The rest of the merge in of the main trunk, a few conflicts resolved.OpenSSL-engine-0_9_6-beta2

32 files changed, 1866 insertions, 498 deletions

diff --git a/Configure b/Configure
index afe66c440e..0c92260a06 100755
--- a/Configure
+++ b/Configure
@@ -221,12 +221,21 @@ my %table=(
 #   crypto/sha/sha_lcl.h.
 #					<appro@fy.chalmers.se>
 #
-"hpux-parisc-cc","cc:-Ae +O3 +ESlit -z -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY:::-ldld:BN_LLONG DES_PTR DES_UNROLL DES_RISC1::::::::::dl",
+#!#"hpux-parisc-cc","cc:-Ae +O3 +ESlit -z -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY:::-ldld:BN_LLONG DES_PTR DES_UNROLL DES_RISC1::::::::::dl",
 # Since there is mention of this in shlib/hpux10-cc.sh
 "hpux-parisc-cc-o4","cc:-Ae +O4 +ESlit -z -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY:::-ldld:BN_LLONG DES_PTR DES_UNROLL DES_RISC1::::::::::dl",
 "hpux-parisc-gcc","gcc:-O3 -DB_ENDIAN -DBN_DIV2W:::-ldld:BN_LLONG DES_PTR DES_UNROLL DES_RISC1::::::::::dl",
 "hpux64-parisc-cc","cc:-Ae +DD64 +O3 +ESlit -z -DB_ENDIAN -DMD32_XARRAY::-D_REENTRANT:-ldld:SIXTY_FOUR_BIT_LONG MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT::::::::::dl",
 
+# More attempts at unified 10.X and 11.X targets for HP C compiler.
+#
+# Chris Ruemmler <ruemmler@cup.hp.com>
+# Kevin Steves <ks@hp.se>
+"hpux-parisc-cc","cc:+O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY::-D_REENTRANT:-ldl:MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT::::::::::dl",
+"hpux-parisc2-cc","cc:+DA2.0 +DS2.0 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY::-D_REENTRANT:-ldl:SIXTY_FOUR_BIT MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT:asm/pa-risc2.o:::::::::dl",
+"hpux64-parisc2-cc","cc:+DD64 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY::-D_REENTRANT:-ldl:SIXTY_FOUR_BIT_LONG MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT:asm/pa-risc2W.o:::::::::dl",
+"hpux-parisc1_1-cc","cc:+DA1.1 +DS1.1 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY::-D_REENTRANT:-ldl:MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT::::::::::dl",
+
 # HPUX 9.X config.
 # Don't use the bundled cc.  It is broken.  Use HP ANSI C if possible, or
 # egcs.  gcc 2.8.1 is also broken.
diff --git a/README b/README
index 739dd0d62f..46252db971 100644
--- a/README
+++ b/README
@@ -1,5 +1,5 @@
 
- OpenSSL 0.9.6-beta1 [engine] 11 Sep 2000
+ OpenSSL 0.9.6-beta2 [engine] 17 Sep 2000
 
  Copyright (c) 1998-2000 The OpenSSL Project
  Copyright (c) 1995-1998 Eric A. Young, Tim J. Hudson
diff --git a/TABLE b/TABLE
index eb561a69b0..d2d06f3b12 100644
--- a/TABLE
+++ b/TABLE
@@ -1162,11 +1162,11 @@ $shared_cflag =
 
 *** hpux-parisc-cc
 $cc           = cc
-$cflags       = -Ae +O3 +ESlit -z -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY
+$cflags       = +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DBN_DIV2W -DMD32_XARRAY
 $unistd       = 
-$thread_cflag = 
-$lflags       = -ldld
-$bn_ops       = BN_LLONG DES_PTR DES_UNROLL DES_RISC1
+$thread_cflag = -D_REENTRANT
+$lflags       = -ldl
+$bn_ops       = MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT
 $bn_obj       = 
 $des_obj      = 
 $bf_obj       = 
@@ -1220,6 +1220,46 @@ $dso_scheme   = dl
 $shared_target= 
 $shared_cflag = 
 
+*** hpux-parisc1_1-cc
+$cc           = cc
+$cflags       = +DA1.1 +DS1.1 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY
+$unistd       = 
+$thread_cflag = -D_REENTRANT
+$lflags       = -ldl
+$bn_ops       = MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT
+$bn_obj       = 
+$des_obj      = 
+$bf_obj       = 
+$md5_obj      = 
+$sha1_obj     = 
+$cast_obj     = 
+$rc4_obj      = 
+$rmd160_obj   = 
+$rc5_obj      = 
+$dso_scheme   = dl
+$shared_target= 
+$shared_cflag = 
+
+*** hpux-parisc2-cc
+$cc           = cc
+$cflags       = +DA2.0 +DS2.0 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY
+$unistd       = 
+$thread_cflag = -D_REENTRANT
+$lflags       = -ldl
+$bn_ops       = SIXTY_FOUR_BIT MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT
+$bn_obj       = asm/pa-risc2.o
+$des_obj      = 
+$bf_obj       = 
+$md5_obj      = 
+$sha1_obj     = 
+$cast_obj     = 
+$rc4_obj      = 
+$rmd160_obj   = 
+$rc5_obj      = 
+$dso_scheme   = dl
+$shared_target= 
+$shared_cflag = 
+
 *** hpux10-brokencc
 $cc           = cc
 $cflags       = -DB_ENDIAN -DBN_DIV2W -Ae +ESlit +O2 -z
@@ -1320,6 +1360,26 @@ $dso_scheme   = dl
 $shared_target= 
 $shared_cflag = 
 
+*** hpux64-parisc2-cc
+$cc           = cc
+$cflags       = +DD64 +O3 +Optrs_strongly_typed +Olibcalls -Ae +ESlit -DB_ENDIAN -DMD32_XARRAY
+$unistd       = 
+$thread_cflag = -D_REENTRANT
+$lflags       = -ldl
+$bn_ops       = SIXTY_FOUR_BIT_LONG MD2_CHAR RC4_INDEX RC4_CHAR DES_UNROLL DES_RISC1 DES_INT
+$bn_obj       = asm/pa-risc2W.o
+$des_obj      = 
+$bf_obj       = 
+$md5_obj      = 
+$sha1_obj     = 
+$cast_obj     = 
+$rc4_obj      = 
+$rmd160_obj   = 
+$rc5_obj      = 
+$dso_scheme   = dl
+$shared_target= 
+$shared_cflag = 
+
 *** irix-cc
 $cc           = cc
 $cflags       = -O2 -use_readonly_const -DTERMIOS -DB_ENDIAN
diff --git a/apps/apps.c b/apps/apps.c
index 50ea09baca..167c319ebe 100644
--- a/apps/apps.c
+++ b/apps/apps.c
@@ -662,7 +662,7 @@ STACK_OF(X509) *load_certs(BIO *err, char *file, int format)
 
 	if      (format == FORMAT_PEM)
 		{
-		othercerts = sk_X509_new(NULL);
+		othercerts = sk_X509_new_null();
 		if(!othercerts)
 			{
 			sk_X509_free(othercerts);
diff --git a/apps/crl2p7.c b/apps/crl2p7.c
index ebf5fe90ec..d02862710d 100644
--- a/apps/crl2p7.c
+++ b/apps/crl2p7.c
@@ -141,7 +141,7 @@ int MAIN(int argc, char **argv)
 		else if (strcmp(*argv,"-certfile") == 0)
 			{
 			if (--argc < 1) goto bad;
-			if(!certflst) certflst = sk_new(NULL);
+			if(!certflst) certflst = sk_new_null();
 			sk_push(certflst,*(++argv));
 			}
 		else
@@ -215,7 +215,7 @@ bad:
 	p7s->contents->type=OBJ_nid2obj(NID_pkcs7_data);
 
 	if (!ASN1_INTEGER_set(p7s->version,1)) goto end;
-	if ((crl_stack=sk_X509_CRL_new(NULL)) == NULL) goto end;
+	if ((crl_stack=sk_X509_CRL_new_null()) == NULL) goto end;
 	p7s->crl=crl_stack;
 	if (crl != NULL)
 		{
@@ -223,7 +223,7 @@ bad:
 		crl=NULL; /* now part of p7 for OPENSSL_freeing */
 		}
 
-	if ((cert_stack=sk_X509_new(NULL)) == NULL) goto end;
+	if ((cert_stack=sk_X509_new_null()) == NULL) goto end;
 	p7s->cert=cert_stack;
 
 	if(certflst) for(i = 0; i < sk_num(certflst); i++) {
diff --git a/apps/dgst.c b/apps/dgst.c
index f0e5f954da..dbfa387f33 100644
--- a/apps/dgst.c
+++ b/apps/dgst.c
@@ -307,7 +307,7 @@ int MAIN(int argc, char **argv)
 		}
 		siglen = BIO_read(sigbio, sigbuf, siglen);
 		BIO_free(sigbio);
-		if(siglen <= 0) {
+		if(siglen == 0) {
 			BIO_printf(bio_err, "Error reading signature file %s\n",
 								sigfile);
 			ERR_print_errors(bio_err);
diff --git a/apps/nseq.c b/apps/nseq.c
index cc88d50ceb..7210fbdb5e 100644
--- a/apps/nseq.c
+++ b/apps/nseq.c
@@ -123,7 +123,7 @@ int MAIN(int argc, char **argv)
 
 	if (toseq) {
 		seq = NETSCAPE_CERT_SEQUENCE_new();
-		seq->certs = sk_X509_new(NULL);
+		seq->certs = sk_X509_new_null();
 		while((x509 = PEM_read_bio_X509(in, NULL, NULL, NULL))) 
 		    sk_X509_push(seq->certs,x509);
 
diff --git a/apps/pkcs12.c b/apps/pkcs12.c
index b57cf5f829..2cc9a0caee 100644
--- a/apps/pkcs12.c
+++ b/apps/pkcs12.c
@@ -200,7 +200,7 @@ int MAIN(int argc, char **argv)
 		} else if (!strcmp (*args, "-caname")) {
 		    if (args[1]) {
 			args++;	
-			if (!canames) canames = sk_new(NULL);
+			if (!canames) canames = sk_new_null();
 			sk_push(canames, *args);
 		    } else badarg = 1;
 		} else if (!strcmp (*args, "-in")) {
@@ -427,7 +427,7 @@ int MAIN(int argc, char **argv)
 	CRYPTO_push_info("reading certs from input");
 #endif
 
-	certs = sk_X509_new(NULL);
+	certs = sk_X509_new_null();
 
 	/* Load in all certs in input file */
 	if(!cert_load(in, certs)) {
@@ -459,7 +459,7 @@ int MAIN(int argc, char **argv)
 	CRYPTO_push_info("reading certs from certfile");
 #endif
 
-	bags = sk_PKCS12_SAFEBAG_new (NULL);
+	bags = sk_PKCS12_SAFEBAG_new_null ();
 
 	/* Add any more certificates asked for */
 	if (certsin) {
@@ -550,7 +550,7 @@ int MAIN(int argc, char **argv)
 		goto export_end;
 	}
 
-	safes = sk_PKCS7_new (NULL);
+	safes = sk_PKCS7_new_null ();
 	sk_PKCS7_push (safes, authsafe);
 
 #ifdef CRYPTO_MDEBUG
@@ -566,7 +566,7 @@ int MAIN(int argc, char **argv)
 	p8 = NULL;
         if (name) PKCS12_add_friendlyname (bag, name, -1);
 	PKCS12_add_localkeyid (bag, keyid, keyidlen);
-	bags = sk_PKCS12_SAFEBAG_new(NULL);
+	bags = sk_PKCS12_SAFEBAG_new_null();
 	sk_PKCS12_SAFEBAG_push (bags, bag);
 
 #ifdef CRYPTO_MDEBUG
diff --git a/crypto/asn1/a_set.c b/crypto/asn1/a_set.c
index 1921f5eaa1..caf5a1419c 100644
--- a/crypto/asn1/a_set.c
+++ b/crypto/asn1/a_set.c
@@ -158,7 +158,7 @@ STACK *d2i_ASN1_SET(STACK **a, unsigned char **pp, long length,
 	STACK *ret=NULL;
 
 	if ((a == NULL) || ((*a) == NULL))
-		{ if ((ret=sk_new(NULL)) == NULL) goto err; }
+		{ if ((ret=sk_new_null()) == NULL) goto err; }
 	else
 		ret=(*a);
 
diff --git a/crypto/bn/asm/README b/crypto/bn/asm/README
index 86bf64cfc2..a0fe58a677 100644
--- a/crypto/bn/asm/README
+++ b/crypto/bn/asm/README
@@ -15,9 +15,9 @@ On the 2 alpha C compilers I had access to, it was not possible to do
 were 64 bits).  So the hand assember gives access to the 128 bit result and
 a 2 times speedup :-).
 
-There are 2 versions of assember for the HP PA-RISC.
-pa-risc.s is the origional one which works fine.
-pa-risc2.s is a new version that often generates warnings but if the
-tests pass, it gives performance that is over 2 times faster than
-pa-risc.s.
-Both were generated using gcc :-)
+There are 3 versions of assember for the HP PA-RISC.
+
+pa-risc.s is the origional one which works fine and generated using gcc :-)
+
+pa-risc2W.s and pa-risc2.s are 64 and 32-bit PA-RISC 2.0 implementations
+by Chris Ruemmler from HP (with some help from the HP C compiler).
diff --git a/crypto/bn/asm/pa-risc2.s b/crypto/bn/asm/pa-risc2.s
index c2725996a4..7239aa2c76 100644
--- a/crypto/bn/asm/pa-risc2.s
+++ b/crypto/bn/asm/pa-risc2.s
@@ -1,416 +1,1618 @@
-	.SPACE $PRIVATE$
-	.SUBSPA $DATA$,QUAD=1,ALIGN=8,ACCESS=31
-	.SUBSPA $BSS$,QUAD=1,ALIGN=8,ACCESS=31,ZERO,SORT=82
-	.SPACE $TEXT$
-	.SUBSPA $LIT$,QUAD=0,ALIGN=8,ACCESS=44
-	.SUBSPA $CODE$,QUAD=0,ALIGN=8,ACCESS=44,CODE_ONLY
-	.IMPORT $global$,DATA
-	.IMPORT $$dyncall,MILLICODE
-; gcc_compiled.:
-	.SPACE $TEXT$
-	.SUBSPA $CODE$
-
-	.align 4
-	.EXPORT bn_mul_add_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR
+;
+; PA-RISC 2.0 implementation of bn_asm code, based on the
+; 64-bit version of the code.  This code is effectively the
+; same as the 64-bit version except the register model is
+; slightly different given all values must be 32-bit between
+; function calls.  Thus the 64-bit return values are returned
+; in %ret0 and %ret1 vs just %ret0 as is done in 64-bit
+;
+;
+; This code is approximately 2x faster than the C version
+; for RSA/DSA.
+;
+; See http://devresource.hp.com/  for more details on the PA-RISC
+; architecture.  Also see the book "PA-RISC 2.0 Architecture"
+; by Gerry Kane for information on the instruction set architecture.
+;
+; Code written by Chris Ruemmler (with some help from the HP C
+; compiler).
+;
+; The code compiles with HP's assembler
+;
+
+	.level	2.0N
+	.space	$TEXT$
+	.subspa	$CODE$,QUAD=0,ALIGN=8,ACCESS=0x2c,CODE_ONLY
+
+;
+; Global Register definitions used for the routines.
+;
+; Some information about HP's runtime architecture for 32-bits.
+;
+; "Caller save" means the calling function must save the register
+; if it wants the register to be preserved.
+; "Callee save" means if a function uses the register, it must save
+; the value before using it.
+;
+; For the floating point registers 
+;
+;    "caller save" registers: fr4-fr11, fr22-fr31
+;    "callee save" registers: fr12-fr21
+;    "special" registers: fr0-fr3 (status and exception registers)
+;
+; For the integer registers
+;     value zero             :  r0
+;     "caller save" registers: r1,r19-r26
+;     "callee save" registers: r3-r18
+;     return register        :  r2  (rp)
+;     return values          ; r28,r29  (ret0,ret1)
+;     Stack pointer          ; r30  (sp) 
+;     millicode return ptr   ; r31  (also a caller save register)
+
+
+;
+; Arguments to the routines
+;
+r_ptr       .reg %r26
+a_ptr       .reg %r25
+b_ptr       .reg %r24
+num         .reg %r24
+n           .reg %r23
+
+;
+; Note that the "w" argument for bn_mul_add_words and bn_mul_words
+; is passed on the stack at a delta of -56 from the top of stack
+; as the routine is entered.
+;
+
+;
+; Globals used in some routines
+;
+
+top_overflow .reg %r23
+high_mask    .reg %r22    ; value 0xffffffff80000000L
+
+
+;------------------------------------------------------------------------------
+;
+; bn_mul_add_words
+;
+;BN_ULONG bn_mul_add_words(BN_ULONG *r_ptr, BN_ULONG *a_ptr, 
+;								int num, BN_ULONG w)
+;
+; arg0 = r_ptr
+; arg1 = a_ptr
+; arg3 = num
+; -56(sp) =  w
+;
+; Local register definitions
+;
+
+fm1          .reg %fr22
+fm           .reg %fr23
+ht_temp      .reg %fr24
+ht_temp_1    .reg %fr25
+lt_temp      .reg %fr26
+lt_temp_1    .reg %fr27
+fm1_1        .reg %fr28
+fm_1         .reg %fr29
+
+fw_h         .reg %fr7L
+fw_l         .reg %fr7R
+fw           .reg %fr7
+
+fht_0        .reg %fr8L
+flt_0        .reg %fr8R
+t_float_0    .reg %fr8
+
+fht_1        .reg %fr9L
+flt_1        .reg %fr9R
+t_float_1    .reg %fr9
+
+tmp_0        .reg %r31
+tmp_1        .reg %r21
+m_0          .reg %r20 
+m_1          .reg %r19 
+ht_0         .reg %r1  
+ht_1         .reg %r3
+lt_0         .reg %r4
+lt_1         .reg %r5
+m1_0         .reg %r6 
+m1_1         .reg %r7 
+rp_val       .reg %r8
+rp_val_1     .reg %r9
+
 bn_mul_add_words
-	.PROC
-	.CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=4
-	.ENTRY
-	stw %r2,-20(0,%r30)
-	stwm %r4,64(0,%r30)
-	copy %r24,%r31
-	stw %r3,-60(0,%r30)
-	ldi 0,%r20
-	ldo 12(%r26),%r2
-	stw %r23,-16(0,%r30)
-	copy %r25,%r3
-	ldo 12(%r3),%r1
-	fldws -16(0,%r30),%fr8L
-L$0010
-	copy %r20,%r25
-	ldi 0,%r24
-	fldws 0(0,%r3),%fr9L
-	ldw 0(0,%r26),%r19
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r19,%r23
-	ldw -16(0,%r30),%r28
-	ldw -12(0,%r30),%r29
-	ldi 0,%r22
-	add %r23,%r29,%r29
-	addc %r22,%r28,%r28
-	add %r25,%r29,%r29
-	addc %r24,%r28,%r28
-	copy %r28,%r21
-	ldi 0,%r20
-	copy %r21,%r20
-	addib,= -1,%r31,L$0011
-	stw %r29,0(0,%r26)
-	copy %r20,%r25
-	ldi 0,%r24
-	fldws -8(0,%r1),%fr9L
-	ldw -8(0,%r2),%r19
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r19,%r23
-	ldw -16(0,%r30),%r28
-	ldw -12(0,%r30),%r29
-	ldi 0,%r22
-	add %r23,%r29,%r29
-	addc %r22,%r28,%r28
-	add %r25,%r29,%r29
-	addc %r24,%r28,%r28
-	copy %r28,%r21
-	ldi 0,%r20
-	copy %r21,%r20
-	addib,= -1,%r31,L$0011
-	stw %r29,-8(0,%r2)
-	copy %r20,%r25
-	ldi 0,%r24
-	fldws -4(0,%r1),%fr9L
-	ldw -4(0,%r2),%r19
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r19,%r23
-	ldw -16(0,%r30),%r28
-	ldw -12(0,%r30),%r29
-	ldi 0,%r22
-	add %r23,%r29,%r29
-	addc %r22,%r28,%r28
-	add %r25,%r29,%r29
-	addc %r24,%r28,%r28
-	copy %r28,%r21
-	ldi 0,%r20
-	copy %r21,%r20
-	addib,= -1,%r31,L$0011
-	stw %r29,-4(0,%r2)
-	copy %r20,%r25
-	ldi 0,%r24
-	fldws 0(0,%r1),%fr9L
-	ldw 0(0,%r2),%r19
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r19,%r23
-	ldw -16(0,%r30),%r28
-	ldw -12(0,%r30),%r29
-	ldi 0,%r22
-	add %r23,%r29,%r29
-	addc %r22,%r28,%r28
-	add %r25,%r29,%r29
-	addc %r24,%r28,%r28
-	copy %r28,%r21
-	ldi 0,%r20
-	copy %r21,%r20
-	addib,= -1,%r31,L$0011
-	stw %r29,0(0,%r2)
-	ldo 16(%r1),%r1
-	ldo 16(%r3),%r3
-	ldo 16(%r2),%r2
-	bl L$0010,0
-	ldo 16(%r26),%r26
-L$0011
-	copy %r20,%r28
-	ldw -84(0,%r30),%r2
-	ldw -60(0,%r30),%r3
-	bv 0(%r2)
-	ldwm -64(0,%r30),%r4
-	.EXIT
-	.PROCEND
-	.align 4
-	.EXPORT bn_mul_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR
+	.export	bn_mul_add_words,entry,NO_RELOCATION,LONG_RETURN
+	.proc
+	.callinfo frame=128
+    .entry
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3  
+    STD     %r4,8(%sp)          ; save r4  
+	NOP                         ; Needed to make the loop 16-byte aligned
+	NOP                         ; needed to make the loop 16-byte aligned
+
+    STD     %r5,16(%sp)         ; save r5  
+	NOP
+    STD     %r6,24(%sp)         ; save r6  
+    STD     %r7,32(%sp)         ; save r7  
+
+    STD     %r8,40(%sp)         ; save r8  
+    STD     %r9,48(%sp)         ; save r9  
+    COPY    %r0,%ret1           ; return 0 by default
+    DEPDI,Z 1,31,1,top_overflow ; top_overflow = 1 << 32    
+
+    CMPIB,>= 0,num,bn_mul_add_words_exit  ; if (num <= 0) then exit
+	LDO     128(%sp),%sp        ; bump stack
+
+	;
+	; The loop is unrolled twice, so if there is only 1 number
+    ; then go straight to the cleanup code.
+	;
+	CMPIB,= 1,num,bn_mul_add_words_single_top
+	FLDD    -184(%sp),fw        ; (-56-128) load up w into fw (fw_h/fw_l)
+
+	;
+	; This loop is unrolled 2 times (64-byte aligned as well)
+	;
+	; PA-RISC 2.0 chips have two fully pipelined multipliers, thus
+    ; two 32-bit mutiplies can be issued per cycle.
+    ; 
+bn_mul_add_words_unroll2
+
+    FLDD    0(a_ptr),t_float_0       ; load up 64-bit value (fr8L) ht(L)/lt(R)
+    FLDD    8(a_ptr),t_float_1       ; load up 64-bit value (fr8L) ht(L)/lt(R)
+    LDD     0(r_ptr),rp_val          ; rp[0]
+    LDD     8(r_ptr),rp_val_1        ; rp[1]
+
+    XMPYU   fht_0,fw_l,fm1           ; m1[0] = fht_0*fw_l
+    XMPYU   fht_1,fw_l,fm1_1         ; m1[1] = fht_1*fw_l
+    FSTD    fm1,-16(%sp)             ; -16(sp) = m1[0]
+    FSTD    fm1_1,-48(%sp)           ; -48(sp) = m1[1]
+
+    XMPYU   flt_0,fw_h,fm            ; m[0] = flt_0*fw_h
+    XMPYU   flt_1,fw_h,fm_1          ; m[1] = flt_1*fw_h
+    FSTD    fm,-8(%sp)               ; -8(sp) = m[0]
+    FSTD    fm_1,-40(%sp)            ; -40(sp) = m[1]
+
+    XMPYU   fht_0,fw_h,ht_temp       ; ht_temp   = fht_0*fw_h
+    XMPYU   fht_1,fw_h,ht_temp_1     ; ht_temp_1 = fht_1*fw_h
+    FSTD    ht_temp,-24(%sp)         ; -24(sp)   = ht_temp
+    FSTD    ht_temp_1,-56(%sp)       ; -56(sp)   = ht_temp_1
+
+    XMPYU   flt_0,fw_l,lt_temp       ; lt_temp = lt*fw_l
+    XMPYU   flt_1,fw_l,lt_temp_1     ; lt_temp = lt*fw_l
+    FSTD    lt_temp,-32(%sp)         ; -32(sp) = lt_temp 
+    FSTD    lt_temp_1,-64(%sp)       ; -64(sp) = lt_temp_1 
+
+    LDD     -8(%sp),m_0              ; m[0] 
+    LDD     -40(%sp),m_1             ; m[1]
+    LDD     -16(%sp),m1_0            ; m1[0]
+    LDD     -48(%sp),m1_1            ; m1[1]
+
+    LDD     -24(%sp),ht_0            ; ht[0]
+    LDD     -56(%sp),ht_1            ; ht[1]
+    ADD,L   m1_0,m_0,tmp_0           ; tmp_0 = m[0] + m1[0]; 
+    ADD,L   m1_1,m_1,tmp_1           ; tmp_1 = m[1] + m1[1]; 
+
+    LDD     -32(%sp),lt_0            
+    LDD     -64(%sp),lt_1            
+    CMPCLR,*>>= tmp_0,m1_0, %r0      ; if (m[0] < m1[0])
+    ADD,L   ht_0,top_overflow,ht_0   ; ht[0] += (1<<32)
+
+    CMPCLR,*>>= tmp_1,m1_1,%r0       ; if (m[1] < m1[1])
+    ADD,L   ht_1,top_overflow,ht_1   ; ht[1] += (1<<32)
+    EXTRD,U tmp_0,31,32,m_0          ; m[0]>>32  
+    DEPD,Z  tmp_0,31,32,m1_0         ; m1[0] = m[0]<<32 
+
+    EXTRD,U tmp_1,31,32,m_1          ; m[1]>>32  
+    DEPD,Z  tmp_1,31,32,m1_1         ; m1[1] = m[1]<<32 
+    ADD,L   ht_0,m_0,ht_0            ; ht[0]+= (m[0]>>32)
+    ADD,L   ht_1,m_1,ht_1            ; ht[1]+= (m[1]>>32)
+
+    ADD     lt_0,m1_0,lt_0           ; lt[0] = lt[0]+m1[0];
+	ADD,DC  ht_0,%r0,ht_0            ; ht[0]++
+    ADD     lt_1,m1_1,lt_1           ; lt[1] = lt[1]+m1[1];
+    ADD,DC  ht_1,%r0,ht_1            ; ht[1]++
+
+    ADD    %ret1,lt_0,lt_0           ; lt[0] = lt[0] + c;
+	ADD,DC  ht_0,%r0,ht_0            ; ht[0]++
+    ADD     lt_0,rp_val,lt_0         ; lt[0] = lt[0]+rp[0]
+    ADD,DC  ht_0,%r0,ht_0            ; ht[0]++
+
+	LDO    -2(num),num               ; num = num - 2;
+    ADD     ht_0,lt_1,lt_1           ; lt[1] = lt[1] + ht_0 (c);
+    ADD,DC  ht_1,%r0,ht_1            ; ht[1]++
+    STD     lt_0,0(r_ptr)            ; rp[0] = lt[0]
+
+    ADD     lt_1,rp_val_1,lt_1       ; lt[1] = lt[1]+rp[1]
+    ADD,DC  ht_1,%r0,%ret1           ; ht[1]++
+    LDO     16(a_ptr),a_ptr          ; a_ptr += 2
+
+    STD     lt_1,8(r_ptr)            ; rp[1] = lt[1]
+	CMPIB,<= 2,num,bn_mul_add_words_unroll2 ; go again if more to do
+    LDO     16(r_ptr),r_ptr          ; r_ptr += 2
+
+    CMPIB,=,N 0,num,bn_mul_add_words_exit ; are we done, or cleanup last one
+
+	;
+	; Top of loop aligned on 64-byte boundary
+	;
+bn_mul_add_words_single_top
+    FLDD    0(a_ptr),t_float_0        ; load up 64-bit value (fr8L) ht(L)/lt(R)
+    LDD     0(r_ptr),rp_val           ; rp[0]
+    LDO     8(a_ptr),a_ptr            ; a_ptr++
+    XMPYU   fht_0,fw_l,fm1            ; m1 = ht*fw_l
+    FSTD    fm1,-16(%sp)              ; -16(sp) = m1
+    XMPYU   flt_0,fw_h,fm             ; m = lt*fw_h
+    FSTD    fm,-8(%sp)                ; -8(sp) = m
+    XMPYU   fht_0,fw_h,ht_temp        ; ht_temp = ht*fw_h
+    FSTD    ht_temp,-24(%sp)          ; -24(sp) = ht
+    XMPYU   flt_0,fw_l,lt_temp        ; lt_temp = lt*fw_l
+    FSTD    lt_temp,-32(%sp)          ; -32(sp) = lt 
+
+    LDD     -8(%sp),m_0               
+    LDD    -16(%sp),m1_0              ; m1 = temp1 
+    ADD,L   m_0,m1_0,tmp_0            ; tmp_0 = m + m1; 
+    LDD     -24(%sp),ht_0             
+    LDD     -32(%sp),lt_0             
+
+    CMPCLR,*>>= tmp_0,m1_0,%r0        ; if (m < m1)
+    ADD,L   ht_0,top_overflow,ht_0    ; ht += (1<<32)
+
+    EXTRD,U tmp_0,31,32,m_0           ; m>>32  
+    DEPD,Z  tmp_0,31,32,m1_0          ; m1 = m<<32 
+
+    ADD,L   ht_0,m_0,ht_0             ; ht+= (m>>32)
+    ADD     lt_0,m1_0,tmp_0           ; tmp_0 = lt+m1;
+    ADD,DC  ht_0,%r0,ht_0             ; ht++
+    ADD     %ret1,tmp_0,lt_0          ; lt = lt + c;
+    ADD,DC  ht_0,%r0,ht_0             ; ht++
+    ADD     lt_0,rp_val,lt_0          ; lt = lt+rp[0]
+    ADD,DC  ht_0,%r0,%ret1            ; ht++
+    STD     lt_0,0(r_ptr)             ; rp[0] = lt
+
+bn_mul_add_words_exit
+    .EXIT
+	
+    EXTRD,U %ret1,31,32,%ret0         ; for 32-bit, return in ret0/ret1
+    LDD     -80(%sp),%r9              ; restore r9  
+    LDD     -88(%sp),%r8              ; restore r8  
+    LDD     -96(%sp),%r7              ; restore r7  
+    LDD     -104(%sp),%r6             ; restore r6  
+    LDD     -112(%sp),%r5             ; restore r5  
+    LDD     -120(%sp),%r4             ; restore r4  
+    BVE     (%rp)
+    LDD,MB  -128(%sp),%r3             ; restore r3
+	.PROCEND	;in=23,24,25,26,29;out=28;
+
+;----------------------------------------------------------------------------
+;
+;BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+;
+; arg0 = rp
+; arg1 = ap
+; arg3 = num
+; w on stack at -56(sp)
+
 bn_mul_words
-	.PROC
-	.CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
-	.ENTRY
-	stw %r2,-20(0,%r30)
-	copy %r25,%r2
-	stwm %r4,64(0,%r30)
-	copy %r24,%r19
-	ldi 0,%r28
-	stw %r23,-16(0,%r30)
-	ldo 12(%r26),%r31
-	ldo 12(%r2),%r29
-	fldws -16(0,%r30),%fr8L
-L$0026
-	fldws 0(0,%r2),%fr9L
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r28,%r21
-	ldi 0,%r20
-	ldw -16(0,%r30),%r24
-	ldw -12(0,%r30),%r25
-	add %r21,%r25,%r25
-	addc %r20,%r24,%r24
-	copy %r24,%r23
-	ldi 0,%r22
-	copy %r23,%r28
-	addib,= -1,%r19,L$0027
-	stw %r25,0(0,%r26)
-	fldws -8(0,%r29),%fr9L
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r28,%r21
-	ldi 0,%r20
-	ldw -16(0,%r30),%r24
-	ldw -12(0,%r30),%r25
-	add %r21,%r25,%r25
-	addc %r20,%r24,%r24
-	copy %r24,%r23
-	ldi 0,%r22
-	copy %r23,%r28
-	addib,= -1,%r19,L$0027
-	stw %r25,-8(0,%r31)
-	fldws -4(0,%r29),%fr9L
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r28,%r21
-	ldi 0,%r20
-	ldw -16(0,%r30),%r24
-	ldw -12(0,%r30),%r25
-	add %r21,%r25,%r25
-	addc %r20,%r24,%r24
-	copy %r24,%r23
-	ldi 0,%r22
-	copy %r23,%r28
-	addib,= -1,%r19,L$0027
-	stw %r25,-4(0,%r31)
-	fldws 0(0,%r29),%fr9L
-	xmpyu %fr8L,%fr9L,%fr9
-	fstds %fr9,-16(0,%r30)
-	copy %r28,%r21
-	ldi 0,%r20
-	ldw -16(0,%r30),%r24
-	ldw -12(0,%r30),%r25
-	add %r21,%r25,%r25
-	addc %r20,%r24,%r24
-	copy %r24,%r23
-	ldi 0,%r22
-	copy %r23,%r28
-	addib,= -1,%r19,L$0027
-	stw %r25,0(0,%r31)
-	ldo 16(%r29),%r29
-	ldo 16(%r2),%r2
-	ldo 16(%r31),%r31
-	bl L$0026,0
-	ldo 16(%r26),%r26
-L$0027
-	ldw -84(0,%r30),%r2
-	bv 0(%r2)
-	ldwm -64(0,%r30),%r4
-	.EXIT
-	.PROCEND
-	.align 4
-	.EXPORT bn_sqr_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR
+	.proc
+	.callinfo frame=128
+    .entry
+	.EXPORT	bn_mul_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3  
+    STD     %r4,8(%sp)          ; save r4  
+	NOP
+    STD     %r5,16(%sp)         ; save r5  
+
+    STD     %r6,24(%sp)         ; save r6  
+    STD     %r7,32(%sp)         ; save r7  
+    COPY    %r0,%ret1           ; return 0 by default
+    DEPDI,Z 1,31,