integer_8cpp-source.html

Crypto++是一个非常强大的密码学库,主要是功能全

00807         sum = _mm_add_epi64(a2b1, a0b3);
00808         C[3] = _mm_add_epi64(sum, sum1);
00809 
00810         __m128i a3b1_a1b3 = _mm_mul_epu32(a2301, b2103);
00811         __m128i a2b2 = _mm_unpackhi_epi32(a2b2_a0b0, z);
00812         __m128i a3b1 = _mm_unpackhi_epi32(a3b1_a1b3, z);
00813         __m128i a1b3 = _mm_unpacklo_epi32(a3b1_a1b3, z);
00814         sum = _mm_add_epi64(a2b2, a3b1);
00815         C[4] = _mm_add_epi64(sum, a1b3);
00816 
00817         __m128i a1302 = _mm_shuffle_epi32(a3210, _MM_SHUFFLE(1, 3, 0, 2));
00818         __m128i b1203 = _mm_shuffle_epi32(b3210, _MM_SHUFFLE(1, 2, 0, 3));
00819         __m128i a3b2_a2b3 = _mm_mul_epu32(a1302, b1203);
00820         __m128i a3b2 = _mm_unpackhi_epi32(a3b2_a2b3, z);
00821         __m128i a2b3 = _mm_unpacklo_epi32(a3b2_a2b3, z);
00822         C[5] = _mm_add_epi64(a3b2, a2b3);
00823 }
00824 
00825 <span class="keywordtype">void</span> P4Optimized::Multiply4(word *C, <span class="keyword">const</span> word *A, <span class="keyword">const</span> word *B)
00826 {
00827         __m128i temp[7];
00828         <span class="keyword">const</span> word *w = (word *)temp;
00829         <span class="keyword">const</span> __m64 *mw = (__m64 *)w;
00830 
00831         P4_Mul(temp, (__m128i *)A, (__m128i *)B);
00832 
00833         C[0] = w[0];
00834 
00835         __m64 s1, s2;
00836 
00837         __m64 w1 = _m_from_int(w[1]);
00838         __m64 w4 = mw[2];
00839         __m64 w6 = mw[3];
00840         __m64 w8 = mw[4];
00841         __m64 w10 = mw[5];
00842         __m64 w12 = mw[6];
00843         __m64 w14 = mw[7];
00844         __m64 w16 = mw[8];
00845         __m64 w18 = mw[9];
00846         __m64 w20 = mw[10];
00847         __m64 w22 = mw[11];
00848         __m64 w26 = _m_from_int(w[26]);
00849 
00850         s1 = _mm_add_si64(w1, w4);
00851         C[1] = _m_to_int(s1);
00852         s1 = _m_psrlqi(s1, 32);
00853 
00854         s2 = _mm_add_si64(w6, w8);
00855         s1 = _mm_add_si64(s1, s2);
00856         C[2] = _m_to_int(s1);
00857         s1 = _m_psrlqi(s1, 32);
00858 
00859         s2 = _mm_add_si64(w10, w12);
00860         s1 = _mm_add_si64(s1, s2);
00861         C[3] = _m_to_int(s1);
00862         s1 = _m_psrlqi(s1, 32);
00863 
00864         s2 = _mm_add_si64(w14, w16);
00865         s1 = _mm_add_si64(s1, s2);
00866         C[4] = _m_to_int(s1);
00867         s1 = _m_psrlqi(s1, 32);
00868 
00869         s2 = _mm_add_si64(w18, w20);
00870         s1 = _mm_add_si64(s1, s2);
00871         C[5] = _m_to_int(s1);
00872         s1 = _m_psrlqi(s1, 32);
00873 
00874         s2 = _mm_add_si64(w22, w26);
00875         s1 = _mm_add_si64(s1, s2);
00876         C[6] = _m_to_int(s1);
00877         s1 = _m_psrlqi(s1, 32);
00878 
00879         C[7] = _m_to_int(s1) + w[27];
00880         _mm_empty();
00881 }
00882 
00883 <span class="keywordtype">void</span> P4Optimized::Multiply8(word *C, <span class="keyword">const</span> word *A, <span class="keyword">const</span> word *B)
00884 {
00885         __m128i temp[28];
00886         <span class="keyword">const</span> word *w = (word *)temp;
00887         <span class="keyword">const</span> __m64 *mw = (__m64 *)w;
00888         <span class="keyword">const</span> word *x = (word *)temp+7*4;
00889         <span class="keyword">const</span> __m64 *mx = (__m64 *)x;
00890         <span class="keyword">const</span> word *y = (word *)temp+7*4*2;
00891         <span class="keyword">const</span> __m64 *my = (__m64 *)y;
00892         <span class="keyword">const</span> word *z = (word *)temp+7*4*3;
00893         <span class="keyword">const</span> __m64 *mz = (__m64 *)z;
00894 
00895         P4_Mul(temp, (__m128i *)A, (__m128i *)B);
00896 
00897         P4_Mul(temp+7, (__m128i *)A+1, (__m128i *)B);
00898 
00899         P4_Mul(temp+14, (__m128i *)A, (__m128i *)B+1);
00900 
00901         P4_Mul(temp+21, (__m128i *)A+1, (__m128i *)B+1);
00902 
00903         C[0] = w[0];
00904 
00905         __m64 s1, s2, s3, s4;
00906 
00907         __m64 w1 = _m_from_int(w[1]);
00908         __m64 w4 = mw[2];
00909         __m64 w6 = mw[3];
00910         __m64 w8 = mw[4];
00911         __m64 w10 = mw[5];
00912         __m64 w12 = mw[6];
00913         __m64 w14 = mw[7];
00914         __m64 w16 = mw[8];
00915         __m64 w18 = mw[9];
00916         __m64 w20 = mw[10];
00917         __m64 w22 = mw[11];
00918         __m64 w26 = _m_from_int(w[26]);
00919         __m64 w27 = _m_from_int(w[27]);
00920 
00921         __m64 x0 = _m_from_int(x[0]);
00922         __m64 x1 = _m_from_int(x[1]);
00923         __m64 x4 = mx[2];
00924         __m64 x6 = mx[3];
00925         __m64 x8 = mx[4];
00926         __m64 x10 = mx[5];
00927         __m64 x12 = mx[6];
00928         __m64 x14 = mx[7];
00929         __m64 x16 = mx[8];
00930         __m64 x18 = mx[9];
00931         __m64 x20 = mx[10];
00932         __m64 x22 = mx[11];
00933         __m64 x26 = _m_from_int(x[26]);
00934         __m64 x27 = _m_from_int(x[27]);
00935 
00936         __m64 y0 = _m_from_int(y[0]);
00937         __m64 y1 = _m_from_int(y[1]);
00938         __m64 y4 = my[2];
00939         __m64 y6 = my[3];
00940         __m64 y8 = my[4];
00941         __m64 y10 = my[5];
00942         __m64 y12 = my[6];
00943         __m64 y14 = my[7];
00944         __m64 y16 = my[8];
00945         __m64 y18 = my[9];
00946         __m64 y20 = my[10];
00947         __m64 y22 = my[11];
00948         __m64 y26 = _m_from_int(y[26]);
00949         __m64 y27 = _m_from_int(y[27]);
00950 
00951         __m64 z0 = _m_from_int(z[0]);
00952         __m64 z1 = _m_from_int(z[1]);
00953         __m64 z4 = mz[2];
00954         __m64 z6 = mz[3];
00955         __m64 z8 = mz[4];
00956         __m64 z10 = mz[5];
00957         __m64 z12 = mz[6];
00958         __m64 z14 = mz[7];
00959         __m64 z16 = mz[8];
00960         __m64 z18 = mz[9];
00961         __m64 z20 = mz[10];
00962         __m64 z22 = mz[11];
00963         __m64 z26 = _m_from_int(z[26]);
00964 
00965         s1 = _mm_add_si64(w1, w4);
00966         C[1] = _m_to_int(s1);
00967         s1 = _m_psrlqi(s1, 32);
00968 
00969         s2 = _mm_add_si64(w6, w8);
00970         s1 = _mm_add_si64(s1, s2);
00971         C[2] = _m_to_int(s1);
00972         s1 = _m_psrlqi(s1, 32);
00973 
00974         s2 = _mm_add_si64(w10, w12);
00975         s1 = _mm_add_si64(s1, s2);
00976         C[3] = _m_to_int(s1);
00977         s1 = _m_psrlqi(s1, 32);
00978 
00979         s3 = _mm_add_si64(x0, y0);
00980         s2 = _mm_add_si64(w14, w16);
00981         s1 = _mm_add_si64(s1, s3);
00982         s1 = _mm_add_si64(s1, s2);
00983         C[4] = _m_to_int(s1);
00984         s1 = _m_psrlqi(s1, 32);
00985 
00986         s3 = _mm_add_si64(x1, y1);
00987         s4 = _mm_add_si64(x4, y4);
00988         s1 = _mm_add_si64(s1, w18);
00989         s3 = _mm_add_si64(s3, s4);
00990         s1 = _mm_add_si64(s1, w20);
00991         s1 = _mm_add_si64(s1, s3);
00992         C[5] = _m_to_int(s1);
00993         s1 = _m_psrlqi(s1, 32);
00994 
00995         s3 = _mm_add_si64(x6, y6);
00996         s4 = _mm_add_si64(x8, y8);
00997         s1 = _mm_add_si64(s1, w22);
00998         s3 = _mm_add_si64(s3, s4);
00999         s1 = _mm_add_si64(s1, w26);
01000         s1 = _mm_add_si64(s1, s3);
01001         C[6] = _m_to_int(s1);
01002         s1 = _m_psrlqi(s1, 32);
01003 
01004         s3 = _mm_add_si64(x10, y10);
01005         s4 = _mm_add_si64(x12, y12);
01006         s1 = _mm_add_si64(s1, w27);
01007         s3 = _mm_add_si64(s3, s4);
01008         s1 = _mm_add_si64(s1, s3);
01009         C[7] = _m_to_int(s1);
01010         s1 = _m_psrlqi(s1, 32);
01011 
01012         s3 = _mm_add_si64(x14, y14);
01013         s4 = _mm_add_si64(x16, y16);
01014         s1 = _mm_add_si64(s1, z0);
01015         s3 = _mm_add_si64(s3, s4);
01016         s1 = _mm_add_si64(s1, s3);
01017         C[8] = _m_to_int(s1);
01018         s1 = _m_psrlqi(s1, 32);
01019 
01020         s3 = _mm_add_si64(x18, y18);
01021         s4 = _mm_add_si64(x20, y20);
01022         s1 = _mm_add_si64(s1, z1);
01023         s3 = _mm_add_si64(s3, s4);
01024         s1 = _mm_add_si64(s1, z4);
01025         s1 = _mm_add_si64(s1, s3);
01026         C[9] = _m_to_int(s1);
01027         s1 = _m_psrlqi(s1, 32);
01028 
01029         s3 = _mm_add_si64(x22, y22);
01030         s4 = _mm_add_si64(x26, y26);
01031         s1 = _mm_add_si64(s1, z6);
01032         s3 = _mm_add_si64(s3, s4);
01033         s1 = _mm_add_si64(s1, z8);
01034         s1 = _mm_add_si64(s1, s3);
01035         C[10] = _m_to_int(s1);
01036         s1 = _m_psrlqi(s1, 32);
01037 
01038         s3 = _mm_add_si64(x27, y27);
01039         s1 = _mm_add_si64(s1, z10);
01040         s1 = _mm_add_si64(s1, z12);
01041         s1 = _mm_add_si64(s1, s3);
01042         C[11] = _m_to_int(s1);
01043         s1 = _m_psrlqi(s1, 32);
01044 
01045         s3 = _mm_add_si64(z14, z16);
01046         s1 = _mm_add_si64(s1, s3);
01047         C[12] = _m_to_int(s1);
01048         s1 = _m_psrlqi(s1, 32);
01049 
01050         s3 = _mm_add_si64(z18, z20);
01051         s1 = _mm_add_si64(s1, s3);
01052         C[13] = _m_to_int(s1);
01053         s1 = _m_psrlqi(s1, 32);
01054 
01055         s3 = _mm_add_si64(z22, z26);
01056         s1 = _mm_add_si64(s1, s3);
01057         C[14] = _m_to_int(s1);
01058         s1 = _m_psrlqi(s1, 32);
01059 
01060         C[15] = z[27] + _m_to_int(s1);
01061         _mm_empty();
01062 }
01063 
01064 <span class="keywordtype">void</span> P4Optimized::Multiply8Bottom(word *C, <span class="keyword">const</span> word *A, <span class="keyword">const</span> word *B)
01065 {
01066         __m128i temp[21];
01067         <span class="keyword">const</span> word *w = (word *)temp;
01068         <span class="keyword">const</span> __m64 *mw = (__m64 *)w;
01069         <span class="keyword">const</span> word *x = (word *)temp+7*4;
01070         <span class="keyword">const</span> __m64 *mx = (__m64 *)x;
01071         <span class="keyword">const</span> word *y = (word *)temp+7*4*2;
01072         <span class="keyword">const</span> __m64 *my = (__m64 *)y;
01073 
01074         P4_Mul(temp, (__m128i *)A, (__m128i *)B);
01075 
01076         P4_Mul(temp+7, (__m128i *)A+1, (__m128i *)B);
01077