Synchronize message digests in 098-fips with 098.

Configure

@@ -201,11 +201,11 @@ my %table=(
 "solaris-sparcv7-gcc","gcc:-O3 -fomit-frame-pointer -Wall -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR:${no_asm}:dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "solaris-sparcv8-gcc","gcc:-mv8 -O3 -fomit-frame-pointer -Wall -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 # -m32 should be safe to add as long as driver recognizes -mcpu=ultrasparc
-"solaris-sparcv9-gcc","gcc:-m32 -mcpu=ultrasparc -O3 -fomit-frame-pointer -Wall -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::md5-sparcv8plus.o::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"solaris-sparcv9-gcc","gcc:-m32 -mcpu=ultrasparc -O3 -fomit-frame-pointer -Wall -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
-"solaris64-sparcv9-gcc","gcc:-m64 -mcpu=ultrasparc -O3 -Wall -DB_ENDIAN::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR:::des_enc-sparc.o fcrypt_b.o:::md5-sparcv9.o::::::dlfcn:solaris-shared:-fPIC:-m64 -shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"solaris64-sparcv9-gcc","gcc:-m64 -mcpu=ultrasparc -O3 -Wall -DB_ENDIAN::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR:::des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-fPIC:-m64 -shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 ####
 "debug-solaris-sparcv8-gcc","gcc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -O -g -mv8 -Wall -DB_ENDIAN::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8.o::::::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
-"debug-solaris-sparcv9-gcc","gcc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -DPEDANTIC -O -g -mcpu=ultrasparc -pedantic -ansi -Wall -Wshadow -Wno-long-long -D__EXTENSIONS__ -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::md5-sparcv8plus.o::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"debug-solaris-sparcv9-gcc","gcc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -DPEDANTIC -O -g -mcpu=ultrasparc -pedantic -ansi -Wall -Wshadow -Wno-long-long -D__EXTENSIONS__ -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-fPIC:-shared:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 
 #### SPARC Solaris with Sun C setups
 # SC4.0 doesn't pass 'make test', upgrade to SC5.0 or SC4.2.
@@ -213,11 +213,11 @@ my %table=(
 # SC5.0 note: Compiler common patch 107357-01 or later is required!
 "solaris-sparcv7-cc","cc:-xO5 -xstrconst -xdepend -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_PTR DES_RISC1 DES_UNROLL BF_PTR:${no_asm}:dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "solaris-sparcv8-cc","cc:-xarch=v8 -xO5 -xstrconst -xdepend -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
-"solaris-sparcv9-cc","cc:-xtarget=ultra -xarch=v8plus -xO5 -xstrconst -xdepend -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK_LL DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::md5-sparcv8plus.o::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"solaris-sparcv9-cc","cc:-xtarget=ultra -xarch=v8plus -xO5 -xstrconst -xdepend -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK_LL DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
-"solaris64-sparcv9-cc","cc:-xtarget=ultra -xarch=v9 -xO5 -xstrconst -xdepend -Xa -DB_ENDIAN::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR:::des_enc-sparc.o fcrypt_b.o:::md5-sparcv9.o::::::dlfcn:solaris-shared:-KPIC:-xarch=v9 -G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):/usr/ccs/bin/ar rs",
+"solaris64-sparcv9-cc","cc:-xtarget=ultra -xarch=v9 -xO5 -xstrconst -xdepend -Xa -DB_ENDIAN::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC1 DES_UNROLL BF_PTR:::des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:solaris-shared:-KPIC:-xarch=v9 -G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):/usr/ccs/bin/ar rs",
 ####
 "debug-solaris-sparcv8-cc","cc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -xarch=v8 -g -O -xstrconst -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT::-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8.o::::::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
-"debug-solaris-sparcv9-cc","cc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -xtarget=ultra -xarch=v8plus -g -O -xstrconst -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK_LL DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8plus.o::::md5-sparcv8plus.o::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)", 
+"debug-solaris-sparcv9-cc","cc:-DBN_DEBUG -DREF_CHECK -DCONF_DEBUG -DCRYPTO_MDEBUG_ALL -xtarget=ultra -xarch=v8plus -g -O -xstrconst -Xa -DB_ENDIAN -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-lsocket -lnsl -ldl:BN_LLONG RC4_CHAR RC4_CHUNK_LL DES_PTR DES_RISC1 DES_UNROLL BF_PTR::sparcv8plus.o::::::::::dlfcn:solaris-shared:-KPIC:-G -dy -z text:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)", 
 
 #### SunOS configs, assuming sparc for the gcc one.
 #"sunos-cc", "cc:-O4 -DNOPROTO -DNOCONST::(unknown):SUNOS::DES_UNROLL:${no_asm}::",
@@ -332,9 +332,9 @@ my %table=(
 "linux-sparcv8","gcc:-mv8 -DB_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall -DBN_DIV2W::-D_REENTRANT::-ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 # it's a real mess with -mcpu=ultrasparc option under Linux, but
 # -Wa,-Av8plus should do the trick no matter what.
-"linux-sparcv9","gcc:-m32 -mcpu=ultrasparc -DB_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall -Wa,-Av8plus -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::md5-sparcv8plus.o::::::dlfcn:linux-shared:-fPIC:-m32:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"linux-sparcv9","gcc:-m32 -mcpu=ultrasparc -DB_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall -Wa,-Av8plus -DBN_DIV2W::-D_REENTRANT:ULTRASPARC:-ldl:BN_LLONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::sparcv8plus.o:des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:linux-shared:-fPIC:-m32:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 # GCC 3.1 is a requirement
-"linux64-sparcv9","gcc:-m64 -mcpu=ultrasparc -DB_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall::-D_REENTRANT:ULTRASPARC:-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::::::md5-sparcv9.o::::::dlfcn:linux-shared:-fPIC:-m64:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"linux64-sparcv9","gcc:-m64 -mcpu=ultrasparc -DB_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall::-D_REENTRANT:ULTRASPARC:-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_UNROLL BF_PTR::::::::::::dlfcn:linux-shared:-fPIC:-m64:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 #### Alpha Linux with GNU C and Compaq C setups
 # Special notes:
 # - linux-alpha+bwx-gcc is ment to be used from ./config only. If you
@@ -364,7 +364,7 @@ my %table=(
 # -DMD32_REG_T=int doesn't actually belong in sparc64 target, it
 # simply *happens* to work around a compiler bug in gcc 3.3.3,
 # triggered by RIPEMD160 code.
-"BSD-sparc64",	"gcc:-DB_ENDIAN -DTERMIOS -O3 -DMD32_REG_T=int -Wall::${BSDthreads}:::SIXTY_FOUR_BIT_LONG RC2_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC2 BF_PTR:::des_enc-sparc.o fcrypt_b.o:::md5-sparcv9.o::::::dlfcn:bsd-gcc-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
+"BSD-sparc64",	"gcc:-DB_ENDIAN -DTERMIOS -O3 -DMD32_REG_T=int -Wall::${BSDthreads}:::SIXTY_FOUR_BIT_LONG RC2_CHAR RC4_CHUNK DES_INT DES_PTR DES_RISC2 BF_PTR:::des_enc-sparc.o fcrypt_b.o:::::::::dlfcn:bsd-gcc-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "BSD-ia64",	"gcc:-DL_ENDIAN -DTERMIOS -O3 -Wall::${BSDthreads}:::SIXTY_FOUR_BIT_LONG RC4_CHUNK:${ia64_asm}:dlfcn:bsd-gcc-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "BSD-x86_64",	"gcc:-DL_ENDIAN -DTERMIOS -O3 -DMD32_REG_T=int -Wall::${BSDthreads}:::SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT DES_UNROLL:${x86_64_asm}:dlfcn:bsd-gcc-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 

TABLE

@@ -148,7 +148,7 @@ $bn_obj       =
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv9.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -1660,7 +1660,7 @@ $bn_obj       = sparcv8plus.o
 $des_obj      = 
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv8plus.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -1687,7 +1687,7 @@ $bn_obj       = sparcv8plus.o
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv8plus.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -2929,7 +2929,7 @@ $bn_obj       = sparcv8plus.o
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv8plus.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -2983,7 +2983,7 @@ $bn_obj       =
 $des_obj      = 
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv9.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -3577,7 +3577,7 @@ $bn_obj       = sparcv8plus.o
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv8plus.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -3604,7 +3604,7 @@ $bn_obj       = sparcv8plus.o
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv8plus.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -3685,7 +3685,7 @@ $bn_obj       =
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv9.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 
@@ -3712,7 +3712,7 @@ $bn_obj       =
 $des_obj      = des_enc-sparc.o fcrypt_b.o
 $aes_obj      = 
 $bf_obj       = 
-$md5_obj      = md5-sparcv9.o
+$md5_obj      = 
 $sha1_obj     = 
 $cast_obj     = 
 $rc4_obj      = 

crypto/md32_common.h

@@ -1,6 +1,6 @@
 /* crypto/md32_common.h */
 /* ====================================================================
- * Copyright (c) 1999-2002 The OpenSSL Project.  All rights reserved.
+ * Copyright (c) 1999-2007 The OpenSSL Project.  All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
@@ -47,10 +47,6 @@
  * OF THE POSSIBILITY OF SUCH DAMAGE.
  * ====================================================================
  *
- * This product includes cryptographic software written by Eric Young
- * (eay@cryptsoft.com).  This product includes software written by Tim
- * Hudson (tjh@cryptsoft.com).
- *
  */
 
 /*
@@ -76,40 +72,27 @@
  *		typedef struct {
  *			...
  *			HASH_LONG	Nl,Nh;
+ *			either {
  *			HASH_LONG	data[HASH_LBLOCK];
+ *			unsigned char	data[HASH_CBLOCK];
+ *			};
  *			unsigned int	num;
  *			...
  *			} HASH_CTX;
+ *	data[] vector is expected to be zeroed upon first call to
+ *	HASH_UPDATE.
  * HASH_UPDATE
  *	name of "Update" function, implemented here.
  * HASH_TRANSFORM
  *	name of "Transform" function, implemented here.
  * HASH_FINAL
  *	name of "Final" function, implemented here.
- * HASH_BLOCK_HOST_ORDER
- *	name of "block" function treating *aligned* input message
- *	in host byte order, implemented externally.
  * HASH_BLOCK_DATA_ORDER
- *	name of "block" function treating *unaligned* input message
+ *	name of "block" function capable of treating *unaligned* input
- *	in original (data) byte order, implemented externally (it
+ *	message in original (data) byte order, implemented externally.
- *	actually is optional if data and host are of the same
- *	"endianess").
  * HASH_MAKE_STRING
  *	macro convering context variables to an ASCII hash string.
  *
- * Optional macros:
- *
- * B_ENDIAN or L_ENDIAN
- *	defines host byte-order.
- * HASH_LONG_LOG2
- *	defaults to 2 if not states otherwise.
- * HASH_LBLOCK
- *	assumed to be HASH_CBLOCK/4 if not stated otherwise.
- * HASH_BLOCK_DATA_ORDER_ALIGNED
- *	alternative "block" function capable of treating
- *	aligned input message in original (data) order,
- *	implemented externally.
- *
  * MD5 example:
  *
  *	#define DATA_ORDER_IS_LITTLE_ENDIAN
@@ -118,11 +101,9 @@
  *	#define HASH_LONG_LOG2		MD5_LONG_LOG2
  *	#define HASH_CTX		MD5_CTX
  *	#define HASH_CBLOCK		MD5_CBLOCK
- *	#define HASH_LBLOCK		MD5_LBLOCK
  *	#define HASH_UPDATE		MD5_Update
  *	#define HASH_TRANSFORM		MD5_Transform
  *	#define HASH_FINAL		MD5_Final
- *	#define HASH_BLOCK_HOST_ORDER	md5_block_host_order
  *	#define HASH_BLOCK_DATA_ORDER	md5_block_data_order
  *
  *					<appro@fy.chalmers.se>
@@ -152,27 +133,9 @@
 #error "HASH_FINAL must be defined!"
 #endif
 
-#ifndef HASH_BLOCK_HOST_ORDER
-#error "HASH_BLOCK_HOST_ORDER must be defined!"
-#endif
-
-#if 0
-/*
- * Moved below as it's required only if HASH_BLOCK_DATA_ORDER_ALIGNED
- * isn't defined.
- */
 #ifndef HASH_BLOCK_DATA_ORDER
 #error "HASH_BLOCK_DATA_ORDER must be defined!"
 #endif
-#endif
-
-#ifndef HASH_LBLOCK
-#define HASH_LBLOCK	(HASH_CBLOCK/4)
-#endif
-
-#ifndef HASH_LONG_LOG2
-#define HASH_LONG_LOG2	2
-#endif
 
 /*
  * Engage compiler specific rotate intrinsic function if available.
@@ -206,7 +169,8 @@
 				: "cc");		\
 			   ret;				\
 			})
-#  elif defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
+#  elif defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
+	defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
 #   define ROTATE(a,n)	({ register unsigned int ret;	\
 				asm (			\
 				"rlwinm %0,%1,%2,0,31"	\
@@ -214,80 +178,28 @@
 				: "r"(a), "I"(n));	\
 			   ret;				\
 			})
+#  elif defined(__s390x__)
+#   define ROTATE(a,n) ({ register unsigned int ret;	\
+				asm ("rll %0,%1,%2"	\
+				: "=r"(ret)		\
+				: "r"(a), "I"(n));	\
+			  ret;				\
+			})
 #  endif
 # endif
 #endif /* PEDANTIC */
 
-#if HASH_LONG_LOG2==2	/* Engage only if sizeof(HASH_LONG)== 4 */
-/* A nice byte order reversal from Wei Dai <weidai@eskimo.com> */
-#ifdef ROTATE
-/* 5 instructions with rotate instruction, else 9 */
-#define REVERSE_FETCH32(a,l)	(					\
-		l=*(const HASH_LONG *)(a),				\
-		((ROTATE(l,8)&0x00FF00FF)|(ROTATE((l&0x00FF00FF),24)))	\
-				)
-#else
-/* 6 instructions with rotate instruction, else 8 */
-#define REVERSE_FETCH32(a,l)	(				\
-		l=*(const HASH_LONG *)(a),			\
-		l=(((l>>8)&0x00FF00FF)|((l&0x00FF00FF)<<8)),	\
-		ROTATE(l,16)					\
-				)
-/*
- * Originally the middle line started with l=(((l&0xFF00FF00)>>8)|...
- * It's rewritten as above for two reasons:
- *	- RISCs aren't good at long constants and have to explicitely
- *	  compose 'em with several (well, usually 2) instructions in a
- *	  register before performing the actual operation and (as you
- *	  already realized:-) having same constant should inspire the
- *	  compiler to permanently allocate the only register for it;
- *	- most modern CPUs have two ALUs, but usually only one has
- *	  circuitry for shifts:-( this minor tweak inspires compiler
- *	  to schedule shift instructions in a better way...
- *
- *				<appro@fy.chalmers.se>
- */
-#endif
-#endif
-
 #ifndef ROTATE
 #define ROTATE(a,n)     (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
 #endif
 
-/*
- * Make some obvious choices. E.g., HASH_BLOCK_DATA_ORDER_ALIGNED
- * and HASH_BLOCK_HOST_ORDER ought to be the same if input data
- * and host are of the same "endianess". It's possible to mask
- * this with blank #define HASH_BLOCK_DATA_ORDER though...
- *
- *				<appro@fy.chalmers.se>
- */
-#if defined(B_ENDIAN)
-#  if defined(DATA_ORDER_IS_BIG_ENDIAN)
-#    if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED) && HASH_LONG_LOG2==2
-#      define HASH_BLOCK_DATA_ORDER_ALIGNED	HASH_BLOCK_HOST_ORDER
-#    endif
-#  endif
-#elif defined(L_ENDIAN)
-#  if defined(DATA_ORDER_IS_LITTLE_ENDIAN)
-#    if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED) && HASH_LONG_LOG2==2
-#      define HASH_BLOCK_DATA_ORDER_ALIGNED	HASH_BLOCK_HOST_ORDER
-#    endif
-#  endif
-#endif
-
-#if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
-#ifndef HASH_BLOCK_DATA_ORDER
-#error "HASH_BLOCK_DATA_ORDER must be defined!"
-#endif
-#endif
-
 #if defined(DATA_ORDER_IS_BIG_ENDIAN)
 
 #ifndef PEDANTIC
 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
 #  if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
       (defined(__x86_64) || defined(__x86_64__))
+#   if !defined(B_ENDIAN)
     /*
      * This gives ~30-40% performance improvement in SHA-256 compiled
      * with gcc [on P4]. Well, first macro to be frank. We can pull
@@ -302,6 +214,11 @@
 				   *((unsigned int *)(c))=r; (c)+=4; r;	})
 #   endif
 #  endif
+# endif
+#endif
+#if defined(__s390__) || defined(__s390x__)
+# define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, (l))
+# define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, (l))
 #endif
 
 #ifndef HOST_c2l
@@ -311,29 +228,6 @@
 			 l|=(((unsigned long)(*((c)++)))    ),		\
 			 l)
 #endif
-#define HOST_p_c2l(c,l,n)	{					\
-			switch (n) {					\
-			case 0: l =((unsigned long)(*((c)++)))<<24;	\
-			case 1: l|=((unsigned long)(*((c)++)))<<16;	\
-			case 2: l|=((unsigned long)(*((c)++)))<< 8;	\
-			case 3: l|=((unsigned long)(*((c)++)));		\
-				} }
-#define HOST_p_c2l_p(c,l,sc,len) {					\
-			switch (sc) {					\
-			case 0: l =((unsigned long)(*((c)++)))<<24;	\
-				if (--len == 0) break;			\
-			case 1: l|=((unsigned long)(*((c)++)))<<16;	\
-				if (--len == 0) break;			\
-			case 2: l|=((unsigned long)(*((c)++)))<< 8;	\
-				} }
-/* NOTE the pointer is not incremented at the end of this */
-#define HOST_c2l_p(c,l,n)	{					\
-			l=0; (c)+=n;					\
-			switch (n) {					\
-			case 3: l =((unsigned long)(*(--(c))))<< 8;	\
-			case 2: l|=((unsigned long)(*(--(c))))<<16;	\
-			case 1: l|=((unsigned long)(*(--(c))))<<24;	\
-				} }
 #ifndef HOST_l2c
 #define HOST_l2c(l,c)	(*((c)++)=(unsigned char)(((l)>>24)&0xff),	\
 			 *((c)++)=(unsigned char)(((l)>>16)&0xff),	\
@@ -344,6 +238,18 @@
 
 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
 
+#ifndef PEDANTIC
+# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
+#  if defined(__s390x__)
+#   define HOST_c2l(c,l)	({ asm ("lrv	%0,0(%1)"		\
+					:"=r"(l) : "r"(c));		\
+				   (c)+=4; (l);				})
+#   define HOST_l2c(l,c)	({ asm ("strv	%0,0(%1)"		\
+					: : "r"(l),"r"(c) : "memory");	\
+				   (c)+=4; (l);				})
+#  endif
+# endif
+#endif
 #if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
 # ifndef B_ENDIAN
    /* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
@@ -359,29 +265,6 @@
 			 l|=(((unsigned long)(*((c)++)))<<24),		\
 			 l)
 #endif
-#define HOST_p_c2l(c,l,n)	{					\
-			switch (n) {					\
-			case 0: l =((unsigned long)(*((c)++)));		\
-			case 1: l|=((unsigned long)(*((c)++)))<< 8;	\
-			case 2: l|=((unsigned long)(*((c)++)))<<16;	\
-			case 3: l|=((unsigned long)(*((c)++)))<<24;	\
-				} }
-#define HOST_p_c2l_p(c,l,sc,len) {					\
-			switch (sc) {					\
-			case 0: l =((unsigned long)(*((c)++)));		\
-				if (--len == 0) break;			\
-			case 1: l|=((unsigned long)(*((c)++)))<< 8;	\
-				if (--len == 0) break;			\
-			case 2: l|=((unsigned long)(*((c)++)))<<16;	\
-				} }
-/* NOTE the pointer is not incremented at the end of this */
-#define HOST_c2l_p(c,l,n)	{					\
-			l=0; (c)+=n;					\
-			switch (n) {					\
-			case 3: l =((unsigned long)(*(--(c))))<<16;	\
-			case 2: l|=((unsigned long)(*(--(c))))<< 8;	\
-			case 1: l|=((unsigned long)(*(--(c))));		\
-				} }
 #ifndef HOST_l2c
 #define HOST_l2c(l,c)	(*((c)++)=(unsigned char)(((l)    )&0xff),	\
 			 *((c)++)=(unsigned char)(((l)>> 8)&0xff),	\
@@ -399,9 +282,9 @@
 int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
 	{
 	const unsigned char *data=data_;
-	register HASH_LONG * p;
+	unsigned char *p;
-	register HASH_LONG l;
+	HASH_LONG l;
-	size_t sw,sc,ew,ec;
+	size_t n;
 
 	if (len==0) return 1;
 
@@ -413,101 +296,43 @@ int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
 	c->Nh+=(len>>29);	/* might cause compiler warning on 16-bit */
 	c->Nl=l;
 
-	if (c->num != 0)
+	n = c->num;
+	if (n != 0)
 		{
-		p=c->data;
+		p=(unsigned char *)c->data;
-		sw=c->num>>2;
-		sc=c->num&0x03;
 
-		if ((c->num+len) >= HASH_CBLOCK)
+		if ((n+len) >= HASH_CBLOCK)
 			{
-			l=p[sw]; HOST_p_c2l(data,l,sc); p[sw++]=l;
+			memcpy (p+n,data,HASH_CBLOCK-n);
-			for (; sw<HASH_LBLOCK; sw++)
+			HASH_BLOCK_DATA_ORDER (c,p,1);
-				{
+			n      = HASH_CBLOCK-n;
-				HOST_c2l(data,l); p[sw]=l;
+			data  += n;
-				}
+			len   -= n;
-			HASH_BLOCK_HOST_ORDER (c,p,1);
+			c->num = 0;
-			len-=(HASH_CBLOCK-c->num);
+			memset (p,0,HASH_CBLOCK);	/* keep it zeroed */
-			c->num=0;
-			/* drop through and do the rest */
 			}
 		else
 			{
-			c->num+=(unsigned int)len;
+			memcpy (p+n,data,len);
-			if ((sc+len) < 4) /* ugly, add char's to a word */
+			c->num += (unsigned int)len;
-				{
-				l=p[sw]; HOST_p_c2l_p(data,l,sc,len); p[sw]=l;
-				}
-			else
-				{
-				ew=(c->num>>2);
-				ec=(c->num&0x03);
-				if (sc)
-					l=p[sw];
-				HOST_p_c2l(data,l,sc);
-				p[sw++]=l;
-				for (; sw < ew; sw++)
-					{
-					HOST_c2l(data,l); p[sw]=l;
-					}
-				if (ec)
-					{
-					HOST_c2l_p(data,l,ec); p[sw]=l;
-					}
-				}
 			return 1;
 			}
 		}
 
-	sw=len/HASH_CBLOCK;
+	n = len/HASH_CBLOCK;
-	if (sw > 0)
+	if (n > 0)
 		{
-#if defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
+		HASH_BLOCK_DATA_ORDER (c,data,n);
-		/*
+		n    *= HASH_CBLOCK;
-		 * Note that HASH_BLOCK_DATA_ORDER_ALIGNED gets defined
+		data += n;
-		 * only if sizeof(HASH_LONG)==4.
+		len  -= n;
-		 */
-		if ((((size_t)data)%4) == 0)
-			{
-			/* data is properly aligned so that we can cast it: */
-			HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,sw);
-			sw*=HASH_CBLOCK;
-			data+=sw;
-			len-=sw;
-			}
-		else
-#if !defined(HASH_BLOCK_DATA_ORDER)
-			while (sw--)
-				{
-				memcpy (p=c->data,data,HASH_CBLOCK);
-				HASH_BLOCK_DATA_ORDER_ALIGNED(c,p,1);
-				data+=HASH_CBLOCK;
-				len-=HASH_CBLOCK;
-				}
-#endif
-#endif
-#if defined(HASH_BLOCK_DATA_ORDER)
-			{
-			HASH_BLOCK_DATA_ORDER(c,data,sw);
-			sw*=HASH_CBLOCK;
-			data+=sw;
-			len-=sw;
-			}
-#endif
 		}
 
-	if (len!=0)
+	if (len != 0)
 		{
-		p = c->data;
+		p = (unsigned char *)c->data;
 		c->num = len;
-		ew=len>>2;	/* words to copy */
+		memcpy (p,data,len);
-		ec=len&0x03;
-		for (; ew; ew--,p++)
-			{
-			HOST_c2l(data,l); *p=l;
-			}
-		HOST_c2l_p(data,l,ec);
-		*p=l;
 		}
 	return 1;
 	}
@@ -515,73 +340,38 @@ int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
 
 void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
 	{
-#if defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
-	if ((((size_t)data)%4) == 0)
-		/* data is properly aligned so that we can cast it: */
-		HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,1);
-	else
-#if !defined(HASH_BLOCK_DATA_ORDER)
-		{
-		memcpy (c->data,data,HASH_CBLOCK);
-		HASH_BLOCK_DATA_ORDER_ALIGNED (c,c->data,1);
-		}
-#endif
-#endif
-#if defined(HASH_BLOCK_DATA_ORDER)
 	HASH_BLOCK_DATA_ORDER (c,data,1);
-#endif
 	}
 
 
 int HASH_FINAL (unsigned char *md, HASH_CTX *c)
 	{
-	register HASH_LONG *p;
+	unsigned char *p = (unsigned char *)c->data;
-	register unsigned long l;
+	size_t n = c->num;
-	register int i,j;
-	static const unsigned char end[4]={0x80,0x00,0x00,0x00};
-	const unsigned char *cp=end;
 
-	/* c->num should definitly have room for at least one more byte. */
+	p[n] = 0x80; /* there is always room for one */
-	p=c->data;
+	n++;
-	i=c->num>>2;
-	j=c->num&0x03;
 
-#if 0
+	if (n > (HASH_CBLOCK-8))
-	/* purify often complains about the following line as an
-	 * Uninitialized Memory Read.  While this can be true, the
-	 * following p_c2l macro will reset l when that case is true.
-	 * This is because j&0x03 contains the number of 'valid' bytes
-	 * already in p[i].  If and only if j&0x03 == 0, the UMR will
-	 * occur but this is also the only time p_c2l will do
-	 * l= *(cp++) instead of l|= *(cp++)
-	 * Many thanks to Alex Tang <altitude@cic.net> for pickup this
-	 * 'potential bug' */
-#ifdef PURIFY
-	if (j==0) p[i]=0; /* Yeah, but that's not the way to fix it:-) */
-#endif
-	l=p[i];
-#else
-	l = (j==0) ? 0 : p[i];
-#endif
-	HOST_p_c2l(cp,l,j); p[i++]=l; /* i is the next 'undefined word' */
-
-	if (i>(HASH_LBLOCK-2)) /* save room for Nl and Nh */
 		{
-		if (i<HASH_LBLOCK) p[i]=0;
+		memset (p+n,0,HASH_CBLOCK-n);
-		HASH_BLOCK_HOST_ORDER (c,p,1);
+		n=0;
-		i=0;
+		HASH_BLOCK_DATA_ORDER (c,p,1);
 		}
-	for (; i<(HASH_LBLOCK-2); i++)
+	memset (p+n,0,HASH_CBLOCK-8-n);
-		p[i]=0;
 
+	p += HASH_CBLOCK-8;
 #if   defined(DATA_ORDER_IS_BIG_ENDIAN)
-	p[HASH_LBLOCK-2]=c->Nh;
+	(void)HOST_l2c(c->Nh,p);
-	p[HASH_LBLOCK-1]=c->Nl;
+	(void)HOST_l2c(c->Nl,p);
 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
-	p[HASH_LBLOCK-2]=c->Nl;
+	(void)HOST_l2c(c->Nl,p);
-	p[HASH_LBLOCK-1]=c->Nh;
+	(void)HOST_l2c(c->Nh,p);
 #endif
-	HASH_BLOCK_HOST_ORDER (c,p,1);
+	p -= HASH_CBLOCK;
+	HASH_BLOCK_DATA_ORDER (c,p,1);
+	c->num=0;
+	memset (p,0,HASH_CBLOCK);
 
 #ifndef HASH_MAKE_STRING
 #error "HASH_MAKE_STRING must be defined!"
@@ -589,11 +379,6 @@ int HASH_FINAL (unsigned char *md, HASH_CTX *c)
 	HASH_MAKE_STRING(c,md);
 #endif
 
-	c->num=0;
-	/* clear stuff, HASH_BLOCK may be leaving some stuff on the stack
-	 * but I'm not worried :-)
-	OPENSSL_cleanse((void *)c,sizeof(HASH_CTX));
-	 */
 	return 1;
 	}
 

crypto/md4/md4_dgst.c

@@ -82,79 +82,6 @@ int MD4_Init(MD4_CTX *c)
 	return 1;
 	}
 
-#ifndef md4_block_host_order
-void md4_block_host_order (MD4_CTX *c, const void *data, size_t num)
-	{
-	const MD4_LONG *X=data;
-	register unsigned MD32_REG_T A,B,C,D;
-
-	A=c->A;
-	B=c->B;
-	C=c->C;
-	D=c->D;
-
-	for (;num--;X+=HASH_LBLOCK)
-		{
-	/* Round 0 */
-	R0(A,B,C,D,X[ 0], 3,0);
-	R0(D,A,B,C,X[ 1], 7,0);
-	R0(C,D,A,B,X[ 2],11,0);
-	R0(B,C,D,A,X[ 3],19,0);
-	R0(A,B,C,D,X[ 4], 3,0);
-	R0(D,A,B,C,X[ 5], 7,0);
-	R0(C,D,A,B,X[ 6],11,0);
-	R0(B,C,D,A,X[ 7],19,0);
-	R0(A,B,C,D,X[ 8], 3,0);
-	R0(D,A,B,C,X[ 9], 7,0);
-	R0(C,D,A,B,X[10],11,0);
-	R0(B,C,D,A,X[11],19,0);
-	R0(A,B,C,D,X[12], 3,0);
-	R0(D,A,B,C,X[13], 7,0);
-	R0(C,D,A,B,X[14],11,0);
-	R0(B,C,D,A,X[15],19,0);
-	/* Round 1 */
-	R1(A,B,C,D,X[ 0], 3,0x5A827999L);
-	R1(D,A,B,C,X[ 4], 5,0x5A827999L);
-	R1(C,D,A,B,X[ 8], 9,0x5A827999L);
-	R1(B,C,D,A,X[12],13,0x5A827999L);
-	R1(A,B,C,D,X[ 1], 3,0x5A827999L);
-	R1(D,A,B,C,X[ 5], 5,0x5A827999L);
-	R1(C,D,A,B,X[ 9], 9,0x5A827999L);
-	R1(B,C,D,A,X[13],13,0x5A827999L);
-	R1(A,B,C,D,X[ 2], 3,0x5A827999L);
-	R1(D,A,B,C,X[ 6], 5,0x5A827999L);
-	R1(C,D,A,B,X[10], 9,0x5A827999L);
-	R1(B,C,D,A,X[14],13,0x5A827999L);
-	R1(A,B,C,D,X[ 3], 3,0x5A827999L);
-	R1(D,A,B,C,X[ 7], 5,0x5A827999L);
-	R1(C,D,A,B,X[11], 9,0x5A827999L);
-	R1(B,C,D,A,X[15],13,0x5A827999L);
-	/* Round 2 */
-	R2(A,B,C,D,X[ 0], 3,0x6ED9EBA1);
-	R2(D,A,B,C,X[ 8], 9,0x6ED9EBA1);
-	R2(C,D,A,B,X[ 4],11,0x6ED9EBA1);
-	R2(B,C,D,A,X[12],15,0x6ED9EBA1);
-	R2(A,B,C,D,X[ 2], 3,0x6ED9EBA1);
-	R2(D,A,B,C,X[10], 9,0x6ED9EBA1);
-	R2(C,D,A,B,X[ 6],11,0x6ED9EBA1);
-	R2(B,C,D,A,X[14],15,0x6ED9EBA1);
-	R2(A,B,C,D,X[ 1], 3,0x6ED9EBA1);
-	R2(D,A,B,C,X[ 9], 9,0x6ED9EBA1);
-	R2(C,D,A,B,X[ 5],11,0x6ED9EBA1);
-	R2(B,C,D,A,X[13],15,0x6ED9EBA1);
-	R2(A,B,C,D,X[ 3], 3,0x6ED9EBA1);
-	R2(D,A,B,C,X[11], 9,0x6ED9EBA1);
-	R2(C,D,A,B,X[ 7],11,0x6ED9EBA1);
-	R2(B,C,D,A,X[15],15,0x6ED9EBA1);
-
-	A = c->A += A;
-	B = c->B += B;
-	C = c->C += C;
-	D = c->D += D;
-		}
-	}
-#endif
-
 #ifndef md4_block_data_order
 #ifdef X
 #undef X
@@ -240,19 +167,3 @@ void md4_block_data_order (MD4_CTX *c, const void *data_, size_t num)
 		}
 	}
 #endif
-
-#ifdef undef
-int printit(unsigned long *l)
-	{
-	int i,ii;
-
-	for (i=0; i<2; i++)
-		{
-		for (ii=0; ii<8; ii++)
-			{
-			fprintf(stderr,"%08lx ",l[i*8+ii]);
-			}
-		fprintf(stderr,"\n");
-		}
-	}
-#endif

crypto/md4/md4_locl.h

@@ -65,43 +65,13 @@
 #define MD4_LONG_LOG2 2 /* default to 32 bits */
 #endif
 
-void md4_block_host_order (MD4_CTX *c, const void *p,size_t num);
 void md4_block_data_order (MD4_CTX *c, const void *p,size_t num);
 
-#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
-# if !defined(B_ENDIAN)
-/*
- * *_block_host_order is expected to handle aligned data while
- * *_block_data_order - unaligned. As algorithm and host (x86)
- * are in this case of the same "endianness" these two are
- * otherwise indistinguishable. But normally you don't want to
- * call the same function because unaligned access in places
- * where alignment is expected is usually a "Bad Thing". Indeed,
- * on RISCs you get punished with BUS ERROR signal or *severe*
- * performance degradation. Intel CPUs are in turn perfectly
- * capable of loading unaligned data without such drastic side
- * effect. Yes, they say it's slower than aligned load, but no
- * exception is generated and therefore performance degradation
- * is *incomparable* with RISCs. What we should weight here is
- * costs of unaligned access against costs of aligning data.
- * According to my measurements allowing unaligned access results
- * in ~9% performance improvement on Pentium II operating at
- * 266MHz. I won't be surprised if the difference will be higher
- * on faster systems:-)
- *
- *				<appro@fy.chalmers.se>
- */
-# define md4_block_data_order md4_block_host_order
-# endif
-#endif
-
 #define DATA_ORDER_IS_LITTLE_ENDIAN
 
 #define HASH_LONG		MD4_LONG
-#define HASH_LONG_LOG2		MD4_LONG_LOG2
 #define HASH_CTX		MD4_CTX
 #define HASH_CBLOCK		MD4_CBLOCK
-#define HASH_LBLOCK		MD4_LBLOCK
 #define HASH_UPDATE		MD4_Update
 #define HASH_TRANSFORM		MD4_Transform
 #define HASH_FINAL		MD4_Final
@@ -112,21 +82,7 @@ void md4_block_data_order (MD4_CTX *c, const void *p,size_t num);
 	ll=(c)->C; HOST_l2c(ll,(s));	\
 	ll=(c)->D; HOST_l2c(ll,(s));	\
 	} while (0)
-#define HASH_BLOCK_HOST_ORDER	md4_block_host_order
-#if !defined(L_ENDIAN) || defined(md4_block_data_order)
 #define	HASH_BLOCK_DATA_ORDER	md4_block_data_order
-/*
- * Little-endians (Intel and Alpha) feel better without this.
- * It looks like memcpy does better job than generic
- * md4_block_data_order on copying-n-aligning input data.
- * But frankly speaking I didn't expect such result on Alpha.
- * On the other hand I've got this with egcs-1.0.2 and if
- * program is compiled with another (better?) compiler it
- * might turn out other way around.
- *
- *				<appro@fy.chalmers.se>
- */
-#endif
 
 #include "md32_common.h"
 

crypto/md5/Makefile

@@ -52,24 +52,6 @@ mx86-cof.s: asm/md5-586.pl ../perlasm/x86asm.pl
 mx86-out.s: asm/md5-586.pl ../perlasm/x86asm.pl
 	(cd asm; $(PERL) md5-586.pl a.out $(CFLAGS) > ../$@)
 
-md5-sparcv8plus.o: asm/md5-sparcv9.S
-	$(CC) $(ASFLAGS) -DMD5_BLOCK_DATA_ORDER -c \
-		-o md5-sparcv8plus.o asm/md5-sparcv9.S
-
-# Old GNU assembler doesn't understand V9 instructions, so we
-# hire /usr/ccs/bin/as to do the job. Note that option is called
-# *-gcc27, but even gcc 2>=8 users may experience similar problem
-# if they didn't bother to upgrade GNU assembler. Such users should
-# not choose this option, but be adviced to *remove* GNU assembler
-# or upgrade it.
-md5-sparcv8plus-gcc27.o: asm/md5-sparcv9.S
-	$(CC) $(ASFLAGS) -DMD5_BLOCK_DATA_ORDER -E asm/md5-sparcv9.S | \
-		/usr/ccs/bin/as -xarch=v8plus - -o md5-sparcv8plus-gcc27.o
-
-md5-sparcv9.o: asm/md5-sparcv9.S
-	$(CC) $(ASFLAGS) -DMD5_BLOCK_DATA_ORDER -c \
-		-o md5-sparcv9.o asm/md5-sparcv9.S
-
 md5-x86_64.s:	asm/md5-x86_64.pl;	$(PERL) asm/md5-x86_64.pl $@
 
 files:

crypto/md5/asm/md5-586.pl

@@ -29,7 +29,7 @@ $X="esi";
  0, 7, 14, 5, 12, 3, 10, 1, 8, 15, 6, 13, 4, 11, 2, 9,	# R3
  );
 
-&md5_block("md5_block_asm_host_order");
+&md5_block("md5_block_asm_data_order");
 &asm_finish();
 
 sub Np

crypto/md5/asm/md5-x86_64.pl

@@ -111,9 +111,9 @@ $code .= <<EOF;
 .text
 .align 16
 
-.globl md5_block_asm_host_order
+.globl md5_block_asm_data_order
-.type md5_block_asm_host_order,\@function,3
+.type md5_block_asm_data_order,\@function,3
-md5_block_asm_host_order:
+md5_block_asm_data_order:
 	push	%rbp
 	push	%rbx
 	push	%r14
@@ -237,7 +237,7 @@ $code .= <<EOF;
 	pop	%rbx
 	pop	%rbp
 	ret
-.size md5_block_asm_host_order,.-md5_block_asm_host_order
+.size md5_block_asm_data_order,.-md5_block_asm_data_order
 EOF
 
 print $code;

crypto/md5/md5_dgst.c

@@ -82,96 +82,6 @@ int MD5_Init(MD5_CTX *c)
 	return 1;
 	}
 
-#ifndef md5_block_host_order
-void md5_block_host_order (MD5_CTX *c, const void *data, size_t num)
-	{
-	const MD5_LONG *X=data;
-	register unsigned MD32_REG_T A,B,C,D;
-
-	A=c->A;
-	B=c->B;
-	C=c->C;
-	D=c->D;
-
-	for (;num--;X+=HASH_LBLOCK)
-		{
-	/* Round 0 */
-	R0(A,B,C,D,X[ 0], 7,0xd76aa478L);
-	R0(D,A,B,C,X[ 1],12,0xe8c7b756L);
-	R0(C,D,A,B,X[ 2],17,0x242070dbL);
-	R0(B,C,D,A,X[ 3],22,0xc1bdceeeL);
-	R0(A,B,C,D,X[ 4], 7,0xf57c0fafL);
-	R0(D,A,B,C,X[ 5],12,0x4787c62aL);
-	R0(C,D,A,B,X[ 6],17,0xa8304613L);
-	R0(B,C,D,A,X[ 7],22,0xfd469501L);
-	R0(A,B,C,D,X[ 8], 7,0x698098d8L);
-	R0(D,A,B,C,X[ 9],12,0x8b44f7afL);
-	R0(C,D,A,B,X[10],17,0xffff5bb1L);
-	R0(B,C,D,A,X[11],22,0x895cd7beL);
-	R0(A,B,C,D,X[12], 7,0x6b901122L);
-	R0(D,A,B,C,X[13],12,0xfd987193L);
-	R0(C,D,A,B,X[14],17,0xa679438eL);
-	R0(B,C,D,A,X[15],22,0x49b40821L);
-	/* Round 1 */
-	R1(A,B,C,D,X[ 1], 5,0xf61e2562L);
-	R1(D,A,B,C,X[ 6], 9,0xc040b340L);
-	R1(C,D,A,B,X[11],14,0x265e5a51L);
-	R1(B,C,D,A,X[ 0],20,0xe9b6c7aaL);
-	R1(A,B,C,D,X[ 5], 5,0xd62f105dL);
-	R1(D,A,B,C,X[10], 9,0x02441453L);
-	R1(C,D,A,B,X[15],14,0xd8a1e681L);
-	R1(B,C,D,A,X[ 4],20,0xe7d3fbc8L);
-	R1(A,B,C,D,X[ 9], 5,0x21e1cde6L);
-	R1(D,A,B,C,X[14], 9,0xc33707d6L);
-	R1(C,D,A,B,X[ 3],14,0xf4d50d87L);
-	R1(B,C,D,A,X[ 8],20,0x455a14edL);
-	R1(A,B,C,D,X[13], 5,0xa9e3e905L);
-	R1(D,A,B,C,X[ 2], 9,0xfcefa3f8L);
-	R1(C,D,A,B,X[ 7],14,0x676f02d9L);
-	R1(B,C,D,A,X[12],20,0x8d2a4c8aL);
-	/* Round 2 */
-	R2(A,B,C,D,X[ 5], 4,0xfffa3942L);
-	R2(D,A,B,C,X[ 8],11,0x8771f681L);
-	R2(C,D,A,B,X[11],16,0x6d9d6122L);
-	R2(B,C,D,A,X[14],23,0xfde5380cL);
-	R2(A,B,C,D,X[ 1], 4,0xa4beea44L);
-	R2(D,A,B,C,X[ 4],11,0x4bdecfa9L);
-	R2(C,D,A,B,X[ 7],16,0xf6bb4b60L);
-	R2(B,C,D,A,X[10],23,0xbebfbc70L);
-	R2(A,B,C,D,X[13], 4,0x289b7ec6L);
-	R2(D,A,B,C,X[ 0],11,0xeaa127faL);
-	R2(C,D,A,B,X[ 3],16,0xd4ef3085L);
-	R2(B,C,D,A,X[ 6],23,0x04881d05L);
-	R2(A,B,C,D,X[ 9], 4,0xd9d4d039L);
-	R2(D,A,B,C,X[12],11,0xe6db99e5L);
-	R2(C,D,A,B,X[15],16,0x1fa27cf8L);
-	R2(B,C,D,A,X[ 2],23,0xc4ac5665L);
-	/* Round 3 */
-	R3(A,B,C,D,X[ 0], 6,0xf4292244L);
-	R3(D,A,B,C,X[ 7],10,0x432aff97L);
-	R3(C,D,A,B,X[14],15,0xab9423a7L);
-	R3(B,C,D,A,X[ 5],21,0xfc93a039L);
-	R3(A,B,C,D,X[12], 6,0x655b59c3L);
-	R3(D,A,B,C,X[ 3],10,0x8f0ccc92L);
-	R3(C,D,A,B,X[10],15,0xffeff47dL);
-	R3(B,C,D,A,X[ 1],21,0x85845dd1L);
-	R3(A,B,C,D,X[ 8], 6,0x6fa87e4fL);
-	R3(D,A,B,C,X[15],10,0xfe2ce6e0L);
-	R3(C,D,A,B,X[ 6],15,0xa3014314L);
-	R3(B,C,D,A,X[13],21,0x4e0811a1L);
-	R3(A,B,C,D,X[ 4], 6,0xf7537e82L);
-	R3(D,A,B,C,X[11],10,0xbd3af235L);
-	R3(C,D,A,B,X[ 2],15,0x2ad7d2bbL);
-	R3(B,C,D,A,X[ 9],21,0xeb86d391L);
-
-	A = c->A += A;
-	B = c->B += B;
-	C = c->C += C;
-	D = c->D += D;
-		}
-	}
-#endif
-
 #ifndef md5_block_data_order
 #ifdef X
 #undef X
@@ -274,19 +184,3 @@ void md5_block_data_order (MD5_CTX *c, const void *data_, size_t num)
 		}
 	}
 #endif
-
-#ifdef undef
-int printit(unsigned long *l)
-	{
-	int i,ii;
-
-	for (i=0; i<2; i++)
-		{
-		for (ii=0; ii<8; ii++)
-			{
-			fprintf(stderr,"%08lx ",l[i*8+ii]);
-			}
-		fprintf(stderr,"\n");
-		}
-	}
-#endif

crypto/md5/md5_locl.h

@@ -66,53 +66,19 @@
 #endif
 
 #ifdef MD5_ASM
-# if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__) || defined(__x86_64) || defined(__x86_64__)
+# if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__) || \
-#  if !defined(B_ENDIAN)
+     defined(__x86_64) || defined(__x86_64__) || defined(_M_AMD64) || defined(_M_X64)
-#   define md5_block_host_order md5_block_asm_host_order
+#  define md5_block_data_order md5_block_asm_data_order
-#  endif
-# elif defined(__sparc) && defined(OPENSSL_SYS_ULTRASPARC)
-   void md5_block_asm_data_order_aligned (MD5_CTX *c, const MD5_LONG *p,size_t num);
-#  define HASH_BLOCK_DATA_ORDER_ALIGNED md5_block_asm_data_order_aligned
 # endif
 #endif
 
-void md5_block_host_order (MD5_CTX *c, const void *p,size_t num);
 void md5_block_data_order (MD5_CTX *c, const void *p,size_t num);
 
-#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__) || defined(__x86_64) || defined(__x86_64__)
-# if !defined(B_ENDIAN)
-/*
- * *_block_host_order is expected to handle aligned data while
- * *_block_data_order - unaligned. As algorithm and host (x86)
- * are in this case of the same "endianness" these two are
- * otherwise indistinguishable. But normally you don't want to
- * call the same function because unaligned access in places
- * where alignment is expected is usually a "Bad Thing". Indeed,
- * on RISCs you get punished with BUS ERROR signal or *severe*
- * performance degradation. Intel CPUs are in turn perfectly
- * capable of loading unaligned data without such drastic side
- * effect. Yes, they say it's slower than aligned load, but no
- * exception is generated and therefore performance degradation
- * is *incomparable* with RISCs. What we should weight here is
- * costs of unaligned access against costs of aligning data.
- * According to my measurements allowing unaligned access results
- * in ~9% performance improvement on Pentium II operating at
- * 266MHz. I won't be surprised if the difference will be higher
- * on faster systems:-)
- *
- *				<appro@fy.chalmers.se>
- */
-# define md5_block_data_order md5_block_host_order
-# endif
-#endif
-
 #define DATA_ORDER_IS_LITTLE_ENDIAN
 
 #define HASH_LONG		MD5_LONG
-#define HASH_LONG_LOG2		MD5_LONG_LOG2
 #define HASH_CTX		MD5_CTX
 #define HASH_CBLOCK		MD5_CBLOCK
-#define HASH_LBLOCK		MD5_LBLOCK
 #define HASH_UPDATE		MD5_Update
 #define HASH_TRANSFORM		MD5_Transform
 #define HASH_FINAL		MD5_Final
@@ -123,21 +89,7 @@ void md5_block_data_order (MD5_CTX *c, const void *p,size_t num);
 	ll=(c)->C; HOST_l2c(ll,(s));	\
 	ll=(c)->D; HOST_l2c(ll,(s));	\
 	} while (0)
-#define HASH_BLOCK_HOST_ORDER	md5_block_host_order
-#if !defined(L_ENDIAN) || defined(md5_block_data_order)
 #define	HASH_BLOCK_DATA_ORDER	md5_block_data_order
-/*
- * Little-endians (Intel and Alpha) feel better without this.
- * It looks like memcpy does better job than generic
- * md5_block_data_order on copying-n-aligning input data.
- * But frankly speaking I didn't expect such result on Alpha.
- * On the other hand I've got this with egcs-1.0.2 and if
- * program is compiled with another (better?) compiler it
- * might turn out other way around.
- *
- *				<appro@fy.chalmers.se>
- */
-#endif
 
 #include "md32_common.h"
 

crypto/ripemd/asm/rmd-586.pl

@@ -1,7 +1,7 @@
 #!/usr/local/bin/perl
 
 # Normal is the
-# ripemd160_block_asm_host_order(RIPEMD160_CTX *c, ULONG *X,int blocks);
+# ripemd160_block_asm_data_order(RIPEMD160_CTX *c, ULONG *X,int blocks);
 
 $normal=0;
 
@@ -56,7 +56,7 @@ $KR3=0x7A6D76E9;
 	 8, 5,12, 9,12, 5,14, 6, 8,13, 6, 5,15,13,11,11,
  	);
 
-&ripemd160_block("ripemd160_block_asm_host_order");
+&ripemd160_block("ripemd160_block_asm_data_order");
 &asm_finish();
 
 sub Xv

crypto/ripemd/rmd_dgst.c

@@ -82,207 +82,6 @@ int RIPEMD160_Init(RIPEMD160_CTX *c)
 	return 1;
 	}
 
-#ifndef ripemd160_block_host_order
-#ifdef X
-#undef X
-#endif
-#define X(i)	XX[i]
-void ripemd160_block_host_order (RIPEMD160_CTX *ctx, const void *p, size_t num)
-	{
-	const RIPEMD160_LONG *XX=p;
-	register unsigned MD32_REG_T A,B,C,D,E;
-	register unsigned MD32_REG_T a,b,c,d,e;
-
-	for (;num--;XX+=HASH_LBLOCK)
-		{
-
-	A=ctx->A; B=ctx->B; C=ctx->C; D=ctx->D; E=ctx->E;
-
-	RIP1(A,B,C,D,E,WL00,SL00);
-	RIP1(E,A,B,C,D,WL01,SL01);
-	RIP1(D,E,A,B,C,WL02,SL02);
-	RIP1(C,D,E,A,B,WL03,SL03);
-	RIP1(B,C,D,E,A,WL04,SL04);
-	RIP1(A,B,C,D,E,WL05,SL05);
-	RIP1(E,A,B,C,D,WL06,SL06);
-	RIP1(D,E,A,B,C,WL07,SL07);
-	RIP1(C,D,E,A,B,WL08,SL08);
-	RIP1(B,C,D,E,A,WL09,SL09);
-	RIP1(A,B,C,D,E,WL10,SL10);
-	RIP1(E,A,B,C,D,WL11,SL11);
-	RIP1(D,E,A,B,C,WL12,SL12);
-	RIP1(C,D,E,A,B,WL13,SL13);
-	RIP1(B,C,D,E,A,WL14,SL14);
-	RIP1(A,B,C,D,E,WL15,SL15);
-
-	RIP2(E,A,B,C,D,WL16,SL16,KL1);
-	RIP2(D,E,A,B,C,WL17,SL17,KL1);
-	RIP2(C,D,E,A,B,WL18,SL18,KL1);
-	RIP2(B,C,D,E,A,WL19,SL19,KL1);
-	RIP2(A,B,C,D,E,WL20,SL20,KL1);
-	RIP2(E,A,B,C,D,WL21,SL21,KL1);
-	RIP2(D,E,A,B,C,WL22,SL22,KL1);
-	RIP2(C,D,E,A,B,WL23,SL23,KL1);
-	RIP2(B,C,D,E,A,WL24,SL24,KL1);
-	RIP2(A,B,C,D,E,WL25,SL25,KL1);
-	RIP2(E,A,B,C,D,WL26,SL26,KL1);
-	RIP2(D,E,A,B,C,WL27,SL27,KL1);
-	RIP2(C,D,E,A,B,WL28,SL28,KL1);
-	RIP2(B,C,D,E,A,WL29,SL29,KL1);
-	RIP2(A,B,C,D,E,WL30,SL30,KL1);
-	RIP2(E,A,B,C,D,WL31,SL31,KL1);
-
-	RIP3(D,E,A,B,C,WL32,SL32,KL2);
-	RIP3(C,D,E,A,B,WL33,SL33,KL2);
-	RIP3(B,C,D,E,A,WL34,SL34,KL2);
-	RIP3(A,B,C,D,E,WL35,SL35,KL2);
-	RIP3(E,A,B,C,D,WL36,SL36,KL2);
-	RIP3(D,E,A,B,C,WL37,SL37,KL2);
-	RIP3(C,D,E,A,B,WL38,SL38,KL2);
-	RIP3(B,C,D,E,A,WL39,SL39,KL2);
-	RIP3(A,B,C,D,E,WL40,SL40,KL2);
-	RIP3(E,A,B,C,D,WL41,SL41,KL2);
-	RIP3(D,E,A,B,C,WL42,SL42,KL2);
-	RIP3(C,D,E,A,B,WL43,SL43,KL2);
-	RIP3(B,C,D,E,A,WL44,SL44,KL2);
-	RIP3(A,B,C,D,E,WL45,SL45,KL2);
-	RIP3(E,A,B,C,D,WL46,SL46,KL2);
-	RIP3(D,E,A,B,C,WL47,SL47,KL2);
-
-	RIP4(C,D,E,A,B,WL48,SL48,KL3);
-	RIP4(B,C,D,E,A,WL49,SL49,KL3);
-	RIP4(A,B,C,D,E,WL50,SL50,KL3);
-	RIP4(E,A,B,C,D,WL51,SL51,KL3);
-	RIP4(D,E,A,B,C,WL52,SL52,KL3);
-	RIP4(C,D,E,A,B,WL53,SL53,KL3);
-	RIP4(B,C,D,E,A,WL54,SL54,KL3);
-	RIP4(A,B,C,D,E,WL55,SL55,KL3);
-	RIP4(E,A,B,C,D,WL56,SL56,KL3);
-	RIP4(D,E,A,B,C,WL57,SL57,KL3);
-	RIP4(C,D,E,A,B,WL58,SL58,KL3);
-	RIP4(B,C,D,E,A,WL59,SL59,KL3);
-	RIP4(A,B,C,D,E,WL60,SL60,KL3);
-	RIP4(E,A,B,C,D,WL61,SL61,KL3);
-	RIP4(D,E,A,B,C,WL62,SL62,KL3);
-	RIP4(C,D,E,A,B,WL63,SL63,KL3);
-
-	RIP5(B,C,D,E,A,WL64,SL64,KL4);
-	RIP5(A,B,C,D,E,WL65,SL65,KL4);
-	RIP5(E,A,B,C,D,WL66,SL66,KL4);
-	RIP5(D,E,A,B,C,WL67,SL67,KL4);
-	RIP5(C,D,E,A,B,WL68,SL68,KL4);
-	RIP5(B,C,D,E,A,WL69,SL69,KL4);
-	RIP5(A,B,C,D,E,WL70,SL70,KL4);
-	RIP5(E,A,B,C,D,WL71,SL71,KL4);
-	RIP5(D,E,A,B,C,WL72,SL72,KL4);
-	RIP5(C,D,E,A,B,WL73,SL73,KL4);
-	RIP5(B,C,D,E,A,WL74,SL74,KL4);
-	RIP5(A,B,C,D,E,WL75,SL75,KL4);
-	RIP5(E,A,B,C,D,WL76,SL76,KL4);
-	RIP5(D,E,A,B,C,WL77,SL77,KL4);
-	RIP5(C,D,E,A,B,WL78,SL78,KL4);
-	RIP5(B,C,D,E,A,WL79,SL79,KL4);
-
-	a=A; b=B; c=C; d=D; e=E;
-	/* Do other half */
-	A=ctx->A; B=ctx->B; C=ctx->C; D=ctx->D; E=ctx->E;
-
-	RIP5(A,B,C,D,E,WR00,SR00,KR0);
-	RIP5(E,A,B,C,D,WR01,SR01,KR0);
-	RIP5(D,E,A,B,C,WR02,SR02,KR0);
-	RIP5(C,D,E,A,B,WR03,SR03,KR0);
-	RIP5(B,C,D,E,A,WR04,SR04,KR0);
-	RIP5(A,B,C,D,E,WR05,SR05,KR0);
-	RIP5(E,A,B,C,D,WR06,SR06,KR0);
-	RIP5(D,E,A,B,C,WR07,SR07,KR0);
-	RIP5(C,D,E,A,B,WR08,SR08,KR0);
-	RIP5(B,C,D,E,A,WR09,SR09,KR0);
-	RIP5(A,B,C,D,E,WR10,SR10,KR0);
-	RIP5(E,A,B,C,D,WR11,SR11,KR0);
-	RIP5(D,E,A,B,C,WR12,SR12,KR0);
-	RIP5(C,D,E,A,B,WR13,SR13,KR0);
-	RIP5(B,C,D,E,A,WR14,SR14,KR0);
-	RIP5(A,B,C,D,E,WR15,SR15,KR0);
-
-	RIP4(E,A,B,C,D,WR16,SR16,KR1);
-	RIP4(D,E,A,B,C,WR17,SR17,KR1);
-	RIP4(C,D,E,A,B,WR18,SR18,KR1);
-	RIP4(B,C,D,E,A,WR19,SR19,KR1);
-	RIP4(A,B,C,D,E,WR20,SR20,KR1);
-	RIP4(E,A,B,C,D,WR21,SR21,KR1);
-	RIP4(D,E,A,B,C,WR22,SR22,KR1);
-	RIP4(C,D,E,A,B,WR23,SR23,KR1);
-	RIP4(B,C,D,E,A,WR24,SR24,KR1);
-	RIP4(A,B,C,D,E,WR25,SR25,KR1);
-	RIP4(E,A,B,C,D,WR26,SR26,KR1);
-	RIP4(D,E,A,B,C,WR27,SR27,KR1);
-	RIP4(C,D,E,A,B,WR28,SR28,KR1);
-	RIP4(B,C,D,E,A,WR29,SR29,KR1);
-	RIP4(A,B,C,D,E,WR30,SR30,KR1);
-	RIP4(E,A,B,C,D,WR31,SR31,KR1);
-
-	RIP3(D,E,A,B,C,WR32,SR32,KR2);
-	RIP3(C,D,E,A,B,WR33,SR33,KR2);
-	RIP3(B,C,D,E,A,WR34,SR34,KR2);
-	RIP3(A,B,C,D,E,WR35,SR35,KR2);
-	RIP3(E,A,B,C,D,WR36,SR36,KR2);
-	RIP3(D,E,A,B,C,WR37,SR37,KR2);
-	RIP3(C,D,E,A,B,WR38,SR38,KR2);
-	RIP3(B,C,D,E,A,WR39,SR39,KR2);
-	RIP3(A,B,C,D,E,WR40,SR40,KR2);
-	RIP3(E,A,B,C,D,WR41,SR41,KR2);
-	RIP3(D,E,A,B,C,WR42,SR42,KR2);
-	RIP3(C,D,E,A,B,WR43,SR43,KR2);
-	RIP3(B,C,D,E,A,WR44,SR44,KR2);
-	RIP3(A,B,C,D,E,WR45,SR45,KR2);
-	RIP3(E,A,B,C,D,WR46,SR46,KR2);
-	RIP3(D,E,A,B,C,WR47,SR47,KR2);
-
-	RIP2(C,D,E,A,B,WR48,SR48,KR3);
-	RIP2(B,C,D,E,A,WR49,SR49,KR3);
-	RIP2(A,B,C,D,E,WR50,SR50,KR3);
-	RIP2(E,A,B,C,D,WR51,SR51,KR3);
-	RIP2(D,E,A,B,C,WR52,SR52,KR3);
-	RIP2(C,D,E,A,B,WR53,SR53,KR3);
-	RIP2(B,C,D,E,A,WR54,SR54,KR3);
-	RIP2(A,B,C,D,E,WR55,SR55,KR3);
-	RIP2(E,A,B,C,D,WR56,SR56,KR3);
-	RIP2(D,E,A,B,C,WR57,SR57,KR3);
-	RIP2(C,D,E,A,B,WR58,SR58,KR3);
-	RIP2(B,C,D,E,A,WR59,SR59,KR3);
-	RIP2(A,B,C,D,E,WR60,SR60,KR3);
-	RIP2(E,A,B,C,D,WR61,SR61,KR3);
-	RIP2(D,E,A,B,C,WR62,SR62,KR3);
-	RIP2(C,D,E,A,B,WR63,SR63,KR3);
-
-	RIP1(B,C,D,E,A,WR64,SR64);
-	RIP1(A,B,C,D,E,WR65,SR65);
-	RIP1(E,A,B,C,D,WR66,SR66);
-	RIP1(D,E,A,B,C,WR67,SR67);
-	RIP1(C,D,E,A,B,WR68,SR68);
-	RIP1(B,C,D,E,A,WR69,SR69);
-	RIP1(A,B,C,D,E,WR70,SR70);
-	RIP1(E,A,B,C,D,WR71,SR71);
-	RIP1(D,E,A,B,C,WR72,SR72);
-	RIP1(C,D,E,A,B,WR73,SR73);
-	RIP1(B,C,D,E,A,WR74,SR74);
-	RIP1(A,B,C,D,E,WR75,SR75);
-	RIP1(E,A,B,C,D,WR76,SR76);
-	RIP1(D,E,A,B,C,WR77,SR77);
-	RIP1(C,D,E,A,B,WR78,SR78);
-	RIP1(B,C,D,E,A,WR79,SR79);
-
-	D     =ctx->B+c+D;
-	ctx->B=ctx->C+d+E;
-	ctx->C=ctx->D+e+A;
-	ctx->D=ctx->E+a+B;
-	ctx->E=ctx->A+b+C;
-	ctx->A=D;
-
-		}
-	}
-#endif
-
 #ifndef ripemd160_block_data_order
 #ifdef X
 #undef X

crypto/ripemd/rmd_locl.h

@@ -72,32 +72,20 @@
  */
 #ifdef RMD160_ASM
 # if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
-#  if !defined(B_ENDIAN)
+#  define ripemd160_block_data_order ripemd160_block_asm_data_order
-#   define ripemd160_block_host_order ripemd160_block_asm_host_order
-#  endif
 # endif
 #endif
 
-void ripemd160_block_host_order (RIPEMD160_CTX *c, const void *p,size_t num);
 void ripemd160_block_data_order (RIPEMD160_CTX *c, const void *p,size_t num);
 
-#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
-# if !defined(B_ENDIAN)
-#  define ripemd160_block_data_order ripemd160_block_host_order
-# endif
-#endif
-
 #define DATA_ORDER_IS_LITTLE_ENDIAN
 
 #define HASH_LONG               RIPEMD160_LONG
-#define HASH_LONG_LOG2          RIPEMD160_LONG_LOG2
 #define HASH_CTX                RIPEMD160_CTX
 #define HASH_CBLOCK             RIPEMD160_CBLOCK
-#define HASH_LBLOCK             RIPEMD160_LBLOCK
 #define HASH_UPDATE             RIPEMD160_Update
 #define HASH_TRANSFORM          RIPEMD160_Transform
 #define HASH_FINAL              RIPEMD160_Final
-#define HASH_BLOCK_HOST_ORDER   ripemd160_block_host_order
 #define	HASH_MAKE_STRING(c,s)	do {	\
 	unsigned long ll;		\
 	ll=(c)->A; HOST_l2c(ll,(s));	\
@@ -106,9 +94,7 @@ void ripemd160_block_data_order (RIPEMD160_CTX *c, const void *p,size_t num);
 	ll=(c)->D; HOST_l2c(ll,(s));	\
 	ll=(c)->E; HOST_l2c(ll,(s));	\
 	} while (0)
-#if !defined(L_ENDIAN) || defined(ripemd160_block_data_order)
 #define HASH_BLOCK_DATA_ORDER   ripemd160_block_data_order
-#endif
 
 #include "md32_common.h"
 

crypto/sha/asm/sha1-586.pl

@@ -1,4 +1,16 @@
-#!/usr/local/bin/perl
+#!/usr/bin/env perl
+
+# ====================================================================
+# [Re]written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# "[Re]written" was achieved in two major overhauls. In 2004 BODY_*
+# functions were re-implemented to address P4 performance issue [see
+# commentary below], and in 2006 the rest was rewritten in order to
+# gain freedom to liberate licensing terms.
 
 # It was noted that Intel IA-32 C compiler generates code which
 # performs ~30% *faster* on P4 CPU than original *hand-coded*
@@ -17,90 +29,27 @@
 # improvement on P4 outweights the loss and incorporate this
 # re-tuned code to 0.9.7 and later.
 # ----------------------------------------------------------------
-# Those who for any particular reason absolutely must score on
-# Pentium can replace this module with one from 0.9.6 distribution.
-# This "offer" shall be revoked the moment programming interface to
-# this module is changed, in which case this paragraph should be
-# removed.
-# ----------------------------------------------------------------
 #					<appro@fy.chalmers.se>
 
-$normal=0;
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
-
+push(@INC,"${dir}","${dir}../../perlasm");
-push(@INC,"perlasm","../../perlasm");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386");
 
 $A="eax";
-$B="ecx";
+$B="ebx";
-$C="ebx";
+$C="ecx";
 $D="edx";
 $E="edi";
 $T="esi";
 $tmp1="ebp";
 
-$off=9*4;
+@V=($A,$B,$C,$D,$E,$T);
-
-@K=(0x5a827999,0x6ed9eba1,0x8f1bbcdc,0xca62c1d6);
-
-&sha1_block_data("sha1_block_asm_data_order");
-
-&asm_finish();
-
-sub Nn
-	{
-	local($p)=@_;
-	local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E);
-	return($n{$p});
-	}
-
-sub Np
-	{
-	local($p)=@_;
-	local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E);
-	local(%n)=($A,$B,$B,$C,$C,$D,$D,$E,$E,$T,$T,$A);
-	return($n{$p});
-	}
-
-sub Na
-	{
-	local($n)=@_;
-	return( (($n   )&0x0f),
-		(($n+ 2)&0x0f),
-		(($n+ 8)&0x0f),
-		(($n+13)&0x0f),
-		(($n+ 1)&0x0f));
-	}
-
-sub X_expand
-	{
-	local($in)=@_;
-
-	&comment("First, load the words onto the stack in network byte order");
-	for ($i=0; $i<16; $i+=2)
-		{
-		&mov($A,&DWP(($i+0)*4,$in,"",0));# unless $i == 0;
-		 &mov($B,&DWP(($i+1)*4,$in,"",0));
-		&bswap($A);
-		 &bswap($B);
-		&mov(&swtmp($i+0),$A);
-		 &mov(&swtmp($i+1),$B);
-		}
-
-	&comment("We now have the X array on the stack");
-	&comment("starting at sp-4");
-	}
-
-# Rules of engagement
-# F is always trashable at the start, the running total.
-# E becomes the next F so it can be trashed after it has been 'accumulated'
-# F becomes A in the next round.  We don't need to access it much.
-# During the X update part, the result ends up in $X[$n0].
 
 sub BODY_00_15
 	{
-	local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
+	local($n,$a,$b,$c,$d,$e,$f)=@_;
 
 	&comment("00_15 $n");
 
@@ -109,37 +58,37 @@ sub BODY_00_15
 	 else        { &mov($a,$tmp1); }
 	&rotl($tmp1,5);			# tmp1=ROTATE(a,5)
 	 &xor($f,$d);
-	&and($f,$b);
 	&add($tmp1,$e);			# tmp1+=e;
-	&mov($e,&swtmp($n));		# e becomes volatile and
+	 &and($f,$b);
-	 				# is loaded with xi
+	&mov($e,&swtmp($n%16));		# e becomes volatile and is loaded
+	 				# with xi, also note that e becomes
+					# f in next round...
 	 &xor($f,$d);			# f holds F_00_19(b,c,d)
 	&rotr($b,2);			# b=ROTATE(b,30)
-	 &lea($tmp1,&DWP($K,$tmp1,$e,1));# tmp1+=K_00_19+xi
+	 &lea($tmp1,&DWP(0x5a827999,$tmp1,$e));	# tmp1+=K_00_19+xi
 
 	if ($n==15) { &add($f,$tmp1); }	# f+=tmp1
-	else        { &add($tmp1,$f); }
+	else        { &add($tmp1,$f); }	# f becomes a in next round
 	}
 
 sub BODY_16_19
 	{
-	local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
+	local($n,$a,$b,$c,$d,$e,$f)=@_;
-	local($n0,$n1,$n2,$n3,$np)=&Na($n);
 
 	&comment("16_19 $n");
 
-	&mov($f,&swtmp($n1));		# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&mov($f,&swtmp($n%16));		# f to hold Xupdate(xi,xa,xb,xc,xd)
 	 &mov($tmp1,$c);		# tmp1 to hold F_00_19(b,c,d)
-	&xor($f,&swtmp($n0));
+	&xor($f,&swtmp(($n+2)%16));
 	 &xor($tmp1,$d);
-	&xor($f,&swtmp($n2));
+	&xor($f,&swtmp(($n+8)%16));
 	 &and($tmp1,$b);		# tmp1 holds F_00_19(b,c,d)
 	&rotr($b,2);			# b=ROTATE(b,30)
-	 &xor($f,&swtmp($n3));		# f holds xa^xb^xc^xd
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
-	&rotl($f,1);			# f=ROATE(f,1)
+	&rotl($f,1);			# f=ROTATE(f,1)
 	 &xor($tmp1,$d);		# tmp1=F_00_19(b,c,d)
-	&mov(&swtmp($n0),$f);		# xi=f
+	&mov(&swtmp($n%16),$f);		# xi=f
-	&lea($f,&DWP($K,$f,$e,1));	# f+=K_00_19+e
+	&lea($f,&DWP(0x5a827999,$f,$e));# f+=K_00_19+e
 	 &mov($e,$a);			# e becomes volatile
 	&rotl($e,5);			# e=ROTATE(a,5)
 	 &add($f,$tmp1);		# f+=F_00_19(b,c,d)
@@ -148,48 +97,47 @@ sub BODY_16_19
 
 sub BODY_20_39
 	{
-	local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
+	local($n,$a,$b,$c,$d,$e,$f)=@_;
+	local $K=($n<40)?0x6ed9eba1:0xca62c1d6;
 
 	&comment("20_39 $n");
-	local($n0,$n1,$n2,$n3,$np)=&Na($n);
 
 	&mov($tmp1,$b);			# tmp1 to hold F_20_39(b,c,d)
-	 &mov($f,&swtmp($n0));		# f to hold Xupdate(xi,xa,xb,xc,xd)
+	 &mov($f,&swtmp($n%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
 	&rotr($b,2);			# b=ROTATE(b,30)
-	 &xor($f,&swtmp($n1));
+	 &xor($f,&swtmp(($n+2)%16));
 	&xor($tmp1,$c);
-	 &xor($f,&swtmp($n2));
+	 &xor($f,&swtmp(($n+8)%16));
 	&xor($tmp1,$d);			# tmp1 holds F_20_39(b,c,d)
-	 &xor($f,&swtmp($n3));		# f holds xa^xb^xc^xd
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
 	&rotl($f,1);			# f=ROTATE(f,1)
 	 &add($tmp1,$e);
-	&mov(&swtmp($n0),$f);		# xi=f
+	&mov(&swtmp($n%16),$f);		# xi=f
 	 &mov($e,$a);			# e becomes volatile
 	&rotl($e,5);			# e=ROTATE(a,5)
-	 &lea($f,&DWP($K,$f,$tmp1,1));	# f+=K_20_39+e
+	 &lea($f,&DWP($K,$f,$tmp1));	# f+=K_20_39+e
 	&add($f,$e);			# f+=ROTATE(a,5)
 	}
 
 sub BODY_40_59
 	{
-	local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
+	local($n,$a,$b,$c,$d,$e,$f)=@_;
 
 	&comment("40_59 $n");
-	local($n0,$n1,$n2,$n3,$np)=&Na($n);
 
-	&mov($f,&swtmp($n0));		# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&mov($f,&swtmp($n%16));		# f to hold Xupdate(xi,xa,xb,xc,xd)
-	 &mov($tmp1,&swtmp($n1));
+	 &mov($tmp1,&swtmp(($n+2)%16));
 	&xor($f,$tmp1);
-	 &mov($tmp1,&swtmp($n2));
+	 &mov($tmp1,&swtmp(($n+8)%16));
 	&xor($f,$tmp1);
-	 &mov($tmp1,&swtmp($n3));
+	 &mov($tmp1,&swtmp(($n+13)%16));
 	&xor($f,$tmp1);			# f holds xa^xb^xc^xd
 	 &mov($tmp1,$b);		# tmp1 to hold F_40_59(b,c,d)
 	&rotl($f,1);			# f=ROTATE(f,1)
 	 &or($tmp1,$c);
-	&mov(&swtmp($n0),$f);		# xi=f
+	&mov(&swtmp($n%16),$f);		# xi=f
 	 &and($tmp1,$d);
-	&lea($f,&DWP($K,$f,$e,1));	# f+=K_40_59+e
+	&lea($f,&DWP(0x8f1bbcdc,$f,$e));# f+=K_40_59+e
 	 &mov($e,$b);			# e becomes volatile and is used
 					# to calculate F_40_59(b,c,d)
 	&rotr($b,2);			# b=ROTATE(b,30)
@@ -201,230 +149,71 @@ sub BODY_40_59
 	&add($f,$e);			# f+=ROTATE(a,5)
 	}
 
-sub BODY_60_79
+&function_begin("sha1_block_data_order",16);
+	&mov($tmp1,&wparam(0));	# SHA_CTX *c
+	&mov($T,&wparam(1));	# const void *input
+	&mov($A,&wparam(2));	# size_t num
+	&stack_push(16);	# allocate X[16]
+	&shl($A,6);
+	&add($A,$T);
+	&mov(&wparam(2),$A);	# pointer beyond the end of input
+	&mov($E,&DWP(16,$tmp1));# pre-load E
+
+	&set_label("loop",16);
+
+	# copy input chunk to X, but reversing byte order!
+	for ($i=0; $i<16; $i+=4)
 		{
-	&BODY_20_39(@_);
+		&mov($A,&DWP(4*($i+0),$T));
-	}
+		&mov($B,&DWP(4*($i+1),$T));
-
+		&mov($C,&DWP(4*($i+2),$T));
-sub sha1_block_host
+		&mov($D,&DWP(4*($i+3),$T));
-	{
+		&bswap($A);
-	local($name, $sclabel)=@_;
+		&bswap($B);
-
+		&bswap($C);
-	&function_begin_B($name,"");
+		&bswap($D);
-
-	# parameter 1 is the MD5_CTX structure.
-	# A	0
-	# B	4
-	# C	8
-	# D 	12
-	# E 	16
-
-	&mov("ecx",	&wparam(2));
-	 &push("esi");
-	&shl("ecx",6);
-	 &mov("esi",	&wparam(1));
-	&push("ebp");
-	 &add("ecx","esi");	# offset to leave on
-	&push("ebx");
-	 &mov("ebp",	&wparam(0));
-	&push("edi");
-	 &mov($D,	&DWP(12,"ebp","",0));
-	&stack_push(18+9);
-	 &mov($E,	&DWP(16,"ebp","",0));
-	&mov($C,	&DWP( 8,"ebp","",0));
-	 &mov(&swtmp(17),"ecx");
-
-	&comment("First we need to setup the X array");
-
-	for ($i=0; $i<16; $i+=2)
-		{
-		&mov($A,&DWP(($i+0)*4,"esi","",0));# unless $i == 0;
-		 &mov($B,&DWP(($i+1)*4,"esi","",0));
 		&mov(&swtmp($i+0),$A);
 		&mov(&swtmp($i+1),$B);
+		&mov(&swtmp($i+2),$C);
+		&mov(&swtmp($i+3),$D);
 		}
-	&jmp($sclabel);
+	&mov(&wparam(1),$T);	# redundant in 1st spin
-	&function_end_B($name);
-	}
 
+	&mov($A,&DWP(0,$tmp1));	# load SHA_CTX
+	&mov($B,&DWP(4,$tmp1));
+	&mov($C,&DWP(8,$tmp1));
+	&mov($D,&DWP(12,$tmp1));
+	# E is pre-loaded
 
-sub sha1_block_data
+	for($i=0;$i<16;$i++)	{ &BODY_00_15($i,@V); unshift(@V,pop(@V)); }
-	{
+	for(;$i<20;$i++)	{ &BODY_16_19($i,@V); unshift(@V,pop(@V)); }
-	local($name)=@_;
+	for(;$i<40;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+	for(;$i<60;$i++)	{ &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
+	for(;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
 
-	&function_begin_B($name,"");
+	(($V[5] eq $D) and ($V[0] eq $E)) or die;	# double-check
 
-	# parameter 1 is the MD5_CTX structure.
+	&mov($tmp1,&wparam(0));	# re-load SHA_CTX*
-	# A	0
+	&mov($D,&wparam(1));	# D is last "T" and is discarded
-	# B	4
-	# C	8
-	# D 	12
-	# E 	16
 
-	&mov("ecx",	&wparam(2));
+	&add($E,&DWP(0,$tmp1));	# E is last "A"...
-	 &push("esi");
+	&add($T,&DWP(4,$tmp1));
-	&shl("ecx",6);
+	&add($A,&DWP(8,$tmp1));
-	 &mov("esi",	&wparam(1));
+	&add($B,&DWP(12,$tmp1));
-	&push("ebp");
+	&add($C,&DWP(16,$tmp1));
-	 &add("ecx","esi");	# offset to leave on
-	&push("ebx");
-	 &mov("ebp",	&wparam(0));
-	&push("edi");
-	 &mov($D,	&DWP(12,"ebp","",0));
-	&stack_push(18+9);
-	 &mov($E,	&DWP(16,"ebp","",0));
-	&mov($C,	&DWP( 8,"ebp","",0));
-	 &mov(&swtmp(17),"ecx");
 
-	&comment("First we need to setup the X array");
+	&mov(&DWP(0,$tmp1),$E);	# update SHA_CTX
+	 &add($D,64);		# advance input pointer
+	&mov(&DWP(4,$tmp1),$T);
+	 &cmp($D,&wparam(2));	# have we reached the end yet?
+	&mov(&DWP(8,$tmp1),$A);
+	 &mov($E,$C);		# C is last "E" which needs to be "pre-loaded"
+	&mov(&DWP(12,$tmp1),$B);
+	 &mov($T,$D);		# input pointer
+	&mov(&DWP(16,$tmp1),$C);
+	&jb(&label("loop"));
 
-	&set_label("start") unless $normal;
+	&stack_pop(16);
-
+&function_end("sha1_block_data_order");
-	&X_expand("esi");
-	 &mov(&wparam(1),"esi");
-
-	&set_label("shortcut", 0, 1);
-	&comment("");
-	&comment("Start processing");
-
-	# odd start
-	&mov($A,	&DWP( 0,"ebp","",0));
-	 &mov($B,	&DWP( 4,"ebp","",0));
-	$X="esp";
-	&BODY_00_15(-2,$K[0],$X, 0,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15( 0,$K[0],$X, 1,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15( 0,$K[0],$X, 2,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15( 0,$K[0],$X, 3,$D,$E,$T,$A,$B,$C);
-	&BODY_00_15( 0,$K[0],$X, 4,$C,$D,$E,$T,$A,$B);
-	&BODY_00_15( 0,$K[0],$X, 5,$B,$C,$D,$E,$T,$A);
-	&BODY_00_15( 0,$K[0],$X, 6,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15( 0,$K[0],$X, 7,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15( 0,$K[0],$X, 8,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15( 0,$K[0],$X, 9,$D,$E,$T,$A,$B,$C);
-	&BODY_00_15( 0,$K[0],$X,10,$C,$D,$E,$T,$A,$B);
-	&BODY_00_15( 0,$K[0],$X,11,$B,$C,$D,$E,$T,$A);
-	&BODY_00_15( 0,$K[0],$X,12,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15( 0,$K[0],$X,13,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15( 0,$K[0],$X,14,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15( 1,$K[0],$X,15,$D,$E,$T,$A,$B,$C);
-	&BODY_16_19(-1,$K[0],$X,16,$C,$D,$E,$T,$A,$B);
-	&BODY_16_19( 0,$K[0],$X,17,$B,$C,$D,$E,$T,$A);
-	&BODY_16_19( 0,$K[0],$X,18,$A,$B,$C,$D,$E,$T);
-	&BODY_16_19( 1,$K[0],$X,19,$T,$A,$B,$C,$D,$E);
-
-	&BODY_20_39(-1,$K[1],$X,20,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39( 0,$K[1],$X,21,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39( 0,$K[1],$X,22,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39( 0,$K[1],$X,23,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39( 0,$K[1],$X,24,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39( 0,$K[1],$X,25,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39( 0,$K[1],$X,26,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39( 0,$K[1],$X,27,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39( 0,$K[1],$X,28,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39( 0,$K[1],$X,29,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39( 0,$K[1],$X,30,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39( 0,$K[1],$X,31,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39( 0,$K[1],$X,32,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39( 0,$K[1],$X,33,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39( 0,$K[1],$X,34,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39( 0,$K[1],$X,35,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39( 0,$K[1],$X,36,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39( 0,$K[1],$X,37,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39( 0,$K[1],$X,38,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39( 1,$K[1],$X,39,$D,$E,$T,$A,$B,$C);
-
-	&BODY_40_59(-1,$K[2],$X,40,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59( 0,$K[2],$X,41,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59( 0,$K[2],$X,42,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59( 0,$K[2],$X,43,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59( 0,$K[2],$X,44,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59( 0,$K[2],$X,45,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59( 0,$K[2],$X,46,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59( 0,$K[2],$X,47,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59( 0,$K[2],$X,48,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59( 0,$K[2],$X,49,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59( 0,$K[2],$X,50,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59( 0,$K[2],$X,51,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59( 0,$K[2],$X,52,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59( 0,$K[2],$X,53,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59( 0,$K[2],$X,54,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59( 0,$K[2],$X,55,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59( 0,$K[2],$X,56,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59( 0,$K[2],$X,57,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59( 0,$K[2],$X,58,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59( 1,$K[2],$X,59,$B,$C,$D,$E,$T,$A);
-
-	&BODY_60_79(-1,$K[3],$X,60,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79( 0,$K[3],$X,61,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79( 0,$K[3],$X,62,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79( 0,$K[3],$X,63,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79( 0,$K[3],$X,64,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79( 0,$K[3],$X,65,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79( 0,$K[3],$X,66,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79( 0,$K[3],$X,67,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79( 0,$K[3],$X,68,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79( 0,$K[3],$X,69,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79( 0,$K[3],$X,70,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79( 0,$K[3],$X,71,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79( 0,$K[3],$X,72,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79( 0,$K[3],$X,73,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79( 0,$K[3],$X,74,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79( 0,$K[3],$X,75,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79( 0,$K[3],$X,76,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79( 0,$K[3],$X,77,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79( 0,$K[3],$X,78,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79( 2,$K[3],$X,79,$T,$A,$B,$C,$D,$E);
-
-	&comment("End processing");
-	&comment("");
-	# D is the tmp value
-
-	# E -> A
-	# T -> B
-	# A -> C
-	# B -> D
-	# C -> E
-	# D -> T
-
-	&mov($tmp1,&wparam(0));
-
-	 &mov($D,	&DWP(12,$tmp1,"",0));
-	&add($D,$B);
-	 &mov($B,	&DWP( 4,$tmp1,"",0));
-	&add($B,$T);
-	 &mov($T,	$A);
-	&mov($A,	&DWP( 0,$tmp1,"",0));
-	 &mov(&DWP(12,$tmp1,"",0),$D);
-
-	&add($A,$E);
-	 &mov($E,	&DWP(16,$tmp1,"",0));
-	&add($E,$C);
-	 &mov($C,	&DWP( 8,$tmp1,"",0));
-	&add($C,$T);
-
-	 &mov(&DWP( 0,$tmp1,"",0),$A);
-	&mov("esi",&wparam(1));
-	 &mov(&DWP( 8,$tmp1,"",0),$C);
- 	&add("esi",64);
-	 &mov("eax",&swtmp(17));
-	&mov(&DWP(16,$tmp1,"",0),$E);
-	 &cmp("esi","eax");
-	&mov(&DWP( 4,$tmp1,"",0),$B);
-	 &jb(&label("start"));
-
-	&stack_pop(18+9);
-	 &pop("edi");
-	&pop("ebx");
-	 &pop("ebp");
-	&pop("esi");
-	 &ret();
-
-	# keep a note of shortcut label so it can be used outside
-	# block.
-	my $sclabel = &label("shortcut");
-
-	&function_end_B($name);
-	# Putting this here avoids problems with MASM in debugging mode
-	&sha1_block_host("sha1_block_asm_host_order", $sclabel);
-	}
 
+&asm_finish();

crypto/sha/asm/sha1-ia64.pl

@@ -2,8 +2,9 @@
 #
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
+# project. The module is, however, dual licensed under OpenSSL and
-# forms are granted according to the OpenSSL license.
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 #
 # Eternal question is what's wrong with compiler generated code? The
@@ -11,15 +12,10 @@
 # to perform rotations by maintaining copy of 32-bit value in upper
 # bits of 64-bit register. Just follow mux2 and shrp instructions...
 # Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which
-# is >50% better than HP C and >2x better than gcc. As of this moment
+# is >50% better than HP C and >2x better than gcc.
-# performance under little-endian OS such as Linux and Windows will be
-# a bit lower, because data has to be picked in reverse byte-order.
-# It's possible to resolve this issue by implementing third function,
-# sha1_block_asm_data_order_aligned, which would temporarily flip
-# BE field in User Mask register...
 
 $code=<<___;
-.ident  \"sha1-ia64.s, version 1.0\"
+.ident  \"sha1-ia64.s, version 1.2\"
 .ident  \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
 .explicit
 
@@ -55,63 +51,55 @@ else {
 
 sub BODY_00_15 {
 local	*code=shift;
-local	($i,$a,$b,$c,$d,$e,$f,$unaligned)=@_;
+local	($i,$a,$b,$c,$d,$e,$f)=@_;
 
-if ($unaligned) {
+$code.=<<___ if ($i==0);
-	$code.=<<___;
+{ .mmi;	ld1	$X[$i&0xf]=[inp],2	    // MSB
-{ .mmi;	ld1	tmp0=[inp],2		    // MSB
+	ld1	tmp2=[tmp3],2		};;
-	ld1	tmp1=[tmp3],2		};;
+{ .mmi;	ld1	tmp0=[inp],2
-{ .mmi;	ld1	tmp2=[inp],2
+	ld1	tmp4=[tmp3],2		    // LSB
-	ld1	$X[$i&0xf]=[tmp3],2	    // LSB
+	dep	$X[$i&0xf]=$X[$i&0xf],tmp2,8,8	};;
-	dep	tmp1=tmp0,tmp1,8,8	};;
-{ .mii;	cmp.ne	p16,p0=r0,r0		    // no misaligned prefetch
-	dep	$X[$i&0xf]=tmp2,$X[$i&0xf],8,8;;
-	dep	$X[$i&0xf]=tmp1,$X[$i&0xf],16,16	};;
-{ .mmi;	nop.m	0
 ___
-	}
-elsif ($i<15) {
-	$code.=<<___;
-{ .mmi;	ld4	$X[($i+1)&0xf]=[inp],4	// prefetch
-___
-	}
-else	{
-	$code.=<<___;
-{ .mmi;	nop.m	0
-___
-	}
 if ($i<15) {
 	$code.=<<___;
-	and	tmp0=$c,$b
+{ .mmi;	ld1	$X[($i+1)&0xf]=[inp],2	    // +1
-	dep.z	tmp5=$a,5,27		}   // a<<5
+	dep	tmp1=tmp0,tmp4,8,8	};;
+{ .mmi;	ld1	tmp2=[tmp3],2		    // +1
+	and	tmp4=$c,$b
+	dep	$X[$i&0xf]=$X[$i&0xf],tmp1,16,16	} //;;
 { .mmi;	andcm	tmp1=$d,$b
-	add	tmp4=$e,$K_00_19	};;
+	add	tmp0=$e,$K_00_19
-{ .mmi;	or	tmp0=tmp0,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
+	dep.z	tmp5=$a,5,27		};; // a<<5
-	add	$f=tmp4,$X[$i&0xf]	    // f=xi+e+K_00_19
+{ .mmi;	or	tmp4=tmp4,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
+	add	$f=tmp0,$X[$i&0xf]	    // f=xi+e+K_00_19
 	extr.u	tmp1=$a,27,5		};; // a>>27
-{ .mib;	add	$f=$f,tmp0		    // f+=F_00_19(b,c,d)
+{ .mmi;	ld1	tmp0=[inp],2		    // +1
+	add	$f=$f,tmp4		    // f+=F_00_19(b,c,d)
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
-{ .mib;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
+{ .mmi;	ld1	tmp4=[tmp3],2		    // +1
+	or	tmp5=tmp1,tmp5		    // ROTATE(a,5)
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp1		    // f+=ROTATE(a,5)
+{ .mii;	add	$f=$f,tmp5		    // f+=ROTATE(a,5)
-	mux2	$X[$i&0xf]=$X[$i&0xf],0x44
+	dep	$X[($i+1)&0xf]=$X[($i+1)&0xf],tmp2,8,8	// +1
-	nop.i	0			};;
+	mux2	$X[$i&0xf]=$X[$i&0xf],0x44	} //;;
 
 ___
 	}
 else	{
 	$code.=<<___;
-	and	tmp0=$c,$b
+{ .mii;	and	tmp3=$c,$b
-	dep.z	tmp5=$a,5,27		}   // a<<5 ;;?
+	dep	tmp1=tmp0,tmp4,8,8;;
+	dep	$X[$i&0xf]=$X[$i&0xf],tmp1,16,16	} //;;
 { .mmi;	andcm	tmp1=$d,$b
-	add	tmp4=$e,$K_00_19	};;
+	add	tmp0=$e,$K_00_19
-{ .mmi;	or	tmp0=tmp0,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
+	dep.z	tmp5=$a,5,27		};; // a<<5
-	add	$f=tmp4,$X[$i&0xf]	    // f=xi+e+K_00_19
+{ .mmi;	or	tmp4=tmp3,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
+	add	$f=tmp0,$X[$i&0xf]	    // f=xi+e+K_00_19
 	extr.u	tmp1=$a,27,5		}   // a>>27
 { .mmi;	xor	tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf]	// +1
 	xor	tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
 	nop.i	0			};;
-{ .mmi;	add	$f=$f,tmp0		    // f+=F_00_19(b,c,d)
+{ .mmi;	add	$f=$f,tmp4		    // f+=F_00_19(b,c,d)
 	xor	tmp2=tmp2,tmp3		    // +1
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
 { .mmi; or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
@@ -190,9 +178,7 @@ $code.=<<___;
 	extr.u	tmp1=$a,27,5		}   // a>>27
 { .mib;	add	$f=$f,tmp4		    // f+=e+K_20_39
 	add	$h1=$h1,$a		};; // wrap up
-{ .mmi;
+{ .mmi;	add	$f=$f,tmp0		    // f+=F_20_39(b,c,d)
-(p16)	ld4.s	$X[0]=[inp],4		    // non-faulting prefetch
-	add	$f=$f,tmp0		    // f+=F_20_39(b,c,d)
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30) ;;?
 { .mmi;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
 	add	$h3=$h3,$c		};; // wrap up
@@ -245,164 +231,11 @@ tmp3=r11;
 ctx=r32;	// in0
 inp=r33;	// in1
 
-// void sha1_block_asm_host_order(SHA_CTX *c,const void *p,size_t num);
+// void sha1_block_data_order(SHA_CTX *c,const void *p,size_t num);
-.global	sha1_block_asm_host_order#
+.global	sha1_block_data_order#
-.proc	sha1_block_asm_host_order#
+.proc	sha1_block_data_order#
 .align	32
-sha1_block_asm_host_order:
+sha1_block_data_order:
-	.prologue
-{ .mmi;	alloc	tmp1=ar.pfs,3,15,0,0
-	$ADDP	tmp0=4,ctx
-	.save	ar.lc,r3
-	mov	r3=ar.lc		}
-{ .mmi;	$ADDP	ctx=0,ctx
-	$ADDP	inp=0,inp
-	mov	r2=pr			};;
-tmp4=in2;
-tmp5=loc13;
-tmp6=loc14;
-	.body
-{ .mlx;	ld4	$h0=[ctx],8
-	movl	$K_00_19=0x5a827999	}
-{ .mlx;	ld4	$h1=[tmp0],8
-	movl	$K_20_39=0x6ed9eba1	};;
-{ .mlx;	ld4	$h2=[ctx],8
-	movl	$K_40_59=0x8f1bbcdc	}
-{ .mlx;	ld4	$h3=[tmp0]
-	movl	$K_60_79=0xca62c1d6	};;
-{ .mmi;	ld4	$h4=[ctx],-16
-	add	in2=-1,in2		    // adjust num for ar.lc
-	mov	ar.ec=1			};;
-{ .mmi;	ld4	$X[0]=[inp],4		    // prefetch
-	cmp.ne	p16,p0=r0,in2		    // prefecth at loop end
-	mov	ar.lc=in2		};; // brp.loop.imp: too far
-
-.Lhtop:
-{ .mmi;	mov	$A=$h0
-	mov	$B=$h1
-	mux2	tmp6=$h1,0x44		}
-{ .mmi;	mov	$C=$h2
-	mov	$D=$h3
-	mov	$E=$h4			};;
-
-___
-
-	&BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C);
-	&BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B);
-	&BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A);
-	&BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C);
-	&BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B);
-	&BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A);
-	&BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T);
-	&BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E);
-	&BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D);
-	&BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C);
-
-	&BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
-	&BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
-	&BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
-	&BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
-
-	&BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
-
-	&BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
-
-	&BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
-
-$code.=<<___;
-{ .mmb;	add	$h0=$h0,$E
-	nop.m	0
-	br.ctop.dptk.many	.Lhtop	};;
-.Lhend:
-{ .mmi;	add	tmp0=4,ctx
-	mov	ar.lc=r3		};;
-{ .mmi;	st4	[ctx]=$h0,8
-	st4	[tmp0]=$h1,8		};;
-{ .mmi;	st4	[ctx]=$h2,8
-	st4	[tmp0]=$h3		};;
-{ .mib;	st4	[ctx]=$h4,-16
-	mov	pr=r2,0x1ffff
-	br.ret.sptk.many	b0	};;
-.endp	sha1_block_asm_host_order#
-___
-
-
-$code.=<<___;
-// void sha1_block_asm_data_order(SHA_CTX *c,const void *p,size_t num);
-.global	sha1_block_asm_data_order#
-.proc	sha1_block_asm_data_order#
-.align	32
-sha1_block_asm_data_order:
-___
-$code.=<<___ if ($big_endian);
-{ .mmi;	and	r2=3,inp				};;
-{ .mib;	cmp.eq	p6,p0=r0,r2
-(p6)	br.dptk.many	sha1_block_asm_host_order	};;
-___
-$code.=<<___;
 	.prologue
 { .mmi;	alloc	tmp1=ar.pfs,3,15,0,0
 	$ADDP	tmp0=4,ctx
@@ -440,90 +273,16 @@ tmp6=loc14;
 
 ___
 
-	&BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T,1);
+{ my $i,@V=($A,$B,$C,$D,$E,$T);
-	&BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E,1);
-	&BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D,1);
-	&BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C,1);
-	&BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B,1);
-	&BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A,1);
-	&BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T,1);
-	&BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E,1);
-	&BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D,1);
-	&BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C,1);
-	&BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B,1);
-	&BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A,1);
-	&BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T,1);
-	&BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E,1);
-	&BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D,1);
-	&BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C,1);
 
-	&BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
+	for($i=0;$i<16;$i++)	{ &BODY_00_15(\$code,$i,@V); unshift(@V,pop(@V)); }
-	&BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
+	for(;$i<20;$i++)	{ &BODY_16_19(\$code,$i,@V); unshift(@V,pop(@V)); }
-	&BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
+	for(;$i<40;$i++)	{ &BODY_20_39(\$code,$i,@V); unshift(@V,pop(@V)); }
-	&BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
+	for(;$i<60;$i++)	{ &BODY_40_59(\$code,$i,@V); unshift(@V,pop(@V)); }
+	for(;$i<80;$i++)	{ &BODY_60_79(\$code,$i,@V); unshift(@V,pop(@V)); }
 
-	&BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
+	(($V[5] eq $D) and ($V[0] eq $E)) or die;	# double-check
-	&BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
+}
-	&BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
-	&BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
-	&BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
-	&BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
-	&BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
-	&BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
-	&BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
-
-	&BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
-	&BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
-	&BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
-	&BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
-	&BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
-	&BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
-	&BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
-
-	&BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
-	&BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
-	&BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
-	&BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
-	&BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
-	&BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
-	&BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
 
 $code.=<<___;
 { .mmb;	add	$h0=$h0,$E
@@ -539,7 +298,8 @@ $code.=<<___;
 { .mib;	st4	[ctx]=$h4,-16
 	mov	pr=r2,0x1ffff
 	br.ret.sptk.many	b0	};;
-.endp	sha1_block_asm_data_order#
+.endp	sha1_block_data_order#
+stringz	"SHA1 block transform for IA64, CRYPTOGAMS by <appro\@openssl.org>"
 ___
 
 print $code;

crypto/sha/asm/sha512-ia64.pl

@@ -2,8 +2,9 @@
 #
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
+# project. The module is, however, dual licensed under OpenSSL and
-# forms are granted according to the OpenSSL license.
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 #
 # SHA256/512_Transform for Itanium.
@@ -71,7 +72,7 @@ if ($output =~ /512.*\.[s|asm]/) {
 	$ADD="add";
 	$SHRU="shr.u";
 	$TABLE="K512";
-	$func="sha512_block";
+	$func="sha512_block_data_order";
 	@Sigma0=(28,34,39);
 	@Sigma1=(14,18,41);
 	@sigma0=(1,  8, 7);
@@ -85,7 +86,7 @@ if ($output =~ /512.*\.[s|asm]/) {
 	$ADD="padd4";
 	$SHRU="pshr4.u";
 	$TABLE="K256";
-	$func="sha256_block";
+	$func="sha256_block_data_order";
 	@Sigma0=( 2,13,22);
 	@Sigma1=( 6,11,25);
 	@sigma0=( 7,18, 3);
@@ -105,11 +106,13 @@ if (!defined($big_endian))
              {	$big_endian=(unpack('L',pack('N',1))==1);  }
 
 $code=<<___;
-.ident  \"$output, version 1.0\"
+.ident  \"$output, version 1.1\"
 .ident  \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
 .explicit
 .text
 
+pfssave=r2;
+lcsave=r3;
 prsave=r14;
 K=r15;
 A=r16;	B=r17;	C=r18;	D=r19;
@@ -121,6 +124,8 @@ ctx=r31;	// 1st arg
 input=r48;	// 2nd arg
 num=r49;	// 3rd arg
 sgm0=r50;	sgm1=r51;	// small constants
+A_=r54;	B_=r55;	C_=r56;	D_=r57;
+E_=r58;	F_=r59;	G_=r60;	H_=r61;
 
 // void $func (SHA_CTX *ctx, const void *in,size_t num[,int host])
 .global	$func#
@@ -128,81 +133,319 @@ sgm0=r50;	sgm1=r51;	// small constants
 .align	32
 $func:
 	.prologue
-	.save	ar.pfs,r2
+	.save	ar.pfs,pfssave
-{ .mmi;	alloc	r2=ar.pfs,3,17,0,16
+{ .mmi;	alloc	pfssave=ar.pfs,3,27,0,16
 	$ADDP	ctx=0,r32		// 1st arg
-	.save	ar.lc,r3
+	.save	ar.lc,lcsave
-	mov	r3=ar.lc	}
+	mov	lcsave=ar.lc	}
 { .mmi;	$ADDP	input=0,r33		// 2nd arg
-	addl	Ktbl=\@ltoff($TABLE#),gp
+	mov	num=r34			// 3rd arg
 	.save	pr,prsave
 	mov	prsave=pr	};;
 
 	.body
-{ .mii;	ld8	Ktbl=[Ktbl]
-	mov	num=r34		};;	// 3rd arg
-
 { .mib;	add	r8=0*$SZ,ctx
 	add	r9=1*$SZ,ctx
-	brp.loop.imp	.L_first16,.L_first16_ctop
+	brp.loop.imp	.L_first16,.L_first16_end-16	}
-				}
 { .mib;	add	r10=2*$SZ,ctx
 	add	r11=3*$SZ,ctx
-	brp.loop.imp	.L_rest,.L_rest_ctop
+	brp.loop.imp	.L_rest,.L_rest_end-16		};;
-				};;
-// load A-H
-{ .mmi;	$LDW	A=[r8],4*$SZ
-	$LDW	B=[r9],4*$SZ
-	mov	sgm0=$sigma0[2]	}
-{ .mmi;	$LDW	C=[r10],4*$SZ
-	$LDW	D=[r11],4*$SZ
-	mov	sgm1=$sigma1[2]	};;
-{ .mmi;	$LDW	E=[r8]
-	$LDW	F=[r9]		}
-{ .mmi;	$LDW	G=[r10]
-	$LDW	H=[r11]
-	cmp.ne	p15,p14=0,r35	};;	// used in sha256_block
 
+// load A-H
+.Lpic_point:
+{ .mmi;	$LDW	A_=[r8],4*$SZ
+	$LDW	B_=[r9],4*$SZ
+	mov	Ktbl=ip		}
+{ .mmi;	$LDW	C_=[r10],4*$SZ
+	$LDW	D_=[r11],4*$SZ
+	mov	sgm0=$sigma0[2]	};;
+{ .mmi;	$LDW	E_=[r8]
+	$LDW	F_=[r9]
+	add	Ktbl=($TABLE#-.Lpic_point),Ktbl		}
+{ .mmi;	$LDW	G_=[r10]
+	$LDW	H_=[r11]
+	cmp.ne	p0,p16=0,r0	};;	// used in sha256_block
+___
+$code.=<<___ if ($BITS==64);
+{ .mii;	and	r8=7,input
+	and	input=~7,input;;
+	cmp.eq	p9,p0=1,r8	}
+{ .mmi;	cmp.eq	p10,p0=2,r8
+	cmp.eq	p11,p0=3,r8
+	cmp.eq	p12,p0=4,r8	}
+{ .mmi;	cmp.eq	p13,p0=5,r8
+	cmp.eq	p14,p0=6,r8
+	cmp.eq	p15,p0=7,r8	};;
+___
+$code.=<<___;
 .L_outer:
-{ .mii;	mov	ar.lc=15
-	mov	ar.ec=1		};;
-.align	32
-.L_first16:
 .rotr	X[16]
+{ .mmi;	mov	A=A_
+	mov	B=B_
+	mov	ar.lc=14	}
+{ .mmi;	mov	C=C_
+	mov	D=D_
+	mov	E=E_		}
+{ .mmi;	mov	F=F_
+	mov	G=G_
+	mov	ar.ec=2		}
+{ .mmi;	ld1	X[15]=[input],$SZ		// eliminated in 64-bit
+	mov	H=H_
+	mov	sgm1=$sigma1[2]	};;
+
 ___
 $t0="t0", $t1="t1", $code.=<<___ if ($BITS==32);
-{ .mib;	(p14)	add	r9=1,input
+.align	32
-	(p14)	add	r10=2,input	}
+.L_first16:
-{ .mib;	(p14)	add	r11=3,input
+{ .mmi;		add	r9=1-$SZ,input
-	(p15)	br.dptk.few	.L_host	};;
+		add	r10=2-$SZ,input
-{ .mmi;	(p14)	ld1	r8=[input],$SZ
+		add	r11=3-$SZ,input	};;
-	(p14)	ld1	r9=[r9]		}
+{ .mmi;		ld1	r9=[r9]
-{ .mmi;	(p14)	ld1	r10=[r10]
+		ld1	r10=[r10]
-	(p14)	ld1	r11=[r11]	};;
-{ .mii;	(p14)	dep	r9=r8,r9,8,8
-	(p14)	dep	r11=r10,r11,8,8	};;
-{ .mib;	(p14)	dep	X[15]=r9,r11,16,16 };;
-.L_host:
-{ .mib;	(p15)	$LDW	X[15]=[input],$SZ	// X[i]=*input++
 		dep.z	$t1=E,32,32	}
-{ .mib;		$LDW	K=[Ktbl],$SZ
+{ .mmi;		$LDW	K=[Ktbl],$SZ
+		ld1	r11=[r11]
 		zxt4	E=E		};;
-{ .mmi;		or	$t1=$t1,E
+{ .mii;		or	$t1=$t1,E
-		and	T1=F,E
+		dep	X[15]=X[15],r9,8,8
-		and	T2=A,B		}
+		dep	r11=r10,r11,8,8	};;
+{ .mmi;		and	T1=F,E
+		and	T2=A,B
+		dep	X[15]=X[15],r11,16,16	}
 { .mmi;		andcm	r8=G,E
 		and	r9=A,C
 		mux2	$t0=A,0x44	};;	// copy lower half to upper
-{ .mib;		xor	T1=T1,r8		// T1=((e & f) ^ (~e & g))
+{ .mmi;	(p16)	ld1	X[15-1]=[input],$SZ	// prefetch
+		xor	T1=T1,r8		// T1=((e & f) ^ (~e & g))
 		_rotr	r11=$t1,$Sigma1[0] }	// ROTR(e,14)
 { .mib;		and	r10=B,C
 		xor	T2=T2,r9	};;
 ___
 $t0="A", $t1="E", $code.=<<___ if ($BITS==64);
-{ .mmi;		$LDW	X[15]=[input],$SZ	// X[i]=*input++
+// in 64-bit mode I load whole X[16] at once and take care of alignment...
+{ .mmi;	add	r8=1*$SZ,input
+	add	r9=2*$SZ,input
+	add	r10=3*$SZ,input		};;
+{ .mmb;	$LDW	X[15]=[input],4*$SZ
+	$LDW	X[14]=[r8],4*$SZ
+(p9)	br.cond.dpnt.many	.L1byte	};;
+{ .mmb;	$LDW	X[13]=[r9],4*$SZ
+	$LDW	X[12]=[r10],4*$SZ
+(p10)	br.cond.dpnt.many	.L2byte	};;
+{ .mmb;	$LDW	X[11]=[input],4*$SZ
+	$LDW	X[10]=[r8],4*$SZ
+(p11)	br.cond.dpnt.many	.L3byte	};;
+{ .mmb;	$LDW	X[ 9]=[r9],4*$SZ
+	$LDW	X[ 8]=[r10],4*$SZ
+(p12)	br.cond.dpnt.many	.L4byte	};;
+{ .mmb;	$LDW	X[ 7]=[input],4*$SZ
+	$LDW	X[ 6]=[r8],4*$SZ
+(p13)	br.cond.dpnt.many	.L5byte	};;
+{ .mmb;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+(p14)	br.cond.dpnt.many	.L6byte	};;
+{ .mmb;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+(p15)	br.cond.dpnt.many	.L7byte	};;
+{ .mmb;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	br.many	.L_first16		};;
+.L1byte:
+{ .mmi;	$LDW	X[13]=[r9],4*$SZ
+	$LDW	X[12]=[r10],4*$SZ
+	shrp	X[15]=X[15],X[14],56	};;
+{ .mmi;	$LDW	X[11]=[input],4*$SZ
+	$LDW	X[10]=[r8],4*$SZ
+	shrp	X[14]=X[14],X[13],56	}
+{ .mmi;	$LDW	X[ 9]=[r9],4*$SZ
+	$LDW	X[ 8]=[r10],4*$SZ
+	shrp	X[13]=X[13],X[12],56	};;
+{ .mmi;	$LDW	X[ 7]=[input],4*$SZ
+	$LDW	X[ 6]=[r8],4*$SZ
+	shrp	X[12]=X[12],X[11],56	}
+{ .mmi;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+	shrp	X[11]=X[11],X[10],56	};;
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[10]=X[10],X[ 9],56	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[ 9]=X[ 9],X[ 8],56	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[ 8]=X[ 8],X[ 7],56
+	shrp	X[ 7]=X[ 7],X[ 6],56	}
+{ .mii;	shrp	X[ 6]=X[ 6],X[ 5],56
+	shrp	X[ 5]=X[ 5],X[ 4],56	};;
+{ .mii;	shrp	X[ 4]=X[ 4],X[ 3],56
+	shrp	X[ 3]=X[ 3],X[ 2],56	}
+{ .mii;	shrp	X[ 2]=X[ 2],X[ 1],56
+	shrp	X[ 1]=X[ 1],X[ 0],56	}
+{ .mib;	shrp	X[ 0]=X[ 0],T1,56
+	br.many	.L_first16		};;
+.L2byte:
+{ .mmi;	$LDW	X[11]=[input],4*$SZ
+	$LDW	X[10]=[r8],4*$SZ
+	shrp	X[15]=X[15],X[14],48	}
+{ .mmi;	$LDW	X[ 9]=[r9],4*$SZ
+	$LDW	X[ 8]=[r10],4*$SZ
+	shrp	X[14]=X[14],X[13],48	};;
+{ .mmi;	$LDW	X[ 7]=[input],4*$SZ
+	$LDW	X[ 6]=[r8],4*$SZ
+	shrp	X[13]=X[13],X[12],48	}
+{ .mmi;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+	shrp	X[12]=X[12],X[11],48	};;
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[11]=X[11],X[10],48	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[10]=X[10],X[ 9],48	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[ 9]=X[ 9],X[ 8],48
+	shrp	X[ 8]=X[ 8],X[ 7],48	}
+{ .mii;	shrp	X[ 7]=X[ 7],X[ 6],48
+	shrp	X[ 6]=X[ 6],X[ 5],48	};;
+{ .mii;	shrp	X[ 5]=X[ 5],X[ 4],48
+	shrp	X[ 4]=X[ 4],X[ 3],48	}
+{ .mii;	shrp	X[ 3]=X[ 3],X[ 2],48
+	shrp	X[ 2]=X[ 2],X[ 1],48	}
+{ .mii;	shrp	X[ 1]=X[ 1],X[ 0],48
+	shrp	X[ 0]=X[ 0],T1,48	}
+{ .mfb;	br.many	.L_first16		};;
+.L3byte:
+{ .mmi;	$LDW	X[ 9]=[r9],4*$SZ
+	$LDW	X[ 8]=[r10],4*$SZ
+	shrp	X[15]=X[15],X[14],40	};;
+{ .mmi;	$LDW	X[ 7]=[input],4*$SZ
+	$LDW	X[ 6]=[r8],4*$SZ
+	shrp	X[14]=X[14],X[13],40	}
+{ .mmi;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+	shrp	X[13]=X[13],X[12],40	};;
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[12]=X[12],X[11],40	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[11]=X[11],X[10],40	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[10]=X[10],X[ 9],40
+	shrp	X[ 9]=X[ 9],X[ 8],40	}
+{ .mii;	shrp	X[ 8]=X[ 8],X[ 7],40
+	shrp	X[ 7]=X[ 7],X[ 6],40	};;
+{ .mii;	shrp	X[ 6]=X[ 6],X[ 5],40
+	shrp	X[ 5]=X[ 5],X[ 4],40	}
+{ .mii;	shrp	X[ 4]=X[ 4],X[ 3],40
+	shrp	X[ 3]=X[ 3],X[ 2],40	}
+{ .mii;	shrp	X[ 2]=X[ 2],X[ 1],40
+	shrp	X[ 1]=X[ 1],X[ 0],40	}
+{ .mib;	shrp	X[ 0]=X[ 0],T1,40
+	br.many	.L_first16		};;
+.L4byte:
+{ .mmi;	$LDW	X[ 7]=[input],4*$SZ
+	$LDW	X[ 6]=[r8],4*$SZ
+	shrp	X[15]=X[15],X[14],32	}
+{ .mmi;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+	shrp	X[14]=X[14],X[13],32	};;
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[13]=X[13],X[12],32	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[12]=X[12],X[11],32	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[11]=X[11],X[10],32
+	shrp	X[10]=X[10],X[ 9],32	}
+{ .mii;	shrp	X[ 9]=X[ 9],X[ 8],32
+	shrp	X[ 8]=X[ 8],X[ 7],32	};;
+{ .mii;	shrp	X[ 7]=X[ 7],X[ 6],32
+	shrp	X[ 6]=X[ 6],X[ 5],32	}
+{ .mii;	shrp	X[ 5]=X[ 5],X[ 4],32
+	shrp	X[ 4]=X[ 4],X[ 3],32	}
+{ .mii;	shrp	X[ 3]=X[ 3],X[ 2],32
+	shrp	X[ 2]=X[ 2],X[ 1],32	}
+{ .mii;	shrp	X[ 1]=X[ 1],X[ 0],32
+	shrp	X[ 0]=X[ 0],T1,32	}
+{ .mfb;	br.many	.L_first16		};;
+.L5byte:
+{ .mmi;	$LDW	X[ 5]=[r9],4*$SZ
+	$LDW	X[ 4]=[r10],4*$SZ
+	shrp	X[15]=X[15],X[14],24	};;
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[14]=X[14],X[13],24	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[13]=X[13],X[12],24	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[12]=X[12],X[11],24
+	shrp	X[11]=X[11],X[10],24	}
+{ .mii;	shrp	X[10]=X[10],X[ 9],24
+	shrp	X[ 9]=X[ 9],X[ 8],24	};;
+{ .mii;	shrp	X[ 8]=X[ 8],X[ 7],24
+	shrp	X[ 7]=X[ 7],X[ 6],24	}
+{ .mii;	shrp	X[ 6]=X[ 6],X[ 5],24
+	shrp	X[ 5]=X[ 5],X[ 4],24	}
+{ .mii;	shrp	X[ 4]=X[ 4],X[ 3],24
+	shrp	X[ 3]=X[ 3],X[ 2],24	}
+{ .mii;	shrp	X[ 2]=X[ 2],X[ 1],24
+	shrp	X[ 1]=X[ 1],X[ 0],24	}
+{ .mib;	shrp	X[ 0]=X[ 0],T1,24
+	br.many	.L_first16		};;
+.L6byte:
+{ .mmi;	$LDW	X[ 3]=[input],4*$SZ
+	$LDW	X[ 2]=[r8],4*$SZ
+	shrp	X[15]=X[15],X[14],16	}
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[14]=X[14],X[13],16	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[13]=X[13],X[12],16
+	shrp	X[12]=X[12],X[11],16	}
+{ .mii;	shrp	X[11]=X[11],X[10],16
+	shrp	X[10]=X[10],X[ 9],16	};;
+{ .mii;	shrp	X[ 9]=X[ 9],X[ 8],16
+	shrp	X[ 8]=X[ 8],X[ 7],16	}
+{ .mii;	shrp	X[ 7]=X[ 7],X[ 6],16
+	shrp	X[ 6]=X[ 6],X[ 5],16	}
+{ .mii;	shrp	X[ 5]=X[ 5],X[ 4],16
+	shrp	X[ 4]=X[ 4],X[ 3],16	}
+{ .mii;	shrp	X[ 3]=X[ 3],X[ 2],16
+	shrp	X[ 2]=X[ 2],X[ 1],16	}
+{ .mii;	shrp	X[ 1]=X[ 1],X[ 0],16
+	shrp	X[ 0]=X[ 0],T1,16	}
+{ .mfb;	br.many	.L_first16		};;
+.L7byte:
+{ .mmi;	$LDW	X[ 1]=[r9],4*$SZ
+	$LDW	X[ 0]=[r10],4*$SZ
+	shrp	X[15]=X[15],X[14],8	};;
+{ .mii;	$LDW	T1=[input]
+	shrp	X[14]=X[14],X[13],8
+	shrp	X[13]=X[13],X[12],8	}
+{ .mii;	shrp	X[12]=X[12],X[11],8
+	shrp	X[11]=X[11],X[10],8	};;
+{ .mii;	shrp	X[10]=X[10],X[ 9],8
+	shrp	X[ 9]=X[ 9],X[ 8],8	}
+{ .mii;	shrp	X[ 8]=X[ 8],X[ 7],8
+	shrp	X[ 7]=X[ 7],X[ 6],8	}
+{ .mii;	shrp	X[ 6]=X[ 6],X[ 5],8
+	shrp	X[ 5]=X[ 5],X[ 4],8	}
+{ .mii;	shrp	X[ 4]=X[ 4],X[ 3],8
+	shrp	X[ 3]=X[ 3],X[ 2],8	}
+{ .mii;	shrp	X[ 2]=X[ 2],X[ 1],8
+	shrp	X[ 1]=X[ 1],X[ 0],8	}
+{ .mib;	shrp	X[ 0]=X[ 0],T1,8
+	br.many	.L_first16		};;
+
+.align	32
+.L_first16:
+{ .mmi;		$LDW	K=[Ktbl],$SZ
 		and	T1=F,E
 		and	T2=A,B		}
-{ .mmi;		$LDW	K=[Ktbl],$SZ
+{ .mmi;		//$LDW	X[15]=[input],$SZ	// X[i]=*input++
 		andcm	r8=G,E
 		and	r9=A,C		};;
 { .mmi;		xor	T1=T1,r8		//T1=((e & f) ^ (~e & g))
@@ -235,13 +478,14 @@ $code.=<<___;
 { .mmi;		xor	r10=r8,r10		// r10=Sigma0(a)
 		mov	B=A
 		add	A=T1,T2		};;
-.L_first16_ctop:
 { .mib;		add	E=E,T1
 		add	A=A,r10			// T2=Maj(a,b,c)+Sigma0(a)
 	br.ctop.sptk	.L_first16	};;
+.L_first16_end:
+
+{ .mii;	mov	ar.lc=$rounds-17
+	mov	ar.ec=1			};;
 
-{ .mib;	mov	ar.lc=$rounds-17	}
-{ .mib;	mov	ar.ec=1			};;
 .align	32
 .L_rest:
 .rotr	X[16]
@@ -310,46 +554,38 @@ $code.=<<___;
 { .mmi;		xor	r10=r8,r10		// r10=Sigma0(a)
 		mov	B=A
 		add	A=T1,T2		};;
-.L_rest_ctop:
 { .mib;		add	E=E,T1
 		add	A=A,r10			// T2=Maj(a,b,c)+Sigma0(a)
 	br.ctop.sptk	.L_rest	};;
+.L_rest_end:
+
+{ .mmi;	add	A_=A_,A
+	add	B_=B_,B
+	add	C_=C_,C			}
+{ .mmi;	add	D_=D_,D
+	add	E_=E_,E
+	cmp.ltu	p16,p0=1,num		};;
+{ .mmi;	add	F_=F_,F
+	add	G_=G_,G
+	add	H_=H_,H			}
+{ .mmb;	add	Ktbl=-$SZ*$rounds,Ktbl
+(p16)	add	num=-1,num
+(p16)	br.dptk.many	.L_outer	};;
 
 { .mib;	add	r8=0*$SZ,ctx
 	add	r9=1*$SZ,ctx		}
 { .mib;	add	r10=2*$SZ,ctx
 	add	r11=3*$SZ,ctx		};;
-{ .mmi;	$LDW	r32=[r8],4*$SZ
+{ .mmi;	$STW	[r8]=A_,4*$SZ
-	$LDW	r33=[r9],4*$SZ		}
+	$STW	[r9]=B_,4*$SZ
-{ .mmi;	$LDW	r34=[r10],4*$SZ
+	mov	ar.lc=lcsave		}
-	$LDW	r35=[r11],4*$SZ
+{ .mmi;	$STW	[r10]=C_,4*$SZ
-	cmp.ltu	p6,p7=1,num		};;
+	$STW	[r11]=D_,4*$SZ
-{ .mmi;	$LDW	r36=[r8],-4*$SZ
+	mov	pr=prsave,0x1ffff	};;
-	$LDW	r37=[r9],-4*$SZ
+{ .mmb;	$STW	[r8]=E_
-(p6)	add	Ktbl=-$SZ*$rounds,Ktbl	}
+	$STW	[r9]=F_			}
-{ .mmi;	$LDW	r38=[r10],-4*$SZ
+{ .mmb;	$STW	[r10]=G_
-	$LDW	r39=[r11],-4*$SZ
+	$STW	[r11]=H_
-(p7)	mov	ar.lc=r3		};;
-{ .mmi;	add	A=A,r32
-	add	B=B,r33
-	add	C=C,r34			}
-{ .mmi;	add	D=D,r35
-	add	E=E,r36
-	add	F=F,r37			};;
-{ .mmi;	$STW	[r8]=A,4*$SZ
-	$STW	[r9]=B,4*$SZ
-	add	G=G,r38			}
-{ .mmi;	$STW	[r10]=C,4*$SZ
-	$STW	[r11]=D,4*$SZ
-	add	H=H,r39			};;
-{ .mmi;	$STW	[r8]=E
-	$STW	[r9]=F
-(p6)	add	num=-1,num		}
-{ .mmb;	$STW	[r10]=G
-	$STW	[r11]=H
-(p6)	br.dptk.many	.L_outer	};;
-
-{ .mib;	mov	pr=prsave,0x1ffff
 	br.ret.sptk.many	b0	};;
 .endp	$func#
 ___
@@ -359,6 +595,9 @@ $code =~ s/_rotr(\s+)([^=]+)=([^,]+),([0-9]+)/shrp$1$2=$3,$3,$4/gm;
 if ($BITS==64) {
     $code =~ s/mux2(\s+)\S+/nop.i$1 0x0/gm;
     $code =~ s/mux1(\s+)\S+/nop.i$1 0x0/gm	if ($big_endian);
+    $code =~ s/(shrp\s+X\[[^=]+)=([^,]+),([^,]+),([1-9]+)/$1=$3,$2,64-$4/gm
+    						if (!$big_endian);
+    $code =~ s/ld1(\s+)X\[\S+/nop.m$1 0x0/gm;
 }
 
 print $code;
@@ -383,6 +622,7 @@ K256:	data4	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
 	data4	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
 	data4	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
 .size	K256#,$SZ*$rounds
+stringz	"SHA256 block transform for IA64, CRYPTOGAMS by <appro\@openssl.org>"
 ___
 print<<___ if ($BITS==64);
 .align	64
@@ -428,4 +668,5 @@ K512:	data8	0x428a2f98d728ae22,0x7137449123ef65cd
 	data8	0x4cc5d4becb3e42b6,0x597f299cfc657e2a
 	data8	0x5fcb6fab3ad6faec,0x6c44198c4a475817
 .size	K512#,$SZ*$rounds
+stringz	"SHA512 block transform for IA64, CRYPTOGAMS by <appro\@openssl.org>"
 ___

crypto/sha/sha256.c

@@ -69,17 +69,11 @@ int SHA224_Update(SHA256_CTX *c, const void *data, size_t len)
 int SHA224_Final (unsigned char *md, SHA256_CTX *c)
 {   return SHA256_Final (md,c);   }
 
-#ifndef	SHA_LONG_LOG2
-#define	SHA_LONG_LOG2	2	/* default to 32 bits */
-#endif
-
 #define	DATA_ORDER_IS_BIG_ENDIAN
 
 #define	HASH_LONG		SHA_LONG
-#define	HASH_LONG_LOG2		SHA_LONG_LOG2
 #define	HASH_CTX		SHA256_CTX
 #define	HASH_CBLOCK		SHA_CBLOCK
-#define	HASH_LBLOCK		SHA_LBLOCK
 /*
  * Note that FIPS180-2 discusses "Truncation of the Hash Function Output."
  * default: case below covers for it. It's not clear however if it's
@@ -112,16 +106,15 @@ int SHA224_Final (unsigned char *md, SHA256_CTX *c)
 #define	HASH_UPDATE		SHA256_Update
 #define	HASH_TRANSFORM		SHA256_Transform
 #define	HASH_FINAL		SHA256_Final
-#define	HASH_BLOCK_HOST_ORDER	sha256_block_host_order
 #define	HASH_BLOCK_DATA_ORDER	sha256_block_data_order
-void sha256_block_host_order (SHA256_CTX *ctx, const void *in, size_t num);
+#ifndef SHA256_ASM
+static
+#endif
 void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num);
 
 #include "md32_common.h"
 
-#ifdef SHA256_ASM
+#ifndef SHA256_ASM
-void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host);
-#else
 static const SHA_LONG K256[64] = {
 	0x428a2f98UL,0x71374491UL,0xb5c0fbcfUL,0xe9b5dba5UL,
 	0x3956c25bUL,0x59f111f1UL,0x923f82a4UL,0xab1c5ed5UL,
@@ -155,10 +148,10 @@ static const SHA_LONG K256[64] = {
 
 #ifdef OPENSSL_SMALL_FOOTPRINT
 
-static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
+static void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num)
 	{
 	unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1,T2;
-	SHA_LONG	X[16];
+	SHA_LONG	X[16],l;
 	int i;
 	const unsigned char *data=in;
 
@@ -167,25 +160,6 @@ static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
 	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
 	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
 
-	if (host)
-		{
-		const SHA_LONG *W=(const SHA_LONG *)data;
-
-		for (i=0;i<16;i++)
-			{
-			T1 = X[i] = W[i];
-			T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i];
-			T2 = Sigma0(a) + Maj(a,b,c);
-			h = g;	g = f;	f = e;	e = d + T1;
-			d = c;	c = b;	b = a;	a = T1 + T2;
-			}
-
-		data += SHA256_CBLOCK;
-		}
-	else
-		{
-		SHA_LONG l;
-
 	for (i=0;i<16;i++)
 		{
 		HOST_c2l(data,l); T1 = X[i] = l;
@@ -194,7 +168,6 @@ static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
 		h = g;	g = f;	f = e;	e = d + T1;
 		d = c;	c = b;	b = a;	a = T1 + T2;
 		}
-		}
 
 	for (;i<64;i++)
 		{
@@ -227,19 +200,20 @@ static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
 	T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f];	\
 	ROUND_00_15(i,a,b,c,d,e,f,g,h);		} while (0)
 
-static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
+static void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num)
 	{
 	unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1;
 	SHA_LONG	X[16];
 	int i;
 	const unsigned char *data=in;
+	const union { long one; char little; } is_endian = {1};
 
 			while (num--) {
 
 	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
 	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
 
-	if (host)
+	if (!is_endian.little && sizeof(SHA_LONG)==4 && ((size_t)in%4)==0)
 		{
 		const SHA_LONG *W=(const SHA_LONG *)data;
 
@@ -305,15 +279,4 @@ static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
 #endif
 #endif /* SHA256_ASM */
 
-/*
- * Idea is to trade couple of cycles for some space. On IA-32 we save
- * about 4K in "big footprint" case. In "small footprint" case any gain
- * is appreciated:-)
- */
-void HASH_BLOCK_HOST_ORDER (SHA256_CTX *ctx, const void *in, size_t num)
-{   sha256_block (ctx,in,num,1);   }
-
-void HASH_BLOCK_DATA_ORDER (SHA256_CTX *ctx, const void *in, size_t num)
-{   sha256_block (ctx,in,num,0);   }
-
 #endif /* OPENSSL_NO_SHA256 */

crypto/sha/sha512.c

@@ -52,7 +52,10 @@
 
 const char SHA512_version[]="SHA-512" OPENSSL_VERSION_PTEXT;
 
-#if defined(_M_IX86) || defined(_M_AMD64) || defined(__i386) || defined(__x86_64)
+#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || \
+    defined(__x86_64) || defined(_M_AMD64) || defined(_M_X64) || \
+    defined(__s390__) || defined(__s390x__) || \
+    defined(SHA512_ASM)
 #define SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
 #endif
 
@@ -89,7 +92,7 @@ int SHA512_Init (SHA512_CTX *c)
 #ifndef SHA512_ASM
 static
 #endif
-void sha512_block (SHA512_CTX *ctx, const void *in, size_t num);
+void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num);
 
 int SHA512_Final (unsigned char *md, SHA512_CTX *c)
 	{
@@ -100,7 +103,7 @@ int SHA512_Final (unsigned char *md, SHA512_CTX *c)
 	n++;
 	if (n > (sizeof(c->u)-16))
 		memset (p+n,0,sizeof(c->u)-n), n=0,
-		sha512_block (c,p,1);
+		sha512_block_data_order (c,p,1);
 
 	memset (p+n,0,sizeof(c->u)-16-n);
 #ifdef	B_ENDIAN
@@ -125,7 +128,7 @@ int SHA512_Final (unsigned char *md, SHA512_CTX *c)
 	p[sizeof(c->u)-16] = (unsigned char)(c->Nh>>56);
 #endif
 
-	sha512_block (c,p,1);
+	sha512_block_data_order (c,p,1);
 
 	if (md==0) return 0;
 
@@ -197,7 +200,7 @@ int SHA512_Update (SHA512_CTX *c, const void *_data, size_t len)
 		else	{
 			memcpy (p+c->num,data,n), c->num = 0;
 			len-=n, data+=n;
-			sha512_block (c,p,1);
+			sha512_block_data_order (c,p,1);
 			}
 		}
 
@@ -207,12 +210,12 @@ int SHA512_Update (SHA512_CTX *c, const void *_data, size_t len)
 		if ((size_t)data%sizeof(c->u.d[0]) != 0)
 			while (len >= sizeof(c->u))
 				memcpy (p,data,sizeof(c->u)),
-				sha512_block (c,p,1),
+				sha512_block_data_order (c,p,1),
 				len  -= sizeof(c->u),
 				data += sizeof(c->u);
 		else
 #endif
-			sha512_block (c,data,len/sizeof(c->u)),
+			sha512_block_data_order (c,data,len/sizeof(c->u)),
 			data += len,
 			len  %= sizeof(c->u),
 			data -= len;
@@ -227,7 +230,7 @@ int SHA384_Update (SHA512_CTX *c, const void *data, size_t len)
 {   return SHA512_Update (c,data,len);   }
 
 void SHA512_Transform (SHA512_CTX *c, const unsigned char *data)
-{   sha512_block (c,data,1);  }
+{   sha512_block_data_order (c,data,1);  }
 
 unsigned char *SHA384(const unsigned char *d, size_t n, unsigned char *md)
 	{
@@ -301,11 +304,70 @@ static const SHA_LONG64 K512[80] = {
 #ifndef PEDANTIC
 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
 #  if defined(__x86_64) || defined(__x86_64__)
+#   define ROTR(a,n)	({ unsigned long ret;		\
+				asm ("rorq %1,%0"	\
+				: "=r"(ret)		\
+				: "J"(n),"0"(a)		\
+				: "cc"); ret;		})
+#   if !defined(B_ENDIAN)
 #    define PULL64(x) ({ SHA_LONG64 ret=*((const SHA_LONG64 *)(&(x)));	\
 				asm ("bswapq	%0"		\
 				: "=r"(ret)			\
 				: "0"(ret)); ret;		})
 #   endif
+#  elif (defined(__i386) || defined(__i386__)) && !defined(B_ENDIAN)
+#   if defined(I386_ONLY)
+#    define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\
+			unsigned int hi,lo;			\
+				asm("xchgb %%ah,%%al;xchgb %%dh,%%dl;"\
+				    "roll $16,%%eax; roll $16,%%edx; "\
+				    "xchgb %%ah,%%al;xchgb %%dh,%%dl;" \
+				: "=a"(lo),"=d"(hi)		\
+				: "0"(p[1]),"1"(p[0]) : "cc");	\
+				((SHA_LONG64)hi)<<32|lo;	})
+#   else
+#    define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\
+			unsigned int hi,lo;			\
+				asm ("bswapl %0; bswapl %1;"	\
+				: "=r"(lo),"=r"(hi)		\
+				: "0"(p[1]),"1"(p[0]));		\
+				((SHA_LONG64)hi)<<32|lo;	})
+#   endif
+#  elif (defined(_ARCH_PPC) && defined(__64BIT__)) || defined(_ARCH_PPC64)
+#   define ROTR(a,n)	({ unsigned long ret;		\
+				asm ("rotrdi %0,%1,%2"	\
+				: "=r"(ret)		\
+				: "r"(a),"K"(n)); ret;	})
+#  endif
+# elif defined(_MSC_VER)
+#  if defined(_WIN64)	/* applies to both IA-64 and AMD64 */
+#   define ROTR(a,n)	_rotr64((a),n)
+#  endif
+#  if defined(_M_IX86) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
+#   if defined(I386_ONLY)
+    static SHA_LONG64 __fastcall __pull64be(const void *x)
+    {	_asm	mov	edx, [ecx + 0]
+	_asm	mov	eax, [ecx + 4]
+	_asm	xchg	dh,dl
+	_asm	xchg	ah,al
+	_asm	rol	edx,16
+	_asm	rol	eax,16
+	_asm	xchg	dh,dl
+	_asm	xchg	ah,al
+    }
+#   else
+    static SHA_LONG64 __fastcall __pull64be(const void *x)
+    {	_asm	mov	edx, [ecx + 0]
+	_asm	mov	eax, [ecx + 4]
+	_asm	bswap	edx
+	_asm	bswap	eax
+    }
+#   endif
+#   define PULL64(x) __pull64be(&(x))
+#   if _MSC_VER<=1200
+#    pragma inline_depth(0)
+#   endif
+#  endif
 # endif
 #endif
 
@@ -314,27 +376,6 @@ static const SHA_LONG64 K512[80] = {
 #define PULL64(x) (B(x,0)|B(x,1)|B(x,2)|B(x,3)|B(x,4)|B(x,5)|B(x,6)|B(x,7))
 #endif
 
-#ifndef PEDANTIC
-# if defined(_MSC_VER)
-#  if defined(_WIN64)	/* applies to both IA-64 and AMD64 */
-#   define ROTR(a,n)	_rotr64((a),n)
-#  endif
-# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
-#  if defined(__x86_64) || defined(__x86_64__)
-#   define ROTR(a,n)	({ unsigned long ret;		\
-				asm ("rorq %1,%0"	\
-				: "=r"(ret)		\
-				: "J"(n),"0"(a)		\
-				: "cc"); ret;		})
-#  elif defined(_ARCH_PPC) && defined(__64BIT__)
-#   define ROTR(a,n)	({ unsigned long ret;		\
-				asm ("rotrdi %0,%1,%2"	\
-				: "=r"(ret)		\
-				: "r"(a),"K"(n)); ret;	})
-#  endif
-# endif
-#endif
-
 #ifndef ROTR
 #define ROTR(x,s)	(((x)>>s) | (x)<<(64-s))
 #endif
@@ -357,7 +398,7 @@ static const SHA_LONG64 K512[80] = {
 
 #ifdef OPENSSL_SMALL_FOOTPRINT
 
-static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num)
+static void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num)
 	{
 	const SHA_LONG64 *W=in;
 	SHA_LONG64	a,b,c,d,e,f,g,h,s0,s1,T1,T2;
@@ -418,7 +459,7 @@ static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num)
 	T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f];	\
 	ROUND_00_15(i,a,b,c,d,e,f,g,h);		} while (0)
 
-static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num)
+static void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num)
 	{
 	const SHA_LONG64 *W=in;
 	SHA_LONG64	a,b,c,d,e,f,g,h,s0,s1,T1;

crypto/sha/sha_locl.h

@@ -62,17 +62,11 @@
 #include <openssl/opensslconf.h>
 #include <openssl/sha.h>
 
-#ifndef SHA_LONG_LOG2
-#define SHA_LONG_LOG2	2	/* default to 32 bits */
-#endif
-
 #define DATA_ORDER_IS_BIG_ENDIAN
 
 #define HASH_LONG               SHA_LONG
-#define HASH_LONG_LOG2          SHA_LONG_LOG2
 #define HASH_CTX                SHA_CTX
 #define HASH_CBLOCK             SHA_CBLOCK
-#define HASH_LBLOCK             SHA_LBLOCK
 #define HASH_MAKE_STRING(c,s)   do {	\
 	unsigned long ll;		\
 	ll=(c)->h0; HOST_l2c(ll,(s));	\
@@ -88,12 +82,10 @@
 # define HASH_TRANSFORM          	SHA_Transform
 # define HASH_FINAL              	SHA_Final
 # define HASH_INIT			SHA_Init
-# define HASH_BLOCK_HOST_ORDER   	sha_block_host_order
 # define HASH_BLOCK_DATA_ORDER   	sha_block_data_order
 # define Xupdate(a,ix,ia,ib,ic,id)	(ix=(a)=(ia^ib^ic^id))
 
-  void sha_block_host_order (SHA_CTX *c, const void *p,size_t num);
+static void sha_block_data_order (SHA_CTX *c, const void *p,size_t num);
-  void sha_block_data_order (SHA_CTX *c, const void *p,size_t num);
 
 #elif defined(SHA_1)
 
@@ -101,7 +93,6 @@
 # define HASH_TRANSFORM          	SHA1_Transform
 # define HASH_FINAL              	SHA1_Final
 # define HASH_INIT			SHA1_Init
-# define HASH_BLOCK_HOST_ORDER   	sha1_block_host_order
 # define HASH_BLOCK_DATA_ORDER   	sha1_block_data_order
 # if defined(__MWERKS__) && defined(__MC68K__)
    /* Metrowerks for Motorola fails otherwise:-( <appro@fy.chalmers.se> */
@@ -114,24 +105,10 @@
 					)
 # endif
 
-# ifdef SHA1_ASM
+#ifndef SHA1_ASM
-#  if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
+static
-#   if !defined(B_ENDIAN)
+#endif
-#    define sha1_block_host_order		sha1_block_asm_host_order
+void sha1_block_data_order (SHA_CTX *c, const void *p,size_t num);
-#    define DONT_IMPLEMENT_BLOCK_HOST_ORDER
-#    define sha1_block_data_order		sha1_block_asm_data_order
-#    define DONT_IMPLEMENT_BLOCK_DATA_ORDER
-#    define HASH_BLOCK_DATA_ORDER_ALIGNED	sha1_block_asm_data_order
-#   endif
-#  elif defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
-#   define sha1_block_host_order		sha1_block_asm_host_order
-#   define DONT_IMPLEMENT_BLOCK_HOST_ORDER
-#   define sha1_block_data_order		sha1_block_asm_data_order
-#   define DONT_IMPLEMENT_BLOCK_DATA_ORDER
-#  endif
-# endif
-  void sha1_block_host_order (SHA_CTX *c, const void *p,size_t num);
-  void sha1_block_data_order (SHA_CTX *c, const void *p,size_t num);
 
 #else
 # error "Either SHA_0 or SHA_1 must be defined."
@@ -229,133 +206,8 @@ int HASH_INIT (SHA_CTX *c)
 # define X(i)	XX[i]
 #endif
 
-#ifndef DONT_IMPLEMENT_BLOCK_HOST_ORDER
+#if !defined(SHA_1) || !defined(SHA1_ASM)
-void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, size_t num)
+static void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
-	{
-	const SHA_LONG *W=d;
-	register unsigned MD32_REG_T A,B,C,D,E,T;
-#ifndef MD32_XARRAY
-	unsigned MD32_REG_T	XX0, XX1, XX2, XX3, XX4, XX5, XX6, XX7,
-				XX8, XX9,XX10,XX11,XX12,XX13,XX14,XX15;
-#else
-	SHA_LONG	XX[16];
-#endif
-
-	A=c->h0;
-	B=c->h1;
-	C=c->h2;
-	D=c->h3;
-	E=c->h4;
-
-	for (;;)
-		{
-	BODY_00_15( 0,A,B,C,D,E,T,W[ 0]);
-	BODY_00_15( 1,T,A,B,C,D,E,W[ 1]);
-	BODY_00_15( 2,E,T,A,B,C,D,W[ 2]);
-	BODY_00_15( 3,D,E,T,A,B,C,W[ 3]);
-	BODY_00_15( 4,C,D,E,T,A,B,W[ 4]);
-	BODY_00_15( 5,B,C,D,E,T,A,W[ 5]);
-	BODY_00_15( 6,A,B,C,D,E,T,W[ 6]);
-	BODY_00_15( 7,T,A,B,C,D,E,W[ 7]);
-	BODY_00_15( 8,E,T,A,B,C,D,W[ 8]);
-	BODY_00_15( 9,D,E,T,A,B,C,W[ 9]);
-	BODY_00_15(10,C,D,E,T,A,B,W[10]);
-	BODY_00_15(11,B,C,D,E,T,A,W[11]);
-	BODY_00_15(12,A,B,C,D,E,T,W[12]);
-	BODY_00_15(13,T,A,B,C,D,E,W[13]);
-	BODY_00_15(14,E,T,A,B,C,D,W[14]);
-	BODY_00_15(15,D,E,T,A,B,C,W[15]);
-
-	BODY_16_19(16,C,D,E,T,A,B,X( 0),W[ 0],W[ 2],W[ 8],W[13]);
-	BODY_16_19(17,B,C,D,E,T,A,X( 1),W[ 1],W[ 3],W[ 9],W[14]);
-	BODY_16_19(18,A,B,C,D,E,T,X( 2),W[ 2],W[ 4],W[10],W[15]);
-	BODY_16_19(19,T,A,B,C,D,E,X( 3),W[ 3],W[ 5],W[11],X( 0));
-
-	BODY_20_31(20,E,T,A,B,C,D,X( 4),W[ 4],W[ 6],W[12],X( 1));
-	BODY_20_31(21,D,E,T,A,B,C,X( 5),W[ 5],W[ 7],W[13],X( 2));
-	BODY_20_31(22,C,D,E,T,A,B,X( 6),W[ 6],W[ 8],W[14],X( 3));
-	BODY_20_31(23,B,C,D,E,T,A,X( 7),W[ 7],W[ 9],W[15],X( 4));
-	BODY_20_31(24,A,B,C,D,E,T,X( 8),W[ 8],W[10],X( 0),X( 5));
-	BODY_20_31(25,T,A,B,C,D,E,X( 9),W[ 9],W[11],X( 1),X( 6));
-	BODY_20_31(26,E,T,A,B,C,D,X(10),W[10],W[12],X( 2),X( 7));
-	BODY_20_31(27,D,E,T,A,B,C,X(11),W[11],W[13],X( 3),X( 8));
-	BODY_20_31(28,C,D,E,T,A,B,X(12),W[12],W[14],X( 4),X( 9));
-	BODY_20_31(29,B,C,D,E,T,A,X(13),W[13],W[15],X( 5),X(10));
-	BODY_20_31(30,A,B,C,D,E,T,X(14),W[14],X( 0),X( 6),X(11));
-	BODY_20_31(31,T,A,B,C,D,E,X(15),W[15],X( 1),X( 7),X(12));
-
-	BODY_32_39(32,E,T,A,B,C,D,X( 0),X( 2),X( 8),X(13));
-	BODY_32_39(33,D,E,T,A,B,C,X( 1),X( 3),X( 9),X(14));
-	BODY_32_39(34,C,D,E,T,A,B,X( 2),X( 4),X(10),X(15));
-	BODY_32_39(35,B,C,D,E,T,A,X( 3),X( 5),X(11),X( 0));
-	BODY_32_39(36,A,B,C,D,E,T,X( 4),X( 6),X(12),X( 1));
-	BODY_32_39(37,T,A,B,C,D,E,X( 5),X( 7),X(13),X( 2));
-	BODY_32_39(38,E,T,A,B,C,D,X( 6),X( 8),X(14),X( 3));
-	BODY_32_39(39,D,E,T,A,B,C,X( 7),X( 9),X(15),X( 4));
-
-	BODY_40_59(40,C,D,E,T,A,B,X( 8),X(10),X( 0),X( 5));
-	BODY_40_59(41,B,C,D,E,T,A,X( 9),X(11),X( 1),X( 6));
-	BODY_40_59(42,A,B,C,D,E,T,X(10),X(12),X( 2),X( 7));
-	BODY_40_59(43,T,A,B,C,D,E,X(11),X(13),X( 3),X( 8));
-	BODY_40_59(44,E,T,A,B,C,D,X(12),X(14),X( 4),X( 9));
-	BODY_40_59(45,D,E,T,A,B,C,X(13),X(15),X( 5),X(10));
-	BODY_40_59(46,C,D,E,T,A,B,X(14),X( 0),X( 6),X(11));
-	BODY_40_59(47,B,C,D,E,T,A,X(15),X( 1),X( 7),X(12));
-	BODY_40_59(48,A,B,C,D,E,T,X( 0),X( 2),X( 8),X(13));
-	BODY_40_59(49,T,A,B,C,D,E,X( 1),X( 3),X( 9),X(14));
-	BODY_40_59(50,E,T,A,B,C,D,X( 2),X( 4),X(10),X(15));
-	BODY_40_59(51,D,E,T,A,B,C,X( 3),X( 5),X(11),X( 0));
-	BODY_40_59(52,C,D,E,T,A,B,X( 4),X( 6),X(12),X( 1));
-	BODY_40_59(53,B,C,D,E,T,A,X( 5),X( 7),X(13),X( 2));
-	BODY_40_59(54,A,B,C,D,E,T,X( 6),X( 8),X(14),X( 3));
-	BODY_40_59(55,T,A,B,C,D,E,X( 7),X( 9),X(15),X( 4));
-	BODY_40_59(56,E,T,A,B,C,D,X( 8),X(10),X( 0),X( 5));
-	BODY_40_59(57,D,E,T,A,B,C,X( 9),X(11),X( 1),X( 6));
-	BODY_40_59(58,C,D,E,T,A,B,X(10),X(12),X( 2),X( 7));
-	BODY_40_59(59,B,C,D,E,T,A,X(11),X(13),X( 3),X( 8));
-
-	BODY_60_79(60,A,B,C,D,E,T,X(12),X(14),X( 4),X( 9));
-	BODY_60_79(61,T,A,B,C,D,E,X(13),X(15),X( 5),X(10));
-	BODY_60_79(62,E,T,A,B,C,D,X(14),X( 0),X( 6),X(11));
-	BODY_60_79(63,D,E,T,A,B,C,X(15),X( 1),X( 7),X(12));
-	BODY_60_79(64,C,D,E,T,A,B,X( 0),X( 2),X( 8),X(13));
-	BODY_60_79(65,B,C,D,E,T,A,X( 1),X( 3),X( 9),X(14));
-	BODY_60_79(66,A,B,C,D,E,T,X( 2),X( 4),X(10),X(15));
-	BODY_60_79(67,T,A,B,C,D,E,X( 3),X( 5),X(11),X( 0));
-	BODY_60_79(68,E,T,A,B,C,D,X( 4),X( 6),X(12),X( 1));
-	BODY_60_79(69,D,E,T,A,B,C,X( 5),X( 7),X(13),X( 2));
-	BODY_60_79(70,C,D,E,T,A,B,X( 6),X( 8),X(14),X( 3));
-	BODY_60_79(71,B,C,D,E,T,A,X( 7),X( 9),X(15),X( 4));
-	BODY_60_79(72,A,B,C,D,E,T,X( 8),X(10),X( 0),X( 5));
-	BODY_60_79(73,T,A,B,C,D,E,X( 9),X(11),X( 1),X( 6));
-	BODY_60_79(74,E,T,A,B,C,D,X(10),X(12),X( 2),X( 7));
-	BODY_60_79(75,D,E,T,A,B,C,X(11),X(13),X( 3),X( 8));
-	BODY_60_79(76,C,D,E,T,A,B,X(12),X(14),X( 4),X( 9));
-	BODY_60_79(77,B,C,D,E,T,A,X(13),X(15),X( 5),X(10));
-	BODY_60_79(78,A,B,C,D,E,T,X(14),X( 0),X( 6),X(11));
-	BODY_60_79(79,T,A,B,C,D,E,X(15),X( 1),X( 7),X(12));
-	
-	c->h0=(c->h0+E)&0xffffffffL; 
-	c->h1=(c->h1+T)&0xffffffffL;
-	c->h2=(c->h2+A)&0xffffffffL;
-	c->h3=(c->h3+B)&0xffffffffL;
-	c->h4=(c->h4+C)&0xffffffffL;
-
-	if (--num == 0) break;
-
-	A=c->h0;
-	B=c->h1;
-	C=c->h2;
-	D=c->h3;
-	E=c->h4;
-
-	W+=SHA_LBLOCK;
-		}
-	}
-#endif
-
-#ifndef DONT_IMPLEMENT_BLOCK_DATA_ORDER
-void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
 	{
 	const unsigned char *data=p;
 	register unsigned MD32_REG_T A,B,C,D,E,T,l;
@@ -374,7 +226,34 @@ void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
 
 	for (;;)
 			{
+	const union { long one; char little; } is_endian = {1};
 
+	if (!is_endian.little && sizeof(SHA_LONG)==4 && ((size_t)p%4)==0)
+		{
+		const SHA_LONG *W=(const SHA_LONG *)data;
+
+		X( 0) = W[0];				X( 1) = W[ 1];
+		BODY_00_15( 0,A,B,C,D,E,T,X( 0));	X( 2) = W[ 2];
+		BODY_00_15( 1,T,A,B,C,D,E,X( 1));	X( 3) = W[ 3];
+		BODY_00_15( 2,E,T,A,B,C,D,X( 2));	X( 4) = W[ 4];
+		BODY_00_15( 3,D,E,T,A,B,C,X( 3));	X( 5) = W[ 5];
+		BODY_00_15( 4,C,D,E,T,A,B,X( 4));	X( 6) = W[ 6];
+		BODY_00_15( 5,B,C,D,E,T,A,X( 5));	X( 7) = W[ 7];
+		BODY_00_15( 6,A,B,C,D,E,T,X( 6));	X( 8) = W[ 8];
+		BODY_00_15( 7,T,A,B,C,D,E,X( 7));	X( 9) = W[ 9];
+		BODY_00_15( 8,E,T,A,B,C,D,X( 8));	X(10) = W[10];
+		BODY_00_15( 9,D,E,T,A,B,C,X( 9));	X(11) = W[11];
+		BODY_00_15(10,C,D,E,T,A,B,X(10));	X(12) = W[12];
+		BODY_00_15(11,B,C,D,E,T,A,X(11));	X(13) = W[13];
+		BODY_00_15(12,A,B,C,D,E,T,X(12));	X(14) = W[14];
+		BODY_00_15(13,T,A,B,C,D,E,X(13));	X(15) = W[15];
+		BODY_00_15(14,E,T,A,B,C,D,X(14));
+		BODY_00_15(15,D,E,T,A,B,C,X(15));
+
+		data += SHA_CBLOCK;
+		}
+	else
+		{
 		HOST_c2l(data,l); X( 0)=l;		HOST_c2l(data,l); X( 1)=l;
 		BODY_00_15( 0,A,B,C,D,E,T,X( 0));	HOST_c2l(data,l); X( 2)=l;
 		BODY_00_15( 1,T,A,B,C,D,E,X( 1));	HOST_c2l(data,l); X( 3)=l;
@@ -392,6 +271,7 @@ void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
 		BODY_00_15(13,T,A,B,C,D,E,X(13));	HOST_c2l(data,l); X(15)=l;
 		BODY_00_15(14,E,T,A,B,C,D,X(14));
 		BODY_00_15(15,D,E,T,A,B,C,X(15));
+		}
 
 	BODY_16_19(16,C,D,E,T,A,B,X( 0),X( 0),X( 2),X( 8),X(13));
 	BODY_16_19(17,B,C,D,E,T,A,X( 1),X( 1),X( 3),X( 9),X(14));
@@ -511,54 +391,8 @@ void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
 	E=D, D=C, C=ROTATE(B,30), B=A;	\
 	A=ROTATE(A,5)+T+xa;	    } while(0)
 
-#ifndef DONT_IMPLEMENT_BLOCK_HOST_ORDER
+#if !defined(SHA_1) || !defined(SHA1_ASM)
-void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, size_t num)
+static void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
-	{
-	const SHA_LONG *W=d;
-	register unsigned MD32_REG_T A,B,C,D,E,T;
-	int i;
-	SHA_LONG	X[16];
-
-	A=c->h0;
-	B=c->h1;
-	C=c->h2;
-	D=c->h3;
-	E=c->h4;
-
-	for (;;)
-		{
-	for (i=0;i<16;i++)
-	{ X[i]=W[i]; BODY_00_15(X[i]); }
-	for (i=0;i<4;i++)
-	{ BODY_16_19(X[i],       X[i+2],      X[i+8],     X[(i+13)&15]); }
-	for (;i<24;i++)
-	{ BODY_20_39(X[i&15],    X[(i+2)&15], X[(i+8)&15],X[(i+13)&15]); }
-	for (i=0;i<20;i++)
-	{ BODY_40_59(X[(i+8)&15],X[(i+10)&15],X[i&15],    X[(i+5)&15]);  }
-	for (i=4;i<24;i++)
-	{ BODY_60_79(X[(i+8)&15],X[(i+10)&15],X[i&15],    X[(i+5)&15]);  }
-	
-	c->h0=(c->h0+A)&0xffffffffL; 
-	c->h1=(c->h1+B)&0xffffffffL;
-	c->h2=(c->h2+C)&0xffffffffL;
-	c->h3=(c->h3+D)&0xffffffffL;
-	c->h4=(c->h4+E)&0xffffffffL;
-
-	if (--num == 0) break;
-
-	A=c->h0;
-	B=c->h1;
-	C=c->h2;
-	D=c->h3;
-	E=c->h4;
-
-	W+=SHA_LBLOCK;
-		}
-	}
-#endif
-
-#ifndef DONT_IMPLEMENT_BLOCK_DATA_ORDER
-void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num)
 	{
 	const unsigned char *data=p;
 	register unsigned MD32_REG_T A,B,C,D,E,T,l;