salsa_8cpp-source.html

著名的密码库Crypto++的文档 C++语言的杰作。程序员必备。

<a name="l00102"></a>00102                         ss[0] = _mm_shuffle_epi32(s[0], _MM_SHUFFLE(0, 0, 0, 0));
<a name="l00103"></a>00103                         ss[1] = _mm_shuffle_epi32(s[0], _MM_SHUFFLE(1, 1, 1, 1));
<a name="l00104"></a>00104                         ss[2] = _mm_shuffle_epi32(s[0], _MM_SHUFFLE(2, 2, 2, 2));
<a name="l00105"></a>00105                         ss[3] = _mm_shuffle_epi32(s[0], _MM_SHUFFLE(3, 3, 3, 3));
<a name="l00106"></a>00106                         ss[4] = _mm_shuffle_epi32(s[1], _MM_SHUFFLE(0, 0, 0, 0));
<a name="l00107"></a>00107                         ss[6] = _mm_shuffle_epi32(s[1], _MM_SHUFFLE(2, 2, 2, 2));
<a name="l00108"></a>00108                         ss[7] = _mm_shuffle_epi32(s[1], _MM_SHUFFLE(3, 3, 3, 3));
<a name="l00109"></a>00109                         ss[9] = _mm_shuffle_epi32(s[2], _MM_SHUFFLE(1, 1, 1, 1));
<a name="l00110"></a>00110                         ss[10] = _mm_shuffle_epi32(s[2], _MM_SHUFFLE(2, 2, 2, 2));
<a name="l00111"></a>00111                         ss[11] = _mm_shuffle_epi32(s[2], _MM_SHUFFLE(3, 3, 3, 3));
<a name="l00112"></a>00112                         ss[12] = _mm_shuffle_epi32(s[3], _MM_SHUFFLE(0, 0, 0, 0));
<a name="l00113"></a>00113                         ss[13] = _mm_shuffle_epi32(s[3], _MM_SHUFFLE(1, 1, 1, 1));
<a name="l00114"></a>00114                         ss[14] = _mm_shuffle_epi32(s[3], _MM_SHUFFLE(2, 2, 2, 2));
<a name="l00115"></a>00115                         ss[15] = _mm_shuffle_epi32(s[3], _MM_SHUFFLE(3, 3, 3, 3));
<a name="l00116"></a>00116 
<a name="l00117"></a>00117                         <span class="keywordflow">do</span>
<a name="l00118"></a>00118                         {
<a name="l00119"></a>00119                                 word32 *countersLo = (word32*)&amp;(ss[8]), *countersHi = (word32*)&amp;(ss[5]);
<a name="l00120"></a>00120                                 <span class="keywordflow">for</span> (i=0; i&lt;4; i++)
<a name="l00121"></a>00121                                 {
<a name="l00122"></a>00122                                         countersLo[i] = m_state[8];
<a name="l00123"></a>00123                                         countersHi[i] = m_state[5];
<a name="l00124"></a>00124                                         <span class="keywordflow">if</span> (++m_state[8] == 0)
<a name="l00125"></a>00125                                                 ++m_state[5];
<a name="l00126"></a>00126                                 }
<a name="l00127"></a>00127 
<a name="l00128"></a>00128                                 __m128i x0 = ss[0];
<a name="l00129"></a>00129                                 __m128i x1 = ss[1];
<a name="l00130"></a>00130                                 __m128i x2 = ss[2];
<a name="l00131"></a>00131                                 __m128i x3 = ss[3];
<a name="l00132"></a>00132                                 __m128i x4 = ss[4];
<a name="l00133"></a>00133                                 __m128i x5 = ss[5];
<a name="l00134"></a>00134                                 __m128i x6 = ss[6];
<a name="l00135"></a>00135                                 __m128i x7 = ss[7];
<a name="l00136"></a>00136                                 __m128i x8 = ss[8];
<a name="l00137"></a>00137                                 __m128i x9 = ss[9];
<a name="l00138"></a>00138                                 __m128i x10 = ss[10];
<a name="l00139"></a>00139                                 __m128i x11 = ss[11];
<a name="l00140"></a>00140                                 __m128i x12 = ss[12];
<a name="l00141"></a>00141                                 __m128i x13 = ss[13];
<a name="l00142"></a>00142                                 __m128i x14 = ss[14];
<a name="l00143"></a>00143                                 __m128i x15 = ss[15];
<a name="l00144"></a>00144 
<a name="l00145"></a>00145                                 <span class="keywordflow">for</span> (i=m_rounds; i&gt;0; i-=2)
<a name="l00146"></a>00146                                 {
<a name="l00147"></a>00147                                         #define QUARTER_ROUND(a, b, c, d)       \
<a name="l00148"></a>00148                                                 SSE2_QUARTER_ROUND(a, b, d, 7)  \
<a name="l00149"></a>00149                                                 SSE2_QUARTER_ROUND(b, c, a, 9)  \
<a name="l00150"></a>00150                                                 SSE2_QUARTER_ROUND(c, d, b, 13) \
<a name="l00151"></a>00151                                                 SSE2_QUARTER_ROUND(d, a, c, 18) 
<a name="l00152"></a>00152 
<a name="l00153"></a>00153                                         QUARTER_ROUND(x0, x4, x8, x12)
<a name="l00154"></a>00154                                         QUARTER_ROUND(x1, x5, x9, x13)
<a name="l00155"></a>00155                                         QUARTER_ROUND(x2, x6, x10, x14)
<a name="l00156"></a>00156                                         QUARTER_ROUND(x3, x7, x11, x15)
<a name="l00157"></a>00157 
<a name="l00158"></a>00158                                         QUARTER_ROUND(x0, x13, x10, x7)
<a name="l00159"></a>00159                                         QUARTER_ROUND(x1, x14, x11, x4)
<a name="l00160"></a>00160                                         QUARTER_ROUND(x2, x15, x8, x5)
<a name="l00161"></a>00161                                         QUARTER_ROUND(x3, x12, x9, x6)
<a name="l00162"></a>00162 
<a name="l00163"></a>00163                                         #undef QUARTER_ROUND
<a name="l00164"></a>00164                                 }
<a name="l00165"></a>00165 
<a name="l00166"></a>00166                                 x0 = _mm_add_epi32(x0, ss[0]);
<a name="l00167"></a>00167                                 x1 = _mm_add_epi32(x1, ss[1]);
<a name="l00168"></a>00168                                 x2 = _mm_add_epi32(x2, ss[2]);
<a name="l00169"></a>00169                                 x3 = _mm_add_epi32(x3, ss[3]);
<a name="l00170"></a>00170                                 x4 = _mm_add_epi32(x4, ss[4]);
<a name="l00171"></a>00171                                 x5 = _mm_add_epi32(x5, ss[5]);
<a name="l00172"></a>00172                                 x6 = _mm_add_epi32(x6, ss[6]);
<a name="l00173"></a>00173                                 x7 = _mm_add_epi32(x7, ss[7]);
<a name="l00174"></a>00174                                 x8 = _mm_add_epi32(x8, ss[8]);
<a name="l00175"></a>00175                                 x9 = _mm_add_epi32(x9, ss[9]);
<a name="l00176"></a>00176                                 x10 = _mm_add_epi32(x10, ss[10]);
<a name="l00177"></a>00177                                 x11 = _mm_add_epi32(x11, ss[11]);
<a name="l00178"></a>00178                                 x12 = _mm_add_epi32(x12, ss[12]);
<a name="l00179"></a>00179                                 x13 = _mm_add_epi32(x13, ss[13]);
<a name="l00180"></a>00180                                 x14 = _mm_add_epi32(x14, ss[14]);
<a name="l00181"></a>00181                                 x15 = _mm_add_epi32(x15, ss[15]);
<a name="l00182"></a>00182 
<a name="l00183"></a>00183 <span class="preprocessor">                                #define OUTPUT_4(x, a, b, c, d, e, f, g, h)     {\</span>
<a name="l00184"></a>00184 <span class="preprocessor">                                        __m128i t0 = _mm_unpacklo_epi32(a, b);\</span>
<a name="l00185"></a>00185 <span class="preprocessor">                                        __m128i t1 = _mm_unpacklo_epi32(c, d);\</span>
<a name="l00186"></a>00186 <span class="preprocessor">                                        __m128i t2 = _mm_unpacklo_epi64(t0, t1);\</span>
<a name="l00187"></a>00187 <span class="preprocessor">                                        CRYPTOPP_KEYSTREAM_OUTPUT_XMM(x, e, t2)\</span>
<a name="l00188"></a>00188 <span class="preprocessor">                                        t2 = _mm_unpackhi_epi64(t0, t1);\</span>
<a name="l00189"></a>00189 <span class="preprocessor">                                        CRYPTOPP_KEYSTREAM_OUTPUT_XMM(x, f, t2)\</span>
<a name="l00190"></a>00190 <span class="preprocessor">                                        t0 = _mm_unpackhi_epi32(a, b);\</span>
<a name="l00191"></a>00191 <span class="preprocessor">                                        t1 = _mm_unpackhi_epi32(c, d);\</span>
<a name="l00192"></a>00192 <span class="preprocessor">                                        t2 = _mm_unpacklo_epi64(t0, t1);\</span>
<a name="l00193"></a>00193 <span class="preprocessor">                                        CRYPTOPP_KEYSTREAM_OUTPUT_XMM(x, g, t2)\</span>
<a name="l00194"></a>00194 <span class="preprocessor">                                        t2 = _mm_unpackhi_epi64(t0, t1);\</span>
<a name="l00195"></a>00195 <span class="preprocessor">                                        CRYPTOPP_KEYSTREAM_OUTPUT_XMM(x, h, t2)}</span>
<a name="l00196"></a>00196 <span class="preprocessor"></span>
<a name="l00197"></a>00197 <span class="preprocessor">                                #define SALSA_OUTPUT(x)         \</span>
<a name="l00198"></a>00198 <span class="preprocessor">                                        OUTPUT_4(x, x0, x13, x10, x7, 0, 4, 8, 12)\</span>
<a name="l00199"></a>00199 <span class="preprocessor">                                        OUTPUT_4(x, x4, x1, x14, x11, 1, 5, 9, 13)\</span>
<a name="l00200"></a>00200 <span class="preprocessor">                                        OUTPUT_4(x, x8, x5, x2, x15, 2, 6, 10, 14)\</span>
<a name="l00201"></a>00201 <span class="preprocessor">                                        OUTPUT_4(x, x12, x9, x6, x3, 3, 7, 11, 15)</span>
<a name="l00202"></a>00202 <span class="preprocessor"></span>
<a name="l00203"></a>00203                                 <a class="code" href="strciphr_8h.html#4f82514b5ab874259ac783ff9b96c4e5">CRYPTOPP_KEYSTREAM_OUTPUT_SWITCH</a>(SALSA_OUTPUT, 4*<a class="code" href="struct_additive_cipher_concrete_policy.html#0c584b68c2f3a8208ff245cd8d09fcd5">BYTES_PER_ITERATION</a>)
<a name="l00204"></a>00204 
<a name="l00205"></a>00205                                 <span class="preprocessor">#undef SALSA_OUTPUT</span>
<a name="l00206"></a>00206 <span class="preprocessor"></span>                        } <span class="keywordflow">while</span> ((iterationCount-=4) &gt;= 4);
<a name="l00207"></a>00207                 }
<a name="l00208"></a>00208 <span class="preprocessor">#endif</span>
<a name="l00209"></a>00209 <span class="preprocessor"></span>
<a name="l00210"></a>00210                 <span class="keywordflow">if</span> (!IsP4()) <span class="keywordflow">while</span> (iterationCount)
<a name="l00211"></a>00211                 {
<a name="l00212"></a>00212                         --iterationCount;
<a name="l00213"></a>00213                         __m128i x0 = s[0];
<a name="l00214"></a>00214                         __m128i x1 = s[1];
<a name="l00215"></a>00215                         __m128i x2 = s[2];
<a name="l00216"></a>00216                         __m128i x3 = s[3];
<a name="l00217"></a>00217 
<a name="l00218"></a>00218                         <span class="keywordflow">for</span> (i=m_rounds; i&gt;0; i-=2)
<a name="l00219"></a>00219                         {
<a name="l00220"></a>00220                                 SSE2_QUARTER_ROUND(x0, x1, x3, 7)
<a name="l00221"></a>00221                                 SSE2_QUARTER_ROUND(x1, x2, x0, 9)
<a name="l00222"></a>00222                                 SSE2_QUARTER_ROUND(x2, x3, x1, 13)
<a name="l00223"></a>00223                                 SSE2_QUARTER_ROUND(x3, x0, x2, 18)