. More...
bool	simd::reduce_and (const vbool4 &v)
	Logical reduction across all components. More...
bool	simd::reduce_or (const vbool4 &v)
bool	simd::all (const vbool4 &v)
bool	simd::any (const vbool4 &v)
bool	simd::none (const vbool4 &v)
bool	simd::all (bool v)
template<int i0, int i1, int i2, int i3, int i4, int i5, int i6, int i7>
OIIO_FORCEINLINE vbool8	simd::shuffle (const vbool8 &a)
template<int i>
OIIO_FORCEINLINE vbool8	simd::shuffle (const vbool8 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE bool	simd::extract (const vbool8 &a)
template<int i>
OIIO_FORCEINLINE vbool8	simd::insert (const vbool8 &a, bool val)
	Helper: substitute val for a[i]. More...
bool	simd::reduce_and (const vbool8 &v)
	Logical reduction across all components. More...
bool	simd::reduce_or (const vbool8 &v)
bool	simd::all (const vbool8 &v)
bool	simd::any (const vbool8 &v)
bool	simd::none (const vbool8 &v)
template<int i>
OIIO_FORCEINLINE bool	simd::extract (const vbool16 &a)
template<int i>
OIIO_FORCEINLINE vbool16	simd::insert (const vbool16 &a, bool val)
	Helper: substitute val for a[i]. More...
bool	simd::reduce_and (const vbool16 &v)
	Logical reduction across all components. More...
bool	simd::reduce_or (const vbool16 &v)
bool	simd::all (const vbool16 &v)
bool	simd::any (const vbool16 &v)
bool	simd::none (const vbool16 &v)
vint4	simd::srl (const vint4 &val, const unsigned int bits)
template<int i0, int i1, int i2, int i3>
OIIO_FORCEINLINE vint4	simd::shuffle (const vint4 &a)
template<int i>
OIIO_FORCEINLINE vint4	simd::shuffle (const vint4 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE int	simd::extract (const vint4 &v)
vint4	simd::vreduce_add (const vint4 &v)
	The sum of all components, returned in all components. More...
int	simd::reduce_add (const vint4 &v)
int	simd::reduce_and (const vint4 &v)
int	simd::reduce_or (const vint4 &v)
vint4	simd::blend (const vint4 &a, const vint4 &b, const vbool4 &mask)
vint4	simd::blend0 (const vint4 &a, const vbool4 &mask)
vint4	simd::blend0not (const vint4 &a, const vbool4 &mask)
vint4	simd::select (const vbool4 &mask, const vint4 &a, const vint4 &b)
vint4	simd::abs (const vint4 &a)
vint4	simd::min (const vint4 &a, const vint4 &b)
vint4	simd::max (const vint4 &a, const vint4 &b)
vint4	simd::rotl (const vint4 &x, const int s)
	Circular bit rotate by s bits, for N values at once. More...
vint4	simd::rotl32 (const vint4 &x, const unsigned int k)
vint4	simd::andnot (const vint4 &a, const vint4 &b)
	andnot(a,b) returns ((~a) & b) More...
vint4	simd::bitcast_to_int (const vbool4 &x)
	Bitcast back and forth to intN (not a convert – move the bits!) More...
vint4	simd::bitcast_to_int (const vfloat4 &x)
vfloat4	simd::bitcast_to_float (const vint4 &x)
void	simd::transpose (vint4 &a, vint4 &b, vint4 &c, vint4 &d)
void	simd::transpose (const vint4 &a, const vint4 &b, const vint4 &c, const vint4 &d, vint4 &r0, vint4 &r1, vint4 &r2, vint4 &r3)
vint4	simd::AxBxCxDx (const vint4 &a, const vint4 &b, const vint4 &c, const vint4 &d)
vint4	simd::safe_mod (const vint4 &a, const vint4 &b)
vint4	simd::safe_mod (const vint4 &a, int b)
vint8	simd::srl (const vint8 &val, const unsigned int bits)
template<int i0, int i1, int i2, int i3, int i4, int i5, int i6, int i7>
OIIO_FORCEINLINE vint8	simd::shuffle (const vint8 &a)
template<int i>
OIIO_FORCEINLINE vint8	simd::shuffle (const vint8 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE int	simd::extract (const vint8 &v)
template<int i>
OIIO_FORCEINLINE vint8	simd::insert (const vint8 &a, int val)
	Helper: substitute val for a[i]. More...
vint8	simd::vreduce_add (const vint8 &v)
	The sum of all components, returned in all components. More...
int	simd::reduce_add (const vint8 &v)
int	simd::reduce_and (const vint8 &v)
int	simd::reduce_or (const vint8 &v)
vint8	simd::blend (const vint8 &a, const vint8 &b, const vbool8 &mask)
vint8	simd::blend0 (const vint8 &a, const vbool8 &mask)
vint8	simd::blend0not (const vint8 &a, const vbool8 &mask)
vint8	simd::select (const vbool8 &mask, const vint8 &a, const vint8 &b)
vint8	simd::abs (const vint8 &a)
vint8	simd::min (const vint8 &a, const vint8 &b)
vint8	simd::max (const vint8 &a, const vint8 &b)
vint8	simd::rotl (const vint8 &x, const int s)
	Circular bit rotate by s bits, for N values at once. More...
vint8	simd::rotl32 (const vint8 &x, const unsigned int k)
vint8	simd::andnot (const vint8 &a, const vint8 &b)
	andnot(a,b) returns ((~a) & b) More...
vint8	simd::bitcast_to_int (const vbool8 &x)
	Bitcast back and forth to intN (not a convert – move the bits!) More...
vint8	simd::bitcast_to_int (const vfloat8 &x)
vfloat8	simd::bitcast_to_float (const vint8 &x)
vint8	simd::safe_mod (const vint8 &a, const vint8 &b)
vint8	simd::safe_mod (const vint8 &a, int b)
vint16	simd::srl (const vint16 &val, const unsigned int bits)
template<int i0, int i1, int i2, int i3>
vint16	simd::shuffle4 (const vint16 &a)
	Shuffle groups of 4. More...
template<int i>
vint16	simd::shuffle4 (const vint16 &a)
	shuffle4<i>(a) is the same as shuffle4<i,i,i,i>(a) More...
template<int i0, int i1, int i2, int i3>
vint16	simd::shuffle (const vint16 &a)
	Shuffle within each group of 4. More...
template<int i>
vint16	simd::shuffle (const vint16 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE int	simd::extract (const vint16 &v)
template<int i>
OIIO_FORCEINLINE vint16	simd::insert (const vint16 &a, int val)
	Helper: substitute val for a[i]. More...
vint16	simd::vreduce_add (const vint16 &v)
	The sum of all components, returned in all components. More...
int	simd::reduce_add (const vint16 &v)
int	simd::reduce_and (const vint16 &v)
int	simd::reduce_or (const vint16 &v)
vint16	simd::blend (const vint16 &a, const vint16 &b, const vbool16 &mask)
vint16	simd::blend0 (const vint16 &a, const vbool16 &mask)
vint16	simd::blend0not (const vint16 &a, const vbool16 &mask)
vint16	simd::select (const vbool16 &mask, const vint16 &a, const vint16 &b)
vint16	simd::abs (const vint16 &a)
vint16	simd::min (const vint16 &a, const vint16 &b)
vint16	simd::max (const vint16 &a, const vint16 &b)
vint16	simd::rotl (const vint16 &x, const int s)
	Circular bit rotate by s bits, for N values at once. More...
vint16	simd::rotl32 (const vint16 &x, const unsigned int k)
vint16	simd::andnot (const vint16 &a, const vint16 &b)
	andnot(a,b) returns ((~a) & b) More...
vint16	simd::bitcast_to_int (const vbool16 &x)
	Bitcast back and forth to intN (not a convert – move the bits!) More...
vint16	simd::bitcast_to_int (const vfloat16 &x)
vfloat16	simd::bitcast_to_float (const vint16 &x)
vint16	simd::safe_mod (const vint16 &a, const vint16 &b)
vint16	simd::safe_mod (const vint16 &a, int b)
template<int i0, int i1, int i2, int i3>
OIIO_FORCEINLINE vfloat4	simd::shuffle (const vfloat4 &a)
template<int i>
OIIO_FORCEINLINE vfloat4	simd::shuffle (const vfloat4 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE float	simd::extract (const vfloat4 &a)
template<int i>
OIIO_FORCEINLINE vfloat4	simd::insert (const vfloat4 &a, float val)
	Helper: substitute val for a[i]. More...
vfloat4	simd::vreduce_add (const vfloat4 &v)
	The sum of all components, returned in all components. More...
float	simd::reduce_add (const vfloat4 &v)
	The sum of all components, returned as a scalar. More...
vfloat4	simd::vdot (const vfloat4 &a, const vfloat4 &b)
	Return the float dot (inner) product of a and b in every component. More...
float	simd::dot (const vfloat4 &a, const vfloat4 &b)
	Return the float dot (inner) product of a and b. More...
vfloat4	simd::vdot3 (const vfloat4 &a, const vfloat4 &b)
float	simd::dot3 (const vfloat4 &a, const vfloat4 &b)
	Return the float 3-component dot (inner) product of a and b. More...
vfloat4	simd::blend (const vfloat4 &a, const vfloat4 &b, const vbool4 &mask)
vfloat4	simd::blend0 (const vfloat4 &a, const vbool4 &mask)
vfloat4	simd::blend0not (const vfloat4 &a, const vbool4 &mask)
vfloat4	simd::safe_div (const vfloat4 &a, const vfloat4 &b)
vfloat3	simd::hdiv (const vfloat4 &a)
	Homogeneous divide to turn a vfloat4 into a vfloat3. More...
vfloat4	simd::select (const vbool4 &mask, const vfloat4 &a, const vfloat4 &b)
vfloat4	simd::abs (const vfloat4 &a)
	absolute value (float) More...
vfloat4	simd::sign (const vfloat4 &a)
	1.0 when value >= 0, -1 when negative More...
vfloat4	simd::ceil (const vfloat4 &a)
vfloat4	simd::floor (const vfloat4 &a)
vint4	simd::ifloor (const vfloat4 &a)
	(int)floor More...
vint4	simd::floori (const vfloat4 &a)
vfloat4	simd::round (const vfloat4 &a)
vint4	simd::rint (const vfloat4 &a)
vfloat4	simd::rcp_fast (const vfloat4 &a)
	Fast, approximate 1/a. More...
vfloat4	simd::sqrt (const vfloat4 &a)
vfloat4	simd::rsqrt (const vfloat4 &a)
	Fully accurate 1/sqrt. More...
vfloat4	simd::rsqrt_fast (const vfloat4 &a)
	Fast, approximate 1/sqrt. More...
vfloat4	simd::min (const vfloat4 &a, const vfloat4 &b)
	Per-element min. More...
vfloat4	simd::max (const vfloat4 &a, const vfloat4 &b)
	Per-element max. More...
template<typename T >
OIIO_FORCEINLINE T	simd::exp (const T &v)
template<typename T >
OIIO_FORCEINLINE T	simd::log (const T &v)
vfloat4	simd::andnot (const vfloat4 &a, const vfloat4 &b)
	andnot(a,b) returns ((~a) & b) More...
vfloat4	simd::madd (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c)
vfloat4	simd::msub (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c)
vfloat4	simd::nmadd (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c)
vfloat4	simd::nmsub (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c)
void	simd::transpose (vfloat4 &a, vfloat4 &b, vfloat4 &c, vfloat4 &d)
void	simd::transpose (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c, const vfloat4 &d, vfloat4 &r0, vfloat4 &r1, vfloat4 &r2, vfloat4 &r3)
vfloat4	simd::AxBxCxDx (const vfloat4 &a, const vfloat4 &b, const vfloat4 &c, const vfloat4 &d)
	Make a vfloat4 consisting of the first element of each of 4 vfloat4's. More...
vfloat3	simd::abs (const vfloat3 &a)
vfloat3	simd::sign (const vfloat3 &a)
vfloat3	simd::ceil (const vfloat3 &a)
vfloat3	simd::floor (const vfloat3 &a)
vfloat3	simd::round (const vfloat3 &a)
vfloat3	simd::transformp (const matrix44 &M, const vfloat3 &V)
	Transform 3-point V by 4x4 matrix M. More...
vfloat3	simd::transformp (const Imath::M44f &M, const vfloat3 &V)
vfloat3	simd::transformv (const matrix44 &M, const vfloat3 &V)
	Transform 3-vector V by 4x4 matrix M. More...
vfloat3	simd::transformv (const Imath::M44f &M, const vfloat3 &V)
vfloat3	simd::transformvT (const matrix44 &M, const vfloat3 &V)
vfloat3	simd::transformvT (const Imath::M44f &M, const vfloat3 &V)
template<int i0, int i1, int i2, int i3, int i4, int i5, int i6, int i7>
OIIO_FORCEINLINE vfloat8	simd::shuffle (const vfloat8 &a)
template<int i>
OIIO_FORCEINLINE vfloat8	simd::shuffle (const vfloat8 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i,...>(a) More...
template<int i>
OIIO_FORCEINLINE float	simd::extract (const vfloat8 &a)
template<int i>
OIIO_FORCEINLINE vfloat8	simd::insert (const vfloat8 &a, float val)
	Helper: substitute val for a[i]. More...
vfloat8	simd::vreduce_add (const vfloat8 &v)
	The sum of all components, returned in all components. More...
float	simd::reduce_add (const vfloat8 &v)
	The sum of all components, returned as a scalar. More...
vfloat8	simd::vdot (const vfloat8 &a, const vfloat8 &b)
	Return the float dot (inner) product of a and b in every component. More...
float	simd::dot (const vfloat8 &a, const vfloat8 &b)
	Return the float dot (inner) product of a and b. More...
vfloat8	simd::vdot3 (const vfloat8 &a, const vfloat8 &b)
float	simd::dot3 (const vfloat8 &a, const vfloat8 &b)
	Return the float 3-component dot (inner) product of a and b. More...
vfloat8	simd::blend (const vfloat8 &a, const vfloat8 &b, const vbool8 &mask)
vfloat8	simd::blend0 (const vfloat8 &a, const vbool8 &mask)
vfloat8	simd::blend0not (const vfloat8 &a, const vbool8 &mask)
vfloat8	simd::safe_div (const vfloat8 &a, const vfloat8 &b)
vfloat8	simd::select (const vbool8 &mask, const vfloat8 &a, const vfloat8 &b)
vfloat8	simd::abs (const vfloat8 &a)
	absolute value (float) More...
vfloat8	simd::sign (const vfloat8 &a)
	1.0 when value >= 0, -1 when negative More...
vfloat8	simd::ceil (const vfloat8 &a)
vfloat8	simd::floor (const vfloat8 &a)
vint8	simd::ifloor (const vfloat8 &a)
	(int)floor More...
vint8	simd::floori (const vfloat8 &a)
vfloat8	simd::round (const vfloat8 &a)
vint8	simd::rint (const vfloat8 &a)
vfloat8	simd::rcp_fast (const vfloat8 &a)
	Fast, approximate 1/a. More...
vfloat8	simd::sqrt (const vfloat8 &a)
vfloat8	simd::rsqrt (const vfloat8 &a)
	Fully accurate 1/sqrt. More...
vfloat8	simd::rsqrt_fast (const vfloat8 &a)
	Fast, approximate 1/sqrt. More...
vfloat8	simd::min (const vfloat8 &a, const vfloat8 &b)
	Per-element min. More...
vfloat8	simd::max (const vfloat8 &a, const vfloat8 &b)
	Per-element max. More...
vfloat8	simd::andnot (const vfloat8 &a, const vfloat8 &b)
	andnot(a,b) returns ((~a) & b) More...
vfloat8	simd::madd (const vfloat8 &a, const vfloat8 &b, const vfloat8 &c)
vfloat8	simd::msub (const vfloat8 &a, const vfloat8 &b, const vfloat8 &c)
vfloat8	simd::nmadd (const vfloat8 &a, const vfloat8 &b, const vfloat8 &c)
vfloat8	simd::nmsub (const vfloat8 &a, const vfloat8 &b, const vfloat8 &c)
template<int i0, int i1, int i2, int i3>
OIIO_FORCEINLINE vfloat16	simd::shuffle4 (const vfloat16 &a)
	Shuffle groups of 4. More...
template<int i>
OIIO_FORCEINLINE vfloat16	simd::shuffle4 (const vfloat16 &a)
	shuffle4<i>(a) is the same as shuffle4<i,i,i,i>(a) More...
template<int i0, int i1, int i2, int i3>
OIIO_FORCEINLINE vfloat16	simd::shuffle (const vfloat16 &a)
	Shuffle within each group of 4. More...
template<int i>
vfloat16	simd::shuffle (const vfloat16 &a)
	shuffle<i>(a) is the same as shuffle<i,i,i,i>(a) More...
template<int i>
OIIO_FORCEINLINE float	simd::extract (const vfloat16 &a)
template<int i>
OIIO_FORCEINLINE vfloat16	simd::insert (const vfloat16 &a, float val)
	Helper: substitute val for a[i]. More...
vfloat16	simd::vreduce_add (const vfloat16 &v)
	The sum of all components, returned in all components. More...
float	simd::reduce_add (const vfloat16 &v)
	The sum of all components, returned as a scalar. More...
vfloat16	simd::blend (const vfloat16 &a, const vfloat16 &b, const vbool4 &mask)
vfloat16	simd::blend0 (const vfloat16 &a, const vbool4 &mask)
vfloat16	simd::blend0not (const vfloat16 &a, const vbool4 &mask)
vfloat16	simd::safe_div (const vfloat16 &a, const vfloat16 &b)
vfloat16	simd::select (const vbool16 &mask, const vfloat16 &a, const vfloat16 &b)
vfloat16	simd::abs (const vfloat16 &a)
	absolute value (float) More...
vfloat16	simd::sign (const vfloat16 &a)
	1.0 when value >= 0, -1 when negative More...
vfloat16	simd::ceil (const vfloat16 &a)
vfloat16	simd::floor (const vfloat16 &a)
vint16	simd::ifloor (const vfloat16 &a)
	(int)floor More...
vint16	simd::floori (const vfloat16 &a)
vfloat16	simd::round (const vfloat16 &a)
vint16	simd::rint (const vfloat16 &a)
vfloat16	simd::rcp_fast (const vfloat16 &a)
	Fast, approximate 1/a. More...
vfloat16	simd::sqrt (const vfloat16 &a)
vfloat16	simd::rsqrt (const vfloat16 &a)
	Fully accurate 1/sqrt. More...
vfloat16	simd::rsqrt_fast (const vfloat16 &a)
	Fast, approximate 1/sqrt. More...
vfloat16	simd::min (const vfloat16 &a, const vfloat16 &b)
	Per-element min. More...
vfloat16	simd::max (const vfloat16 &a, const vfloat16 &b)
	Per-element max. More...
vfloat16	simd::andnot (const vfloat16 &a, const vfloat16 &b)
	andnot(a,b) returns ((~a) & b) More...
vfloat16	simd::madd (const vfloat16 &a, const vfloat16 &b, const vfloat16 &c)
vfloat16	simd::msub (const vfloat16 &a, const vfloat16 &b, const vfloat16 &c)
vfloat16	simd::nmadd (const vfloat16 &a, const vfloat16 &b, const vfloat16 &c)
vfloat16	simd::nmsub (const vfloat16 &a, const vfloat16 &b, const vfloat16 &c)
bool	simd::set_flush_zero_mode (bool on)
bool	simd::set_denorms_zero_mode (bool on)
bool	simd::get_flush_zero_mode ()
bool	simd::get_denorms_zero_mode ()
OIIO_FORCEINLINE std::ostream &	simd::operator<< (std::ostream &cout, const vbool4 &a)
OIIO_FORCEINLINE vbool4	simd::operator! (const vbool4 &a)
OIIO_FORCEINLINE vbool4	simd::operator& (const vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE vbool4	simd::operator| (const vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE vbool4	simd::operator^ (const vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE const vbool4 &	simd::operator&= (vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE const vbool4 &	simd::operator|= (vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE const vbool4 &	simd::operator^= (vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE vbool4	simd::operator~ (const vbool4 &a)
OIIO_FORCEINLINE vbool4	simd::operator== (const vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE vbool4	simd::operator!= (const vbool4 &a, const vbool4 &b)
OIIO_FORCEINLINE std::ostream &	simd::operator<< (std::ostream &cout, const vbool8 &a)
OIIO_FORCEINLINE vbool8	simd::operator! (const vbool8 &a)
OIIO_FORCEINLINE vbool8	simd::operator& (const vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE vbool8	simd::operator| (const vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE vbool8	simd::operator^ (const vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE const vbool8 &	simd::operator&= (vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE const vbool8 &	simd::operator|= (vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE const vbool8 &	simd::operator^= (vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE vbool8	simd::operator~ (const vbool8 &a)
OIIO_FORCEINLINE vbool8	simd::operator== (const vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE vbool8	simd::operator!= (const vbool8 &a, const vbool8 &b)
OIIO_FORCEINLINE std::ostream &	simd::operator<< (std::ostream &cout, const vbool16 &a)
OIIO_FORCEINLINE vbool16	simd::operator! (const vbool16 &a)
OIIO_FORCEINLINE vbool16	simd::operator& (const vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE vbool16	simd::operator| (const vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE vbool16	simd::operator^ (const vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE const vbool16 &	simd::operator&= (vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE const vbool16 &	simd::operator|= (vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE const vbool16 &	simd::operator^= (vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE vbool16	simd::operator~ (const vbool16 &a)
OIIO_FORCEINLINE vbool16	simd::operator== (const vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE vbool16	simd::operator!= (const vbool16 &a, const vbool16 &b)
OIIO_FORCEINLINE vint4	simd::operator+ (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator+= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator- (const vint4 &a)
OIIO_FORCEINLINE vint4	simd::operator- (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator-= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator* (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator*= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator*= (vint4 &a, int b)
OIIO_FORCEINLINE vint4	simd::operator/ (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator/= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator% (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator%= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator% (const vint4 &a, int w)
OIIO_FORCEINLINE const vint4 &	simd::operator%= (vint4 &a, int b)
OIIO_FORCEINLINE vint4	simd::operator& (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator&= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator| (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator|= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator^ (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE const vint4 &	simd::operator^= (vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vint4	simd::operator~ (const vint4 &a)
OIIO_FORCEINLINE vint4	simd::operator<< (const vint4 &a, unsigned int bits)
OIIO_FORCEINLINE const vint4 &	simd::operator<<= (vint4 &a, const unsigned int bits)
OIIO_FORCEINLINE vint4	simd::operator>> (const vint4 &a, const unsigned int bits)
OIIO_FORCEINLINE const vint4 &	simd::operator>>= (vint4 &a, const unsigned int bits)
OIIO_FORCEINLINE vbool4	simd::operator== (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vbool4	simd::operator!= (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vbool4	simd::operator> (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vbool4	simd::operator< (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vbool4	simd::operator>= (const vint4 &a, const vint4 &b)
OIIO_FORCEINLINE vbool4	simd::operator<= (const vint4 &a, const vint4 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vint4 &val)
template<int i>
OIIO_FORCEINLINE vint4	simd::insert (const vint4 &a, int val)
	simd::OIIO_DEPRECATED ("use bitcast_to_int() (1.8)") inline vint4 bitcast_to_int4(const vbool4 &x)
OIIO_FORCEINLINE vint8	simd::operator+ (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator+= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator- (const vint8 &a)
OIIO_FORCEINLINE vint8	simd::operator- (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator-= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator* (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator*= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator*= (vint8 &a, int b)
OIIO_FORCEINLINE vint8	simd::operator/ (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator/= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator% (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator%= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator% (const vint8 &a, int w)
OIIO_FORCEINLINE const vint8 &	simd::operator%= (vint8 &a, int b)
OIIO_FORCEINLINE vint8	simd::operator& (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator&= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator| (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator|= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator^ (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE const vint8 &	simd::operator^= (vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vint8	simd::operator~ (const vint8 &a)
OIIO_FORCEINLINE vint8	simd::operator<< (const vint8 &a, unsigned int bits)
OIIO_FORCEINLINE const vint8 &	simd::operator<<= (vint8 &a, const unsigned int bits)
OIIO_FORCEINLINE vint8	simd::operator>> (const vint8 &a, const unsigned int bits)
OIIO_FORCEINLINE const vint8 &	simd::operator>>= (vint8 &a, const unsigned int bits)
OIIO_FORCEINLINE vbool8	simd::operator== (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vbool8	simd::operator!= (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vbool8	simd::operator> (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vbool8	simd::operator< (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vbool8	simd::operator>= (const vint8 &a, const vint8 &b)
OIIO_FORCEINLINE vbool8	simd::operator<= (const vint8 &a, const vint8 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vint8 &val)
OIIO_FORCEINLINE vint16	simd::operator+ (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator+= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator- (const vint16 &a)
OIIO_FORCEINLINE vint16	simd::operator- (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator-= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator* (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator*= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator*= (vint16 &a, int b)
OIIO_FORCEINLINE vint16	simd::operator/ (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator/= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator% (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator%= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator% (const vint16 &a, int w)
OIIO_FORCEINLINE const vint16 &	simd::operator%= (vint16 &a, int b)
OIIO_FORCEINLINE vint16	simd::operator& (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator&= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator| (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator|= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator^ (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE const vint16 &	simd::operator^= (vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vint16	simd::operator~ (const vint16 &a)
OIIO_FORCEINLINE vint16	simd::operator<< (const vint16 &a, const unsigned int bits)
OIIO_FORCEINLINE const vint16 &	simd::operator<<= (vint16 &a, const unsigned int bits)
OIIO_FORCEINLINE vint16	simd::operator>> (const vint16 &a, const unsigned int bits)
OIIO_FORCEINLINE const vint16 &	simd::operator>>= (vint16 &a, const unsigned int bits)
OIIO_FORCEINLINE vbool16	simd::operator== (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vbool16	simd::operator!= (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vbool16	simd::operator> (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vbool16	simd::operator< (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vbool16	simd::operator>= (const vint16 &a, const vint16 &b)
OIIO_FORCEINLINE vbool16	simd::operator<= (const vint16 &a, const vint16 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vint16 &val)
OIIO_FORCEINLINE vfloat4	simd::operator+ (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::operator- (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::operator* (const vfloat4 &a, float b)
OIIO_FORCEINLINE vfloat4	simd::operator* (float a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::operator* (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::operator/ (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator== (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator!= (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator< (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator> (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator>= (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vbool4	simd::operator<= (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::AxyBxy (const vfloat4 &a, const vfloat4 &b)
OIIO_FORCEINLINE vfloat4	simd::AxBxAyBy (const vfloat4 &a, const vfloat4 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vfloat4 &val)
vint4	simd::bitcast_to_int4 (const vfloat4 &x)
vfloat4	simd::bitcast_to_float4 (const vint4 &x)
OIIO_FORCEINLINE vfloat3	simd::operator+ (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat3	simd::operator- (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat3	simd::operator* (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat3	simd::operator* (const vfloat3 &a, float b)
OIIO_FORCEINLINE vfloat3	simd::operator* (float a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat3	simd::operator/ (const vfloat3 &a, const vfloat3 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vfloat3 &val)
OIIO_FORCEINLINE vfloat3	simd::vreduce_add (const vfloat3 &v)
OIIO_FORCEINLINE vfloat3	simd::vdot (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE float	simd::dot (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat3	simd::vdot3 (const vfloat3 &a, const vfloat3 &b)
OIIO_FORCEINLINE vfloat4	simd::operator* (const vfloat4 &V, const matrix44 &M)
OIIO_FORCEINLINE vfloat4	simd::operator* (const matrix44 &M, const vfloat4 &V)
OIIO_FORCEINLINE bool	simd::operator== (const Imath::M44f &a, const matrix44 &b)
OIIO_FORCEINLINE bool	simd::operator!= (const Imath::M44f &a, const matrix44 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const matrix44 &M)
std::ostream &	simd::operator<< (std::ostream &cout, const vfloat8 &val)
OIIO_FORCEINLINE vfloat8	simd::operator+ (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE const vfloat8 &	simd::operator+= (vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vfloat8	simd::operator- (const vfloat8 &a)
OIIO_FORCEINLINE vfloat8	simd::operator- (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE const vfloat8 &	simd::operator-= (vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vfloat8	simd::operator* (const vfloat8 &a, float b)
OIIO_FORCEINLINE vfloat8	simd::operator* (float a, const vfloat8 &b)
OIIO_FORCEINLINE vfloat8	simd::operator* (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE const vfloat8 &	simd::operator*= (vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vfloat8	simd::operator/ (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE const vfloat8 &	simd::operator/= (vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator== (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator!= (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator< (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator> (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator>= (const vfloat8 &a, const vfloat8 &b)
OIIO_FORCEINLINE vbool8	simd::operator<= (const vfloat8 &a, const vfloat8 &b)
std::ostream &	simd::operator<< (std::ostream &cout, const vfloat16 &val)
OIIO_FORCEINLINE vfloat16	simd::operator+ (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE const vfloat16 &	simd::operator+= (vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vfloat16	simd::operator- (const vfloat16 &a)
OIIO_FORCEINLINE vfloat16	simd::operator- (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE const vfloat16 &	simd::operator-= (vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vfloat16	simd::operator* (const vfloat16 &a, float b)
OIIO_FORCEINLINE vfloat16	simd::operator* (float a, const vfloat16 &b)
OIIO_FORCEINLINE vfloat16	simd::operator* (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE const vfloat16 &	simd::operator*= (vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vfloat16	simd::operator/ (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE const vfloat16 &	simd::operator/= (vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator== (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator!= (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator< (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator> (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator>= (const vfloat16 &a, const vfloat16 &b)
OIIO_FORCEINLINE vbool16	simd::operator<= (const vfloat16 &a, const vfloat16 &b)
template<int i0, int i1, int i2, int i3>
vfloat16	simd::shuffle4 (const vfloat16 &a)
	Shuffle groups of 4. More...
template<int i>
vfloat16	simd::shuffle4 (const vfloat16 &a)
	Shuffle groups of 4. More...
template<int i0, int i1, int i2, int i3>
vfloat16	simd::shuffle (const vfloat16 &a)
	Shuffle within each group of 4. More...
OIIO_FORCEINLINE vfloat16	simd::blend (const vfloat16 &a, const vfloat16 &b, const vbool16 &mask)
OIIO_FORCEINLINE vfloat16	simd::blend0 (const vfloat16 &a, const vbool16 &mask)
OIIO_FORCEINLINE vfloat16	simd::blend0not (const vfloat16 &a, const vbool16 &mask)

Detailed Description

Classes for SIMD processing.

Nice references for all the Intel intrinsics (SSE*, AVX*, etc.): https://software.intel.com/sites/landingpage/IntrinsicsGuide/

Similar guide for ARM intrinsics: https://developer.arm.com/architectures/instruction-sets/simd-isas/neon/intrinsics

It helped me a lot to peruse the source of these packages: Syrah: https://github.com/boulos/syrah Embree: https://github.com/embree Vectorial: https://github.com/scoopr/vectorial

To find out which CPU features you have: Linux: cat /proc/cpuinfo OSX: sysctl machdep.cpu.features

Additional web resources: http://www.codersnotes.com/notes/maths-lib-2016/

Definition in file simd.h.

Macro Definition Documentation

#define OIIO_AVX512BW_ENABLED   0

Definition at line 179 of file simd.h.

#define OIIO_AVX512CD_ENABLED   0

Definition at line 178 of file simd.h.

#define OIIO_AVX512DQ_ENABLED   0

Definition at line 175 of file simd.h.

#define OIIO_AVX512ER_ENABLED   0

Definition at line 177 of file simd.h.

#define OIIO_AVX512IFMA_ENABLED   0

Definition at line 190 of file simd.h.

#define OIIO_AVX512PF_ENABLED   0

Definition at line 176 of file simd.h.

#define OIIO_AVX512VL_ENABLED   0

Definition at line 174 of file simd.h.

#define OIIO_F16C_ENABLED   0

Definition at line 196 of file simd.h.

#define OIIO_FMA_ENABLED   0

Definition at line 185 of file simd.h.

#define OIIO_SIMD   0

Definition at line 213 of file simd.h.

#define OIIO_SIMD16_ALIGN   OIIO_SIMD8_ALIGN

Definition at line 222 of file simd.h.

#define OIIO_SIMD4_ALIGN

Definition at line 214 of file simd.h.

#define OIIO_SIMD8_ALIGN   OIIO_SIMD4_ALIGN

Definition at line 219 of file simd.h.

#define OIIO_SIMD_AVX   0

Definition at line 173 of file simd.h.

#define OIIO_SIMD_FLOAT4_CONST	(		name,
			val
	)		static const OIIO_SIMD4_ALIGN float name[4] = { (val), (val), (val), (val) }

Definition at line 365 of file simd.h.

#define OIIO_SIMD_FLOAT4_CONST4	(		name,
			v0,
			v1,
			v2,
			v3
	)		static const OIIO_SIMD4_ALIGN float name[4] = { (v0), (v1), (v2), (v3) }

Definition at line 367 of file simd.h.

#define OIIO_SIMD_FLOAT8_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD8_ALIGN float name[8] = { (val), (val), (val), (val), \
                                                    (val), (val), (val), (val) }

Definition at line 378 of file simd.h.

#define OIIO_SIMD_FLOAT8_CONST8	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7
	)

Value:

static const OIIO_SIMD8_ALIGN float name[8] = { (v0), (v1), (v2), (v3), \
                                                    (v4), (v5), (v6), (v7) }

Definition at line 381 of file simd.h.

#define OIIO_SIMD_HAS_FLOAT8   1 /* DEPRECATED(1.8) */

Definition at line 232 of file simd.h.

#define OIIO_SIMD_HAS_MATRIX4   1 /* matrix44 defined */

Definition at line 231 of file simd.h.

#define OIIO_SIMD_HAS_SIMD16   1 /* vfloat16, vint16, vbool16 defined */

Definition at line 234 of file simd.h.

#define OIIO_SIMD_HAS_SIMD8   1 /* vfloat8, vint8, vbool8 defined */

Definition at line 233 of file simd.h.

#define OIIO_SIMD_INT16_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD16_ALIGN int name[16] = {             \
        (val), (val), (val), (val), (val), (val), (val), (val), \
        (val), (val), (val), (val), (val), (val), (val), (val) }

Definition at line 405 of file simd.h.

#define OIIO_SIMD_INT16_CONST16	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7,
			v8,
			v9,
			v10,
			v11,
			v12,
			v13,
			v14,
			v15
	)

Value:

static const OIIO_SIMD16_ALIGN int name[16] = { \
        (v0), (v1), (v2), (v3), (v4), (v5), (v6), (v7), \
        (v8), (v9), (v10), (v11), (v12), (v13), (v14), (v15) }

Definition at line 409 of file simd.h.

#define OIIO_SIMD_INT4_CONST	(		name,
			val
	)		static const OIIO_SIMD4_ALIGN int name[4] = { (val), (val), (val), (val) }

Definition at line 369 of file simd.h.

#define OIIO_SIMD_INT4_CONST4	(		name,
			v0,
			v1,
			v2,
			v3
	)		static const OIIO_SIMD4_ALIGN int name[4] = { (v0), (v1), (v2), (v3) }

Definition at line 371 of file simd.h.

#define OIIO_SIMD_INT8_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD8_ALIGN int name[8] = { (val), (val), (val), (val), \
                                                  (val), (val), (val), (val) }

Definition at line 384 of file simd.h.

#define OIIO_SIMD_INT8_CONST8	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7
	)

Value:

static const OIIO_SIMD8_ALIGN int name[8] = { (v0), (v1), (v2), (v3), \
                                                  (v4), (v5), (v6), (v7) }

Definition at line 387 of file simd.h.

#define OIIO_SIMD_MAX_SIZE_BYTES   16

Definition at line 215 of file simd.h.

#define OIIO_SIMD_NEON   0

Definition at line 208 of file simd.h.

#define OIIO_SIMD_SSE   0

Definition at line 115 of file simd.h.

#define OIIO_SIMD_UINT16_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD16_ALIGN uint32_t name[16] = {        \
        (val), (val), (val), (val), (val), (val), (val), (val), \
        (val), (val), (val), (val), (val), (val), (val), (val) }

Definition at line 413 of file simd.h.

#define OIIO_SIMD_UINT16_CONST16	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7,
			v8,
			v9,
			v10,
			v11,
			v12,
			v13,
			v14,
			v15
	)

Value:

static const OIIO_SIMD16_ALIGN uint32_t name[16] = {        \
        (val), (val), (val), (val), (val), (val), (val), (val), \
        (val), (val), (val), (val), (val), (val), (val), (val) }

Definition at line 417 of file simd.h.

#define OIIO_SIMD_UINT4_CONST	(		name,
			val
	)		static const OIIO_SIMD4_ALIGN uint32_t name[4] = { (val), (val), (val), (val) }

Definition at line 373 of file simd.h.

#define OIIO_SIMD_UINT4_CONST4	(		name,
			v0,
			v1,
			v2,
			v3
	)		static const OIIO_SIMD4_ALIGN uint32_t name[4] = { (v0), (v1), (v2), (v3) }

Definition at line 375 of file simd.h.

#define OIIO_SIMD_UINT8_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD8_ALIGN uint32_t name[8] = { (val), (val), (val), (val), \
                                                       (val), (val), (val), (val) }

Definition at line 390 of file simd.h.

#define OIIO_SIMD_UINT8_CONST8	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7
	)

Value:

static const OIIO_SIMD8_ALIGN uint32_t name[8] = { (v0), (v1), (v2), (v3), \
                                                       (v4), (v5), (v6), (v7) }

Definition at line 393 of file simd.h.

#define OIIO_SIMD_VFLOAT16_CONST	(		name,
			val
	)

Value:

static const OIIO_SIMD16_ALIGN float name[16] = {           \
        (val), (val), (val), (val), (val), (val), (val), (val), \
        (val), (val), (val), (val), (val), (val), (val), (val) }

Definition at line 397 of file simd.h.

#define OIIO_SIMD_VFLOAT16_CONST16	(		name,
			v0,
			v1,
			v2,
			v3,
			v4,
			v5,
			v6,
			v7,
			v8,
			v9,
			v10,
			v11,
			v12,
			v13,
			v14,
			v15
	)

Value:

static const OIIO_SIMD16_ALIGN float name[16] = {           \
        (v0), (v1), (v2), (v3), (v4), (v5), (v6), (v7),         \
        (v8), (v9), (v10), (v11), (v12), (v13), (v14), (v15) }

Definition at line 401 of file simd.h.

#define SIMD_CONSTRUCT

(

x

)

for (int i = 0; i < elements; ++i) m_val[i] = (x)

Definition at line 428 of file simd.h.

#define SIMD_CONSTRUCT_PAD

(

x

)

Value:

for (int i = 0; i < elements; ++i) m_val[i] = (x); \
                              for (int i = elements; i < paddedelements; ++i) m_val[i] = 0

Definition at line 429 of file simd.h.

#define SIMD_DO

(

x

)

for (int i = 0; i < elements; ++i) x

Definition at line 427 of file simd.h.

#define SIMD_RETURN	(		T,
			x
	)		T r; for (int i = 0; i < r.elements; ++i) r[i] = (x); return r

Definition at line 431 of file simd.h.

#define SIMD_RETURN_REDUCE	(		T,
			init,
			op
	)		T r = init; for (int i = 0; i < v.elements; ++i) op; return r

Definition at line 432 of file simd.h.