aes-x86_64.s 54.4 KB
Newer Older
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
	movl	%eax,36(%rdi)

	cmpl	$7,%ecx
	je	.L12break
	addl	$1,%ecx

	xorl	16(%rdi),%eax
	movl	%eax,40(%rdi)
	xorl	20(%rdi),%eax
	movl	%eax,44(%rdi)

	leaq	24(%rdi),%rdi
	jmp	.L12loop
.L12break:
	movl	$12,72(%rdi)
	xorq	%rax,%rax
	jmp	.Lexit

.L14rounds:
	movq	0(%rsi),%rax
	movq	8(%rsi),%rbx
	movq	16(%rsi),%rcx
	movq	24(%rsi),%rdx
	movq	%rax,0(%rdi)
	movq	%rbx,8(%rdi)
	movq	%rcx,16(%rdi)
	movq	%rdx,24(%rdi)

	shrq	$32,%rdx
	xorl	%ecx,%ecx
	jmp	.L14shortcut
.align	4
.L14loop:
	movl	0(%rdi),%eax
	movl	28(%rdi),%edx
.L14shortcut:
	movzbl	%dl,%esi
	movzbl	-128(%rbp,%rsi,1),%ebx
	movzbl	%dh,%esi
	shll	$24,%ebx
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	shrl	$16,%edx
	movzbl	%dl,%esi
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	movzbl	%dh,%esi
	shll	$8,%ebx
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	shll	$16,%ebx
	xorl	%ebx,%eax

	xorl	1024-128(%rbp,%rcx,4),%eax
	movl	%eax,32(%rdi)
	xorl	4(%rdi),%eax
	movl	%eax,36(%rdi)
	xorl	8(%rdi),%eax
	movl	%eax,40(%rdi)
	xorl	12(%rdi),%eax
	movl	%eax,44(%rdi)

	cmpl	$6,%ecx
	je	.L14break
	addl	$1,%ecx

	movl	%eax,%edx
	movl	16(%rdi),%eax
	movzbl	%dl,%esi
	movzbl	-128(%rbp,%rsi,1),%ebx
	movzbl	%dh,%esi
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	shrl	$16,%edx
	shll	$8,%ebx
	movzbl	%dl,%esi
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	movzbl	%dh,%esi
	shll	$16,%ebx
	xorl	%ebx,%eax

	movzbl	-128(%rbp,%rsi,1),%ebx
	shll	$24,%ebx
	xorl	%ebx,%eax

	movl	%eax,48(%rdi)
	xorl	20(%rdi),%eax
	movl	%eax,52(%rdi)
	xorl	24(%rdi),%eax
	movl	%eax,56(%rdi)
	xorl	28(%rdi),%eax
	movl	%eax,60(%rdi)

	leaq	32(%rdi),%rdi
	jmp	.L14loop
.L14break:
	movl	$14,48(%rdi)
	xorq	%rax,%rax
	jmp	.Lexit

.Lbadpointer:
	movq	$-1,%rax
.Lexit:
.byte	0xf3,0xc3			
.size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
.globl	private_AES_set_decrypt_key
.type	private_AES_set_decrypt_key,@function
.align	16
private_AES_set_decrypt_key:
	pushq	%rbx
	pushq	%rbp
	pushq	%r12
	pushq	%r13
	pushq	%r14
	pushq	%r15
	pushq	%rdx
.Ldec_key_prologue:

	call	_x86_64_AES_set_encrypt_key
	movq	(%rsp),%r8
	cmpl	$0,%eax
	jne	.Labort

	movl	240(%r8),%r14d
	xorq	%rdi,%rdi
	leaq	(%rdi,%r14,4),%rcx
	movq	%r8,%rsi
	leaq	(%r8,%rcx,4),%rdi
.align	4
.Linvert:
	movq	0(%rsi),%rax
	movq	8(%rsi),%rbx
	movq	0(%rdi),%rcx
	movq	8(%rdi),%rdx
	movq	%rax,0(%rdi)
	movq	%rbx,8(%rdi)
	movq	%rcx,0(%rsi)
	movq	%rdx,8(%rsi)
	leaq	16(%rsi),%rsi
	leaq	-16(%rdi),%rdi
	cmpq	%rsi,%rdi
	jne	.Linvert

	leaq	.LAES_Te+2048+1024(%rip),%rax

	movq	40(%rax),%rsi
	movq	48(%rax),%rdi
	movq	56(%rax),%rbp

	movq	%r8,%r15
	subl	$1,%r14d
.align	4
.Lpermute:
	leaq	16(%r15),%r15
	movq	0(%r15),%rax
	movq	8(%r15),%rcx
	movq	%rax,%rbx
	movq	%rcx,%rdx
	andq	%rsi,%rbx
	andq	%rsi,%rdx
	movq	%rbx,%r9
	movq	%rdx,%r12
	shrq	$7,%r9
	leaq	(%rax,%rax,1),%r8
	shrq	$7,%r12
	leaq	(%rcx,%rcx,1),%r11
	subq	%r9,%rbx
	subq	%r12,%rdx
	andq	%rdi,%r8
	andq	%rdi,%r11
	andq	%rbp,%rbx
	andq	%rbp,%rdx
	xorq	%r8,%rbx
	xorq	%r11,%rdx
	movq	%rbx,%r8
	movq	%rdx,%r11

	andq	%rsi,%rbx
	andq	%rsi,%rdx
	movq	%rbx,%r10
	movq	%rdx,%r13
	shrq	$7,%r10
	leaq	(%r8,%r8,1),%r9
	shrq	$7,%r13
	leaq	(%r11,%r11,1),%r12
	subq	%r10,%rbx
	subq	%r13,%rdx
	andq	%rdi,%r9
	andq	%rdi,%r12
	andq	%rbp,%rbx
	andq	%rbp,%rdx
	xorq	%r9,%rbx
	xorq	%r12,%rdx
	movq	%rbx,%r9
	movq	%rdx,%r12

	andq	%rsi,%rbx
	andq	%rsi,%rdx
	movq	%rbx,%r10
	movq	%rdx,%r13
	shrq	$7,%r10
	xorq	%rax,%r8
	shrq	$7,%r13
	xorq	%rcx,%r11
	subq	%r10,%rbx
	subq	%r13,%rdx
	leaq	(%r9,%r9,1),%r10
	leaq	(%r12,%r12,1),%r13
	xorq	%rax,%r9
	xorq	%rcx,%r12
	andq	%rdi,%r10
	andq	%rdi,%r13
	andq	%rbp,%rbx
	andq	%rbp,%rdx
	xorq	%rbx,%r10
	xorq	%rdx,%r13

	xorq	%r10,%rax
	xorq	%r13,%rcx
	xorq	%r10,%r8
	xorq	%r13,%r11
	movq	%rax,%rbx
	movq	%rcx,%rdx
	xorq	%r10,%r9
	xorq	%r13,%r12
	shrq	$32,%rbx
	shrq	$32,%rdx
	xorq	%r8,%r10
	xorq	%r11,%r13
	roll	$8,%eax
	roll	$8,%ecx
	xorq	%r9,%r10
	xorq	%r12,%r13

	roll	$8,%ebx
	roll	$8,%edx
	xorl	%r10d,%eax
	xorl	%r13d,%ecx
	shrq	$32,%r10
	shrq	$32,%r13
	xorl	%r10d,%ebx
	xorl	%r13d,%edx

	movq	%r8,%r10
	movq	%r11,%r13
	shrq	$32,%r10
	shrq	$32,%r13
	roll	$24,%r8d
	roll	$24,%r11d
	roll	$24,%r10d
	roll	$24,%r13d
	xorl	%r8d,%eax
	xorl	%r11d,%ecx
	movq	%r9,%r8
	movq	%r12,%r11
	xorl	%r10d,%ebx
	xorl	%r13d,%edx


	shrq	$32,%r8
	shrq	$32,%r11

	roll	$16,%r9d
	roll	$16,%r12d

	roll	$16,%r8d
	roll	$16,%r11d

	xorl	%r9d,%eax
	xorl	%r12d,%ecx

	xorl	%r8d,%ebx
	xorl	%r11d,%edx
	movl	%eax,0(%r15)
	movl	%ebx,4(%r15)
	movl	%ecx,8(%r15)
	movl	%edx,12(%r15)
	subl	$1,%r14d
	jnz	.Lpermute

	xorq	%rax,%rax
.Labort:
	movq	8(%rsp),%r15
	movq	16(%rsp),%r14
	movq	24(%rsp),%r13
	movq	32(%rsp),%r12
	movq	40(%rsp),%rbp
	movq	48(%rsp),%rbx
	addq	$56,%rsp
.Ldec_key_epilogue:
	.byte	0xf3,0xc3
.size	private_AES_set_decrypt_key,.-private_AES_set_decrypt_key
.globl	AES_cbc_encrypt
.type	AES_cbc_encrypt,@function
.align	16

.globl	asm_AES_cbc_encrypt
.hidden	asm_AES_cbc_encrypt
asm_AES_cbc_encrypt:
AES_cbc_encrypt:
	cmpq	$0,%rdx
	je	.Lcbc_epilogue
	pushfq
	pushq	%rbx
	pushq	%rbp
	pushq	%r12
	pushq	%r13
	pushq	%r14
	pushq	%r15
.Lcbc_prologue:

	cld
	movl	%r9d,%r9d

	leaq	.LAES_Te(%rip),%r14
	cmpq	$0,%r9
	jne	.Lcbc_picked_te
	leaq	.LAES_Td(%rip),%r14
.Lcbc_picked_te:

	movl	OPENSSL_ia32cap_P(%rip),%r10d
	cmpq	$512,%rdx
	jb	.Lcbc_slow_prologue
	testq	$15,%rdx
	jnz	.Lcbc_slow_prologue
	btl	$28,%r10d
	jc	.Lcbc_slow_prologue


	leaq	-88-248(%rsp),%r15
	andq	$-64,%r15


	movq	%r14,%r10
	leaq	2304(%r14),%r11
	movq	%r15,%r12
	andq	$4095,%r10
	andq	$4095,%r11
	andq	$4095,%r12

	cmpq	%r11,%r12
	jb	.Lcbc_te_break_out
	subq	%r11,%r12
	subq	%r12,%r15
	jmp	.Lcbc_te_ok
.Lcbc_te_break_out:
	subq	%r10,%r12
	andq	$4095,%r12
	addq	$320,%r12
	subq	%r12,%r15
.align	4
.Lcbc_te_ok:

	xchgq	%rsp,%r15

	movq	%r15,16(%rsp)
.Lcbc_fast_body:
	movq	%rdi,24(%rsp)
	movq	%rsi,32(%rsp)
	movq	%rdx,40(%rsp)
	movq	%rcx,48(%rsp)
	movq	%r8,56(%rsp)
	movl	$0,80+240(%rsp)
	movq	%r8,%rbp
	movq	%r9,%rbx
	movq	%rsi,%r9
	movq	%rdi,%r8
	movq	%rcx,%r15

	movl	240(%r15),%eax

	movq	%r15,%r10
	subq	%r14,%r10
	andq	$4095,%r10
	cmpq	$2304,%r10
	jb	.Lcbc_do_ecopy
	cmpq	$4096-248,%r10
	jb	.Lcbc_skip_ecopy
.align	4
.Lcbc_do_ecopy:
	movq	%r15,%rsi
	leaq	80(%rsp),%rdi
	leaq	80(%rsp),%r15
	movl	$30,%ecx
.long	0x90A548F3	
	movl	%eax,(%rdi)
.Lcbc_skip_ecopy:
	movq	%r15,0(%rsp)

	movl	$18,%ecx
.align	4
.Lcbc_prefetch_te:
	movq	0(%r14),%r10
	movq	32(%r14),%r11
	movq	64(%r14),%r12
	movq	96(%r14),%r13
	leaq	128(%r14),%r14
	subl	$1,%ecx
	jnz	.Lcbc_prefetch_te
	leaq	-2304(%r14),%r14

	cmpq	$0,%rbx
	je	.LFAST_DECRYPT


	movl	0(%rbp),%eax
	movl	4(%rbp),%ebx
	movl	8(%rbp),%ecx
	movl	12(%rbp),%edx

.align	4
.Lcbc_fast_enc_loop:
	xorl	0(%r8),%eax
	xorl	4(%r8),%ebx
	xorl	8(%r8),%ecx
	xorl	12(%r8),%edx
	movq	0(%rsp),%r15
	movq	%r8,24(%rsp)

	call	_x86_64_AES_encrypt

	movq	24(%rsp),%r8
	movq	40(%rsp),%r10
	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	leaq	16(%r8),%r8
	leaq	16(%r9),%r9
	subq	$16,%r10
	testq	$-16,%r10
	movq	%r10,40(%rsp)
	jnz	.Lcbc_fast_enc_loop
	movq	56(%rsp),%rbp
	movl	%eax,0(%rbp)
	movl	%ebx,4(%rbp)
	movl	%ecx,8(%rbp)
	movl	%edx,12(%rbp)

	jmp	.Lcbc_fast_cleanup


.align	16
.LFAST_DECRYPT:
	cmpq	%r8,%r9
	je	.Lcbc_fast_dec_in_place

	movq	%rbp,64(%rsp)
.align	4
.Lcbc_fast_dec_loop:
	movl	0(%r8),%eax
	movl	4(%r8),%ebx
	movl	8(%r8),%ecx
	movl	12(%r8),%edx
	movq	0(%rsp),%r15
	movq	%r8,24(%rsp)

	call	_x86_64_AES_decrypt

	movq	64(%rsp),%rbp
	movq	24(%rsp),%r8
	movq	40(%rsp),%r10
	xorl	0(%rbp),%eax
	xorl	4(%rbp),%ebx
	xorl	8(%rbp),%ecx
	xorl	12(%rbp),%edx
	movq	%r8,%rbp

	subq	$16,%r10
	movq	%r10,40(%rsp)
	movq	%rbp,64(%rsp)

	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	leaq	16(%r8),%r8
	leaq	16(%r9),%r9
	jnz	.Lcbc_fast_dec_loop
	movq	56(%rsp),%r12
	movq	0(%rbp),%r10
	movq	8(%rbp),%r11
	movq	%r10,0(%r12)
	movq	%r11,8(%r12)
	jmp	.Lcbc_fast_cleanup

.align	16
.Lcbc_fast_dec_in_place:
	movq	0(%rbp),%r10
	movq	8(%rbp),%r11
	movq	%r10,0+64(%rsp)
	movq	%r11,8+64(%rsp)
.align	4
.Lcbc_fast_dec_in_place_loop:
	movl	0(%r8),%eax
	movl	4(%r8),%ebx
	movl	8(%r8),%ecx
	movl	12(%r8),%edx
	movq	0(%rsp),%r15
	movq	%r8,24(%rsp)

	call	_x86_64_AES_decrypt

	movq	24(%rsp),%r8
	movq	40(%rsp),%r10
	xorl	0+64(%rsp),%eax
	xorl	4+64(%rsp),%ebx
	xorl	8+64(%rsp),%ecx
	xorl	12+64(%rsp),%edx

	movq	0(%r8),%r11
	movq	8(%r8),%r12
	subq	$16,%r10
	jz	.Lcbc_fast_dec_in_place_done

	movq	%r11,0+64(%rsp)
	movq	%r12,8+64(%rsp)

	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	leaq	16(%r8),%r8
	leaq	16(%r9),%r9
	movq	%r10,40(%rsp)
	jmp	.Lcbc_fast_dec_in_place_loop
.Lcbc_fast_dec_in_place_done:
	movq	56(%rsp),%rdi
	movq	%r11,0(%rdi)
	movq	%r12,8(%rdi)

	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

.align	4
.Lcbc_fast_cleanup:
	cmpl	$0,80+240(%rsp)
	leaq	80(%rsp),%rdi
	je	.Lcbc_exit
	movl	$30,%ecx
	xorq	%rax,%rax
.long	0x90AB48F3	

	jmp	.Lcbc_exit


.align	16
.Lcbc_slow_prologue:

	leaq	-88(%rsp),%rbp
	andq	$-64,%rbp

	leaq	-88-63(%rcx),%r10
	subq	%rbp,%r10
	negq	%r10
	andq	$960,%r10
	subq	%r10,%rbp

	xchgq	%rsp,%rbp

	movq	%rbp,16(%rsp)
.Lcbc_slow_body:




	movq	%r8,56(%rsp)
	movq	%r8,%rbp
	movq	%r9,%rbx
	movq	%rsi,%r9
	movq	%rdi,%r8
	movq	%rcx,%r15
	movq	%rdx,%r10

	movl	240(%r15),%eax
	movq	%r15,0(%rsp)
	shll	$4,%eax
	leaq	(%r15,%rax,1),%rax
	movq	%rax,8(%rsp)


	leaq	2048(%r14),%r14
	leaq	768-8(%rsp),%rax
	subq	%r14,%rax
	andq	$768,%rax
	leaq	(%r14,%rax,1),%r14

	cmpq	$0,%rbx
	je	.LSLOW_DECRYPT


	testq	$-16,%r10
	movl	0(%rbp),%eax
	movl	4(%rbp),%ebx
	movl	8(%rbp),%ecx
	movl	12(%rbp),%edx
	jz	.Lcbc_slow_enc_tail	

.align	4
.Lcbc_slow_enc_loop:
	xorl	0(%r8),%eax
	xorl	4(%r8),%ebx
	xorl	8(%r8),%ecx
	xorl	12(%r8),%edx
	movq	0(%rsp),%r15
	movq	%r8,24(%rsp)
	movq	%r9,32(%rsp)
	movq	%r10,40(%rsp)

	call	_x86_64_AES_encrypt_compact

	movq	24(%rsp),%r8
	movq	32(%rsp),%r9
	movq	40(%rsp),%r10
	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	leaq	16(%r8),%r8
	leaq	16(%r9),%r9
	subq	$16,%r10
	testq	$-16,%r10
	jnz	.Lcbc_slow_enc_loop
	testq	$15,%r10
	jnz	.Lcbc_slow_enc_tail
	movq	56(%rsp),%rbp
	movl	%eax,0(%rbp)
	movl	%ebx,4(%rbp)
	movl	%ecx,8(%rbp)
	movl	%edx,12(%rbp)

	jmp	.Lcbc_exit

.align	4
.Lcbc_slow_enc_tail:
	movq	%rax,%r11
	movq	%rcx,%r12
	movq	%r10,%rcx
	movq	%r8,%rsi
	movq	%r9,%rdi
.long	0x9066A4F3		
	movq	$16,%rcx
	subq	%r10,%rcx
	xorq	%rax,%rax
.long	0x9066AAF3		
	movq	%r9,%r8
	movq	$16,%r10
	movq	%r11,%rax
	movq	%r12,%rcx
	jmp	.Lcbc_slow_enc_loop	

.align	16
.LSLOW_DECRYPT:
	shrq	$3,%rax
	addq	%rax,%r14

	movq	0(%rbp),%r11
	movq	8(%rbp),%r12
	movq	%r11,0+64(%rsp)
	movq	%r12,8+64(%rsp)

.align	4
.Lcbc_slow_dec_loop:
	movl	0(%r8),%eax
	movl	4(%r8),%ebx
	movl	8(%r8),%ecx
	movl	12(%r8),%edx
	movq	0(%rsp),%r15
	movq	%r8,24(%rsp)
	movq	%r9,32(%rsp)
	movq	%r10,40(%rsp)

	call	_x86_64_AES_decrypt_compact

	movq	24(%rsp),%r8
	movq	32(%rsp),%r9
	movq	40(%rsp),%r10
	xorl	0+64(%rsp),%eax
	xorl	4+64(%rsp),%ebx
	xorl	8+64(%rsp),%ecx
	xorl	12+64(%rsp),%edx

	movq	0(%r8),%r11
	movq	8(%r8),%r12
	subq	$16,%r10
	jc	.Lcbc_slow_dec_partial
	jz	.Lcbc_slow_dec_done

	movq	%r11,0+64(%rsp)
	movq	%r12,8+64(%rsp)

	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	leaq	16(%r8),%r8
	leaq	16(%r9),%r9
	jmp	.Lcbc_slow_dec_loop
.Lcbc_slow_dec_done:
	movq	56(%rsp),%rdi
	movq	%r11,0(%rdi)
	movq	%r12,8(%rdi)

	movl	%eax,0(%r9)
	movl	%ebx,4(%r9)
	movl	%ecx,8(%r9)
	movl	%edx,12(%r9)

	jmp	.Lcbc_exit

.align	4
.Lcbc_slow_dec_partial:
	movq	56(%rsp),%rdi
	movq	%r11,0(%rdi)
	movq	%r12,8(%rdi)

	movl	%eax,0+64(%rsp)
	movl	%ebx,4+64(%rsp)
	movl	%ecx,8+64(%rsp)
	movl	%edx,12+64(%rsp)

	movq	%r9,%rdi
	leaq	64(%rsp),%rsi
	leaq	16(%r10),%rcx
.long	0x9066A4F3	
	jmp	.Lcbc_exit

.align	16
.Lcbc_exit:
	movq	16(%rsp),%rsi
	movq	(%rsi),%r15
	movq	8(%rsi),%r14
	movq	16(%rsi),%r13
	movq	24(%rsi),%r12
	movq	32(%rsi),%rbp
	movq	40(%rsi),%rbx
	leaq	48(%rsi),%rsp
.Lcbc_popfq:
	popfq
.Lcbc_epilogue:
	.byte	0xf3,0xc3
.size	AES_cbc_encrypt,.-AES_cbc_encrypt
.align	64
.LAES_Te:
.long	0xa56363c6,0xa56363c6
.long	0x847c7cf8,0x847c7cf8
.long	0x997777ee,0x997777ee
.long	0x8d7b7bf6,0x8d7b7bf6
.long	0x0df2f2ff,0x0df2f2ff
.long	0xbd6b6bd6,0xbd6b6bd6
.long	0xb16f6fde,0xb16f6fde
.long	0x54c5c591,0x54c5c591
.long	0x50303060,0x50303060
.long	0x03010102,0x03010102
.long	0xa96767ce,0xa96767ce
.long	0x7d2b2b56,0x7d2b2b56
.long	0x19fefee7,0x19fefee7
.long	0x62d7d7b5,0x62d7d7b5
.long	0xe6abab4d,0xe6abab4d
.long	0x9a7676ec,0x9a7676ec
.long	0x45caca8f,0x45caca8f
.long	0x9d82821f,0x9d82821f
.long	0x40c9c989,0x40c9c989
.long	0x877d7dfa,0x877d7dfa
.long	0x15fafaef,0x15fafaef
.long	0xeb5959b2,0xeb5959b2
.long	0xc947478e,0xc947478e
.long	0x0bf0f0fb,0x0bf0f0fb
.long	0xecadad41,0xecadad41
.long	0x67d4d4b3,0x67d4d4b3
.long	0xfda2a25f,0xfda2a25f
.long	0xeaafaf45,0xeaafaf45
.long	0xbf9c9c23,0xbf9c9c23
.long	0xf7a4a453,0xf7a4a453
.long	0x967272e4,0x967272e4
.long	0x5bc0c09b,0x5bc0c09b
.long	0xc2b7b775,0xc2b7b775
.long	0x1cfdfde1,0x1cfdfde1
.long	0xae93933d,0xae93933d
.long	0x6a26264c,0x6a26264c
.long	0x5a36366c,0x5a36366c
.long	0x413f3f7e,0x413f3f7e
.long	0x02f7f7f5,0x02f7f7f5
.long	0x4fcccc83,0x4fcccc83
.long	0x5c343468,0x5c343468
.long	0xf4a5a551,0xf4a5a551
.long	0x34e5e5d1,0x34e5e5d1
.long	0x08f1f1f9,0x08f1f1f9
.long	0x937171e2,0x937171e2
.long	0x73d8d8ab,0x73d8d8ab
.long	0x53313162,0x53313162
.long	0x3f15152a,0x3f15152a
.long	0x0c040408,0x0c040408
.long	0x52c7c795,0x52c7c795
.long	0x65232346,0x65232346
.long	0x5ec3c39d,0x5ec3c39d
.long	0x28181830,0x28181830
.long	0xa1969637,0xa1969637
.long	0x0f05050a,0x0f05050a
.long	0xb59a9a2f,0xb59a9a2f
.long	0x0907070e,0x0907070e
.long	0x36121224,0x36121224
.long	0x9b80801b,0x9b80801b
.long	0x3de2e2df,0x3de2e2df
.long	0x26ebebcd,0x26ebebcd
.long	0x6927274e,0x6927274e
.long	0xcdb2b27f,0xcdb2b27f
.long	0x9f7575ea,0x9f7575ea
.long	0x1b090912,0x1b090912
.long	0x9e83831d,0x9e83831d
.long	0x742c2c58,0x742c2c58
.long	0x2e1a1a34,0x2e1a1a34
.long	0x2d1b1b36,0x2d1b1b36
.long	0xb26e6edc,0xb26e6edc
.long	0xee5a5ab4,0xee5a5ab4
.long	0xfba0a05b,0xfba0a05b
.long	0xf65252a4,0xf65252a4
.long	0x4d3b3b76,0x4d3b3b76
.long	0x61d6d6b7,0x61d6d6b7
.long	0xceb3b37d,0xceb3b37d
.long	0x7b292952,0x7b292952
.long	0x3ee3e3dd,0x3ee3e3dd
.long	0x712f2f5e,0x712f2f5e
.long	0x97848413,0x97848413
.long	0xf55353a6,0xf55353a6
.long	0x68d1d1b9,0x68d1d1b9
.long	0x00000000,0x00000000
.long	0x2cededc1,0x2cededc1
.long	0x60202040,0x60202040
.long	0x1ffcfce3,0x1ffcfce3
.long	0xc8b1b179,0xc8b1b179
.long	0xed5b5bb6,0xed5b5bb6
.long	0xbe6a6ad4,0xbe6a6ad4
.long	0x46cbcb8d,0x46cbcb8d
.long	0xd9bebe67,0xd9bebe67
.long	0x4b393972,0x4b393972
.long	0xde4a4a94,0xde4a4a94
.long	0xd44c4c98,0xd44c4c98
.long	0xe85858b0,0xe85858b0
.long	0x4acfcf85,0x4acfcf85
.long	0x6bd0d0bb,0x6bd0d0bb
.long	0x2aefefc5,0x2aefefc5
.long	0xe5aaaa4f,0xe5aaaa4f
.long	0x16fbfbed,0x16fbfbed
.long	0xc5434386,0xc5434386
.long	0xd74d4d9a,0xd74d4d9a
.long	0x55333366,0x55333366
.long	0x94858511,0x94858511
.long	0xcf45458a,0xcf45458a
.long	0x10f9f9e9,0x10f9f9e9
.long	0x06020204,0x06020204
.long	0x817f7ffe,0x817f7ffe
.long	0xf05050a0,0xf05050a0
.long	0x443c3c78,0x443c3c78
.long	0xba9f9f25,0xba9f9f25
.long	0xe3a8a84b,0xe3a8a84b
.long	0xf35151a2,0xf35151a2
.long	0xfea3a35d,0xfea3a35d
.long	0xc0404080,0xc0404080
.long	0x8a8f8f05,0x8a8f8f05
.long	0xad92923f,0xad92923f
.long	0xbc9d9d21,0xbc9d9d21
.long	0x48383870,0x48383870
.long	0x04f5f5f1,0x04f5f5f1
.long	0xdfbcbc63,0xdfbcbc63
.long	0xc1b6b677,0xc1b6b677
.long	0x75dadaaf,0x75dadaaf
.long	0x63212142,0x63212142
.long	0x30101020,0x30101020
.long	0x1affffe5,0x1affffe5
.long	0x0ef3f3fd,0x0ef3f3fd
.long	0x6dd2d2bf,0x6dd2d2bf
.long	0x4ccdcd81,0x4ccdcd81
.long	0x140c0c18,0x140c0c18
.long	0x35131326,0x35131326
.long	0x2fececc3,0x2fececc3
.long	0xe15f5fbe,0xe15f5fbe
.long	0xa2979735,0xa2979735
.long	0xcc444488,0xcc444488
.long	0x3917172e,0x3917172e
.long	0x57c4c493,0x57c4c493
.long	0xf2a7a755,0xf2a7a755
.long	0x827e7efc,0x827e7efc
.long	0x473d3d7a,0x473d3d7a
.long	0xac6464c8,0xac6464c8
.long	0xe75d5dba,0xe75d5dba
.long	0x2b191932,0x2b191932
.long	0x957373e6,0x957373e6
.long	0xa06060c0,0xa06060c0
.long	0x98818119,0x98818119
.long	0xd14f4f9e,0xd14f4f9e
.long	0x7fdcdca3,0x7fdcdca3
.long	0x66222244,0x66222244
.long	0x7e2a2a54,0x7e2a2a54
.long	0xab90903b,0xab90903b
.long	0x8388880b,0x8388880b
.long	0xca46468c,0xca46468c
.long	0x29eeeec7,0x29eeeec7
.long	0xd3b8b86b,0xd3b8b86b
.long	0x3c141428,0x3c141428
.long	0x79dedea7,0x79dedea7
.long	0xe25e5ebc,0xe25e5ebc
.long	0x1d0b0b16,0x1d0b0b16
.long	0x76dbdbad,0x76dbdbad
.long	0x3be0e0db,0x3be0e0db
.long	0x56323264,0x56323264
.long	0x4e3a3a74,0x4e3a3a74
.long	0x1e0a0a14,0x1e0a0a14
.long	0xdb494992,0xdb494992
.long	0x0a06060c,0x0a06060c
.long	0x6c242448,0x6c242448
.long	0xe45c5cb8,0xe45c5cb8
.long	0x5dc2c29f,0x5dc2c29f
.long	0x6ed3d3bd,0x6ed3d3bd
.long	0xefacac43,0xefacac43
.long	0xa66262c4,0xa66262c4
.long	0xa8919139,0xa8919139
.long	0xa4959531,0xa4959531
.long	0x37e4e4d3,0x37e4e4d3
.long	0x8b7979f2,0x8b7979f2
.long	0x32e7e7d5,0x32e7e7d5
.long	0x43c8c88b,0x43c8c88b
.long	0x5937376e,0x5937376e
.long	0xb76d6dda,0xb76d6dda
.long	0x8c8d8d01,0x8c8d8d01
.long	0x64d5d5b1,0x64d5d5b1
.long	0xd24e4e9c,0xd24e4e9c
.long	0xe0a9a949,0xe0a9a949
.long	0xb46c6cd8,0xb46c6cd8
.long	0xfa5656ac,0xfa5656ac
.long	0x07f4f4f3,0x07f4f4f3
.long	0x25eaeacf,0x25eaeacf
.long	0xaf6565ca,0xaf6565ca
.long	0x8e7a7af4,0x8e7a7af4
.long	0xe9aeae47,0xe9aeae47
.long	0x18080810,0x18080810
.long	0xd5baba6f,0xd5baba6f
.long	0x887878f0,0x887878f0
.long	0x6f25254a,0x6f25254a
.long	0x722e2e5c,0x722e2e5c
.long	0x241c1c38,0x241c1c38
.long	0xf1a6a657,0xf1a6a657
.long	0xc7b4b473,0xc7b4b473
.long	0x51c6c697,0x51c6c697
.long	0x23e8e8cb,0x23e8e8cb
.long	0x7cdddda1,0x7cdddda1
.long	0x9c7474e8,0x9c7474e8
.long	0x211f1f3e,0x211f1f3e
.long	0xdd4b4b96,0xdd4b4b96
.long	0xdcbdbd61,0xdcbdbd61
.long	0x868b8b0d,0x868b8b0d
.long	0x858a8a0f,0x858a8a0f
.long	0x907070e0,0x907070e0
.long	0x423e3e7c,0x423e3e7c
.long	0xc4b5b571,0xc4b5b571
.long	0xaa6666cc,0xaa6666cc
.long	0xd8484890,0xd8484890
.long	0x05030306,0x05030306
.long	0x01f6f6f7,0x01f6f6f7
.long	0x120e0e1c,0x120e0e1c
.long	0xa36161c2,0xa36161c2
.long	0x5f35356a,0x5f35356a
.long	0xf95757ae,0xf95757ae
.long	0xd0b9b969,0xd0b9b969
.long	0x91868617,0x91868617
.long	0x58c1c199,0x58c1c199
.long	0x271d1d3a,0x271d1d3a
.long	0xb99e9e27,0xb99e9e27
.long	0x38e1e1d9,0x38e1e1d9
.long	0x13f8f8eb,0x13f8f8eb
.long	0xb398982b,0xb398982b
.long	0x33111122,0x33111122
.long	0xbb6969d2,0xbb6969d2
.long	0x70d9d9a9,0x70d9d9a9
.long	0x898e8e07,0x898e8e07
.long	0xa7949433,0xa7949433
.long	0xb69b9b2d,0xb69b9b2d
.long	0x221e1e3c,0x221e1e3c
.long	0x92878715,0x92878715
.long	0x20e9e9c9,0x20e9e9c9
.long	0x49cece87,0x49cece87
.long	0xff5555aa,0xff5555aa
.long	0x78282850,0x78282850
.long	0x7adfdfa5,0x7adfdfa5
.long	0x8f8c8c03,0x8f8c8c03
.long	0xf8a1a159,0xf8a1a159
.long	0x80898909,0x80898909