ref: b8024d53e77374d4a5b7868aa17cac3f11c33581
dir: /sse/blake2b-round.h/
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2B_ROUND_H__
#define __BLAKE2B_ROUND_H__
#define LOAD(p) _mm_load_si128( (const __m128i *)(p) )
#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
#define LOADU(p) _mm_loadu_si128( (const __m128i *)(p) )
#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
#define TOF(reg) _mm_castsi128_ps((reg))
#define TOI(reg) _mm_castps_si128((reg))
#define LIKELY(x) __builtin_expect((x),1)
/* Microarchitecture-specific macros */
#ifndef HAVE_XOP
#ifdef HAVE_SSSE3
#define _mm_roti_epi64(x, c) \
(-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1)) \
: (-(c) == 24) ? _mm_shuffle_epi8((x), r24) \
: (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
: (-(c) == 63) ? _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_add_epi64((x), (x))) \
: _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x), 64-(-(c))))
#else
#define _mm_roti_epi64(r, c) _mm_xor_si128(_mm_srli_epi64( (r), -(c) ),_mm_slli_epi64( (r), 64-(-c) ))
#endif
#else
/* ... */
#endif
#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, -32); \
row4h = _mm_roti_epi64(row4h, -32); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, -24); \
row2h = _mm_roti_epi64(row2h, -24); \
#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, -16); \
row4h = _mm_roti_epi64(row4h, -16); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, -63); \
row2h = _mm_roti_epi64(row2h, -63); \
#if defined(HAVE_SSSE3)
#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2h, row2l, 8); \
t1 = _mm_alignr_epi8(row2l, row2h, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4h, row4l, 8); \
t1 = _mm_alignr_epi8(row4l, row4h, 8); \
row4l = t1; \
row4h = t0;
#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2l, row2h, 8); \
t1 = _mm_alignr_epi8(row2h, row2l, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4l, row4h, 8); \
t1 = _mm_alignr_epi8(row4h, row4l, 8); \
row4l = t1; \
row4h = t0;
#else
#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = row4l;\
t1 = row2l;\
row4l = row3l;\
row3l = row3h;\
row3h = row4l;\
row4l = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t0, t0)); \
row4h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row4h, row4h)); \
row2l = _mm_unpackhi_epi64(row2l, _mm_unpacklo_epi64(row2h, row2h)); \
row2h = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(t1, t1))
#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = row3l;\
row3l = row3h;\
row3h = t0;\
t0 = row2l;\
t1 = row4l;\
row2l = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(row2l, row2l)); \
row2h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row2h, row2h)); \
row4l = _mm_unpackhi_epi64(row4l, _mm_unpacklo_epi64(row4h, row4h)); \
row4h = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t1, t1))
#endif
#if defined(HAVE_SSE41)
#include "blake2b-load-sse41.h"
#else
#include "blake2b-load-sse2.h"
#endif
#define ROUND(r) \
LOAD_MSG_ ##r ##_1(b0, b1); \
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
LOAD_MSG_ ##r ##_2(b0, b1); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
LOAD_MSG_ ##r ##_3(b0, b1); \
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
LOAD_MSG_ ##r ##_4(b0, b1); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
#endif