Adjust various bignum functions to use BN_CTX for variables instead of

[openssl.git] / crypto / md32_common.h

diff --git a/crypto/md32_common.h b/crypto/md32_common.h
index 275b93618bed1107778b6600249b0691966907d9..0cdc06e31ec00954d506e2db2905cbbcb6685aca 100644 (file)
--- a/crypto/md32_common.h
+++ b/crypto/md32_common.h
@@ -184,6 +184,8 @@
 # elif defined(__MWERKS__)
 #  if defined(__POWERPC__)
 #   define ROTATE(a,n) __rlwinm(a,n,0,31)
+#  elif defined(OPENSSL_SYSNAME_NETWARE)
+#   define ROTATE(a,n)  _lrotl(a,n)
 #  elif defined(__MC68K__)
     /* Motorola specific tweak. <appro@fy.chalmers.se> */
 #   define ROTATE(a,n) ( n<24 ? __rol(a,n) : __ror(a,32-n) )
@@ -198,7 +200,7 @@
    *
    *                                   <appro@fy.chalmers.se>
    */
-#  if defined(__i386) || defined(__i386__)
+#  if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
 #   define ROTATE(a,n) ({ register unsigned int ret;   \
                                asm (                   \
                                "roll %1,%0"            \
@@ -224,7 +226,7 @@
  */
 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
   /* some GNU C inline assembler templates by <appro@fy.chalmers.se> */
-#  if (defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)
+#  if (defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)) && !defined(I386_ONLY)
 #   define BE_FETCH32(a)       ({ register unsigned int l=(a);\
                                asm (                   \
                                "bswapl %0"             \
@@ -484,7 +486,7 @@ int HASH_UPDATE (HASH_CTX *c, const void *data_, unsigned long len)
                if ((((unsigned long)data)%4) == 0)
                        {
                        /* data is properly aligned so that we can cast it: */
-                       HASH_BLOCK_DATA_ORDER_ALIGNED (c,(HASH_LONG *)data,sw);
+                       HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,sw);
                        sw*=HASH_CBLOCK;
                        data+=sw;
                        len-=sw;
@@ -532,7 +534,7 @@ void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
 #if defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
        if ((((unsigned long)data)%4) == 0)
                /* data is properly aligned so that we can cast it: */
-               HASH_BLOCK_DATA_ORDER_ALIGNED (c,(HASH_LONG *)data,1);
+               HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,1);
        else
 #if !defined(HASH_BLOCK_DATA_ORDER)
                {
@@ -610,3 +612,28 @@ int HASH_FINAL (unsigned char *md, HASH_CTX *c)
         */
        return 1;
        }
+
+#ifndef MD32_REG_T
+#define MD32_REG_T long
+/*
+ * This comment was originaly written for MD5, which is why it
+ * discusses A-D. But it basically applies to all 32-bit digests,
+ * which is why it was moved to common header file.
+ *
+ * In case you wonder why A-D are declared as long and not
+ * as MD5_LONG. Doing so results in slight performance
+ * boost on LP64 architectures. The catch is we don't
+ * really care if 32 MSBs of a 64-bit register get polluted
+ * with eventual overflows as we *save* only 32 LSBs in
+ * *either* case. Now declaring 'em long excuses the compiler
+ * from keeping 32 MSBs zeroed resulting in 13% performance
+ * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
+ * Well, to be honest it should say that this *prevents* 
+ * performance degradation.
+ *                             <appro@fy.chalmers.se>
+ * Apparently there're LP64 compilers that generate better
+ * code if A-D are declared int. Most notably GCC-x86_64
+ * generates better code.
+ *                             <appro@fy.chalmers.se>
+ */
+#endif