MMX/SSE2 optimizations

git-svn-id: svn://svn.code.sf.net/p/cryptopp/code/trunk/c5@290 57ff6487-cd31-0410-9ec3-f628ee90f5f0

1 files changed, 77 insertions, 87 deletions

diff --git a/rdtables.cpp b/rdtables.cpp
index 13a3614..af0a072 100644
--- a/rdtables.cpp
+++ b/rdtables.cpp
@@ -94,7 +94,74 @@ const byte Rijndael::Base::Se[256] = {
     0xb0, 0x54, 0xbb, 0x16,
 };
 
-const word32 Rijndael::Base::Te0[256] = {
+const byte Rijndael::Base::Sd[256] = {
+    0x52, 0x09, 0x6a, 0xd5,
+    0x30, 0x36, 0xa5, 0x38,
+    0xbf, 0x40, 0xa3, 0x9e,
+    0x81, 0xf3, 0xd7, 0xfb,
+    0x7c, 0xe3, 0x39, 0x82,
+    0x9b, 0x2f, 0xff, 0x87,
+    0x34, 0x8e, 0x43, 0x44,
+    0xc4, 0xde, 0xe9, 0xcb,
+    0x54, 0x7b, 0x94, 0x32,
+    0xa6, 0xc2, 0x23, 0x3d,
+    0xee, 0x4c, 0x95, 0x0b,
+    0x42, 0xfa, 0xc3, 0x4e,
+    0x08, 0x2e, 0xa1, 0x66,
+    0x28, 0xd9, 0x24, 0xb2,
+    0x76, 0x5b, 0xa2, 0x49,
+    0x6d, 0x8b, 0xd1, 0x25,
+    0x72, 0xf8, 0xf6, 0x64,
+    0x86, 0x68, 0x98, 0x16,
+    0xd4, 0xa4, 0x5c, 0xcc,
+    0x5d, 0x65, 0xb6, 0x92,
+    0x6c, 0x70, 0x48, 0x50,
+    0xfd, 0xed, 0xb9, 0xda,
+    0x5e, 0x15, 0x46, 0x57,
+    0xa7, 0x8d, 0x9d, 0x84,
+    0x90, 0xd8, 0xab, 0x00,
+    0x8c, 0xbc, 0xd3, 0x0a,
+    0xf7, 0xe4, 0x58, 0x05,
+    0xb8, 0xb3, 0x45, 0x06,
+    0xd0, 0x2c, 0x1e, 0x8f,
+    0xca, 0x3f, 0x0f, 0x02,
+    0xc1, 0xaf, 0xbd, 0x03,
+    0x01, 0x13, 0x8a, 0x6b,
+    0x3a, 0x91, 0x11, 0x41,
+    0x4f, 0x67, 0xdc, 0xea,
+    0x97, 0xf2, 0xcf, 0xce,
+    0xf0, 0xb4, 0xe6, 0x73,
+    0x96, 0xac, 0x74, 0x22,
+    0xe7, 0xad, 0x35, 0x85,
+    0xe2, 0xf9, 0x37, 0xe8,
+    0x1c, 0x75, 0xdf, 0x6e,
+    0x47, 0xf1, 0x1a, 0x71,
+    0x1d, 0x29, 0xc5, 0x89,
+    0x6f, 0xb7, 0x62, 0x0e,
+    0xaa, 0x18, 0xbe, 0x1b,
+    0xfc, 0x56, 0x3e, 0x4b,
+    0xc6, 0xd2, 0x79, 0x20,
+    0x9a, 0xdb, 0xc0, 0xfe,
+    0x78, 0xcd, 0x5a, 0xf4,
+    0x1f, 0xdd, 0xa8, 0x33,
+    0x88, 0x07, 0xc7, 0x31,
+    0xb1, 0x12, 0x10, 0x59,
+    0x27, 0x80, 0xec, 0x5f,
+    0x60, 0x51, 0x7f, 0xa9,
+    0x19, 0xb5, 0x4a, 0x0d,
+    0x2d, 0xe5, 0x7a, 0x9f,
+    0x93, 0xc9, 0x9c, 0xef,
+    0xa0, 0xe0, 0x3b, 0x4d,
+    0xae, 0x2a, 0xf5, 0xb0,
+    0xc8, 0xeb, 0xbb, 0x3c,
+    0x83, 0x53, 0x99, 0x61,
+    0x17, 0x2b, 0x04, 0x7e,
+    0xba, 0x77, 0xd6, 0x26,
+    0xe1, 0x69, 0x14, 0x63,
+    0x55, 0x21, 0x0c, 0x7d,
+};
+
+const word32 Rijndael::Base::Te[4*256] = {
     0xc66363a5U, 0xf87c7c84U, 0xee777799U, 0xf67b7b8dU,
     0xfff2f20dU, 0xd66b6bbdU, 0xde6f6fb1U, 0x91c5c554U,
     0x60303050U, 0x02010103U, 0xce6767a9U, 0x562b2b7dU,
@@ -159,9 +226,8 @@ const word32 Rijndael::Base::Te0[256] = {
     0x65bfbfdaU, 0xd7e6e631U, 0x844242c6U, 0xd06868b8U,
     0x824141c3U, 0x299999b0U, 0x5a2d2d77U, 0x1e0f0f11U,
     0x7bb0b0cbU, 0xa85454fcU, 0x6dbbbbd6U, 0x2c16163aU,
-};
-const word32 Rijndael::Base::Te1[256] = {
-    0xa5c66363U, 0x84f87c7cU, 0x99ee7777U, 0x8df67b7bU,
+
+	0xa5c66363U, 0x84f87c7cU, 0x99ee7777U, 0x8df67b7bU,
     0x0dfff2f2U, 0xbdd66b6bU, 0xb1de6f6fU, 0x5491c5c5U,
     0x50603030U, 0x03020101U, 0xa9ce6767U, 0x7d562b2bU,
     0x19e7fefeU, 0x62b5d7d7U, 0xe64dababU, 0x9aec7676U,
@@ -225,9 +291,8 @@ const word32 Rijndael::Base::Te1[256] = {
     0xda65bfbfU, 0x31d7e6e6U, 0xc6844242U, 0xb8d06868U,
     0xc3824141U, 0xb0299999U, 0x775a2d2dU, 0x111e0f0fU,
     0xcb7bb0b0U, 0xfca85454U, 0xd66dbbbbU, 0x3a2c1616U,
-};
-const word32 Rijndael::Base::Te2[256] = {
-    0x63a5c663U, 0x7c84f87cU, 0x7799ee77U, 0x7b8df67bU,
+
+	0x63a5c663U, 0x7c84f87cU, 0x7799ee77U, 0x7b8df67bU,
     0xf20dfff2U, 0x6bbdd66bU, 0x6fb1de6fU, 0xc55491c5U,
     0x30506030U, 0x01030201U, 0x67a9ce67U, 0x2b7d562bU,
     0xfe19e7feU, 0xd762b5d7U, 0xabe64dabU, 0x769aec76U,
@@ -291,10 +356,8 @@ const word32 Rijndael::Base::Te2[256] = {
     0xbfda65bfU, 0xe631d7e6U, 0x42c68442U, 0x68b8d068U,
     0x41c38241U, 0x99b02999U, 0x2d775a2dU, 0x0f111e0fU,
     0xb0cb7bb0U, 0x54fca854U, 0xbbd66dbbU, 0x163a2c16U,
-};
 
-const word32 Rijndael::Base::Te3[256] = {
-    0x6363a5c6U, 0x7c7c84f8U, 0x777799eeU, 0x7b7b8df6U,
+	0x6363a5c6U, 0x7c7c84f8U, 0x777799eeU, 0x7b7b8df6U,
     0xf2f20dffU, 0x6b6bbdd6U, 0x6f6fb1deU, 0xc5c55491U,
     0x30305060U, 0x01010302U, 0x6767a9ceU, 0x2b2b7d56U,
     0xfefe19e7U, 0xd7d762b5U, 0xababe64dU, 0x76769aecU,
@@ -360,74 +423,7 @@ const word32 Rijndael::Base::Te3[256] = {
     0xb0b0cb7bU, 0x5454fca8U, 0xbbbbd66dU, 0x16163a2cU,
 };
 
-const byte Rijndael::Base::Sd[256] = {
-    0x52, 0x09, 0x6a, 0xd5,
-    0x30, 0x36, 0xa5, 0x38,
-    0xbf, 0x40, 0xa3, 0x9e,
-    0x81, 0xf3, 0xd7, 0xfb,
-    0x7c, 0xe3, 0x39, 0x82,
-    0x9b, 0x2f, 0xff, 0x87,
-    0x34, 0x8e, 0x43, 0x44,
-    0xc4, 0xde, 0xe9, 0xcb,
-    0x54, 0x7b, 0x94, 0x32,
-    0xa6, 0xc2, 0x23, 0x3d,
-    0xee, 0x4c, 0x95, 0x0b,
-    0x42, 0xfa, 0xc3, 0x4e,
-    0x08, 0x2e, 0xa1, 0x66,
-    0x28, 0xd9, 0x24, 0xb2,
-    0x76, 0x5b, 0xa2, 0x49,
-    0x6d, 0x8b, 0xd1, 0x25,
-    0x72, 0xf8, 0xf6, 0x64,
-    0x86, 0x68, 0x98, 0x16,
-    0xd4, 0xa4, 0x5c, 0xcc,
-    0x5d, 0x65, 0xb6, 0x92,
-    0x6c, 0x70, 0x48, 0x50,
-    0xfd, 0xed, 0xb9, 0xda,
-    0x5e, 0x15, 0x46, 0x57,
-    0xa7, 0x8d, 0x9d, 0x84,
-    0x90, 0xd8, 0xab, 0x00,
-    0x8c, 0xbc, 0xd3, 0x0a,
-    0xf7, 0xe4, 0x58, 0x05,
-    0xb8, 0xb3, 0x45, 0x06,
-    0xd0, 0x2c, 0x1e, 0x8f,
-    0xca, 0x3f, 0x0f, 0x02,
-    0xc1, 0xaf, 0xbd, 0x03,
-    0x01, 0x13, 0x8a, 0x6b,
-    0x3a, 0x91, 0x11, 0x41,
-    0x4f, 0x67, 0xdc, 0xea,
-    0x97, 0xf2, 0xcf, 0xce,
-    0xf0, 0xb4, 0xe6, 0x73,
-    0x96, 0xac, 0x74, 0x22,
-    0xe7, 0xad, 0x35, 0x85,
-    0xe2, 0xf9, 0x37, 0xe8,
-    0x1c, 0x75, 0xdf, 0x6e,
-    0x47, 0xf1, 0x1a, 0x71,
-    0x1d, 0x29, 0xc5, 0x89,
-    0x6f, 0xb7, 0x62, 0x0e,
-    0xaa, 0x18, 0xbe, 0x1b,
-    0xfc, 0x56, 0x3e, 0x4b,
-    0xc6, 0xd2, 0x79, 0x20,
-    0x9a, 0xdb, 0xc0, 0xfe,
-    0x78, 0xcd, 0x5a, 0xf4,
-    0x1f, 0xdd, 0xa8, 0x33,
-    0x88, 0x07, 0xc7, 0x31,
-    0xb1, 0x12, 0x10, 0x59,
-    0x27, 0x80, 0xec, 0x5f,
-    0x60, 0x51, 0x7f, 0xa9,
-    0x19, 0xb5, 0x4a, 0x0d,
-    0x2d, 0xe5, 0x7a, 0x9f,
-    0x93, 0xc9, 0x9c, 0xef,
-    0xa0, 0xe0, 0x3b, 0x4d,
-    0xae, 0x2a, 0xf5, 0xb0,
-    0xc8, 0xeb, 0xbb, 0x3c,
-    0x83, 0x53, 0x99, 0x61,
-    0x17, 0x2b, 0x04, 0x7e,
-    0xba, 0x77, 0xd6, 0x26,
-    0xe1, 0x69, 0x14, 0x63,
-    0x55, 0x21, 0x0c, 0x7d,
-};
-
-const word32 Rijndael::Base::Td0[256] = {
+const word32 Rijndael::Base::Td[4*256] = {
     0x51f4a750U, 0x7e416553U, 0x1a17a4c3U, 0x3a275e96U,
     0x3bab6bcbU, 0x1f9d45f1U, 0xacfa58abU, 0x4be30393U,
     0x2030fa55U, 0xad766df6U, 0x88cc7691U, 0xf5024c25U,
@@ -492,10 +488,8 @@ const word32 Rijndael::Base::Td0[256] = {
     0x161dc372U, 0xbce2250cU, 0x283c498bU, 0xff0d9541U,
     0x39a80171U, 0x080cb3deU, 0xd8b4e49cU, 0x6456c190U,
     0x7bcb8461U, 0xd532b670U, 0x486c5c74U, 0xd0b85742U,
-};
 
-const word32 Rijndael::Base::Td1[256] = {
-    0x5051f4a7U, 0x537e4165U, 0xc31a17a4U, 0x963a275eU,
+	0x5051f4a7U, 0x537e4165U, 0xc31a17a4U, 0x963a275eU,
     0xcb3bab6bU, 0xf11f9d45U, 0xabacfa58U, 0x934be303U,
     0x552030faU, 0xf6ad766dU, 0x9188cc76U, 0x25f5024cU,
     0xfc4fe5d7U, 0xd7c52acbU, 0x80263544U, 0x8fb562a3U,
@@ -559,10 +553,8 @@ const word32 Rijndael::Base::Td1[256] = {
     0x72161dc3U, 0x0cbce225U, 0x8b283c49U, 0x41ff0d95U,
     0x7139a801U, 0xde080cb3U, 0x9cd8b4e4U, 0x906456c1U,
     0x617bcb84U, 0x70d532b6U, 0x74486c5cU, 0x42d0b857U,
-};
 
-const word32 Rijndael::Base::Td2[256] = {
-    0xa75051f4U, 0x65537e41U, 0xa4c31a17U, 0x5e963a27U,
+	0xa75051f4U, 0x65537e41U, 0xa4c31a17U, 0x5e963a27U,
     0x6bcb3babU, 0x45f11f9dU, 0x58abacfaU, 0x03934be3U,
     0xfa552030U, 0x6df6ad76U, 0x769188ccU, 0x4c25f502U,
     0xd7fc4fe5U, 0xcbd7c52aU, 0x44802635U, 0xa38fb562U,
@@ -627,10 +619,8 @@ const word32 Rijndael::Base::Td2[256] = {
     0xc372161dU, 0x250cbce2U, 0x498b283cU, 0x9541ff0dU,
     0x017139a8U, 0xb3de080cU, 0xe49cd8b4U, 0xc1906456U,
     0x84617bcbU, 0xb670d532U, 0x5c74486cU, 0x5742d0b8U,
-};
 
-const word32 Rijndael::Base::Td3[256] = {
-    0xf4a75051U, 0x4165537eU, 0x17a4c31aU, 0x275e963aU,
+	0xf4a75051U, 0x4165537eU, 0x17a4c31aU, 0x275e963aU,
     0xab6bcb3bU, 0x9d45f11fU, 0xfa58abacU, 0xe303934bU,
     0x30fa5520U, 0x766df6adU, 0xcc769188U, 0x024c25f5U,
     0xe5d7fc4fU, 0x2acbd7c5U, 0x35448026U, 0x62a38fb5U,