sosemanuk_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="l00262"></a>00262         0x0CD320B2, 0x14DC607F, 0x3CCDA081, 0x24C2E04C,
<a name="l00263"></a>00263         0xAC977218, 0xB49832D5, 0x9C89F22B, 0x8486B2E6,
<a name="l00264"></a>00264         0xCCABDB7E, 0xD4A49BB3, 0xFCB55B4D, 0xE4BA1B80,
<a name="l00265"></a>00265         0x17566887, 0x0F59284A, 0x2748E8B4, 0x3F47A879,
<a name="l00266"></a>00266         0x776AC1E1, 0x6F65812C, 0x477441D2, 0x5F7B011F,
<a name="l00267"></a>00267         0xD72E934B, 0xCF21D386, 0xE7301378, 0xFF3F53B5,
<a name="l00268"></a>00268         0xB7123A2D, 0xAF1D7AE0, 0x870CBA1E, 0x9F03FAD3,
<a name="l00269"></a>00269         0x3EA637B6, 0x26A9777B, 0x0EB8B785, 0x16B7F748,
<a name="l00270"></a>00270         0x5E9A9ED0, 0x4695DE1D, 0x6E841EE3, 0x768B5E2E,
<a name="l00271"></a>00271         0xFEDECC7A, 0xE6D18CB7, 0xCEC04C49, 0xD6CF0C84,
<a name="l00272"></a>00272         0x9EE2651C, 0x86ED25D1, 0xAEFCE52F, 0xB6F3A5E2
<a name="l00273"></a>00273 };
<a name="l00274"></a>00274 
<a name="l00275"></a>00275 
<a name="l00276"></a>00276 <span class="preprocessor">#if CRYPTOPP_BOOL_X86 || CRYPTOPP_BOOL_X64</span>
<a name="l00277"></a>00277 <span class="preprocessor"></span><span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <a class="code" href="struct_additive_cipher_concrete_policy.html#5bdc1cb44b0ddc8df0fb6953aec93602">SosemanukPolicy::GetAlignment</a>()<span class="keyword"> const</span>
<a name="l00278"></a>00278 <span class="keyword"></span>{
<a name="l00279"></a>00279 <span class="preprocessor">#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE</span>
<a name="l00280"></a>00280 <span class="preprocessor"></span><span class="preprocessor">#ifdef __INTEL_COMPILER</span>
<a name="l00281"></a>00281 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2() &amp;&amp; !IsP4())       <span class="comment">// Intel compiler produces faster code for this algorithm on the P4</span>
<a name="l00282"></a>00282 <span class="preprocessor">#else</span>
<a name="l00283"></a>00283 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2())
<a name="l00284"></a>00284 <span class="preprocessor">#endif</span>
<a name="l00285"></a>00285 <span class="preprocessor"></span>                <span class="keywordflow">return</span> 16;
<a name="l00286"></a>00286         <span class="keywordflow">else</span>
<a name="l00287"></a>00287 <span class="preprocessor">#endif</span>
<a name="l00288"></a>00288 <span class="preprocessor"></span>                <span class="keywordflow">return</span> 1;
<a name="l00289"></a>00289 }
<a name="l00290"></a>00290 
<a name="l00291"></a>00291 <span class="keywordtype">unsigned</span> <span class="keywordtype">int</span> <a class="code" href="struct_additive_cipher_abstract_policy.html#32bbafa12b59e77d4d8bc67e9d5a4004">SosemanukPolicy::GetOptimalBlockSize</a>()<span class="keyword"> const</span>
<a name="l00292"></a>00292 <span class="keyword"></span>{
<a name="l00293"></a>00293 <span class="preprocessor">#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE</span>
<a name="l00294"></a>00294 <span class="preprocessor"></span><span class="preprocessor">#ifdef __INTEL_COMPILER</span>
<a name="l00295"></a>00295 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2() &amp;&amp; !IsP4())       <span class="comment">// Intel compiler produces faster code for this algorithm on the P4</span>
<a name="l00296"></a>00296 <span class="preprocessor">#else</span>
<a name="l00297"></a>00297 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2())
<a name="l00298"></a>00298 <span class="preprocessor">#endif</span>
<a name="l00299"></a>00299 <span class="preprocessor"></span>                <span class="keywordflow">return</span> 4*<a class="code" href="struct_additive_cipher_concrete_policy.html#0c584b68c2f3a8208ff245cd8d09fcd5">BYTES_PER_ITERATION</a>;
<a name="l00300"></a>00300         <span class="keywordflow">else</span>
<a name="l00301"></a>00301 <span class="preprocessor">#endif</span>
<a name="l00302"></a>00302 <span class="preprocessor"></span>                <span class="keywordflow">return</span> <a class="code" href="struct_additive_cipher_concrete_policy.html#0c584b68c2f3a8208ff245cd8d09fcd5">BYTES_PER_ITERATION</a>;
<a name="l00303"></a>00303 }
<a name="l00304"></a>00304 <span class="preprocessor">#endif</span>
<a name="l00305"></a>00305 <span class="preprocessor"></span>
<a name="l00306"></a>00306 <span class="preprocessor">#pragma warning(disable: 4731)  // frame pointer register 'ebp' modified by inline assembly code</span>
<a name="l00307"></a>00307 <span class="preprocessor"></span>
<a name="l00308"></a><a class="code" href="class_sosemanuk_policy.html#8417e2849165e287db8f75cb1ddaea7c">00308</a> <span class="keywordtype">void</span> <a class="code" href="class_sosemanuk_policy.html#8417e2849165e287db8f75cb1ddaea7c">SosemanukPolicy::OperateKeystream</a>(<a class="code" href="strciphr_8h.html#b4a226527d2bd01ff19bfa14d0974227">KeystreamOperation</a> operation, byte *output, <span class="keyword">const</span> byte *input, <span class="keywordtype">size_t</span> iterationCount)
<a name="l00309"></a>00309 {
<a name="l00310"></a>00310 <span class="preprocessor">#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE</span>
<a name="l00311"></a>00311 <span class="preprocessor"></span><span class="preprocessor">#ifdef __INTEL_COMPILER</span>
<a name="l00312"></a>00312 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2() &amp;&amp; !IsP4())       <span class="comment">// Intel compiler produces faster code for this algorithm on the P4</span>
<a name="l00313"></a>00313 <span class="preprocessor">#else</span>
<a name="l00314"></a>00314 <span class="preprocessor"></span>        <span class="keywordflow">if</span> (HasSSE2())
<a name="l00315"></a>00315 <span class="preprocessor">#endif</span>
<a name="l00316"></a>00316 <span class="preprocessor"></span>        {
<a name="l00317"></a>00317 <span class="preprocessor">#ifdef __GNUC__</span>
<a name="l00318"></a>00318 <span class="preprocessor"></span>                __asm__ __volatile__
<a name="l00319"></a>00319                 (
<a name="l00320"></a>00320                 <span class="stringliteral">".intel_syntax noprefix;"</span>
<a name="l00321"></a>00321                 AS_PUSH(                bx)
<a name="l00322"></a>00322 #<span class="keywordflow">else</span>
<a name="l00323"></a>00323                 word32 *state = <a class="code" href="class_sosemanuk_policy.html#996c06115c9de685f45a11789d3f3b76">m_state</a>;
<a name="l00324"></a>00324                 AS2(    mov             WORD_REG(ax), state)
<a name="l00325"></a>00325                 AS2(    mov             WORD_REG(di), output)
<a name="l00326"></a>00326                 AS2(    mov             WORD_REG(dx), input)
<a name="l00327"></a>00327                 AS2(    mov             WORD_REG(cx), iterationCount)
<a name="l00328"></a>00328 <span class="preprocessor">#endif</span>
<a name="l00329"></a>00329 <span class="preprocessor"></span>
<a name="l00330"></a>00330 <span class="preprocessor">#define SSE2_output                     WORD_PTR [WORD_REG(sp)+1*WORD_SZ]</span>
<a name="l00331"></a>00331 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_input                      WORD_PTR [WORD_REG(sp)+2*WORD_SZ]</span>
<a name="l00332"></a>00332 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_wordsLeft          WORD_PTR [WORD_REG(sp)+3*WORD_SZ]</span>
<a name="l00333"></a>00333 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_diEnd                      WORD_PTR [WORD_REG(sp)+4*WORD_SZ]</span>
<a name="l00334"></a>00334 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_pMulTables         WORD_PTR [WORD_REG(sp)+5*WORD_SZ]</span>
<a name="l00335"></a>00335 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_state                      WORD_PTR [WORD_REG(sp)+6*WORD_SZ]</span>
<a name="l00336"></a>00336 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_wordsLeft2         WORD_PTR [WORD_REG(sp)+7*WORD_SZ]</span>
<a name="l00337"></a>00337 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_stateCopy          WORD_REG(sp) + 8*WORD_SZ</span>
<a name="l00338"></a>00338 <span class="preprocessor"></span><span class="preprocessor">#define SSE2_uvStart            SSE2_stateCopy + 12*4</span>
<a name="l00339"></a>00339 <span class="preprocessor"></span>
<a name="l00340"></a>00340                 AS_PUSH(                bp)
<a name="l00341"></a>00341                 AS2(    mov             WORD_REG(bx), WORD_REG(sp))
<a name="l00342"></a>00342                 AS2(    and             WORD_REG(sp), -16)
<a name="l00343"></a>00343                 AS2(    sub             WORD_REG(sp), 80*4*2+12*4+8*WORD_SZ)    <span class="comment">// 80 v's, 80 u's, 12 state, 8 locals</span>
<a name="l00344"></a>00344                 AS2(    mov             [WORD_REG(sp)], WORD_REG(bx))
<a name="l00345"></a>00345                 AS2(    mov             SSE2_output, WORD_REG(di))
<a name="l00346"></a>00346                 AS2(    mov             SSE2_input, WORD_REG(dx))
<a name="l00347"></a>00347                 AS2(    mov             SSE2_state, WORD_REG(ax))
<a name="l00348"></a>00348 <span class="preprocessor">#ifndef _MSC_VER</span>
<a name="l00349"></a>00349 <span class="preprocessor"></span>                AS2(    mov             SSE2_pMulTables, WORD_REG(si))
<a name="l00350"></a>00350 <span class="preprocessor">#endif</span>
<a name="l00351"></a>00351 <span class="preprocessor"></span>                AS2(    lea             WORD_REG(cx), [4*WORD_REG(cx)+WORD_REG(cx)])
<a name="l00352"></a>00352                 AS2(    lea             WORD_REG(si), [4*WORD_REG(cx)])
<a name="l00353"></a>00353                 AS2(    mov             SSE2_wordsLeft, WORD_REG(si))
<a name="l00354"></a>00354                 AS2(    movdqa  xmm0, [WORD_REG(ax)+0*16])              <span class="comment">// copy state to stack to save a register</span>
<a name="l00355"></a>00355                 AS2(    movdqa  [SSE2_stateCopy+0*16], xmm0)
<a name="l00356"></a>00356                 AS2(    movdqa  xmm0, [WORD_REG(ax)+1*16])
<a name="l00357"></a>00357                 AS2(    movdqa  [SSE2_stateCopy+1*16], xmm0)
<a name="l00358"></a>00358                 AS2(    movq    xmm0, QWORD PTR [WORD_REG(ax)+2*16])
<a name="l00359"></a>00359                 AS2(    movq    QWORD PTR [SSE2_stateCopy+2*16], xmm0)
<a name="l00360"></a>00360                 AS2(    psrlq   xmm0, 32)
<a name="l00361"></a>00361                 AS2(    movd    ebx, xmm0)                              <span class="comment">// s(9)</span>
<a name="l00362"></a>00362                 AS2(    mov             ecx, [WORD_REG(ax)+10*4])
<a name="l00363"></a>00363                 AS2(    mov             edx, [WORD_REG(ax)+11*4])
<a name="l00364"></a>00364                 AS2(    pcmpeqb xmm7, xmm7)                             <span class="comment">// all ones</span>
<a name="l00365"></a>00365 
<a name="l00366"></a>00366 <span class="preprocessor">#define s(i)    SSE2_stateCopy + ASM_MOD(i,10)*4</span>
<a name="l00367"></a>00367 <span class="preprocessor"></span><span class="preprocessor">#define u(j)    WORD_REG(di) + (ASM_MOD(j,4)*20 + (j/4)) * 4</span>
<a name="l00368"></a>00368 <span class="preprocessor"></span><span class="preprocessor">#define v(j)    WORD_REG(di) + (ASM_MOD(j,4)*20 + (j/4)) * 4 + 80*4</span>
<a name="l00369"></a>00369 <span class="preprocessor"></span>
<a name="l00370"></a>00370 <span class="preprocessor">#define r10 ecx</span>
<a name="l00371"></a>00371 <span class="preprocessor"></span><span class="preprocessor">#define r11 edx</span>
<a name="l00372"></a>00372 <span class="preprocessor"></span><span class="preprocessor">#define r20 edx</span>
<a name="l00373"></a>00373 <span class="preprocessor"></span><span class="preprocessor">#define r21 ecx</span>
<a name="l00374"></a>00374 <span class="preprocessor"></span>
<a name="l00375"></a>00375 <span class="preprocessor">#define SSE2_STEP(i, j) \</span>
<a name="l00376"></a>00376 <span class="preprocessor">        AS2(    mov             eax, [s(i+0)])\</span>
<a name="l00377"></a>00377 <span class="preprocessor">        AS2(    mov             [v(i)], eax)\</span>
<a name="l00378"></a>00378 <span class="preprocessor">        AS2(    rol             eax, 8)\</span>
<a name="l00379"></a>00379 <span class="preprocessor">        AS2(    lea             ebp, [ebx + r2##j])\</span>
<a name="l00380"></a>00380 <span class="preprocessor">        AS2(    xor             ebp, r1##j)\</span>
<a name="l00381"></a>00381 <span class="preprocessor">        AS2(    mov             [u(i)], ebp)\</span>
<a name="l00382"></a>00382 <span class="preprocessor">        AS2(    mov             ebp, 1)\</span>
<a name="l00383"></a>00383 <span class="preprocessor">        AS2(    and             ebp, r2##j)\</span>
<a name="l00384"></a>00384 <span class="preprocessor">        AS1(    neg             ebp)\</span>
<a name="l00385"></a>00385 <span class="preprocessor">        AS2(    and             ebp, ebx)\</span>
<a name="l00386"></a>00386 <span class="preprocessor">        AS2(    xor             ebx, eax)\</span>
<a name="l00387"></a>00387 <span class="preprocessor">        AS2(    movzx   eax, al)\</span>
<a name="l00388"></a>00388 <span class="preprocessor">        AS2(    xor             ebx, [WORD_REG(si)+WORD_REG(ax)*4])\</span>
<a name="l00389"></a>00389 <span class="preprocessor">        AS2(    mov             eax, [s(i+3)])\</span>
<a name="l00390"></a>00390 <span class="preprocessor">        AS2(    xor             ebp, [s(i+2)])\</span>
<a name="l00391"></a>00391 <span class="preprocessor">        AS2(    add             r1##j, ebp)\</span>
<a name="l00392"></a>00392 <span class="preprocessor">        AS2(    movzx   ebp, al)\</span>
<a name="l00393"></a>00393 <span class="preprocessor">        AS2(    shr             eax, 8)\</span>
<a name="l00394"></a>00394 <span class="preprocessor">        AS2(    xor             ebx, [WORD_REG(si)+1024+WORD_REG(bp)*4])\</span>
<a name="l00395"></a>00395 <span class="preprocessor">        AS2(    xor             ebx, eax)\</span>
<a name="l00396"></a>00396 <span class="preprocessor">        AS2(    imul    r2##j, 0x54655307)\</span>
<a name="l00397"></a>00397 <span class="preprocessor">        AS2(    rol             r2##j, 7)\</span>
<a name="l00398"></a>00398 <span class="preprocessor">        AS2(    mov             [s(i+0)], ebx)\</span>
<a name="l00399"></a>00399 <span class="preprocessor"></span>
<a name="l00400"></a>00400 <span class="preprocessor"></span>                ASL(2)  <span class="comment">// outer loop, each iteration of this processes 80 words</span>
<a name="l00401"></a>00401                 AS2(    lea             WORD_REG(di), [SSE2_uvStart])   <span class="comment">// start of v and u</span>
<a name="l00402"></a>00402                 AS2(    mov             WORD_REG(ax), 80)