1*0e33efe4SConrad Meyer /* 2*0e33efe4SConrad Meyer BLAKE2 reference source code package - optimized C implementations 3*0e33efe4SConrad Meyer 4*0e33efe4SConrad Meyer Written in 2012 by Samuel Neves <sneves@dei.uc.pt> 5*0e33efe4SConrad Meyer 6*0e33efe4SConrad Meyer To the extent possible under law, the author(s) have dedicated all copyright 7*0e33efe4SConrad Meyer and related and neighboring rights to this software to the public domain 8*0e33efe4SConrad Meyer worldwide. This software is distributed without any warranty. 9*0e33efe4SConrad Meyer 10*0e33efe4SConrad Meyer You should have received a copy of the CC0 Public Domain Dedication along with 11*0e33efe4SConrad Meyer this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>. 12*0e33efe4SConrad Meyer */ 13*0e33efe4SConrad Meyer #pragma once 14*0e33efe4SConrad Meyer #ifndef __BLAKE2S_ROUND_H__ 15*0e33efe4SConrad Meyer #define __BLAKE2S_ROUND_H__ 16*0e33efe4SConrad Meyer 17*0e33efe4SConrad Meyer #define LOAD(p) _mm_load_si128( (__m128i *)(p) ) 18*0e33efe4SConrad Meyer #define STORE(p,r) _mm_store_si128((__m128i *)(p), r) 19*0e33efe4SConrad Meyer 20*0e33efe4SConrad Meyer #define LOADU(p) _mm_loadu_si128( (__m128i *)(p) ) 21*0e33efe4SConrad Meyer #define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r) 22*0e33efe4SConrad Meyer 23*0e33efe4SConrad Meyer #define TOF(reg) _mm_castsi128_ps((reg)) 24*0e33efe4SConrad Meyer #define TOI(reg) _mm_castps_si128((reg)) 25*0e33efe4SConrad Meyer 26*0e33efe4SConrad Meyer #define LIKELY(x) __builtin_expect((x),1) 27*0e33efe4SConrad Meyer 28*0e33efe4SConrad Meyer 29*0e33efe4SConrad Meyer /* Microarchitecture-specific macros */ 30*0e33efe4SConrad Meyer #ifndef HAVE_XOP 31*0e33efe4SConrad Meyer #ifdef HAVE_SSSE3 32*0e33efe4SConrad Meyer #define _mm_roti_epi32(r, c) ( \ 33*0e33efe4SConrad Meyer (8==-(c)) ? _mm_shuffle_epi8(r,r8) \ 34*0e33efe4SConrad Meyer : (16==-(c)) ? _mm_shuffle_epi8(r,r16) \ 35*0e33efe4SConrad Meyer : _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) )) ) 36*0e33efe4SConrad Meyer #else 37*0e33efe4SConrad Meyer #define _mm_roti_epi32(r, c) _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32( (r), 32-(-(c)) )) 38*0e33efe4SConrad Meyer #endif 39*0e33efe4SConrad Meyer #else 40*0e33efe4SConrad Meyer /* ... */ 41*0e33efe4SConrad Meyer #endif 42*0e33efe4SConrad Meyer 43*0e33efe4SConrad Meyer 44*0e33efe4SConrad Meyer #define G1(row1,row2,row3,row4,buf) \ 45*0e33efe4SConrad Meyer row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \ 46*0e33efe4SConrad Meyer row4 = _mm_xor_si128( row4, row1 ); \ 47*0e33efe4SConrad Meyer row4 = _mm_roti_epi32(row4, -16); \ 48*0e33efe4SConrad Meyer row3 = _mm_add_epi32( row3, row4 ); \ 49*0e33efe4SConrad Meyer row2 = _mm_xor_si128( row2, row3 ); \ 50*0e33efe4SConrad Meyer row2 = _mm_roti_epi32(row2, -12); 51*0e33efe4SConrad Meyer 52*0e33efe4SConrad Meyer #define G2(row1,row2,row3,row4,buf) \ 53*0e33efe4SConrad Meyer row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \ 54*0e33efe4SConrad Meyer row4 = _mm_xor_si128( row4, row1 ); \ 55*0e33efe4SConrad Meyer row4 = _mm_roti_epi32(row4, -8); \ 56*0e33efe4SConrad Meyer row3 = _mm_add_epi32( row3, row4 ); \ 57*0e33efe4SConrad Meyer row2 = _mm_xor_si128( row2, row3 ); \ 58*0e33efe4SConrad Meyer row2 = _mm_roti_epi32(row2, -7); 59*0e33efe4SConrad Meyer 60*0e33efe4SConrad Meyer #define DIAGONALIZE(row1,row2,row3,row4) \ 61*0e33efe4SConrad Meyer row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(2,1,0,3) ); \ 62*0e33efe4SConrad Meyer row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \ 63*0e33efe4SConrad Meyer row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(0,3,2,1) ); 64*0e33efe4SConrad Meyer 65*0e33efe4SConrad Meyer #define UNDIAGONALIZE(row1,row2,row3,row4) \ 66*0e33efe4SConrad Meyer row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(0,3,2,1) ); \ 67*0e33efe4SConrad Meyer row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \ 68*0e33efe4SConrad Meyer row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(2,1,0,3) ); 69*0e33efe4SConrad Meyer 70*0e33efe4SConrad Meyer #if defined(HAVE_XOP) 71*0e33efe4SConrad Meyer #include "blake2s-load-xop.h" 72*0e33efe4SConrad Meyer #elif defined(HAVE_SSE4_1) 73*0e33efe4SConrad Meyer #include "blake2s-load-sse41.h" 74*0e33efe4SConrad Meyer #else 75*0e33efe4SConrad Meyer #include "blake2s-load-sse2.h" 76*0e33efe4SConrad Meyer #endif 77*0e33efe4SConrad Meyer 78*0e33efe4SConrad Meyer #define ROUND(r) \ 79*0e33efe4SConrad Meyer LOAD_MSG_ ##r ##_1(buf1); \ 80*0e33efe4SConrad Meyer G1(row1,row2,row3,row4,buf1); \ 81*0e33efe4SConrad Meyer LOAD_MSG_ ##r ##_2(buf2); \ 82*0e33efe4SConrad Meyer G2(row1,row2,row3,row4,buf2); \ 83*0e33efe4SConrad Meyer DIAGONALIZE(row1,row2,row3,row4); \ 84*0e33efe4SConrad Meyer LOAD_MSG_ ##r ##_3(buf3); \ 85*0e33efe4SConrad Meyer G1(row1,row2,row3,row4,buf3); \ 86*0e33efe4SConrad Meyer LOAD_MSG_ ##r ##_4(buf4); \ 87*0e33efe4SConrad Meyer G2(row1,row2,row3,row4,buf4); \ 88*0e33efe4SConrad Meyer UNDIAGONALIZE(row1,row2,row3,row4); \ 89*0e33efe4SConrad Meyer 90*0e33efe4SConrad Meyer #endif 91*0e33efe4SConrad Meyer 92