1 /* 2 BLAKE2 reference source code package - optimized C implementations 3 4 Copyright 2012, Samuel Neves <sneves (at) dei.uc.pt>. You may use this under the 5 terms of the CC0, the OpenSSL Licence, or the Apache Public License 2.0, at 6 your option. The terms of these licenses can be found at: 7 8 - CC0 1.0 Universal : http://creativecommons.org/publicdomain/zero/1.0 9 - OpenSSL license : https://www.openssl.org/source/license.html 10 - Apache 2.0 : http://www.apache.org/licenses/LICENSE-2.0 11 12 More information about the BLAKE2 hash function can be found at 13 https://blake2.net. 14 */ 15 #pragma once 16 #ifndef __BLAKE2S_ROUND_H__ 17 #define __BLAKE2S_ROUND_H__ 18 19 #define LOADU(p) _mm_loadu_si128( (const __m128i *)(p) ) 20 #define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r) 21 22 #define TOF(reg) _mm_castsi128_ps((reg)) 23 #define TOI(reg) _mm_castps_si128((reg)) 24 25 #define LIKELY(x) __builtin_expect((x),1) 26 27 28 /* Microarchitecture-specific macros */ 29 #ifndef HAVE_XOP 30 #ifdef HAVE_SSSE3 31 #define _mm_roti_epi32(r, c) ( \ 32 (8==-(c)) ? _mm_shuffle_epi8(r,r8) \ 33 : (16==-(c)) ? _mm_shuffle_epi8(r,r16) \ 34 : _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) )) ) 35 #else 36 #define _mm_roti_epi32(r, c) _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) )) 37 #endif 38 #else 39 /* ... */ 40 #endif 41 42 43 #define G1(row1,row2,row3,row4,buf) \ 44 row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \ 45 row4 = _mm_xor_si128( row4, row1 ); \ 46 row4 = _mm_roti_epi32(row4, -16); \ 47 row3 = _mm_add_epi32( row3, row4 ); \ 48 row2 = _mm_xor_si128( row2, row3 ); \ 49 row2 = _mm_roti_epi32(row2, -12); 50 51 #define G2(row1,row2,row3,row4,buf) \ 52 row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \ 53 row4 = _mm_xor_si128( row4, row1 ); \ 54 row4 = _mm_roti_epi32(row4, -8); \ 55 row3 = _mm_add_epi32( row3, row4 ); \ 56 row2 = _mm_xor_si128( row2, row3 ); \ 57 row2 = _mm_roti_epi32(row2, -7); 58 59 #define DIAGONALIZE(row1,row2,row3,row4) \ 60 row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(2,1,0,3) ); \ 61 row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \ 62 row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(0,3,2,1) ); 63 64 #define UNDIAGONALIZE(row1,row2,row3,row4) \ 65 row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(0,3,2,1) ); \ 66 row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \ 67 row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(2,1,0,3) ); 68 69 #if defined(HAVE_XOP) 70 #include "blake2s-load-xop.h" 71 #elif defined(HAVE_SSE41) 72 #include "blake2s-load-sse41.h" 73 #else 74 #include "blake2s-load-sse2.h" 75 #endif 76 77 #define ROUND(r) \ 78 LOAD_MSG_ ##r ##_1(buf1); \ 79 G1(row1,row2,row3,row4,buf1); \ 80 LOAD_MSG_ ##r ##_2(buf2); \ 81 G2(row1,row2,row3,row4,buf2); \ 82 DIAGONALIZE(row1,row2,row3,row4); \ 83 LOAD_MSG_ ##r ##_3(buf3); \ 84 G1(row1,row2,row3,row4,buf3); \ 85 LOAD_MSG_ ##r ##_4(buf4); \ 86 G2(row1,row2,row3,row4,buf4); \ 87 UNDIAGONALIZE(row1,row2,row3,row4); \ 88 89 #endif 90 91