blake3_avx512.c File Reference (original) (raw)

#include "[blake3_impl.h](blake3%5F%5Fimpl%5F8h%5Fsource.html)"
#include <immintrin.h>

Macros
#define	_mm_shuffle_ps2(a, b, c)
#define	LO_IMM8 0x88
#define	HI_IMM8 0xdd

Functions
INLINE __m128i	loadu_128 (const uint8_t src[16])
INLINE __m256i	loadu_256 (const uint8_t src[32])
INLINE __m512i	loadu_512 (const uint8_t src[64])
INLINE void	storeu_128 (__m128i src, uint8_t dest[16])
INLINE void	storeu_256 (__m256i src, uint8_t dest[32])
INLINE void	storeu_512 (__m512i src, uint8_t dest[64])
INLINE __m128i	add_128 (__m128i a, __m128i b)
INLINE __m256i	add_256 (__m256i a, __m256i b)
INLINE __m512i	add_512 (__m512i a, __m512i b)
INLINE __m128i	xor_128 (__m128i a, __m128i b)
INLINE __m256i	xor_256 (__m256i a, __m256i b)
INLINE __m512i	xor_512 (__m512i a, __m512i b)
INLINE __m128i	set1_128 (uint32_t x)
INLINE __m256i	set1_256 (uint32_t x)
INLINE __m512i	set1_512 (uint32_t x)
INLINE __m128i	set4 (uint32_t a, uint32_t b, uint32_t c, uint32_t d)
INLINE __m128i	rot16_128 (__m128i x)
INLINE __m256i	rot16_256 (__m256i x)
INLINE __m512i	rot16_512 (__m512i x)
INLINE __m128i	rot12_128 (__m128i x)
INLINE __m256i	rot12_256 (__m256i x)
INLINE __m512i	rot12_512 (__m512i x)
INLINE __m128i	rot8_128 (__m128i x)
INLINE __m256i	rot8_256 (__m256i x)
INLINE __m512i	rot8_512 (__m512i x)
INLINE __m128i	rot7_128 (__m128i x)
INLINE __m256i	rot7_256 (__m256i x)
INLINE __m512i	rot7_512 (__m512i x)
INLINE void	g1 (__m128i row0, __m128i row1, __m128i row2, __m128i row3, __m128i m)
INLINE void	g2 (__m128i row0, __m128i row1, __m128i row2, __m128i row3, __m128i m)
INLINE void	diagonalize (__m128i row0, __m128i row2, __m128i *row3)
INLINE void	undiagonalize (__m128i row0, __m128i row2, __m128i *row3)
INLINE void	compress_pre (__m128i rows[4], const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags)
void	blake3_compress_xof_avx512 (const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags, uint8_t out[64])
void	blake3_compress_in_place_avx512 (uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags)
INLINE void	round_fn4 (__m128i v[16], __m128i m[16], size_t r)
INLINE void	transpose_vecs_128 (__m128i vecs[4])
INLINE void	transpose_msg_vecs4 (const uint8_t const inputs, size_t block_offset, __m128i out[16])
INLINE void	load_counters4 (uint64_t counter, bool increment_counter, __m128i out_lo, __m128i out_hi)
static void	blake3_hash4_avx512 (const uint8_t const inputs, size_t blocks, const uint32_t key[8], uint64_t counter, bool increment_counter, uint8_t flags, uint8_t flags_start, uint8_t flags_end, uint8_t *out)
static void	blake3_xof4_avx512 (const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags, uint8_t out[4 *64])
INLINE void	round_fn8 (__m256i v[16], __m256i m[16], size_t r)
INLINE void	transpose_vecs_256 (__m256i vecs[8])
INLINE void	transpose_msg_vecs8 (const uint8_t const inputs, size_t block_offset, __m256i out[16])
INLINE void	load_counters8 (uint64_t counter, bool increment_counter, __m256i out_lo, __m256i out_hi)
static void	blake3_hash8_avx512 (const uint8_t const inputs, size_t blocks, const uint32_t key[8], uint64_t counter, bool increment_counter, uint8_t flags, uint8_t flags_start, uint8_t flags_end, uint8_t *out)
static void	blake3_xof8_avx512 (const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags, uint8_t out[8 *64])
INLINE void	round_fn16 (__m512i v[16], __m512i m[16], size_t r)
INLINE __m512i	unpack_lo_128 (__m512i a, __m512i b)
INLINE __m512i	unpack_hi_128 (__m512i a, __m512i b)
INLINE void	transpose_vecs_512 (__m512i vecs[16])
INLINE void	transpose_msg_vecs16 (const uint8_t const inputs, size_t block_offset, __m512i out[16])
INLINE void	load_counters16 (uint64_t counter, bool increment_counter, __m512i out_lo, __m512i out_hi)
static void	blake3_hash16_avx512 (const uint8_t const inputs, size_t blocks, const uint32_t key[8], uint64_t counter, bool increment_counter, uint8_t flags, uint8_t flags_start, uint8_t flags_end, uint8_t *out)
static void	blake3_xof16_avx512 (const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags, uint8_t out[16 *64])
INLINE void	hash_one_avx512 (const uint8_t *input, size_t blocks, const uint32_t key[8], uint64_t counter, uint8_t flags, uint8_t flags_start, uint8_t flags_end, uint8_t out[BLAKE3_OUT_LEN])
void	blake3_hash_many_avx512 (const uint8_t const inputs, size_t num_inputs, size_t blocks, const uint32_t key[8], uint64_t counter, bool increment_counter, uint8_t flags, uint8_t flags_start, uint8_t flags_end, uint8_t *out)
void	blake3_xof_many_avx512 (const uint32_t cv[8], const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len, uint64_t counter, uint8_t flags, uint8_t *out, size_t outblocks)

◆ _mm_shuffle_ps2

#define _mm_shuffle_ps2	(	a,
b,
c )

Value:

(_mm_castps_si128( \

_mm_shuffle_ps(_mm_castsi128_ps(a), _mm_castsi128_ps(b), (c))))

Definition at line 5 of file blake3_avx512.c.

Referenced by compress_pre().

◆ HI_IMM8

◆ LO_IMM8

◆ add_128()

INLINE __m128i add_128	(	__m128i	a,
__m128i	b )

◆ add_256()

INLINE __m256i add_256	(	__m256i	a,
__m256i	b )

◆ add_512()

INLINE __m512i add_512	(	__m512i	a,
__m512i	b )

◆ blake3_compress_in_place_avx512()

◆ blake3_compress_xof_avx512()

◆ blake3_hash16_avx512()

◆ blake3_hash4_avx512()

Definition at line 496 of file blake3_avx512.c.

References BLAKE3_BLOCK_LEN, block, blocks, IV, load_counters4(), round_fn4(), set1_128(), storeu_128(), transpose_msg_vecs4(), transpose_vecs_128(), and xor_128().

Referenced by blake3_hash_many_avx512().

◆ blake3_hash8_avx512()

Definition at line 799 of file blake3_avx512.c.

References BLAKE3_BLOCK_LEN, block, blocks, IV, load_counters8(), round_fn8(), set1_256(), storeu_256(), transpose_msg_vecs8(), transpose_vecs_256(), and xor_256().

Referenced by blake3_hash_many_avx512().

◆ blake3_hash_many_avx512()

◆ blake3_xof16_avx512()

◆ blake3_xof4_avx512()

◆ blake3_xof8_avx512()

◆ blake3_xof_many_avx512()

◆ compress_pre()

Definition at line 120 of file blake3_avx512.c.

References _mm_shuffle_ps2, BLAKE3_BLOCK_LEN, block, counter_high(), counter_low(), diagonalize(), g1(), g2(), INLINE, IV, loadu_128(), set4(), and undiagonalize().

Referenced by blake3_compress_in_place_avx512(), and blake3_compress_xof_avx512().

◆ diagonalize()

INLINE void diagonalize	(	__m128i *	row0,
__m128i *	row2,
__m128i *	row3 )

◆ g1()

INLINE void g1	(	__m128i *	row0,
__m128i *	row1,
__m128i *	row2,
__m128i *	row3,
__m128i	m )

◆ g2()

INLINE void g2	(	__m128i *	row0,
__m128i *	row1,
__m128i *	row2,
__m128i *	row3,
__m128i	m )

◆ hash_one_avx512()

◆ load_counters16()

INLINE void load_counters16	(	uint64_t	counter,
bool	increment_counter,
__m512i *	out_lo,
__m512i *	out_hi )

◆ load_counters4()

INLINE void load_counters4	(	uint64_t	counter,
bool	increment_counter,
__m128i *	out_lo,
__m128i *	out_hi )

◆ load_counters8()

INLINE void load_counters8	(	uint64_t	counter,
bool	increment_counter,
__m256i *	out_lo,
__m256i *	out_hi )

◆ loadu_128()

◆ loadu_256()

◆ loadu_512()

◆ rot12_128()

INLINE __m128i rot12_128	(	__m128i	x	)

◆ rot12_256()

INLINE __m256i rot12_256	(	__m256i	x	)

◆ rot12_512()

INLINE __m512i rot12_512	(	__m512i	x	)

◆ rot16_128()

INLINE __m128i rot16_128	(	__m128i	x	)

◆ rot16_256()

INLINE __m256i rot16_256	(	__m256i	x	)

◆ rot16_512()

INLINE __m512i rot16_512	(	__m512i	x	)

◆ rot7_128()

INLINE __m128i rot7_128	(	__m128i	x	)

◆ rot7_256()

INLINE __m256i rot7_256	(	__m256i	x	)

◆ rot7_512()

INLINE __m512i rot7_512	(	__m512i	x	)

◆ rot8_128()

INLINE __m128i rot8_128	(	__m128i	x	)

◆ rot8_256()

INLINE __m256i rot8_256	(	__m256i	x	)

◆ rot8_512()

INLINE __m512i rot8_512	(	__m512i	x	)

◆ round_fn16()

INLINE void round_fn16	(	__m512i	_v_[16],
__m512i	_m_[16],
size_t	r )

◆ round_fn4()

INLINE void round_fn4	(	__m128i	_v_[16],
__m128i	_m_[16],
size_t	r )

◆ round_fn8()

INLINE void round_fn8	(	__m256i	_v_[16],
__m256i	_m_[16],
size_t	r )

◆ set1_128()

◆ set1_256()

◆ set1_512()

◆ set4()

◆ storeu_128()

◆ storeu_256()

◆ storeu_512()

◆ transpose_msg_vecs16()

◆ transpose_msg_vecs4()

◆ transpose_msg_vecs8()

◆ transpose_vecs_128()

INLINE void transpose_vecs_128	(	__m128i	_vecs_[4]	)

◆ transpose_vecs_256()

INLINE void transpose_vecs_256	(	__m256i	_vecs_[8]	)

◆ transpose_vecs_512()

INLINE void transpose_vecs_512	(	__m512i	_vecs_[16]	)

◆ undiagonalize()

INLINE void undiagonalize	(	__m128i *	row0,
__m128i *	row2,
__m128i *	row3 )

◆ unpack_hi_128()

INLINE __m512i unpack_hi_128	(	__m512i	a,
__m512i	b )

◆ unpack_lo_128()

INLINE __m512i unpack_lo_128	(	__m512i	a,
__m512i	b )

◆ xor_128()

INLINE __m128i xor_128	(	__m128i	a,
__m128i	b )

◆ xor_256()

INLINE __m256i xor_256	(	__m256i	a,
__m256i	b )

◆ xor_512()

INLINE __m512i xor_512	(	__m512i	a,
__m512i	b )