wp-mmx.pl 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506
  1. #! /usr/bin/env perl
  2. # Copyright 2005-2020 The OpenSSL Project Authors. All Rights Reserved.
  3. #
  4. # Licensed under the Apache License 2.0 (the "License"). You may not use
  5. # this file except in compliance with the License. You can obtain a copy
  6. # in the file LICENSE in the source distribution or at
  7. # https://www.openssl.org/source/license.html
  8. #
  9. # ====================================================================
  10. # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
  11. # project. Rights for redistribution and usage in source and binary
  12. # forms are granted according to the License.
  13. # ====================================================================
  14. #
  15. # whirlpool_block_mmx implementation.
  16. #
  17. *SCALE=\(2); # 2 or 8, that is the question:-) Value of 8 results
  18. # in 16KB large table, which is tough on L1 cache, but eliminates
  19. # unaligned references to it. Value of 2 results in 4KB table, but
  20. # 7/8 of references to it are unaligned. AMD cores seem to be
  21. # allergic to the latter, while Intel ones - to former [see the
  22. # table]. I stick to value of 2 for two reasons: 1. smaller table
  23. # minimizes cache trashing and thus mitigates the hazard of side-
  24. # channel leakage similar to AES cache-timing one; 2. performance
  25. # gap among different µ-archs is smaller.
  26. #
  27. # Performance table lists rounded amounts of CPU cycles spent by
  28. # whirlpool_block_mmx routine on single 64 byte input block, i.e.
  29. # smaller is better and asymptotic throughput can be estimated by
  30. # multiplying 64 by CPU clock frequency and dividing by relevant
  31. # value from the given table:
  32. #
  33. # $SCALE=2/8 icc8 gcc3
  34. # Intel P4 3200/4600 4600(*) 6400
  35. # Intel PIII 2900/3000 4900 5400
  36. # AMD K[78] 2500/1800 9900 8200(**)
  37. #
  38. # (*) I've sketched even non-MMX assembler, but for the record
  39. # I've failed to beat the Intel compiler on P4, without using
  40. # MMX that is...
  41. # (**) ... on AMD on the other hand non-MMX assembler was observed
  42. # to perform significantly better, but I figured this MMX
  43. # implementation is even faster anyway, so why bother? As for
  44. # pre-MMX AMD core[s], the improvement coefficient is more
  45. # than likely to vary anyway and I don't know how. But the
  46. # least I know is that gcc-generated code compiled with
  47. # -DL_ENDIAN and -DOPENSSL_SMALL_FOOTPRINT [see C module for
  48. # details] and optimized for Pentium was observed to perform
  49. # *better* on Pentium 100 than unrolled non-MMX assembler
  50. # loop... So we just say that I don't know if maintaining
  51. # non-MMX implementation would actually pay off, but till
  52. # opposite is proved "unlikely" is assumed.
  53. $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
  54. push(@INC,"${dir}","${dir}../../perlasm");
  55. require "x86asm.pl";
  56. $output=pop and open STDOUT,">$output";
  57. &asm_init($ARGV[0]);
  58. sub L() { &data_byte(@_); }
  59. sub LL()
  60. { if ($SCALE==2) { &data_byte(@_); &data_byte(@_); }
  61. elsif ($SCALE==8) { for ($i=0;$i<8;$i++) {
  62. &data_byte(@_);
  63. unshift(@_,pop(@_));
  64. }
  65. }
  66. else { die "invalid SCALE value"; }
  67. }
  68. sub scale()
  69. { if ($SCALE==2) { &lea(@_[0],&DWP(0,@_[1],@_[1])); }
  70. elsif ($SCALE==8) { &lea(@_[0],&DWP(0,"",@_[1],8)); }
  71. else { die "invalid SCALE value"; }
  72. }
  73. sub row()
  74. { if ($SCALE==2) { ((8-shift)&7); }
  75. elsif ($SCALE==8) { (8*shift); }
  76. else { die "invalid SCALE value"; }
  77. }
  78. $tbl="ebp";
  79. @mm=("mm0","mm1","mm2","mm3","mm4","mm5","mm6","mm7");
  80. &function_begin_B("whirlpool_block_mmx");
  81. &push ("ebp");
  82. &push ("ebx");
  83. &push ("esi");
  84. &push ("edi");
  85. &mov ("esi",&wparam(0)); # hash value
  86. &mov ("edi",&wparam(1)); # input data stream
  87. &mov ("ebp",&wparam(2)); # number of chunks in input
  88. &mov ("eax","esp"); # copy stack pointer
  89. &sub ("esp",128+20); # allocate frame
  90. &and ("esp",-64); # align for cache-line
  91. &lea ("ebx",&DWP(128,"esp"));
  92. &mov (&DWP(0,"ebx"),"esi"); # save parameter block
  93. &mov (&DWP(4,"ebx"),"edi");
  94. &mov (&DWP(8,"ebx"),"ebp");
  95. &mov (&DWP(16,"ebx"),"eax"); # saved stack pointer
  96. &call (&label("pic_point"));
  97. &set_label("pic_point");
  98. &blindpop($tbl);
  99. &lea ($tbl,&DWP(&label("table")."-".&label("pic_point"),$tbl));
  100. &xor ("ecx","ecx");
  101. &xor ("edx","edx");
  102. for($i=0;$i<8;$i++) { &movq(@mm[$i],&QWP($i*8,"esi")); } # L=H
  103. &set_label("outerloop");
  104. for($i=0;$i<8;$i++) { &movq(&QWP($i*8,"esp"),@mm[$i]); } # K=L
  105. for($i=0;$i<8;$i++) { &pxor(@mm[$i],&QWP($i*8,"edi")); } # L^=inp
  106. for($i=0;$i<8;$i++) { &movq(&QWP(64+$i*8,"esp"),@mm[$i]); } # S=L
  107. &xor ("esi","esi");
  108. &mov (&DWP(12,"ebx"),"esi"); # zero round counter
  109. &set_label("round",16);
  110. &movq (@mm[0],&QWP(2048*$SCALE,$tbl,"esi",8)); # rc[r]
  111. &mov ("eax",&DWP(0,"esp"));
  112. &mov ("ebx",&DWP(4,"esp"));
  113. &movz ("ecx",&LB("eax"));
  114. &movz ("edx",&HB("eax"));
  115. for($i=0;$i<8;$i++) {
  116. my $func = ($i==0)? \&movq : \&pxor;
  117. &shr ("eax",16);
  118. &scale ("esi","ecx");
  119. &movz ("ecx",&LB("eax"));
  120. &scale ("edi","edx");
  121. &movz ("edx",&HB("eax"));
  122. &pxor (@mm[0],&QWP(&row(0),$tbl,"esi",8));
  123. &$func (@mm[1],&QWP(&row(1),$tbl,"edi",8));
  124. &mov ("eax",&DWP(($i+1)*8,"esp"));
  125. &scale ("esi","ecx");
  126. &movz ("ecx",&LB("ebx"));
  127. &scale ("edi","edx");
  128. &movz ("edx",&HB("ebx"));
  129. &$func (@mm[2],&QWP(&row(2),$tbl,"esi",8));
  130. &$func (@mm[3],&QWP(&row(3),$tbl,"edi",8));
  131. &shr ("ebx",16);
  132. &scale ("esi","ecx");
  133. &movz ("ecx",&LB("ebx"));
  134. &scale ("edi","edx");
  135. &movz ("edx",&HB("ebx"));
  136. &$func (@mm[4],&QWP(&row(4),$tbl,"esi",8));
  137. &$func (@mm[5],&QWP(&row(5),$tbl,"edi",8));
  138. &mov ("ebx",&DWP(($i+1)*8+4,"esp"));
  139. &scale ("esi","ecx");
  140. &movz ("ecx",&LB("eax"));
  141. &scale ("edi","edx");
  142. &movz ("edx",&HB("eax"));
  143. &$func (@mm[6],&QWP(&row(6),$tbl,"esi",8));
  144. &$func (@mm[7],&QWP(&row(7),$tbl,"edi",8));
  145. push(@mm,shift(@mm));
  146. }
  147. for($i=0;$i<8;$i++) { &movq(&QWP($i*8,"esp"),@mm[$i]); } # K=L
  148. for($i=0;$i<8;$i++) {
  149. &shr ("eax",16);
  150. &scale ("esi","ecx");
  151. &movz ("ecx",&LB("eax"));
  152. &scale ("edi","edx");
  153. &movz ("edx",&HB("eax"));
  154. &pxor (@mm[0],&QWP(&row(0),$tbl,"esi",8));
  155. &pxor (@mm[1],&QWP(&row(1),$tbl,"edi",8));
  156. &mov ("eax",&DWP(64+($i+1)*8,"esp")) if ($i<7);
  157. &scale ("esi","ecx");
  158. &movz ("ecx",&LB("ebx"));
  159. &scale ("edi","edx");
  160. &movz ("edx",&HB("ebx"));
  161. &pxor (@mm[2],&QWP(&row(2),$tbl,"esi",8));
  162. &pxor (@mm[3],&QWP(&row(3),$tbl,"edi",8));
  163. &shr ("ebx",16);
  164. &scale ("esi","ecx");
  165. &movz ("ecx",&LB("ebx"));
  166. &scale ("edi","edx");
  167. &movz ("edx",&HB("ebx"));
  168. &pxor (@mm[4],&QWP(&row(4),$tbl,"esi",8));
  169. &pxor (@mm[5],&QWP(&row(5),$tbl,"edi",8));
  170. &mov ("ebx",&DWP(64+($i+1)*8+4,"esp")) if ($i<7);
  171. &scale ("esi","ecx");
  172. &movz ("ecx",&LB("eax"));
  173. &scale ("edi","edx");
  174. &movz ("edx",&HB("eax"));
  175. &pxor (@mm[6],&QWP(&row(6),$tbl,"esi",8));
  176. &pxor (@mm[7],&QWP(&row(7),$tbl,"edi",8));
  177. push(@mm,shift(@mm));
  178. }
  179. &lea ("ebx",&DWP(128,"esp"));
  180. &mov ("esi",&DWP(12,"ebx")); # pull round counter
  181. &add ("esi",1);
  182. &cmp ("esi",10);
  183. &je (&label("roundsdone"));
  184. &mov (&DWP(12,"ebx"),"esi"); # update round counter
  185. for($i=0;$i<8;$i++) { &movq(&QWP(64+$i*8,"esp"),@mm[$i]); } # S=L
  186. &jmp (&label("round"));
  187. &set_label("roundsdone",16);
  188. &mov ("esi",&DWP(0,"ebx")); # reload argument block
  189. &mov ("edi",&DWP(4,"ebx"));
  190. &mov ("eax",&DWP(8,"ebx"));
  191. for($i=0;$i<8;$i++) { &pxor(@mm[$i],&QWP($i*8,"edi")); } # L^=inp
  192. for($i=0;$i<8;$i++) { &pxor(@mm[$i],&QWP($i*8,"esi")); } # L^=H
  193. for($i=0;$i<8;$i++) { &movq(&QWP($i*8,"esi"),@mm[$i]); } # H=L
  194. &lea ("edi",&DWP(64,"edi")); # inp+=64
  195. &sub ("eax",1); # num--
  196. &jz (&label("alldone"));
  197. &mov (&DWP(4,"ebx"),"edi"); # update argument block
  198. &mov (&DWP(8,"ebx"),"eax");
  199. &jmp (&label("outerloop"));
  200. &set_label("alldone");
  201. &emms ();
  202. &mov ("esp",&DWP(16,"ebx")); # restore saved stack pointer
  203. &pop ("edi");
  204. &pop ("esi");
  205. &pop ("ebx");
  206. &pop ("ebp");
  207. &ret ();
  208. &align(64);
  209. &set_label("table");
  210. &LL(0x18,0x18,0x60,0x18,0xc0,0x78,0x30,0xd8);
  211. &LL(0x23,0x23,0x8c,0x23,0x05,0xaf,0x46,0x26);
  212. &LL(0xc6,0xc6,0x3f,0xc6,0x7e,0xf9,0x91,0xb8);
  213. &LL(0xe8,0xe8,0x87,0xe8,0x13,0x6f,0xcd,0xfb);
  214. &LL(0x87,0x87,0x26,0x87,0x4c,0xa1,0x13,0xcb);
  215. &LL(0xb8,0xb8,0xda,0xb8,0xa9,0x62,0x6d,0x11);
  216. &LL(0x01,0x01,0x04,0x01,0x08,0x05,0x02,0x09);
  217. &LL(0x4f,0x4f,0x21,0x4f,0x42,0x6e,0x9e,0x0d);
  218. &LL(0x36,0x36,0xd8,0x36,0xad,0xee,0x6c,0x9b);
  219. &LL(0xa6,0xa6,0xa2,0xa6,0x59,0x04,0x51,0xff);
  220. &LL(0xd2,0xd2,0x6f,0xd2,0xde,0xbd,0xb9,0x0c);
  221. &LL(0xf5,0xf5,0xf3,0xf5,0xfb,0x06,0xf7,0x0e);
  222. &LL(0x79,0x79,0xf9,0x79,0xef,0x80,0xf2,0x96);
  223. &LL(0x6f,0x6f,0xa1,0x6f,0x5f,0xce,0xde,0x30);
  224. &LL(0x91,0x91,0x7e,0x91,0xfc,0xef,0x3f,0x6d);
  225. &LL(0x52,0x52,0x55,0x52,0xaa,0x07,0xa4,0xf8);
  226. &LL(0x60,0x60,0x9d,0x60,0x27,0xfd,0xc0,0x47);
  227. &LL(0xbc,0xbc,0xca,0xbc,0x89,0x76,0x65,0x35);
  228. &LL(0x9b,0x9b,0x56,0x9b,0xac,0xcd,0x2b,0x37);
  229. &LL(0x8e,0x8e,0x02,0x8e,0x04,0x8c,0x01,0x8a);
  230. &LL(0xa3,0xa3,0xb6,0xa3,0x71,0x15,0x5b,0xd2);
  231. &LL(0x0c,0x0c,0x30,0x0c,0x60,0x3c,0x18,0x6c);
  232. &LL(0x7b,0x7b,0xf1,0x7b,0xff,0x8a,0xf6,0x84);
  233. &LL(0x35,0x35,0xd4,0x35,0xb5,0xe1,0x6a,0x80);
  234. &LL(0x1d,0x1d,0x74,0x1d,0xe8,0x69,0x3a,0xf5);
  235. &LL(0xe0,0xe0,0xa7,0xe0,0x53,0x47,0xdd,0xb3);
  236. &LL(0xd7,0xd7,0x7b,0xd7,0xf6,0xac,0xb3,0x21);
  237. &LL(0xc2,0xc2,0x2f,0xc2,0x5e,0xed,0x99,0x9c);
  238. &LL(0x2e,0x2e,0xb8,0x2e,0x6d,0x96,0x5c,0x43);
  239. &LL(0x4b,0x4b,0x31,0x4b,0x62,0x7a,0x96,0x29);
  240. &LL(0xfe,0xfe,0xdf,0xfe,0xa3,0x21,0xe1,0x5d);
  241. &LL(0x57,0x57,0x41,0x57,0x82,0x16,0xae,0xd5);
  242. &LL(0x15,0x15,0x54,0x15,0xa8,0x41,0x2a,0xbd);
  243. &LL(0x77,0x77,0xc1,0x77,0x9f,0xb6,0xee,0xe8);
  244. &LL(0x37,0x37,0xdc,0x37,0xa5,0xeb,0x6e,0x92);
  245. &LL(0xe5,0xe5,0xb3,0xe5,0x7b,0x56,0xd7,0x9e);
  246. &LL(0x9f,0x9f,0x46,0x9f,0x8c,0xd9,0x23,0x13);
  247. &LL(0xf0,0xf0,0xe7,0xf0,0xd3,0x17,0xfd,0x23);
  248. &LL(0x4a,0x4a,0x35,0x4a,0x6a,0x7f,0x94,0x20);
  249. &LL(0xda,0xda,0x4f,0xda,0x9e,0x95,0xa9,0x44);
  250. &LL(0x58,0x58,0x7d,0x58,0xfa,0x25,0xb0,0xa2);
  251. &LL(0xc9,0xc9,0x03,0xc9,0x06,0xca,0x8f,0xcf);
  252. &LL(0x29,0x29,0xa4,0x29,0x55,0x8d,0x52,0x7c);
  253. &LL(0x0a,0x0a,0x28,0x0a,0x50,0x22,0x14,0x5a);
  254. &LL(0xb1,0xb1,0xfe,0xb1,0xe1,0x4f,0x7f,0x50);
  255. &LL(0xa0,0xa0,0xba,0xa0,0x69,0x1a,0x5d,0xc9);
  256. &LL(0x6b,0x6b,0xb1,0x6b,0x7f,0xda,0xd6,0x14);
  257. &LL(0x85,0x85,0x2e,0x85,0x5c,0xab,0x17,0xd9);
  258. &LL(0xbd,0xbd,0xce,0xbd,0x81,0x73,0x67,0x3c);
  259. &LL(0x5d,0x5d,0x69,0x5d,0xd2,0x34,0xba,0x8f);
  260. &LL(0x10,0x10,0x40,0x10,0x80,0x50,0x20,0x90);
  261. &LL(0xf4,0xf4,0xf7,0xf4,0xf3,0x03,0xf5,0x07);
  262. &LL(0xcb,0xcb,0x0b,0xcb,0x16,0xc0,0x8b,0xdd);
  263. &LL(0x3e,0x3e,0xf8,0x3e,0xed,0xc6,0x7c,0xd3);
  264. &LL(0x05,0x05,0x14,0x05,0x28,0x11,0x0a,0x2d);
  265. &LL(0x67,0x67,0x81,0x67,0x1f,0xe6,0xce,0x78);
  266. &LL(0xe4,0xe4,0xb7,0xe4,0x73,0x53,0xd5,0x97);
  267. &LL(0x27,0x27,0x9c,0x27,0x25,0xbb,0x4e,0x02);
  268. &LL(0x41,0x41,0x19,0x41,0x32,0x58,0x82,0x73);
  269. &LL(0x8b,0x8b,0x16,0x8b,0x2c,0x9d,0x0b,0xa7);
  270. &LL(0xa7,0xa7,0xa6,0xa7,0x51,0x01,0x53,0xf6);
  271. &LL(0x7d,0x7d,0xe9,0x7d,0xcf,0x94,0xfa,0xb2);
  272. &LL(0x95,0x95,0x6e,0x95,0xdc,0xfb,0x37,0x49);
  273. &LL(0xd8,0xd8,0x47,0xd8,0x8e,0x9f,0xad,0x56);
  274. &LL(0xfb,0xfb,0xcb,0xfb,0x8b,0x30,0xeb,0x70);
  275. &LL(0xee,0xee,0x9f,0xee,0x23,0x71,0xc1,0xcd);
  276. &LL(0x7c,0x7c,0xed,0x7c,0xc7,0x91,0xf8,0xbb);
  277. &LL(0x66,0x66,0x85,0x66,0x17,0xe3,0xcc,0x71);
  278. &LL(0xdd,0xdd,0x53,0xdd,0xa6,0x8e,0xa7,0x7b);
  279. &LL(0x17,0x17,0x5c,0x17,0xb8,0x4b,0x2e,0xaf);
  280. &LL(0x47,0x47,0x01,0x47,0x02,0x46,0x8e,0x45);
  281. &LL(0x9e,0x9e,0x42,0x9e,0x84,0xdc,0x21,0x1a);
  282. &LL(0xca,0xca,0x0f,0xca,0x1e,0xc5,0x89,0xd4);
  283. &LL(0x2d,0x2d,0xb4,0x2d,0x75,0x99,0x5a,0x58);
  284. &LL(0xbf,0xbf,0xc6,0xbf,0x91,0x79,0x63,0x2e);
  285. &LL(0x07,0x07,0x1c,0x07,0x38,0x1b,0x0e,0x3f);
  286. &LL(0xad,0xad,0x8e,0xad,0x01,0x23,0x47,0xac);
  287. &LL(0x5a,0x5a,0x75,0x5a,0xea,0x2f,0xb4,0xb0);
  288. &LL(0x83,0x83,0x36,0x83,0x6c,0xb5,0x1b,0xef);
  289. &LL(0x33,0x33,0xcc,0x33,0x85,0xff,0x66,0xb6);
  290. &LL(0x63,0x63,0x91,0x63,0x3f,0xf2,0xc6,0x5c);
  291. &LL(0x02,0x02,0x08,0x02,0x10,0x0a,0x04,0x12);
  292. &LL(0xaa,0xaa,0x92,0xaa,0x39,0x38,0x49,0x93);
  293. &LL(0x71,0x71,0xd9,0x71,0xaf,0xa8,0xe2,0xde);
  294. &LL(0xc8,0xc8,0x07,0xc8,0x0e,0xcf,0x8d,0xc6);
  295. &LL(0x19,0x19,0x64,0x19,0xc8,0x7d,0x32,0xd1);
  296. &LL(0x49,0x49,0x39,0x49,0x72,0x70,0x92,0x3b);
  297. &LL(0xd9,0xd9,0x43,0xd9,0x86,0x9a,0xaf,0x5f);
  298. &LL(0xf2,0xf2,0xef,0xf2,0xc3,0x1d,0xf9,0x31);
  299. &LL(0xe3,0xe3,0xab,0xe3,0x4b,0x48,0xdb,0xa8);
  300. &LL(0x5b,0x5b,0x71,0x5b,0xe2,0x2a,0xb6,0xb9);
  301. &LL(0x88,0x88,0x1a,0x88,0x34,0x92,0x0d,0xbc);
  302. &LL(0x9a,0x9a,0x52,0x9a,0xa4,0xc8,0x29,0x3e);
  303. &LL(0x26,0x26,0x98,0x26,0x2d,0xbe,0x4c,0x0b);
  304. &LL(0x32,0x32,0xc8,0x32,0x8d,0xfa,0x64,0xbf);
  305. &LL(0xb0,0xb0,0xfa,0xb0,0xe9,0x4a,0x7d,0x59);
  306. &LL(0xe9,0xe9,0x83,0xe9,0x1b,0x6a,0xcf,0xf2);
  307. &LL(0x0f,0x0f,0x3c,0x0f,0x78,0x33,0x1e,0x77);
  308. &LL(0xd5,0xd5,0x73,0xd5,0xe6,0xa6,0xb7,0x33);
  309. &LL(0x80,0x80,0x3a,0x80,0x74,0xba,0x1d,0xf4);
  310. &LL(0xbe,0xbe,0xc2,0xbe,0x99,0x7c,0x61,0x27);
  311. &LL(0xcd,0xcd,0x13,0xcd,0x26,0xde,0x87,0xeb);
  312. &LL(0x34,0x34,0xd0,0x34,0xbd,0xe4,0x68,0x89);
  313. &LL(0x48,0x48,0x3d,0x48,0x7a,0x75,0x90,0x32);
  314. &LL(0xff,0xff,0xdb,0xff,0xab,0x24,0xe3,0x54);
  315. &LL(0x7a,0x7a,0xf5,0x7a,0xf7,0x8f,0xf4,0x8d);
  316. &LL(0x90,0x90,0x7a,0x90,0xf4,0xea,0x3d,0x64);
  317. &LL(0x5f,0x5f,0x61,0x5f,0xc2,0x3e,0xbe,0x9d);
  318. &LL(0x20,0x20,0x80,0x20,0x1d,0xa0,0x40,0x3d);
  319. &LL(0x68,0x68,0xbd,0x68,0x67,0xd5,0xd0,0x0f);
  320. &LL(0x1a,0x1a,0x68,0x1a,0xd0,0x72,0x34,0xca);
  321. &LL(0xae,0xae,0x82,0xae,0x19,0x2c,0x41,0xb7);
  322. &LL(0xb4,0xb4,0xea,0xb4,0xc9,0x5e,0x75,0x7d);
  323. &LL(0x54,0x54,0x4d,0x54,0x9a,0x19,0xa8,0xce);
  324. &LL(0x93,0x93,0x76,0x93,0xec,0xe5,0x3b,0x7f);
  325. &LL(0x22,0x22,0x88,0x22,0x0d,0xaa,0x44,0x2f);
  326. &LL(0x64,0x64,0x8d,0x64,0x07,0xe9,0xc8,0x63);
  327. &LL(0xf1,0xf1,0xe3,0xf1,0xdb,0x12,0xff,0x2a);
  328. &LL(0x73,0x73,0xd1,0x73,0xbf,0xa2,0xe6,0xcc);
  329. &LL(0x12,0x12,0x48,0x12,0x90,0x5a,0x24,0x82);
  330. &LL(0x40,0x40,0x1d,0x40,0x3a,0x5d,0x80,0x7a);
  331. &LL(0x08,0x08,0x20,0x08,0x40,0x28,0x10,0x48);
  332. &LL(0xc3,0xc3,0x2b,0xc3,0x56,0xe8,0x9b,0x95);
  333. &LL(0xec,0xec,0x97,0xec,0x33,0x7b,0xc5,0xdf);
  334. &LL(0xdb,0xdb,0x4b,0xdb,0x96,0x90,0xab,0x4d);
  335. &LL(0xa1,0xa1,0xbe,0xa1,0x61,0x1f,0x5f,0xc0);
  336. &LL(0x8d,0x8d,0x0e,0x8d,0x1c,0x83,0x07,0x91);
  337. &LL(0x3d,0x3d,0xf4,0x3d,0xf5,0xc9,0x7a,0xc8);
  338. &LL(0x97,0x97,0x66,0x97,0xcc,0xf1,0x33,0x5b);
  339. &LL(0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00);
  340. &LL(0xcf,0xcf,0x1b,0xcf,0x36,0xd4,0x83,0xf9);
  341. &LL(0x2b,0x2b,0xac,0x2b,0x45,0x87,0x56,0x6e);
  342. &LL(0x76,0x76,0xc5,0x76,0x97,0xb3,0xec,0xe1);
  343. &LL(0x82,0x82,0x32,0x82,0x64,0xb0,0x19,0xe6);
  344. &LL(0xd6,0xd6,0x7f,0xd6,0xfe,0xa9,0xb1,0x28);
  345. &LL(0x1b,0x1b,0x6c,0x1b,0xd8,0x77,0x36,0xc3);
  346. &LL(0xb5,0xb5,0xee,0xb5,0xc1,0x5b,0x77,0x74);
  347. &LL(0xaf,0xaf,0x86,0xaf,0x11,0x29,0x43,0xbe);
  348. &LL(0x6a,0x6a,0xb5,0x6a,0x77,0xdf,0xd4,0x1d);
  349. &LL(0x50,0x50,0x5d,0x50,0xba,0x0d,0xa0,0xea);
  350. &LL(0x45,0x45,0x09,0x45,0x12,0x4c,0x8a,0x57);
  351. &LL(0xf3,0xf3,0xeb,0xf3,0xcb,0x18,0xfb,0x38);
  352. &LL(0x30,0x30,0xc0,0x30,0x9d,0xf0,0x60,0xad);
  353. &LL(0xef,0xef,0x9b,0xef,0x2b,0x74,0xc3,0xc4);
  354. &LL(0x3f,0x3f,0xfc,0x3f,0xe5,0xc3,0x7e,0xda);
  355. &LL(0x55,0x55,0x49,0x55,0x92,0x1c,0xaa,0xc7);
  356. &LL(0xa2,0xa2,0xb2,0xa2,0x79,0x10,0x59,0xdb);
  357. &LL(0xea,0xea,0x8f,0xea,0x03,0x65,0xc9,0xe9);
  358. &LL(0x65,0x65,0x89,0x65,0x0f,0xec,0xca,0x6a);
  359. &LL(0xba,0xba,0xd2,0xba,0xb9,0x68,0x69,0x03);
  360. &LL(0x2f,0x2f,0xbc,0x2f,0x65,0x93,0x5e,0x4a);
  361. &LL(0xc0,0xc0,0x27,0xc0,0x4e,0xe7,0x9d,0x8e);
  362. &LL(0xde,0xde,0x5f,0xde,0xbe,0x81,0xa1,0x60);
  363. &LL(0x1c,0x1c,0x70,0x1c,0xe0,0x6c,0x38,0xfc);
  364. &LL(0xfd,0xfd,0xd3,0xfd,0xbb,0x2e,0xe7,0x46);
  365. &LL(0x4d,0x4d,0x29,0x4d,0x52,0x64,0x9a,0x1f);
  366. &LL(0x92,0x92,0x72,0x92,0xe4,0xe0,0x39,0x76);
  367. &LL(0x75,0x75,0xc9,0x75,0x8f,0xbc,0xea,0xfa);
  368. &LL(0x06,0x06,0x18,0x06,0x30,0x1e,0x0c,0x36);
  369. &LL(0x8a,0x8a,0x12,0x8a,0x24,0x98,0x09,0xae);
  370. &LL(0xb2,0xb2,0xf2,0xb2,0xf9,0x40,0x79,0x4b);
  371. &LL(0xe6,0xe6,0xbf,0xe6,0x63,0x59,0xd1,0x85);
  372. &LL(0x0e,0x0e,0x38,0x0e,0x70,0x36,0x1c,0x7e);
  373. &LL(0x1f,0x1f,0x7c,0x1f,0xf8,0x63,0x3e,0xe7);
  374. &LL(0x62,0x62,0x95,0x62,0x37,0xf7,0xc4,0x55);
  375. &LL(0xd4,0xd4,0x77,0xd4,0xee,0xa3,0xb5,0x3a);
  376. &LL(0xa8,0xa8,0x9a,0xa8,0x29,0x32,0x4d,0x81);
  377. &LL(0x96,0x96,0x62,0x96,0xc4,0xf4,0x31,0x52);
  378. &LL(0xf9,0xf9,0xc3,0xf9,0x9b,0x3a,0xef,0x62);
  379. &LL(0xc5,0xc5,0x33,0xc5,0x66,0xf6,0x97,0xa3);
  380. &LL(0x25,0x25,0x94,0x25,0x35,0xb1,0x4a,0x10);
  381. &LL(0x59,0x59,0x79,0x59,0xf2,0x20,0xb2,0xab);
  382. &LL(0x84,0x84,0x2a,0x84,0x54,0xae,0x15,0xd0);
  383. &LL(0x72,0x72,0xd5,0x72,0xb7,0xa7,0xe4,0xc5);
  384. &LL(0x39,0x39,0xe4,0x39,0xd5,0xdd,0x72,0xec);
  385. &LL(0x4c,0x4c,0x2d,0x4c,0x5a,0x61,0x98,0x16);
  386. &LL(0x5e,0x5e,0x65,0x5e,0xca,0x3b,0xbc,0x94);
  387. &LL(0x78,0x78,0xfd,0x78,0xe7,0x85,0xf0,0x9f);
  388. &LL(0x38,0x38,0xe0,0x38,0xdd,0xd8,0x70,0xe5);
  389. &LL(0x8c,0x8c,0x0a,0x8c,0x14,0x86,0x05,0x98);
  390. &LL(0xd1,0xd1,0x63,0xd1,0xc6,0xb2,0xbf,0x17);
  391. &LL(0xa5,0xa5,0xae,0xa5,0x41,0x0b,0x57,0xe4);
  392. &LL(0xe2,0xe2,0xaf,0xe2,0x43,0x4d,0xd9,0xa1);
  393. &LL(0x61,0x61,0x99,0x61,0x2f,0xf8,0xc2,0x4e);
  394. &LL(0xb3,0xb3,0xf6,0xb3,0xf1,0x45,0x7b,0x42);
  395. &LL(0x21,0x21,0x84,0x21,0x15,0xa5,0x42,0x34);
  396. &LL(0x9c,0x9c,0x4a,0x9c,0x94,0xd6,0x25,0x08);
  397. &LL(0x1e,0x1e,0x78,0x1e,0xf0,0x66,0x3c,0xee);
  398. &LL(0x43,0x43,0x11,0x43,0x22,0x52,0x86,0x61);
  399. &LL(0xc7,0xc7,0x3b,0xc7,0x76,0xfc,0x93,0xb1);
  400. &LL(0xfc,0xfc,0xd7,0xfc,0xb3,0x2b,0xe5,0x4f);
  401. &LL(0x04,0x04,0x10,0x04,0x20,0x14,0x08,0x24);
  402. &LL(0x51,0x51,0x59,0x51,0xb2,0x08,0xa2,0xe3);
  403. &LL(0x99,0x99,0x5e,0x99,0xbc,0xc7,0x2f,0x25);
  404. &LL(0x6d,0x6d,0xa9,0x6d,0x4f,0xc4,0xda,0x22);
  405. &LL(0x0d,0x0d,0x34,0x0d,0x68,0x39,0x1a,0x65);
  406. &LL(0xfa,0xfa,0xcf,0xfa,0x83,0x35,0xe9,0x79);
  407. &LL(0xdf,0xdf,0x5b,0xdf,0xb6,0x84,0xa3,0x69);
  408. &LL(0x7e,0x7e,0xe5,0x7e,0xd7,0x9b,0xfc,0xa9);
  409. &LL(0x24,0x24,0x90,0x24,0x3d,0xb4,0x48,0x19);
  410. &LL(0x3b,0x3b,0xec,0x3b,0xc5,0xd7,0x76,0xfe);
  411. &LL(0xab,0xab,0x96,0xab,0x31,0x3d,0x4b,0x9a);
  412. &LL(0xce,0xce,0x1f,0xce,0x3e,0xd1,0x81,0xf0);
  413. &LL(0x11,0x11,0x44,0x11,0x88,0x55,0x22,0x99);
  414. &LL(0x8f,0x8f,0x06,0x8f,0x0c,0x89,0x03,0x83);
  415. &LL(0x4e,0x4e,0x25,0x4e,0x4a,0x6b,0x9c,0x04);
  416. &LL(0xb7,0xb7,0xe6,0xb7,0xd1,0x51,0x73,0x66);
  417. &LL(0xeb,0xeb,0x8b,0xeb,0x0b,0x60,0xcb,0xe0);
  418. &LL(0x3c,0x3c,0xf0,0x3c,0xfd,0xcc,0x78,0xc1);
  419. &LL(0x81,0x81,0x3e,0x81,0x7c,0xbf,0x1f,0xfd);
  420. &LL(0x94,0x94,0x6a,0x94,0xd4,0xfe,0x35,0x40);
  421. &LL(0xf7,0xf7,0xfb,0xf7,0xeb,0x0c,0xf3,0x1c);
  422. &LL(0xb9,0xb9,0xde,0xb9,0xa1,0x67,0x6f,0x18);
  423. &LL(0x13,0x13,0x4c,0x13,0x98,0x5f,0x26,0x8b);
  424. &LL(0x2c,0x2c,0xb0,0x2c,0x7d,0x9c,0x58,0x51);
  425. &LL(0xd3,0xd3,0x6b,0xd3,0xd6,0xb8,0xbb,0x05);
  426. &LL(0xe7,0xe7,0xbb,0xe7,0x6b,0x5c,0xd3,0x8c);
  427. &LL(0x6e,0x6e,0xa5,0x6e,0x57,0xcb,0xdc,0x39);
  428. &LL(0xc4,0xc4,0x37,0xc4,0x6e,0xf3,0x95,0xaa);
  429. &LL(0x03,0x03,0x0c,0x03,0x18,0x0f,0x06,0x1b);
  430. &LL(0x56,0x56,0x45,0x56,0x8a,0x13,0xac,0xdc);
  431. &LL(0x44,0x44,0x0d,0x44,0x1a,0x49,0x88,0x5e);
  432. &LL(0x7f,0x7f,0xe1,0x7f,0xdf,0x9e,0xfe,0xa0);
  433. &LL(0xa9,0xa9,0x9e,0xa9,0x21,0x37,0x4f,0x88);
  434. &LL(0x2a,0x2a,0xa8,0x2a,0x4d,0x82,0x54,0x67);
  435. &LL(0xbb,0xbb,0xd6,0xbb,0xb1,0x6d,0x6b,0x0a);
  436. &LL(0xc1,0xc1,0x23,0xc1,0x46,0xe2,0x9f,0x87);
  437. &LL(0x53,0x53,0x51,0x53,0xa2,0x02,0xa6,0xf1);
  438. &LL(0xdc,0xdc,0x57,0xdc,0xae,0x8b,0xa5,0x72);
  439. &LL(0x0b,0x0b,0x2c,0x0b,0x58,0x27,0x16,0x53);
  440. &LL(0x9d,0x9d,0x4e,0x9d,0x9c,0xd3,0x27,0x01);
  441. &LL(0x6c,0x6c,0xad,0x6c,0x47,0xc1,0xd8,0x2b);
  442. &LL(0x31,0x31,0xc4,0x31,0x95,0xf5,0x62,0xa4);
  443. &LL(0x74,0x74,0xcd,0x74,0x87,0xb9,0xe8,0xf3);
  444. &LL(0xf6,0xf6,0xff,0xf6,0xe3,0x09,0xf1,0x15);
  445. &LL(0x46,0x46,0x05,0x46,0x0a,0x43,0x8c,0x4c);
  446. &LL(0xac,0xac,0x8a,0xac,0x09,0x26,0x45,0xa5);
  447. &LL(0x89,0x89,0x1e,0x89,0x3c,0x97,0x0f,0xb5);
  448. &LL(0x14,0x14,0x50,0x14,0xa0,0x44,0x28,0xb4);
  449. &LL(0xe1,0xe1,0xa3,0xe1,0x5b,0x42,0xdf,0xba);
  450. &LL(0x16,0x16,0x58,0x16,0xb0,0x4e,0x2c,0xa6);
  451. &LL(0x3a,0x3a,0xe8,0x3a,0xcd,0xd2,0x74,0xf7);
  452. &LL(0x69,0x69,0xb9,0x69,0x6f,0xd0,0xd2,0x06);
  453. &LL(0x09,0x09,0x24,0x09,0x48,0x2d,0x12,0x41);
  454. &LL(0x70,0x70,0xdd,0x70,0xa7,0xad,0xe0,0xd7);
  455. &LL(0xb6,0xb6,0xe2,0xb6,0xd9,0x54,0x71,0x6f);
  456. &LL(0xd0,0xd0,0x67,0xd0,0xce,0xb7,0xbd,0x1e);
  457. &LL(0xed,0xed,0x93,0xed,0x3b,0x7e,0xc7,0xd6);
  458. &LL(0xcc,0xcc,0x17,0xcc,0x2e,0xdb,0x85,0xe2);
  459. &LL(0x42,0x42,0x15,0x42,0x2a,0x57,0x84,0x68);
  460. &LL(0x98,0x98,0x5a,0x98,0xb4,0xc2,0x2d,0x2c);
  461. &LL(0xa4,0xa4,0xaa,0xa4,0x49,0x0e,0x55,0xed);
  462. &LL(0x28,0x28,0xa0,0x28,0x5d,0x88,0x50,0x75);
  463. &LL(0x5c,0x5c,0x6d,0x5c,0xda,0x31,0xb8,0x86);
  464. &LL(0xf8,0xf8,0xc7,0xf8,0x93,0x3f,0xed,0x6b);
  465. &LL(0x86,0x86,0x22,0x86,0x44,0xa4,0x11,0xc2);
  466. &L(0x18,0x23,0xc6,0xe8,0x87,0xb8,0x01,0x4f); # rc[ROUNDS]
  467. &L(0x36,0xa6,0xd2,0xf5,0x79,0x6f,0x91,0x52);
  468. &L(0x60,0xbc,0x9b,0x8e,0xa3,0x0c,0x7b,0x35);
  469. &L(0x1d,0xe0,0xd7,0xc2,0x2e,0x4b,0xfe,0x57);
  470. &L(0x15,0x77,0x37,0xe5,0x9f,0xf0,0x4a,0xda);
  471. &L(0x58,0xc9,0x29,0x0a,0xb1,0xa0,0x6b,0x85);
  472. &L(0xbd,0x5d,0x10,0xf4,0xcb,0x3e,0x05,0x67);
  473. &L(0xe4,0x27,0x41,0x8b,0xa7,0x7d,0x95,0xd8);
  474. &L(0xfb,0xee,0x7c,0x66,0xdd,0x17,0x47,0x9e);
  475. &L(0xca,0x2d,0xbf,0x07,0xad,0x5a,0x83,0x33);
  476. &function_end_B("whirlpool_block_mmx");
  477. &asm_finish();
  478. close STDOUT or die "error closing STDOUT: $!";