neon.hpp File Reference

Helper class for ARM NEON support. More...

#include <boost/cstdint.hpp>

Include dependency graph for neon.hpp:

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

Go to the source code of this file.

Classes
struct	tvector< T, S >
	Emulates a vector. More...
struct	tmatrix< T, S, D >
	Emulates a matrix. More...

Typedefs
typedef tvector< uint8_t, 8 >	uint8x8_t
typedef tvector< uint16_t, 8 >	uint16x8_t
typedef tmatrix< uint8_t, 8, 4 >	uint8x8x4_t

Functions
template<class Td , unsigned S>
tvector< Td, S >	vdup_n (Td imm)
uint16x8_t	vdupq_n_u16 (uint16_t imm)
uint8x8_t	vdup_n_u8 (uint8_t imm)
template<class Td , class Tm , unsigned S>
tvector< Td, S >	vshr (tvector< Tm, S > m, const unsigned imm)
uint8x8_t	vshrn_n_u16 (uint16x8_t m, const unsigned imm)
template<class Td , class Tn , class Tm , unsigned S>
tvector< Td, S >	vadd (tvector< Tn, S > n, tvector< Tm, S > m)
uint16x8_t	vaddq_u16 (uint16x8_t n, uint16x8_t m)
template<class Td , class Ts , unsigned S>
tvector< Td, S >	vmul (tvector< Ts, S > n, tvector< Ts, S > m)
uint16x8_t	vmull_u8 (uint8x8_t n, uint8x8_t m)
template<class Td , unsigned S, unsigned D>
tmatrix< Td, S, D >	vld (Td *base)
uint8x8x4_t	vld4_u8 (uint8_t *base)
template<class Td , unsigned S, unsigned D>
void	vst (Td *base, tmatrix< Td, S, D > list)
void	vst4_u8 (uint8_t *base, uint8x8x4_t list)

Detailed Description

Helper class for ARM NEON support.

When using g++ on an ARM that support the NEON it uses the gcc intrinsics [1], for all other platforms an emulation is used. The emulation is based on the RealView Compilation Tool Assembler Guide (ARM DUI 0204J (ID101213)) [2]. The emulation follows the latter convensions instead of the former. The numbers in the sections refer to the section numers in [2].

Not everything is implemented, only functions used are implemented.

Common template parameters are:

• Td type of the destination.
• Tn type of the first operand.
• Tm type of the second operand.
• Ts type of the source (Tm and Tn).
• S number of vector elements.
• D number of matrix vectors.

[1] http://gcc.gnu.org/onlinedocs/gcc-4.7.0/gcc/ARM-NEON-Intrinsics.html [2] http://infocenter.arm.com/help/topic/com.arm.doc.dui0204i/DUI0204I_rvct_assembler_guide.pdf

Definition in file neon.hpp.

Typedef Documentation

typedef tvector<uint16_t, 8> uint16x8_t

Definition at line 82 of file neon.hpp.

typedef tvector<uint8_t, 8> uint8x8_t

Definition at line 81 of file neon.hpp.

typedef tmatrix<uint8_t, 8, 4> uint8x8x4_t

Definition at line 97 of file neon.hpp.

Function Documentation

template<class Td , class Tn , class Tm , unsigned S>

tvector<Td, S> vadd ( tvector< Tn, S >  n, tvector< Tm, S >  m  )

inline

Definition at line 149 of file neon.hpp.

References d, i, and S.

uint16x8_t vaddq_u16 ( uint16x8_t  n, uint16x8_t  m  )

inline

Definition at line 159 of file neon.hpp.

Referenced by blend_surface().

template<class Td , unsigned S>

tvector<Td, S> vdup_n ( Td  imm )

inline

Definition at line 105 of file neon.hpp.

References d, i, and S.

uint8x8_t vdup_n_u8 ( uint8_t  imm )

inline

Definition at line 121 of file neon.hpp.

Referenced by blend_surface().

uint16x8_t vdupq_n_u16 ( uint16_t  imm )

inline

Definition at line 115 of file neon.hpp.

Referenced by blend_surface().

template<class Td , unsigned S, unsigned D>

tmatrix<Td, S, D> vld ( Td *  base )

inline

Definition at line 188 of file neon.hpp.

References d, i, S, and tmatrix< T, S, D >::val.

uint8x8x4_t vld4_u8 ( uint8_t *  base )

inline

Definition at line 200 of file neon.hpp.

Referenced by blend_surface().

template<class Td , class Ts , unsigned S>

tvector<Td, S> vmul ( tvector< Ts, S >  n, tvector< Ts, S >  m  )

inline

Definition at line 168 of file neon.hpp.

References d, i, and S.

uint16x8_t vmull_u8 ( uint8x8_t  n, uint8x8_t  m  )

inline

Definition at line 179 of file neon.hpp.

Referenced by blend_surface().

template<class Td , class Tm , unsigned S>

tvector<Td, S> vshr ( tvector< Tm, S >  m, const unsigned  imm  )

inline

Definition at line 130 of file neon.hpp.

References d, i, and S.

uint8x8_t vshrn_n_u16 ( uint16x8_t  m, const unsigned  imm  )

inline

Definition at line 140 of file neon.hpp.

Referenced by blend_surface().

template<class Td , unsigned S, unsigned D>

void vst ( Td *  base, tmatrix< Td, S, D >  list  )

inline

Definition at line 207 of file neon.hpp.

References i, S, and tmatrix< T, S, D >::val.

void vst4_u8 ( uint8_t *  base, uint8x8x4_t  list  )

inline

Definition at line 217 of file neon.hpp.

Referenced by blend_surface().