| 1 |
// |
| 2 |
// arch/arm/rsp/rsp.h |
| 3 |
// |
| 4 |
// Extern declarations for host RSP functions. |
| 5 |
// |
| 6 |
// This file is subject to the terms and conditions defined in |
| 7 |
// 'LICENSE', which is part of this source code package. |
| 8 |
// |
| 9 |
|
| 10 |
#ifndef __arch_rsp_h__ |
| 11 |
#define __arch_rsp_h__ |
| 12 |
#include "common.h" |
| 13 |
#include <arm_neon.h> |
| 14 |
|
| 15 |
typedef uint16x8_t rsp_vect_t; |
| 16 |
|
| 17 |
// Loads and shuffles a 16x8 vector according to element. |
| 18 |
extern const uint8_t shuffle_keys[16][16]; |
| 19 |
|
| 20 |
static inline uint16x8_t rsp_vect_load_and_shuffle_operand( |
| 21 |
const uint8_t *src, unsigned element) { |
| 22 |
const uint8x16_t keys = vld1q_u8(shuffle_keys[element]); |
| 23 |
const uint8x8x2_t packed_operand = vld2_u8(src); |
| 24 |
|
| 25 |
uint8x8_t operand_lo_8x8 = vtbl2_u8(packed_operand, vget_low_u8(keys)); |
| 26 |
uint8x8_t operand_hi_8x8 = vtbl2_u8(packed_operand, vget_high_u8(keys)); |
| 27 |
uint8x16_t operand = vcombine_u8(operand_lo_8x8, operand_hi_8x8); |
| 28 |
|
| 29 |
return vreinterpretq_u16_u8(operand); |
| 30 |
} |
| 31 |
|
| 32 |
// Loads a vector without shuffling its elements. |
| 33 |
static inline uint16x8_t rsp_vect_load_unshuffled_operand(const uint8_t *src) { |
| 34 |
return vreinterpretq_u16_u8(vld1q_u8(src)); |
| 35 |
} |
| 36 |
|
| 37 |
// Writes an operand back to memory. |
| 38 |
static inline void rsp_vect_write_operand(uint8_t *dest, uint16x8_t src) { |
| 39 |
vst1q_u8(dest, vreinterpretq_u8_u16(src)); |
| 40 |
} |
| 41 |
|
| 42 |
#include "arch/arm/rsp/vand.h" |
| 43 |
#include "arch/arm/rsp/vnand.h" |
| 44 |
#include "arch/arm/rsp/vnor.h" |
| 45 |
#include "arch/arm/rsp/vor.h" |
| 46 |
#include "arch/arm/rsp/vnxor.h" |
| 47 |
#include "arch/arm/rsp/vxor.h" |
| 48 |
|
| 49 |
#endif |
| 50 |
|