Add an x86 assembly MD5 implementation; works, but needs optimization

3 files changed, 176 insertions, 0 deletions

diff --git a/modules/alg_ia32/md5.cpp b/modules/alg_ia32/md5.cpp
new file mode 100644
index 000000000..0d35adf49
--- /dev/null
+++ b/modules/alg_ia32/md5.cpp
@@ -0,0 +1,43 @@
+/*************************************************
+* MD5 Source File                                *
+* (C) 1999-2006 The Botan Project                *
+*************************************************/
+
+#include <botan/md5.h>
+#include <botan/bit_ops.h>
+
+namespace Botan {
+
+extern "C" void md5_core(u32bit[4], const byte[64], u32bit[16]);
+
+/*************************************************
+* MD5 Compression Function                       *
+*************************************************/
+void MD5::hash(const byte input[])
+   {
+   md5_core(digest, input, M);
+   }
+
+/*************************************************
+* Copy out the digest                            *
+*************************************************/
+void MD5::copy_out(byte output[])
+   {
+   for(u32bit j = 0; j != OUTPUT_LENGTH; ++j)
+      output[j] = get_byte(3 - (j % 4), digest[j/4]);
+   }
+
+/*************************************************
+* Clear memory of sensitive data                 *
+*************************************************/
+void MD5::clear() throw()
+   {
+   MDx_HashFunction::clear();
+   M.clear();
+   digest[0] = 0x67452301;
+   digest[1] = 0xEFCDAB89;
+   digest[2] = 0x98BADCFE;
+   digest[3] = 0x10325476;
+   }
+
+}
diff --git a/modules/alg_ia32/md5core.S b/modules/alg_ia32/md5core.S
new file mode 100644
index 000000000..a801a3681
--- /dev/null
+++ b/modules/alg_ia32/md5core.S
@@ -0,0 +1,131 @@
+/*************************************************
+* MD5 Source File                                *
+* (C) 1999-2006 The Botan Project                *
+*************************************************/
+
+#include <botan/asm_macr.h>
+
+START_LISTING(md5core.S)
+
+START_FUNCTION(md5_core)
+   SPILL_REGS()
+
+#define PUSHED 4
+
+   ASSIGN(EBP, ARG(2)) /* input block */
+   ASSIGN(EDI, ARG(3)) /* expanded words */
+
+   ZEROIZE(ESI)
+
+START_LOOP(.LOAD_INPUT)
+   ADD_IMM(ESI, 4)
+
+   ASSIGN(EAX, ARRAY4(EBP, 0))
+   ASSIGN(EBX, ARRAY4(EBP, 1))
+   ASSIGN(ECX, ARRAY4(EBP, 2))
+   ASSIGN(EDX, ARRAY4(EBP, 3))
+
+   ADD_IMM(EBP, 16)
+
+   ASSIGN(ARRAY4_INDIRECT(EDI,ESI,-4), EAX)
+   ASSIGN(ARRAY4_INDIRECT(EDI,ESI,-3), EBX)
+   ASSIGN(ARRAY4_INDIRECT(EDI,ESI,-2), ECX)
+   ASSIGN(ARRAY4_INDIRECT(EDI,ESI,-1), EDX)
+LOOP_UNTIL(ESI, IMM(16), .LOAD_INPUT)
+   
+   ASSIGN(EBP, ARG(1))
+   ASSIGN(EAX, ARRAY4(EBP, 0))
+   ASSIGN(EBX, ARRAY4(EBP, 1))
+   ASSIGN(ECX, ARRAY4(EBP, 2))
+   ASSIGN(EDX, ARRAY4(EBP, 3))
+
+#define T EBP
+   
+#define FF(A, B, C, D, MSG, S, MAGIC)    \
+   ASSIGN(T, C)                        ; \
+   XOR(T, D)                           ; \
+   AND(T, B)                           ; \
+   XOR(T, D)                           ; \
+   ADD(A, ARRAY4(EDI, MSG))            ; \
+   ADD3_IMM(A, T, MAGIC)               ; \
+   ROTL_IMM(A, S)                      ; \
+   ADD(A, B)                           ;
+
+#define GG(A, B, C, D, MSG, S, MAGIC)    \
+   ASSIGN(T, B)                        ; \
+   XOR(T, C)                           ; \
+   AND(T, D)                           ; \
+   XOR(T, C)                           ; \
+   ADD(A, ARRAY4(EDI, MSG))            ; \
+   ADD3_IMM(A, T, MAGIC)               ; \
+   ROTL_IMM(A, S)                      ; \
+   ADD(A, B)                           ;
+
+#define HH(A, B, C, D, MSG, S, MAGIC)    \
+   ASSIGN(T, B)                        ; \
+   XOR(T, C)                           ; \
+   XOR(T, D)                           ; \
+   ADD(A, ARRAY4(EDI, MSG))            ; \
+   ADD3_IMM(A, T, MAGIC)               ; \
+   ROTL_IMM(A, S)                      ; \
+   ADD(A, B)                           ;
+
+#define II(A, B, C, D, MSG, S, MAGIC)    \
+   ASSIGN(T, D)                        ; \
+   NOT(T)                              ; \
+   OR(T, B)                            ; \
+   XOR(T, C)                           ; \
+   ADD(A, ARRAY4(EDI, MSG))            ; \
+   ADD3_IMM(A, T, MAGIC)               ; \
+   ROTL_IMM(A, S)                      ; \
+   ADD(A, B)                           ;
+
+#define A EAX
+#define B EBX
+#define C ECX
+#define D EDX
+   
+   FF(A,B,C,D, 0, 7,0xD76AA478);   FF(D,A,B,C, 1,12,0xE8C7B756);
+   FF(C,D,A,B, 2,17,0x242070DB);   FF(B,C,D,A, 3,22,0xC1BDCEEE);
+   FF(A,B,C,D, 4, 7,0xF57C0FAF);   FF(D,A,B,C, 5,12,0x4787C62A);
+   FF(C,D,A,B, 6,17,0xA8304613);   FF(B,C,D,A, 7,22,0xFD469501);
+   FF(A,B,C,D, 8, 7,0x698098D8);   FF(D,A,B,C, 9,12,0x8B44F7AF);
+   FF(C,D,A,B,10,17,0xFFFF5BB1);   FF(B,C,D,A,11,22,0x895CD7BE);
+   FF(A,B,C,D,12, 7,0x6B901122);   FF(D,A,B,C,13,12,0xFD987193);
+   FF(C,D,A,B,14,17,0xA679438E);   FF(B,C,D,A,15,22,0x49B40821);
+
+   GG(A,B,C,D, 1, 5,0xF61E2562);   GG(D,A,B,C, 6, 9,0xC040B340);
+   GG(C,D,A,B,11,14,0x265E5A51);   GG(B,C,D,A, 0,20,0xE9B6C7AA);
+   GG(A,B,C,D, 5, 5,0xD62F105D);   GG(D,A,B,C,10, 9,0x02441453);
+   GG(C,D,A,B,15,14,0xD8A1E681);   GG(B,C,D,A, 4,20,0xE7D3FBC8);
+   GG(A,B,C,D, 9, 5,0x21E1CDE6);   GG(D,A,B,C,14, 9,0xC33707D6);
+   GG(C,D,A,B, 3,14,0xF4D50D87);   GG(B,C,D,A, 8,20,0x455A14ED);
+   GG(A,B,C,D,13, 5,0xA9E3E905);   GG(D,A,B,C, 2, 9,0xFCEFA3F8);
+   GG(C,D,A,B, 7,14,0x676F02D9);   GG(B,C,D,A,12,20,0x8D2A4C8A);
+
+   HH(A,B,C,D, 5, 4,0xFFFA3942);   HH(D,A,B,C, 8,11,0x8771F681);
+   HH(C,D,A,B,11,16,0x6D9D6122);   HH(B,C,D,A,14,23,0xFDE5380C);
+   HH(A,B,C,D, 1, 4,0xA4BEEA44);   HH(D,A,B,C, 4,11,0x4BDECFA9);
+   HH(C,D,A,B, 7,16,0xF6BB4B60);   HH(B,C,D,A,10,23,0xBEBFBC70);
+   HH(A,B,C,D,13, 4,0x289B7EC6);   HH(D,A,B,C, 0,11,0xEAA127FA);
+   HH(C,D,A,B, 3,16,0xD4EF3085);   HH(B,C,D,A, 6,23,0x04881D05);
+   HH(A,B,C,D, 9, 4,0xD9D4D039);   HH(D,A,B,C,12,11,0xE6DB99E5);
+   HH(C,D,A,B,15,16,0x1FA27CF8);   HH(B,C,D,A, 2,23,0xC4AC5665);
+
+   II(A,B,C,D, 0, 6,0xF4292244);   II(D,A,B,C, 7,10,0x432AFF97);
+   II(C,D,A,B,14,15,0xAB9423A7);   II(B,C,D,A, 5,21,0xFC93A039);
+   II(A,B,C,D,12, 6,0x655B59C3);   II(D,A,B,C, 3,10,0x8F0CCC92);
+   II(C,D,A,B,10,15,0xFFEFF47D);   II(B,C,D,A, 1,21,0x85845DD1);
+   II(A,B,C,D, 8, 6,0x6FA87E4F);   II(D,A,B,C,15,10,0xFE2CE6E0);
+   II(C,D,A,B, 6,15,0xA3014314);   II(B,C,D,A,13,21,0x4E0811A1);
+   II(A,B,C,D, 4, 6,0xF7537E82);   II(D,A,B,C,11,10,0xBD3AF235);
+   II(C,D,A,B, 2,15,0x2AD7D2BB);   II(B,C,D,A, 9,21,0xEB86D391);
+
+   ASSIGN(EBP, ARG(1))
+   ADD(ARRAY4(EBP, 0), EAX)
+   ADD(ARRAY4(EBP, 1), EBX)
+   ADD(ARRAY4(EBP, 2), ECX)
+   ADD(ARRAY4(EBP, 3), EDX)
+
+   RESTORE_REGS()
+END_FUNCTION(sha160_core)
diff --git a/modules/alg_ia32/modinfo.txt b/modules/alg_ia32/modinfo.txt
index b36d3c618..7d30fb624 100644
--- a/modules/alg_ia32/modinfo.txt
+++ b/modules/alg_ia32/modinfo.txt
@@ -1,9 +1,11 @@
 realname "Algorithm x86 Assembler"
 
 replace_file sha160.cpp
+replace_file md5.cpp
 
 add_file asm_macr.h
 add_file sha1core.S
+add_file md5core.S
 
 <arch>
 ia32