openssl-engine-0.9.6 merge

22 files changed, 2351 insertions, 957 deletions

diff --git a/lib/libcrypto/bn/Makefile.ssl b/lib/libcrypto/bn/Makefile.ssl
index beb9c1b5231..17b72d577f3 100644
--- a/lib/libcrypto/bn/Makefile.ssl
+++ b/lib/libcrypto/bn/Makefile.ssl
@@ -170,118 +170,143 @@ clean:
 bn_add.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_add.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_add.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_add.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_add.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_add.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_add.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_add.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_add.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_add.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_asm.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_asm.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_asm.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_asm.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_asm.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_asm.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_asm.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_asm.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_asm.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_asm.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_blind.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_blind.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_blind.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_blind.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
+bn_blind.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_blind.o: ../../include/openssl/opensslconf.h
 bn_blind.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_blind.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_blind.o: ../../include/openssl/stack.h ../../include/openssl/symhacks.h
+bn_blind.o: ../cryptlib.h bn_lcl.h
 bn_ctx.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_ctx.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_ctx.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_ctx.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_ctx.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_ctx.o: ../../include/openssl/stack.h ../cryptlib.h
+bn_ctx.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_ctx.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_ctx.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_ctx.o: ../../include/openssl/symhacks.h ../cryptlib.h
 bn_div.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_div.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_div.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_div.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_div.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_div.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
-bn_err.o: ../../include/openssl/bn.h ../../include/openssl/err.h
-bn_err.o: ../../include/openssl/opensslconf.h
+bn_div.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_div.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_div.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_div.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
+bn_err.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
+bn_err.o: ../../include/openssl/crypto.h ../../include/openssl/err.h
+bn_err.o: ../../include/openssl/lhash.h ../../include/openssl/opensslconf.h
+bn_err.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
+bn_err.o: ../../include/openssl/stack.h ../../include/openssl/symhacks.h
 bn_exp.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_exp.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_exp.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_exp.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_exp.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_exp.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_exp.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_exp.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_exp.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_exp.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_exp2.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_exp2.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_exp2.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_exp2.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_exp2.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_exp2.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_exp2.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_exp2.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_exp2.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_exp2.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_gcd.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_gcd.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_gcd.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_gcd.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_gcd.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_gcd.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_gcd.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_gcd.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_gcd.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_gcd.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_lib.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_lib.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_lib.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_lib.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_lib.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_lib.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_lib.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_lib.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_lib.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_lib.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_mont.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_mont.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_mont.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_mont.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_mont.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_mont.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_mont.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_mont.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_mont.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_mont.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_mpi.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_mpi.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_mpi.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_mpi.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_mpi.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_mpi.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_mpi.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_mpi.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_mpi.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_mpi.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_mul.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_mul.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_mul.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_mul.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_mul.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_mul.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_mul.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_mul.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_mul.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_mul.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_prime.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_prime.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_prime.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_prime.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
+bn_prime.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_prime.o: ../../include/openssl/opensslconf.h
 bn_prime.o: ../../include/openssl/opensslv.h ../../include/openssl/rand.h
 bn_prime.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
-bn_prime.o: ../cryptlib.h bn_lcl.h bn_prime.h
+bn_prime.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h bn_prime.h
 bn_print.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_print.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_print.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_print.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
+bn_print.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_print.o: ../../include/openssl/opensslconf.h
 bn_print.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_print.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_print.o: ../../include/openssl/stack.h ../../include/openssl/symhacks.h
+bn_print.o: ../cryptlib.h bn_lcl.h
 bn_rand.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_rand.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_rand.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_rand.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_rand.o: ../../include/openssl/opensslv.h ../../include/openssl/rand.h
-bn_rand.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_rand.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_rand.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_rand.o: ../../include/openssl/rand.h ../../include/openssl/safestack.h
+bn_rand.o: ../../include/openssl/stack.h ../../include/openssl/symhacks.h
 bn_rand.o: ../cryptlib.h bn_lcl.h
 bn_recp.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_recp.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_recp.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_recp.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_recp.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_recp.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_recp.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_recp.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_recp.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_recp.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_shift.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_shift.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_shift.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_shift.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
+bn_shift.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_shift.o: ../../include/openssl/opensslconf.h
 bn_shift.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_shift.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_shift.o: ../../include/openssl/stack.h ../../include/openssl/symhacks.h
+bn_shift.o: ../cryptlib.h bn_lcl.h
 bn_sqr.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_sqr.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_sqr.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_sqr.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_sqr.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_sqr.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_sqr.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_sqr.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_sqr.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_sqr.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
 bn_word.o: ../../include/openssl/bio.h ../../include/openssl/bn.h
 bn_word.o: ../../include/openssl/buffer.h ../../include/openssl/crypto.h
 bn_word.o: ../../include/openssl/e_os.h ../../include/openssl/e_os2.h
-bn_word.o: ../../include/openssl/err.h ../../include/openssl/opensslconf.h
-bn_word.o: ../../include/openssl/opensslv.h ../../include/openssl/safestack.h
-bn_word.o: ../../include/openssl/stack.h ../cryptlib.h bn_lcl.h
+bn_word.o: ../../include/openssl/err.h ../../include/openssl/lhash.h
+bn_word.o: ../../include/openssl/opensslconf.h ../../include/openssl/opensslv.h
+bn_word.o: ../../include/openssl/safestack.h ../../include/openssl/stack.h
+bn_word.o: ../../include/openssl/symhacks.h ../cryptlib.h bn_lcl.h
diff --git a/lib/libcrypto/bn/asm/README b/lib/libcrypto/bn/asm/README
index 86bf64cfc2f..a0fe58a6771 100644
--- a/lib/libcrypto/bn/asm/README
+++ b/lib/libcrypto/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/lib/libcrypto/bn/asm/pa-risc2.s b/lib/libcrypto/bn/asm/pa-risc2.s
index c2725996a45..7239aa2c762 100644
--- a/lib/libcrypto/bn/asm/pa-risc2.s
+++ b/lib/libcrypto/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 1,31,1,top_overflow ; top_overflow = 1 << 32    
+
+    CMPIB,>= 0,num,bn_mul_words_exit
+	LDO     128(%sp),%sp    ; bump stack
+
+	;
+	; See if only 1 word to do, thus just do cleanup
+	;
+	CMPIB,= 1,num,bn_mul_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_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)
+    XMPYU   fht_0,fw_l,fm1            ; m1[0] = fht_0*fw_l
+    XMPYU   fht_1,fw_l,fm1_1          ; m1[1] = ht*fw_l
+
+    FSTD    fm1,-16(%sp)              ; -16(sp) = m1
+    FSTD    fm1_1,-48(%sp)            ; -48(sp) = m1
+    XMPYU   flt_0,fw_h,fm             ; m = lt*fw_h
+    XMPYU   flt_1,fw_h,fm_1           ; m = lt*fw_h
+
+    FSTD    fm,-8(%sp)                ; -8(sp) = m
+    FSTD    fm_1,-40(%sp)             ; -40(sp) = m
+    XMPYU   fht_0,fw_h,ht_temp        ; ht_temp = fht_0*fw_h
+    XMPYU   fht_1,fw_h,ht_temp_1      ; ht_temp = ht*fw_h
+
+    FSTD    ht_temp,-24(%sp)          ; -24(sp) = ht
+    FSTD    ht_temp_1,-56(%sp)        ; -56(sp) = ht
+    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 
+    FSTD    lt_temp_1,-64(%sp)        ; -64(sp) = lt 
+    LDD     -8(%sp),m_0               
+    LDD     -40(%sp),m_1              
+
+    LDD    -16(%sp),m1_0              
+    LDD    -48(%sp),m1_1              
+    LDD     -24(%sp),ht_0             
+    LDD     -56(%sp),ht_1             
+
+    ADD,L   m1_0,m_0,tmp_0            ; tmp_0 = m + m1; 
+    ADD,L   m1_1,m_1,tmp_1            ; tmp_1 = m + m1; 
+    LDD     -32(%sp),lt_0             
+    LDD     -64(%sp),lt_1             
+
+    CMPCLR,*>>= tmp_0,m1_0, %r0       ; if (m < m1)
+    ADD,L   ht_0,top_overflow,ht_0    ; ht += (1<<32)
+    CMPCLR,*>>= tmp_1,m1_1,%r0        ; if (m < m1)
+    ADD,L   ht_1,top_overflow,ht_1    ; ht += (1<<32)
+
+    EXTRD,U tmp_0,31,32,m_0           ; m>>32  
+    DEPD,Z  tmp_0,31,32,m1_0          ; m1 = m<<32 
+    EXTRD,U tmp_1,31,32,m_1           ; m>>32  
+    DEPD,Z  tmp_1,31,32,m1_1          ; m1 = m<<32 
+
+    ADD,L   ht_0,m_0,ht_0             ; ht+= (m>>32)
+    ADD,L   ht_1,m_1,ht_1             ; ht+= (m>>32)
+    ADD     lt_0,m1_0,lt_0            ; lt = lt+m1;
+	ADD,DC  ht_0,%r0,ht_0             ; ht++
+
+    ADD     lt_1,m1_1,lt_1            ; lt = lt+m1;
+    ADD,DC  ht_1,%r0,ht_1             ; ht++
+    ADD    %ret1,lt_0,lt_0            ; lt = lt + c (ret1);
+	ADD,DC  ht_0,%r0,ht_0             ; ht++
+
+    ADD     ht_0,lt_1,lt_1            ; lt = lt + c (ht_0)
+    ADD,DC  ht_1,%r0,ht_1             ; ht++
+    STD     lt_0,0(r_ptr)             ; rp[0] = lt
+    STD     lt_1,8(r_ptr)             ; rp[1] = lt
+
+	COPY    ht_1,%ret1                ; carry = ht
+	LDO    -2(num),num                ; num = num - 2;
+    LDO     16(a_ptr),a_ptr           ; ap += 2
+	CMPIB,<= 2,num,bn_mul_words_unroll2
+    LDO     16(r_ptr),r_ptr           ; rp++
+
+    CMPIB,=,N 0,num,bn_mul_words_exit ; are we done?
+
+	;
+	; Top of loop aligned on 64-byte boundary
+	;
+bn_mul_words_single_top
+    FLDD    0(a_ptr),t_float_0        ; load up 64-bit value (fr8L) ht(L)/lt(R)
+
+    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              
+    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,lt_0            ; lt= lt+m1;
+    ADD,DC  ht_0,%r0,ht_0             ; ht++
+
+    ADD     %ret1,lt_0,lt_0           ; lt = lt + c;
+    ADD,DC  ht_0,%r0,ht_0             ; ht++
+
+    COPY    ht_0,%ret1                ; copy carry
+    STD     lt_0,0(r_ptr)             ; rp[0] = lt
+
+bn_mul_words_exit
+    .EXIT
+    EXTRD,U %ret1,31,32,%ret0           ; for 32-bit, return in ret0/ret1
+    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	
+
+;----------------------------------------------------------------------------
+;
+;void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num)
+;
+; arg0 = rp
+; arg1 = ap
+; arg2 = num
+;
+
 bn_sqr_words
+	.proc
+	.callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
+	.EXPORT	bn_sqr_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .entry
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3  
+    STD     %r4,8(%sp)          ; save r4  
+	NOP
+    STD     %r5,16(%sp)         ; save r5  
+
+    CMPIB,>= 0,num,bn_sqr_words_exit
+	LDO     128(%sp),%sp       ; bump stack
+
+	;
+	; If only 1, the goto straight to cleanup
+	;
+	CMPIB,= 1,num,bn_sqr_words_single_top
+    DEPDI,Z -1,32,33,high_mask   ; Create Mask 0xffffffff80000000L
+
+	;
+	; This loop is unrolled 2 times (64-byte aligned as well)
+	;
+
+bn_sqr_words_unroll2
+    FLDD    0(a_ptr),t_float_0        ; a[0]
+    FLDD    8(a_ptr),t_float_1        ; a[1]
+    XMPYU   fht_0,flt_0,fm            ; m[0]
+    XMPYU   fht_1,flt_1,fm_1          ; m[1]
+
+    FSTD    fm,-24(%sp)               ; store m[0]
+    FSTD    fm_1,-56(%sp)             ; store m[1]
+    XMPYU   flt_0,flt_0,lt_temp       ; lt[0]
+    XMPYU   flt_1,flt_1,lt_temp_1     ; lt[1]
+
+    FSTD    lt_temp,-16(%sp)          ; store lt[0]
+    FSTD    lt_temp_1,-48(%sp)        ; store lt[1]
+    XMPYU   fht_0,fht_0,ht_temp       ; ht[0]
+    XMPYU   fht_1,fht_1,ht_temp_1     ; ht[1]
+
+    FSTD    ht_temp,-8(%sp)           ; store ht[0]
+    FSTD    ht_temp_1,-40(%sp)        ; store ht[1]
+    LDD     -24(%sp),m_0             
+    LDD     -56(%sp),m_1              
+
+    AND     m_0,high_mask,tmp_0       ; m[0] & Mask
+    AND     m_1,high_mask,tmp_1       ; m[1] & Mask
+    DEPD,Z  m_0,30,31,m_0             ; m[0] << 32+1
+    DEPD,Z  m_1,30,31,m_1             ; m[1] << 32+1
+
+    LDD     -16(%sp),lt_0        
+    LDD     -48(%sp),lt_1        
+    EXTRD,U tmp_0,32,33,tmp_0         ; tmp_0 = m[0]&Mask >> 32-1
+    EXTRD,U tmp_1,32,33,tmp_1         ; tmp_1 = m[1]&Mask >> 32-1
+
+    LDD     -8(%sp),ht_0            
+    LDD     -40(%sp),ht_1           
+    ADD,L   ht_0,tmp_0,ht_0           ; ht[0] += tmp_0
+    ADD,L   ht_1,tmp_1,ht_1           ; ht[1] += tmp_1
+
+    ADD     lt_0,m_0,lt_0             ; lt = lt+m
+    ADD,DC  ht_0,%r0,ht_0             ; ht[0]++
+    STD     lt_0,0(r_ptr)             ; rp[0] = lt[0]
+    STD     ht_0,8(r_ptr)             ; rp[1] = ht[1]
+
+    ADD     lt_1,m_1,lt_1             ; lt = lt+m
+    ADD,DC  ht_1,%r0,ht_1             ; ht[1]++
+    STD     lt_1,16(r_ptr)            ; rp[2] = lt[1]
+    STD     ht_1,24(r_ptr)            ; rp[3] = ht[1]
+
+	LDO    -2(num),num                ; num = num - 2;
+    LDO     16(a_ptr),a_ptr           ; ap += 2
+	CMPIB,<= 2,num,bn_sqr_words_unroll2
+    LDO     32(r_ptr),r_ptr           ; rp += 4
+
+    CMPIB,=,N 0,num,bn_sqr_words_exit ; are we done?
+
+	;
+	; Top of loop aligned on 64-byte boundary
+	;
+bn_sqr_words_single_top
+    FLDD    0(a_ptr),t_float_0        ; load up 64-bit value (fr8L) ht(L)/lt(R)
+
+    XMPYU   fht_0,flt_0,fm            ; m
+    FSTD    fm,-24(%sp)               ; store m
+
+    XMPYU   flt_0,flt_0,lt_temp       ; lt
+    FSTD    lt_temp,-16(%sp)          ; store lt
+
+    XMPYU   fht_0,fht_0,ht_temp       ; ht
+    FSTD    ht_temp,-8(%sp)           ; store ht
+
+    LDD     -24(%sp),m_0              ; load m
+    AND     m_0,high_mask,tmp_0       ; m & Mask
+    DEPD,Z  m_0,30,31,m_0             ; m << 32+1
+    LDD     -16(%sp),lt_0             ; lt
+
+    LDD     -8(%sp),ht_0              ; ht
+    EXTRD,U tmp_0,32,33,tmp_0         ; tmp_0 = m&Mask >> 32-1
+    ADD     m_0,lt_0,lt_0             ; lt = lt+m
+    ADD,L   ht_0,tmp_0,ht_0           ; ht += tmp_0
+    ADD,DC  ht_0,%r0,ht_0             ; ht++
+
+    STD     lt_0,0(r_ptr)             ; rp[0] = lt
+    STD     ht_0,8(r_ptr)             ; rp[1] = ht
+
+bn_sqr_words_exit
+    .EXIT
+    LDD     -112(%sp),%r5       ; restore r5  
+    LDD     -120(%sp),%r4       ; restore r4  
+    BVE     (%rp)
+    LDD,MB  -128(%sp),%r3 
+	.PROCEND	;in=23,24,25,26,29;out=28;
+
+
+;----------------------------------------------------------------------------
+;
+;BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+;
+; arg0 = rp 
+; arg1 = ap
+; arg2 = bp 
+; arg3 = n
+
+t  .reg %r22
+b  .reg %r21
+l  .reg %r20
+
+bn_add_words
+	.proc
+    .entry
+	.callinfo
+	.EXPORT	bn_add_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+	.align 64
+
+    CMPIB,>= 0,n,bn_add_words_exit
+    COPY    %r0,%ret1           ; return 0 by default
+
+	;
+	; If 2 or more numbers do the loop
+	;
+	CMPIB,= 1,n,bn_add_words_single_top
+	NOP
+
+	;
+	; This loop is unrolled 2 times (64-byte aligned as well)
+	;
+bn_add_words_unroll2
+	LDD     0(a_ptr),t
+	LDD     0(b_ptr),b
+	ADD     t,%ret1,t                    ; t = t+c;
+	ADD,DC  %r0,%r0,%ret1                ; set c to carry
+	ADD     t,b,l                        ; l = t + b[0]
+	ADD,DC  %ret1,%r0,%ret1              ; c+= carry
+	STD     l,0(r_ptr)
+
+	LDD     8(a_ptr),t
+	LDD     8(b_ptr),b
+	ADD     t,%ret1,t                     ; t = t+c;
+	ADD,DC  %r0,%r0,%ret1                 ; set c to carry
+	ADD     t,b,l                         ; l = t + b[0]
+	ADD,DC  %ret1,%r0,%ret1               ; c+= carry
+	STD     l,8(r_ptr)
+
+	LDO     -2(n),n
+	LDO     16(a_ptr),a_ptr
+	LDO     16(b_ptr),b_ptr
+
+	CMPIB,<= 2,n,bn_add_words_unroll2
+	LDO     16(r_ptr),r_ptr
+
+    CMPIB,=,N 0,n,bn_add_words_exit ; are we done?
+
+bn_add_words_single_top
+	LDD     0(a_ptr),t
+	LDD     0(b_ptr),b
+
+	ADD     t,%ret1,t                 ; t = t+c;
+	ADD,DC  %r0,%r0,%ret1             ; set c to carry (could use CMPCLR??)
+	ADD     t,b,l                     ; l = t + b[0]
+	ADD,DC  %ret1,%r0,%ret1           ; c+= carry
+	STD     l,0(r_ptr)
+
+bn_add_words_exit
+    .EXIT
+    BVE     (%rp)
+    EXTRD,U %ret1,31,32,%ret0           ; for 32-bit, return in ret0/ret1
+	.PROCEND	;in=23,24,25,26,29;out=28;
+
+;----------------------------------------------------------------------------
+;
+;BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+;
+; arg0 = rp 
+; arg1 = ap
+; arg2 = bp 
+; arg3 = n
+
+t1       .reg %r22
+t2       .reg %r21
+sub_tmp1 .reg %r20
+sub_tmp2 .reg %r19
+
+
+bn_sub_words
+	.proc
+	.callinfo 
+	.EXPORT	bn_sub_words,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .entry
+	.align 64
+
+    CMPIB,>=  0,n,bn_sub_words_exit
+    COPY    %r0,%ret1           ; return 0 by default
+
+	;
+	; If 2 or more numbers do the loop
+	;
+	CMPIB,= 1,n,bn_sub_words_single_top
+	NOP
+
+	;
+	; This loop is unrolled 2 times (64-byte aligned as well)
+	;
+bn_sub_words_unroll2
+	LDD     0(a_ptr),t1
+	LDD     0(b_ptr),t2
+	SUB     t1,t2,sub_tmp1           ; t3 = t1-t2; 
+	SUB     sub_tmp1,%ret1,sub_tmp1  ; t3 = t3- c; 
+
+	CMPCLR,*>> t1,t2,sub_tmp2        ; clear if t1 > t2
+	LDO      1(%r0),sub_tmp2
+	
+	CMPCLR,*= t1,t2,%r0
+	COPY    sub_tmp2,%ret1
+	STD     sub_tmp1,0(r_ptr)
+
+	LDD     8(a_ptr),t1
+	LDD     8(b_ptr),t2
+	SUB     t1,t2,sub_tmp1            ; t3 = t1-t2; 
+	SUB     sub_tmp1,%ret1,sub_tmp1   ; t3 = t3- c; 
+	CMPCLR,*>> t1,t2,sub_tmp2         ; clear if t1 > t2
+	LDO      1(%r0),sub_tmp2
+	
+	CMPCLR,*= t1,t2,%r0
+	COPY    sub_tmp2,%ret1
+	STD     sub_tmp1,8(r_ptr)
+
+	LDO     -2(n),n
+	LDO     16(a_ptr),a_ptr
+	LDO     16(b_ptr),b_ptr
+
+	CMPIB,<= 2,n,bn_sub_words_unroll2
+	LDO     16(r_ptr),r_ptr
+
+    CMPIB,=,N 0,n,bn_sub_words_exit ; are we done?
+
+bn_sub_words_single_top
+	LDD     0(a_ptr),t1
+	LDD     0(b_ptr),t2
+	SUB     t1,t2,sub_tmp1            ; t3 = t1-t2; 
+	SUB     sub_tmp1,%ret1,sub_tmp1   ; t3 = t3- c; 
+	CMPCLR,*>> t1,t2,sub_tmp2         ; clear if t1 > t2
+	LDO      1(%r0),sub_tmp2
+	
+	CMPCLR,*= t1,t2,%r0
+	COPY    sub_tmp2,%ret1
+
+	STD     sub_tmp1,0(r_ptr)
+
+bn_sub_words_exit
+    .EXIT
+    BVE     (%rp)
+    EXTRD,U %ret1,31,32,%ret0           ; for 32-bit, return in ret0/ret1
+	.PROCEND	;in=23,24,25,26,29;out=28;
+
+;------------------------------------------------------------------------------
+;
+; unsigned long bn_div_words(unsigned long h, unsigned long l, unsigned long d)
+;
+; arg0 = h
+; arg1 = l
+; arg2 = d
+;
+; This is mainly just output from the HP C compiler.  
+;
+;------------------------------------------------------------------------------
+bn_div_words
 	.PROC
-	.CALLINFO FRAME=0,NO_CALLS
-	.ENTRY
-	ldo 28(%r26),%r19
-	ldo 12(%r25),%r28
-L$0042
-	fldws 0(0,%r25),%fr8L
-	fldws 0(0,%r25),%fr8R
-	xmpyu %fr8L,%fr8R,%fr8
-	fstds %fr8,-16(0,%r30)
-	ldw -16(0,%r30),%r22
-	ldw -12(0,%r30),%r23
-	stw %r23,0(0,%r26)
-	copy %r22,%r21
-	ldi 0,%r20
-	addib,= -1,%r24,L$0049
-	stw %r21,-24(0,%r19)
-	fldws -8(0,%r28),%fr8L
-	fldws -8(0,%r28),%fr8R
-	xmpyu %fr8L,%fr8R,%fr8
-	fstds %fr8,-16(0,%r30)
-	ldw -16(0,%r30),%r22
-	ldw -12(0,%r30),%r23
-	stw %r23,-20(0,%r19)
-	copy %r22,%r21
-	ldi 0,%r20
-	addib,= -1,%r24,L$0049
-	stw %r21,-16(0,%r19)
-	fldws -4(0,%r28),%fr8L
-	fldws -4(0,%r28),%fr8R
-	xmpyu %fr8L,%fr8R,%fr8
-	fstds %fr8,-16(0,%r30)
-	ldw -16(0,%r30),%r22
-	ldw -12(0,%r30),%r23
-	stw %r23,-12(0,%r19)
-	copy %r22,%r21
-	ldi 0,%r20
-	addib,= -1,%r24,L$0049
-	stw %r21,-8(0,%r19)
-	fldws 0(0,%r28),%fr8L
-	fldws 0(0,%r28),%fr8R
-	xmpyu %fr8L,%fr8R,%fr8
-	fstds %fr8,-16(0,%r30)
-	ldw -16(0,%r30),%r22
-	ldw -12(0,%r30),%r23
-	stw %r23,-4(0,%r19)
-	copy %r22,%r21
-	ldi 0,%r20
-	addib,= -1,%r24,L$0049
-	stw %r21,0(0,%r19)
-	ldo 16(%r28),%r28
-	ldo 16(%r25),%r25
-	ldo 32(%r19),%r19
-	bl L$0042,0
-	ldo 32(%r26),%r26
-L$0049
-	bv,n 0(%r2)
-	.EXIT
-	.PROCEND
-	.IMPORT BN_num_bits_word,CODE
-	.IMPORT fprintf,CODE
-	.IMPORT __iob,DATA
-	.SPACE $TEXT$
-	.SUBSPA $LIT$
-
-	.align 4
-L$C0000
-	.STRING "Division would overflow (%d)\x0a\x00"
-	.IMPORT abort,CODE
-	.SPACE $TEXT$
-	.SUBSPA $CODE$
-
-	.align 4
-	.EXPORT bn_div64,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,RTNVAL=GR
-bn_div64
+	.EXPORT	bn_div_words,ENTRY,PRIV_LEV=3,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR,RTNVAL=GR,LONG_RETURN
+	.IMPORT	BN_num_bits_word,CODE
+	.IMPORT	__iob,DATA
+	.IMPORT	fprintf,CODE
+	.IMPORT	abort,CODE
+	.IMPORT	$$div2U,MILLICODE
+	.CALLINFO CALLER,FRAME=144,ENTRY_GR=%r9,SAVE_RP,ARGS_SAVED,ORDERING_AWARE
+        .ENTRY
+        STW     %r2,-20(%r30)   ;offset 0x8ec
+        STW,MA  %r3,192(%r30)   ;offset 0x8f0
+        STW     %r4,-188(%r30)  ;offset 0x8f4
+        DEPD    %r5,31,32,%r6   ;offset 0x8f8
+        STD     %r6,-184(%r30)  ;offset 0x8fc
+        DEPD    %r7,31,32,%r8   ;offset 0x900
+        STD     %r8,-176(%r30)  ;offset 0x904
+        STW     %r9,-168(%r30)  ;offset 0x908
+        LDD     -248(%r30),%r3  ;offset 0x90c
+        COPY    %r26,%r4        ;offset 0x910
+        COPY    %r24,%r5        ;offset 0x914
+        DEPD    %r25,31,32,%r4  ;offset 0x918
+        CMPB,*<>        %r3,%r0,$0006000C       ;offset 0x91c
+        DEPD    %r23,31,32,%r5  ;offset 0x920
+        MOVIB,TR        -1,%r29,$00060002       ;offset 0x924
+        EXTRD,U %r29,31,32,%r28 ;offset 0x928
+$0006002A
+        LDO     -1(%r29),%r29   ;offset 0x92c
+        SUB     %r23,%r7,%r23   ;offset 0x930
+$00060024
+        SUB     %r4,%r31,%r25   ;offset 0x934
+        AND     %r25,%r19,%r26  ;offset 0x938
+        CMPB,*<>,N      %r0,%r26,$00060046      ;offset 0x93c
+        DEPD,Z  %r25,31,32,%r20 ;offset 0x940
+        OR      %r20,%r24,%r21  ;offset 0x944
+        CMPB,*<<,N      %r21,%r23,$0006002A     ;offset 0x948
+        SUB     %r31,%r2,%r31   ;offset 0x94c
+$00060046
+$0006002E
+        DEPD,Z  %r23,31,32,%r25 ;offset 0x950
+        EXTRD,U %r23,31,32,%r26 ;offset 0x954
+        AND     %r25,%r19,%r24  ;offset 0x958
+        ADD,L   %r31,%r26,%r31  ;offset 0x95c
+        CMPCLR,*>>=     %r5,%r24,%r0    ;offset 0x960
+        LDO     1(%r31),%r31    ;offset 0x964
+$00060032
+        CMPB,*<<=,N     %r31,%r4,$00060036      ;offset 0x968
+        LDO     -1(%r29),%r29   ;offset 0x96c
+        ADD,L   %r4,%r3,%r4     ;offset 0x970
+$00060036
+        ADDIB,=,N       -1,%r8,$D0      ;offset 0x974
+        SUB     %r5,%r24,%r28   ;offset 0x978
+$0006003A
+        SUB     %r4,%r31,%r24   ;offset 0x97c
+        SHRPD   %r24,%r28,32,%r4        ;offset 0x980
+        DEPD,Z  %r29,31,32,%r9  ;offset 0x984
+        DEPD,Z  %r28,31,32,%r5  ;offset 0x988
+$0006001C
+        EXTRD,U %r4,31,32,%r31  ;offset 0x98c
+        CMPB,*<>,N      %r31,%r2,$00060020      ;offset 0x990
+        MOVB,TR %r6,%r29,$D1    ;offset 0x994
+        STD     %r29,-152(%r30) ;offset 0x998
+$0006000C
+        EXTRD,U %r3,31,32,%r25  ;offset 0x99c
+        COPY    %r3,%r26        ;offset 0x9a0
+        EXTRD,U %r3,31,32,%r9   ;offset 0x9a4
+        EXTRD,U %r4,31,32,%r8   ;offset 0x9a8
+        .CALL   ARGW0=GR,ARGW1=GR,RTNVAL=GR     ;in=25,26;out=28;
+        B,L     BN_num_bits_word,%r2    ;offset 0x9ac
+        EXTRD,U %r5,31,32,%r7   ;offset 0x9b0
+        LDI     64,%r20 ;offset 0x9b4
+        DEPD    %r7,31,32,%r5   ;offset 0x9b8
+        DEPD    %r8,31,32,%r4   ;offset 0x9bc
+        DEPD    %r9,31,32,%r3   ;offset 0x9c0
+        CMPB,=  %r28,%r20,$00060012     ;offset 0x9c4
+        COPY    %r28,%r24       ;offset 0x9c8
+        MTSARCM %r24    ;offset 0x9cc
+        DEPDI,Z -1,%sar,1,%r19  ;offset 0x9d0
+        CMPB,*>>,N      %r4,%r19,$D2    ;offset 0x9d4
+$00060012
+        SUBI    64,%r24,%r31    ;offset 0x9d8
+        CMPCLR,*<<      %r4,%r3,%r0     ;offset 0x9dc
+        SUB     %r4,%r3,%r4     ;offset 0x9e0
+$00060016
+        CMPB,=  %r31,%r0,$0006001A      ;offset 0x9e4
+        COPY    %r0,%r9 ;offset 0x9e8
+        MTSARCM %r31    ;offset 0x9ec
+        DEPD,Z  %r3,%sar,64,%r3 ;offset 0x9f0
+        SUBI    64,%r31,%r26    ;offset 0x9f4
+        MTSAR   %r26    ;offset 0x9f8
+        SHRPD   %r4,%r5,%sar,%r4        ;offset 0x9fc
+        MTSARCM %r31    ;offset 0xa00
+        DEPD,Z  %r5,%sar,64,%r5 ;offset 0xa04
+$0006001A
+        DEPDI,Z -1,31,32,%r19   ;offset 0xa08
+        AND     %r3,%r19,%r29   ;offset 0xa0c
+        EXTRD,U %r29,31,32,%r2  ;offset 0xa10
+        DEPDI,Z -1,63,32,%r6    ;offset 0xa14
+        MOVIB,TR        2,%r8,$0006001C ;offset 0xa18
+        EXTRD,U %r3,63,32,%r7   ;offset 0xa1c
+$D2
+        ADDIL   LR'__iob-$global$,%r27,%r1      ;offset 0xa20
+        LDIL    LR'C$7,%r21     ;offset 0xa24
+        LDO     RR'__iob-$global$+32(%r1),%r26  ;offset 0xa28
+        .CALL   ARGW0=GR,ARGW1=GR,ARGW2=GR,RTNVAL=GR    ;in=24,25,26;out=28;
+        B,L     fprintf,%r2     ;offset 0xa2c
+        LDO     RR'C$7(%r21),%r25       ;offset 0xa30
+        .CALL           ;
+        B,L     abort,%r2       ;offset 0xa34
+        NOP             ;offset 0xa38
+        B       $D3     ;offset 0xa3c
+        LDW     -212(%r30),%r2  ;offset 0xa40
+$00060020
+        COPY    %r4,%r26        ;offset 0xa44
+        EXTRD,U %r4,31,32,%r25  ;offset 0xa48
+        COPY    %r2,%r24        ;offset 0xa4c
+        .CALL   ;in=23,24,25,26;out=20,21,22,28,29; (MILLICALL)
+        B,L     $$div2U,%r31    ;offset 0xa50
+        EXTRD,U %r2,31,32,%r23  ;offset 0xa54
+        DEPD    %r28,31,32,%r29 ;offset 0xa58
+$00060022
+        STD     %r29,-152(%r30) ;offset 0xa5c
+$D1
+        AND     %r5,%r19,%r24   ;offset 0xa60
+        EXTRD,U %r24,31,32,%r24 ;offset 0xa64
+        STW     %r2,-160(%r30)  ;offset 0xa68
+        STW     %r7,-128(%r30)  ;offset 0xa6c
+        FLDD    -152(%r30),%fr4 ;offset 0xa70
+        FLDD    -152(%r30),%fr7 ;offset 0xa74
+        FLDW    -160(%r30),%fr8L        ;offset 0xa78
+        FLDW    -128(%r30),%fr5L        ;offset 0xa7c
+        XMPYU   %fr8L,%fr7L,%fr10       ;offset 0xa80
+        FSTD    %fr10,-136(%r30)        ;offset 0xa84
+        XMPYU   %fr8L,%fr7R,%fr22       ;offset 0xa88
+        FSTD    %fr22,-144(%r30)        ;offset 0xa8c
+        XMPYU   %fr5L,%fr4L,%fr11       ;offset 0xa90
+        XMPYU   %fr5L,%fr4R,%fr23       ;offset 0xa94
+        FSTD    %fr11,-112(%r30)        ;offset 0xa98
+        FSTD    %fr23,-120(%r30)        ;offset 0xa9c
+        LDD     -136(%r30),%r28 ;offset 0xaa0
+        DEPD,Z  %r28,31,32,%r31 ;offset 0xaa4
+        LDD     -144(%r30),%r20 ;offset 0xaa8
+        ADD,L   %r20,%r31,%r31  ;offset 0xaac
+        LDD     -112(%r30),%r22 ;offset 0xab0
+        DEPD,Z  %r22,31,32,%r22 ;offset 0xab4
+        LDD     -120(%r30),%r21 ;offset 0xab8
+        B       $00060024       ;offset 0xabc
+        ADD,L   %r21,%r22,%r23  ;offset 0xac0
+$D0
+        OR      %r9,%r29,%r29   ;offset 0xac4
+$00060040
+        EXTRD,U %r29,31,32,%r28 ;offset 0xac8
+$00060002
+$L2
+        LDW     -212(%r30),%r2  ;offset 0xacc
+$D3
+        LDW     -168(%r30),%r9  ;offset 0xad0
+        LDD     -176(%r30),%r8  ;offset 0xad4
+        EXTRD,U %r8,31,32,%r7   ;offset 0xad8
+        LDD     -184(%r30),%r6  ;offset 0xadc
+        EXTRD,U %r6,31,32,%r5   ;offset 0xae0
+        LDW     -188(%r30),%r4  ;offset 0xae4
+        BVE     (%r2)   ;offset 0xae8
+        .EXIT
+        LDW,MB  -192(%r30),%r3  ;offset 0xaec
+	.PROCEND	;in=23,25;out=28,29;fpin=105,107;
+
+
+
+
+;----------------------------------------------------------------------------
+;
+; Registers to hold 64-bit values to manipulate.  The "L" part
+; of the register corresponds to the upper 32-bits, while the "R"
+; part corresponds to the lower 32-bits
+; 
+; Note, that when using b6 and b7, the code must save these before
+; using them because they are callee save registers 
+; 
+;
+; Floating point registers to use to save values that
+; are manipulated.  These don't collide with ftemp1-6 and
+; are all caller save registers
+;
+a0        .reg %fr22
+a0L       .reg %fr22L
+a0R       .reg %fr22R
+
+a1        .reg %fr23
+a1L       .reg %fr23L
+a1R       .reg %fr23R
+
+a2        .reg %fr24
+a2L       .reg %fr24L
+a2R       .reg %fr24R
+
+a3        .reg %fr25
+a3L       .reg %fr25L
+a3R       .reg %fr25R
+
+a4        .reg %fr26
+a4L       .reg %fr26L
+a4R       .reg %fr26R
+
+a5        .reg %fr27
+a5L       .reg %fr27L
+a5R       .reg %fr27R
+
+a6        .reg %fr28
+a6L       .reg %fr28L
+a6R       .reg %fr28R
+
+a7        .reg %fr29
+a7L       .reg %fr29L
+a7R       .reg %fr29R
+
+b0        .reg %fr30
+b0L       .reg %fr30L
+b0R       .reg %fr30R
+
+b1        .reg %fr31
+b1L       .reg %fr31L
+b1R       .reg %fr31R
+
+;
+; Temporary floating point variables, these are all caller save
+; registers
+;
+ftemp1    .reg %fr4
+ftemp2    .reg %fr5
+ftemp3    .reg %fr6
+ftemp4    .reg %fr7
+
+;
+; The B set of registers when used.
+;
+
+b2        .reg %fr8
+b2L       .reg %fr8L
+b2R       .reg %fr8R
+
+b3        .reg %fr9
+b3L       .reg %fr9L
+b3R       .reg %fr9R
+
+b4        .reg %fr10
+b4L       .reg %fr10L
+b4R       .reg %fr10R
+
+b5        .reg %fr11
+b5L       .reg %fr11L
+b5R       .reg %fr11R
+
+b6        .reg %fr12
+b6L       .reg %fr12L
+b6R       .reg %fr12R
+
+b7        .reg %fr13
+b7L       .reg %fr13L
+b7R       .reg %fr13R
+
+c1           .reg %r21   ; only reg
+temp1        .reg %r20   ; only reg
+temp2        .reg %r19   ; only reg
+temp3        .reg %r31   ; only reg
+
+m1           .reg %r28   
+c2           .reg %r23   
+high_one     .reg %r1
+ht           .reg %r6
+lt           .reg %r5
+m            .reg %r4
+c3           .reg %r3
+
+SQR_ADD_C  .macro  A0L,A0R,C1,C2,C3
+    XMPYU   A0L,A0R,ftemp1       ; m
+    FSTD    ftemp1,-24(%sp)      ; store m
+
+    XMPYU   A0R,A0R,ftemp2       ; lt
+    FSTD    ftemp2,-16(%sp)      ; store lt
+
+    XMPYU   A0L,A0L,ftemp3       ; ht
+    FSTD    ftemp3,-8(%sp)       ; store ht
+
+    LDD     -24(%sp),m           ; load m
+    AND     m,high_mask,temp2    ; m & Mask
+    DEPD,Z  m,30,31,temp3        ; m << 32+1
+    LDD     -16(%sp),lt          ; lt
+
+    LDD     -8(%sp),ht           ; ht
+    EXTRD,U temp2,32,33,temp1    ; temp1 = m&Mask >> 32-1
+    ADD     temp3,lt,lt          ; lt = lt+m
+    ADD,L   ht,temp1,ht          ; ht += temp1
+    ADD,DC  ht,%r0,ht            ; ht++
+
+    ADD     C1,lt,C1             ; c1=c1+lt
+    ADD,DC  ht,%r0,ht            ; ht++
+
+    ADD     C2,ht,C2             ; c2=c2+ht
+    ADD,DC  C3,%r0,C3            ; c3++
+.endm
+
+SQR_ADD_C2 .macro  A0L,A0R,A1L,A1R,C1,C2,C3
+    XMPYU   A0L,A1R,ftemp1          ; m1 = bl*ht
+    FSTD    ftemp1,-16(%sp)         ;
+    XMPYU   A0R,A1L,ftemp2          ; m = bh*lt
+    FSTD    ftemp2,-8(%sp)          ;
+    XMPYU   A0R,A1R,ftemp3          ; lt = bl*lt
+    FSTD    ftemp3,-32(%sp)
+    XMPYU   A0L,A1L,ftemp4          ; ht = bh*ht
+    FSTD    ftemp4,-24(%sp)         ;
+
+    LDD     -8(%sp),m               ; r21 = m
+    LDD     -16(%sp),m1             ; r19 = m1
+    ADD,L   m,m1,m                  ; m+m1
+
+    DEPD,Z  m,31,32,temp3           ; (m+m1<<32)
+    LDD     -24(%sp),ht             ; r24 = ht
+
+    CMPCLR,*>>= m,m1,%r0            ; if (m < m1)
+    ADD,L   ht,high_one,ht          ; ht+=high_one
+
+    EXTRD,U m,31,32,temp1           ; m >> 32
+    LDD     -32(%sp),lt             ; lt
+    ADD,L   ht,temp1,ht             ; ht+= m>>32
+    ADD     lt,temp3,lt             ; lt = lt+m1
+    ADD,DC  ht,%r0,ht               ; ht++
+
+    ADD     ht,ht,ht                ; ht=ht+ht;
+    ADD,DC  C3,%r0,C3               ; add in carry (c3++)
+
+    ADD     lt,lt,lt                ; lt=lt+lt;
+    ADD,DC  ht,%r0,ht               ; add in carry (ht++)
+
+    ADD     C1,lt,C1                ; c1=c1+lt
+    ADD,DC,*NUV ht,%r0,ht           ; add in carry (ht++)
+    LDO     1(C3),C3              ; bump c3 if overflow,nullify otherwise
+
+    ADD     C2,ht,C2                ; c2 = c2 + ht
+    ADD,DC  C3,%r0,C3             ; add in carry (c3++)
+.endm
+
+;
+;void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+; arg0 = r_ptr
+; arg1 = a_ptr
+;
+
+bn_sqr_comba8
 	.PROC
-	.CALLINFO FRAME=128,CALLS,SAVE_RP,ENTRY_GR=8
-	.ENTRY
-	stw %r2,-20(0,%r30)
-	stwm %r8,128(0,%r30)
-	stw %r7,-124(0,%r30)
-	stw %r4,-112(0,%r30)
-	stw %r3,-108(0,%r30)
-	copy %r26,%r3
-	copy %r25,%r4
-	stw %r6,-120(0,%r30)
-	ldi 0,%r7
-	stw %r5,-116(0,%r30)
-	movb,<> %r24,%r5,L$0051
-	ldi 2,%r6
-	bl L$0068,0
-	ldi -1,%r28
-L$0051
-	.CALL ARGW0=GR
-	bl BN_num_bits_word,%r2
-	copy %r5,%r26
-	copy %r28,%r24
-	ldi 32,%r19
-	comb,= %r19,%r24,L$0052
-	subi 31,%r24,%r19
-	mtsar %r19
-	zvdepi 1,32,%r19
-	comb,>>= %r19,%r3,L$0052
-	addil LR'__iob-$global$+32,%r27
-	ldo RR'__iob-$global$+32(%r1),%r26
-	ldil LR'L$C0000,%r25
-	.CALL ARGW0=GR,ARGW1=GR,ARGW2=GR
-	bl fprintf,%r2
-	ldo RR'L$C0000(%r25),%r25
-	.CALL 
-	bl abort,%r2
-	nop
-L$0052
-	comb,>> %r5,%r3,L$0053
-	subi 32,%r24,%r24
-	sub %r3,%r5,%r3
-L$0053
-	comib,= 0,%r24,L$0054
-	subi 31,%r24,%r19
-	mtsar %r19
-	zvdep %r5,32,%r5
-	zvdep %r3,32,%r21
-	subi 32,%r24,%r20
-	mtsar %r20
-	vshd 0,%r4,%r20
-	or %r21,%r20,%r3
-	mtsar %r19
-	zvdep %r4,32,%r4
-L$0054
-	extru %r5,15,16,%r23
-	extru %r5,31,16,%r28
-L$0055
-	extru %r3,15,16,%r19
-	comb,<> %r23,%r19,L$0058
-	copy %r3,%r26
-	bl L$0059,0
-	zdepi -1,31,16,%r29
-L$0058
-	.IMPORT $$divU,MILLICODE
-	bl $$divU,%r31
-	copy %r23,%r25
-L$0059
-	stw %r29,-16(0,%r30)
-	fldws -16(0,%r30),%fr10L
-	stw %r28,-16(0,%r30)
-	fldws -16(0,%r30),%fr10R
-	stw %r23,-16(0,%r30)
-	xmpyu %fr10L,%fr10R,%fr8
-	fldws -16(0,%r30),%fr10R
-	fstws %fr8R,-16(0,%r30)
-	xmpyu %fr10L,%fr10R,%fr9
-	ldw -16(0,%r30),%r8
-	fstws %fr9R,-16(0,%r30)
-	copy %r8,%r22
-	ldw -16(0,%r30),%r8
-	extru %r4,15,16,%r24
-	copy %r8,%r21
-L$0060
-	sub %r3,%r21,%r20
-	copy %r20,%r19
-	depi 0,31,16,%r19
-	comib,<> 0,%r19,L$0061
-	zdep %r20,15,16,%r19
-	addl %r19,%r24,%r19
-	comb,>>= %r19,%r22,L$0061
-	sub %r22,%r28,%r22
-	sub %r21,%r23,%r21
-	bl L$0060,0
-	ldo -1(%r29),%r29
-L$0061
-	stw %r29,-16(0,%r30)
-	fldws -16(0,%r30),%fr10L
-	stw %r28,-16(0,%r30)
-	fldws -16(0,%r30),%fr10R
-	xmpyu %fr10L,%fr10R,%fr8
-	fstws %fr8R,-16(0,%r30)
-	ldw -16(0,%r30),%r8
-	stw %r23,-16(0,%r30)
-	fldws -16(0,%r30),%fr10R
-	copy %r8,%r19
-	xmpyu %fr10L,%fr10R,%fr8
-	fstws %fr8R,-16(0,%r30)
-	extru %r19,15,16,%r20
-	ldw -16(0,%r30),%r8
-	zdep %r19,15,16,%r19
-	addl %r8,%r20,%r20
-	comclr,<<= %r19,%r4,0
-	addi 1,%r20,%r20
-	comb,<<= %r20,%r3,L$0066
-	sub %r4,%r19,%r4
-	addl %r3,%r5,%r3
-	ldo -1(%r29),%r29
-L$0066
-	addib,= -1,%r6,L$0056
-	sub %r3,%r20,%r3
-	zdep %r29,15,16,%r7
-	shd %r3,%r4,16,%r3
-	bl L$0055,0
-	zdep %r4,15,16,%r4
-L$0056
-	or %r7,%r29,%r28
-L$0068
-	ldw -148(0,%r30),%r2
-	ldw -124(0,%r30),%r7
-	ldw -120(0,%r30),%r6
-	ldw -116(0,%r30),%r5
-	ldw -112(0,%r30),%r4
-	ldw -108(0,%r30),%r3
-	bv 0(%r2)
-	ldwm -128(0,%r30),%r8
-	.EXIT
-	.PROCEND
+	.CALLINFO FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
+	.EXPORT	bn_sqr_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .ENTRY
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3
+    STD     %r4,8(%sp)          ; save r4
+    STD     %r5,16(%sp)         ; save r5
+    STD     %r6,24(%sp)         ; save r6
+
+	;
+	; Zero out carries
+	;
+	COPY     %r0,c1
+	COPY     %r0,c2
+	COPY     %r0,c3
+
+	LDO      128(%sp),%sp       ; bump stack
+    DEPDI,Z -1,32,33,high_mask   ; Create Mask 0xffffffff80000000L
+    DEPDI,Z  1,31,1,high_one     ; Create Value  1 << 32
+
+	;
+	; Load up all of the values we are going to use
+	;
+    FLDD     0(a_ptr),a0       
+    FLDD     8(a_ptr),a1       
+    FLDD    16(a_ptr),a2       
+    FLDD    24(a_ptr),a3       
+    FLDD    32(a_ptr),a4       
+    FLDD    40(a_ptr),a5       
+    FLDD    48(a_ptr),a6       
+    FLDD    56(a_ptr),a7       
+
+	SQR_ADD_C a0L,a0R,c1,c2,c3
+	STD     c1,0(r_ptr)          ; r[0] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1
+	STD     c2,8(r_ptr)          ; r[1] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C a1L,a1R,c3,c1,c2
+	SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2
+	STD     c3,16(r_ptr)            ; r[2] = c3;
+	COPY    %r0,c3
+
+	SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3
+	SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3
+	STD     c1,24(r_ptr)           ; r[3] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C a2L,a2R,c2,c3,c1
+	SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1
+	SQR_ADD_C2 a4L,a4R,a0L,a0R,c2,c3,c1
+	STD     c2,32(r_ptr)          ; r[4] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C2 a5L,a5R,a0L,a0R,c3,c1,c2
+	SQR_ADD_C2 a4L,a4R,a1L,a1R,c3,c1,c2
+	SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2
+	STD     c3,40(r_ptr)          ; r[5] = c3;
+	COPY    %r0,c3
+
+	SQR_ADD_C a3L,a3R,c1,c2,c3
+	SQR_ADD_C2 a4L,a4R,a2L,a2R,c1,c2,c3
+	SQR_ADD_C2 a5L,a5R,a1L,a1R,c1,c2,c3
+	SQR_ADD_C2 a6L,a6R,a0L,a0R,c1,c2,c3
+	STD     c1,48(r_ptr)          ; r[6] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C2 a7L,a7R,a0L,a0R,c2,c3,c1
+	SQR_ADD_C2 a6L,a6R,a1L,a1R,c2,c3,c1
+	SQR_ADD_C2 a5L,a5R,a2L,a2R,c2,c3,c1
+	SQR_ADD_C2 a4L,a4R,a3L,a3R,c2,c3,c1
+	STD     c2,56(r_ptr)          ; r[7] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C a4L,a4R,c3,c1,c2
+	SQR_ADD_C2 a5L,a5R,a3L,a3R,c3,c1,c2
+	SQR_ADD_C2 a6L,a6R,a2L,a2R,c3,c1,c2
+	SQR_ADD_C2 a7L,a7R,a1L,a1R,c3,c1,c2
+	STD     c3,64(r_ptr)          ; r[8] = c3;
+	COPY    %r0,c3
+
+	SQR_ADD_C2 a7L,a7R,a2L,a2R,c1,c2,c3
+	SQR_ADD_C2 a6L,a6R,a3L,a3R,c1,c2,c3
+	SQR_ADD_C2 a5L,a5R,a4L,a4R,c1,c2,c3
+	STD     c1,72(r_ptr)          ; r[9] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C a5L,a5R,c2,c3,c1
+	SQR_ADD_C2 a6L,a6R,a4L,a4R,c2,c3,c1
+	SQR_ADD_C2 a7L,a7R,a3L,a3R,c2,c3,c1
+	STD     c2,80(r_ptr)          ; r[10] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C2 a7L,a7R,a4L,a4R,c3,c1,c2
+	SQR_ADD_C2 a6L,a6R,a5L,a5R,c3,c1,c2
+	STD     c3,88(r_ptr)          ; r[11] = c3;
+	COPY    %r0,c3
+	
+	SQR_ADD_C a6L,a6R,c1,c2,c3
+	SQR_ADD_C2 a7L,a7R,a5L,a5R,c1,c2,c3
+	STD     c1,96(r_ptr)          ; r[12] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C2 a7L,a7R,a6L,a6R,c2,c3,c1
+	STD     c2,104(r_ptr)         ; r[13] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C a7L,a7R,c3,c1,c2
+	STD     c3, 112(r_ptr)       ; r[14] = c3
+	STD     c1, 120(r_ptr)       ; r[15] = c1
+
+    .EXIT
+    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
+
+	.PROCEND	
+
+;-----------------------------------------------------------------------------
+;
+;void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+; arg0 = r_ptr
+; arg1 = a_ptr
+;
+
+bn_sqr_comba4
+	.proc
+	.callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
+	.EXPORT	bn_sqr_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .entry
+	.align 64
+    STD     %r3,0(%sp)          ; save r3
+    STD     %r4,8(%sp)          ; save r4
+    STD     %r5,16(%sp)         ; save r5
+    STD     %r6,24(%sp)         ; save r6
+
+	;
+	; Zero out carries
+	;
+	COPY     %r0,c1
+	COPY     %r0,c2
+	COPY     %r0,c3
+
+	LDO      128(%sp),%sp       ; bump stack
+    DEPDI,Z -1,32,33,high_mask   ; Create Mask 0xffffffff80000000L
+    DEPDI,Z  1,31,1,high_one     ; Create Value  1 << 32
+
+	;
+	; Load up all of the values we are going to use
+	;
+    FLDD     0(a_ptr),a0       
+    FLDD     8(a_ptr),a1       
+    FLDD    16(a_ptr),a2       
+    FLDD    24(a_ptr),a3       
+    FLDD    32(a_ptr),a4       
+    FLDD    40(a_ptr),a5       
+    FLDD    48(a_ptr),a6       
+    FLDD    56(a_ptr),a7       
+
+	SQR_ADD_C a0L,a0R,c1,c2,c3
+
+	STD     c1,0(r_ptr)          ; r[0] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C2 a1L,a1R,a0L,a0R,c2,c3,c1
+
+	STD     c2,8(r_ptr)          ; r[1] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C a1L,a1R,c3,c1,c2
+	SQR_ADD_C2 a2L,a2R,a0L,a0R,c3,c1,c2
+
+	STD     c3,16(r_ptr)            ; r[2] = c3;
+	COPY    %r0,c3
+
+	SQR_ADD_C2 a3L,a3R,a0L,a0R,c1,c2,c3
+	SQR_ADD_C2 a2L,a2R,a1L,a1R,c1,c2,c3
+
+	STD     c1,24(r_ptr)           ; r[3] = c1;
+	COPY    %r0,c1
+
+	SQR_ADD_C a2L,a2R,c2,c3,c1
+	SQR_ADD_C2 a3L,a3R,a1L,a1R,c2,c3,c1
+
+	STD     c2,32(r_ptr)           ; r[4] = c2;
+	COPY    %r0,c2
+
+	SQR_ADD_C2 a3L,a3R,a2L,a2R,c3,c1,c2
+	STD     c3,40(r_ptr)           ; r[5] = c3;
+	COPY    %r0,c3
+
+	SQR_ADD_C a3L,a3R,c1,c2,c3
+	STD     c1,48(r_ptr)           ; r[6] = c1;
+	STD     c2,56(r_ptr)           ; r[7] = c2;
+
+    .EXIT
+    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
+
+	.PROCEND	
+
+
+;---------------------------------------------------------------------------
+
+MUL_ADD_C  .macro  A0L,A0R,B0L,B0R,C1,C2,C3
+    XMPYU   A0L,B0R,ftemp1        ; m1 = bl*ht
+    FSTD    ftemp1,-16(%sp)       ;
+    XMPYU   A0R,B0L,ftemp2        ; m = bh*lt
+    FSTD    ftemp2,-8(%sp)        ;
+    XMPYU   A0R,B0R,ftemp3        ; lt = bl*lt
+    FSTD    ftemp3,-32(%sp)
+    XMPYU   A0L,B0L,ftemp4        ; ht = bh*ht
+    FSTD    ftemp4,-24(%sp)       ;
+
+    LDD     -8(%sp),m             ; r21 = m
+    LDD     -16(%sp),m1           ; r19 = m1
+    ADD,L   m,m1,m                ; m+m1
+
+    DEPD,Z  m,31,32,temp3         ; (m+m1<<32)
+    LDD     -24(%sp),ht           ; r24 = ht
+
+    CMPCLR,*>>= m,m1,%r0          ; if (m < m1)
+    ADD,L   ht,high_one,ht        ; ht+=high_one
+
+    EXTRD,U m,31,32,temp1         ; m >> 32
+    LDD     -32(%sp),lt           ; lt
+    ADD,L   ht,temp1,ht           ; ht+= m>>32
+    ADD     lt,temp3,lt           ; lt = lt+m1
+    ADD,DC  ht,%r0,ht             ; ht++
+
+    ADD     C1,lt,C1              ; c1=c1+lt
+    ADD,DC  ht,%r0,ht             ; bump c3 if overflow,nullify otherwise
+
+    ADD     C2,ht,C2              ; c2 = c2 + ht
+    ADD,DC  C3,%r0,C3             ; add in carry (c3++)
+.endm
+
+
+;
+;void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+; arg0 = r_ptr
+; arg1 = a_ptr
+; arg2 = b_ptr
+;
+
+bn_mul_comba8
+	.proc
+	.callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
+	.EXPORT	bn_mul_comba8,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .entry
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3
+    STD     %r4,8(%sp)          ; save r4
+    STD     %r5,16(%sp)         ; save r5
+    STD     %r6,24(%sp)         ; save r6
+    FSTD    %fr12,32(%sp)       ; save r6
+    FSTD    %fr13,40(%sp)       ; save r7
+
+	;
+	; Zero out carries
+	;
+	COPY     %r0,c1
+	COPY     %r0,c2
+	COPY     %r0,c3
+
+	LDO      128(%sp),%sp       ; bump stack
+    DEPDI,Z  1,31,1,high_one     ; Create Value  1 << 32
+
+	;
+	; Load up all of the values we are going to use
+	;
+    FLDD      0(a_ptr),a0       
+    FLDD      8(a_ptr),a1       
+    FLDD     16(a_ptr),a2       
+    FLDD     24(a_ptr),a3       
+    FLDD     32(a_ptr),a4       
+    FLDD     40(a_ptr),a5       
+    FLDD     48(a_ptr),a6       
+    FLDD     56(a_ptr),a7       
+
+    FLDD      0(b_ptr),b0       
+    FLDD      8(b_ptr),b1       
+    FLDD     16(b_ptr),b2       
+    FLDD     24(b_ptr),b3       
+    FLDD     32(b_ptr),b4       
+    FLDD     40(b_ptr),b5       
+    FLDD     48(b_ptr),b6       
+    FLDD     56(b_ptr),b7       
+
+	MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3
+	STD       c1,0(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1
+	MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1
+	STD       c2,8(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2
+	MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2
+	MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2
+	STD       c3,16(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3
+	MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3
+	MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3
+	MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3
+	STD       c1,24(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a4L,a4R,b0L,b0R,c2,c3,c1
+	MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1
+	MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1
+	MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1
+	MUL_ADD_C a0L,a0R,b4L,b4R,c2,c3,c1
+	STD       c2,32(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a0L,a0R,b5L,b5R,c3,c1,c2
+	MUL_ADD_C a1L,a1R,b4L,b4R,c3,c1,c2
+	MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2
+	MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2
+	MUL_ADD_C a4L,a4R,b1L,b1R,c3,c1,c2
+	MUL_ADD_C a5L,a5R,b0L,b0R,c3,c1,c2
+	STD       c3,40(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a6L,a6R,b0L,b0R,c1,c2,c3
+	MUL_ADD_C a5L,a5R,b1L,b1R,c1,c2,c3
+	MUL_ADD_C a4L,a4R,b2L,b2R,c1,c2,c3
+	MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3
+	MUL_ADD_C a2L,a2R,b4L,b4R,c1,c2,c3
+	MUL_ADD_C a1L,a1R,b5L,b5R,c1,c2,c3
+	MUL_ADD_C a0L,a0R,b6L,b6R,c1,c2,c3
+	STD       c1,48(r_ptr)
+	COPY      %r0,c1
+	
+	MUL_ADD_C a0L,a0R,b7L,b7R,c2,c3,c1
+	MUL_ADD_C a1L,a1R,b6L,b6R,c2,c3,c1
+	MUL_ADD_C a2L,a2R,b5L,b5R,c2,c3,c1
+	MUL_ADD_C a3L,a3R,b4L,b4R,c2,c3,c1
+	MUL_ADD_C a4L,a4R,b3L,b3R,c2,c3,c1
+	MUL_ADD_C a5L,a5R,b2L,b2R,c2,c3,c1
+	MUL_ADD_C a6L,a6R,b1L,b1R,c2,c3,c1
+	MUL_ADD_C a7L,a7R,b0L,b0R,c2,c3,c1
+	STD       c2,56(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a7L,a7R,b1L,b1R,c3,c1,c2
+	MUL_ADD_C a6L,a6R,b2L,b2R,c3,c1,c2
+	MUL_ADD_C a5L,a5R,b3L,b3R,c3,c1,c2
+	MUL_ADD_C a4L,a4R,b4L,b4R,c3,c1,c2
+	MUL_ADD_C a3L,a3R,b5L,b5R,c3,c1,c2
+	MUL_ADD_C a2L,a2R,b6L,b6R,c3,c1,c2
+	MUL_ADD_C a1L,a1R,b7L,b7R,c3,c1,c2
+	STD       c3,64(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a2L,a2R,b7L,b7R,c1,c2,c3
+	MUL_ADD_C a3L,a3R,b6L,b6R,c1,c2,c3
+	MUL_ADD_C a4L,a4R,b5L,b5R,c1,c2,c3
+	MUL_ADD_C a5L,a5R,b4L,b4R,c1,c2,c3
+	MUL_ADD_C a6L,a6R,b3L,b3R,c1,c2,c3
+	MUL_ADD_C a7L,a7R,b2L,b2R,c1,c2,c3
+	STD       c1,72(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a7L,a7R,b3L,b3R,c2,c3,c1
+	MUL_ADD_C a6L,a6R,b4L,b4R,c2,c3,c1
+	MUL_ADD_C a5L,a5R,b5L,b5R,c2,c3,c1
+	MUL_ADD_C a4L,a4R,b6L,b6R,c2,c3,c1
+	MUL_ADD_C a3L,a3R,b7L,b7R,c2,c3,c1
+	STD       c2,80(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a4L,a4R,b7L,b7R,c3,c1,c2
+	MUL_ADD_C a5L,a5R,b6L,b6R,c3,c1,c2
+	MUL_ADD_C a6L,a6R,b5L,b5R,c3,c1,c2
+	MUL_ADD_C a7L,a7R,b4L,b4R,c3,c1,c2
+	STD       c3,88(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a7L,a7R,b5L,b5R,c1,c2,c3
+	MUL_ADD_C a6L,a6R,b6L,b6R,c1,c2,c3
+	MUL_ADD_C a5L,a5R,b7L,b7R,c1,c2,c3
+	STD       c1,96(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a6L,a6R,b7L,b7R,c2,c3,c1
+	MUL_ADD_C a7L,a7R,b6L,b6R,c2,c3,c1
+	STD       c2,104(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a7L,a7R,b7L,b7R,c3,c1,c2
+	STD       c3,112(r_ptr)
+	STD       c1,120(r_ptr)
+
+    .EXIT
+    FLDD    -88(%sp),%fr13 
+    FLDD    -96(%sp),%fr12 
+    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
+
+	.PROCEND	
+
+;-----------------------------------------------------------------------------
+;
+;void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+; arg0 = r_ptr
+; arg1 = a_ptr
+; arg2 = b_ptr
+;
+
+bn_mul_comba4
+	.proc
+	.callinfo FRAME=128,ENTRY_GR=%r3,ARGS_SAVED,ORDERING_AWARE
+	.EXPORT	bn_mul_comba4,ENTRY,PRIV_LEV=3,NO_RELOCATION,LONG_RETURN
+    .entry
+	.align 64
+
+    STD     %r3,0(%sp)          ; save r3
+    STD     %r4,8(%sp)          ; save r4
+    STD     %r5,16(%sp)         ; save r5
+    STD     %r6,24(%sp)         ; save r6
+    FSTD    %fr12,32(%sp)       ; save r6
+    FSTD    %fr13,40(%sp)       ; save r7
+
+	;
+	; Zero out carries
+	;
+	COPY     %r0,c1
+	COPY     %r0,c2
+	COPY     %r0,c3
+
+	LDO      128(%sp),%sp       ; bump stack
+    DEPDI,Z  1,31,1,high_one     ; Create Value  1 << 32
+
+	;
+	; Load up all of the values we are going to use
+	;
+    FLDD      0(a_ptr),a0       
+    FLDD      8(a_ptr),a1       
+    FLDD     16(a_ptr),a2       
+    FLDD     24(a_ptr),a3       
+
+    FLDD      0(b_ptr),b0       
+    FLDD      8(b_ptr),b1       
+    FLDD     16(b_ptr),b2       
+    FLDD     24(b_ptr),b3       
+
+	MUL_ADD_C a0L,a0R,b0L,b0R,c1,c2,c3
+	STD       c1,0(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a0L,a0R,b1L,b1R,c2,c3,c1
+	MUL_ADD_C a1L,a1R,b0L,b0R,c2,c3,c1
+	STD       c2,8(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a2L,a2R,b0L,b0R,c3,c1,c2
+	MUL_ADD_C a1L,a1R,b1L,b1R,c3,c1,c2
+	MUL_ADD_C a0L,a0R,b2L,b2R,c3,c1,c2
+	STD       c3,16(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a0L,a0R,b3L,b3R,c1,c2,c3
+	MUL_ADD_C a1L,a1R,b2L,b2R,c1,c2,c3
+	MUL_ADD_C a2L,a2R,b1L,b1R,c1,c2,c3
+	MUL_ADD_C a3L,a3R,b0L,b0R,c1,c2,c3
+	STD       c1,24(r_ptr)
+	COPY      %r0,c1
+
+	MUL_ADD_C a3L,a3R,b1L,b1R,c2,c3,c1
+	MUL_ADD_C a2L,a2R,b2L,b2R,c2,c3,c1
+	MUL_ADD_C a1L,a1R,b3L,b3R,c2,c3,c1
+	STD       c2,32(r_ptr)
+	COPY      %r0,c2
+
+	MUL_ADD_C a2L,a2R,b3L,b3R,c3,c1,c2
+	MUL_ADD_C a3L,a3R,b2L,b2R,c3,c1,c2
+	STD       c3,40(r_ptr)
+	COPY      %r0,c3
+
+	MUL_ADD_C a3L,a3R,b3L,b3R,c1,c2,c3
+	STD       c1,48(r_ptr)
+	STD       c2,56(r_ptr)
+
+    .EXIT
+    FLDD    -88(%sp),%fr13 
+    FLDD    -96(%sp),%fr12 
+    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
+
+	.PROCEND	
+
+
+	.SPACE	$TEXT$
+	.SUBSPA	$CODE$
+	.SPACE	$PRIVATE$,SORT=16
+	.IMPORT	$global$,DATA
+	.SPACE	$TEXT$
+	.SUBSPA	$CODE$
+	.SUBSPA	$LIT$,QUAD=0,ALIGN=8,ACCESS=0x2c,SORT=16
+C$7
+	.ALIGN	8
+	.STRINGZ	"Division would overflow (%d)\n"
+	.END
diff --git a/lib/libcrypto/bn/bn.h b/lib/libcrypto/bn/bn.h
index 009b0eb6856..1eb8395b25c 100644
--- a/lib/libcrypto/bn/bn.h
+++ b/lib/libcrypto/bn/bn.h
@@ -59,7 +59,7 @@
 #ifndef HEADER_BN_H
 #define HEADER_BN_H
 
-#ifndef WIN16
+#ifndef NO_FP_API
 #include <stdio.h> /* FILE */
 #endif
 #include <openssl/opensslconf.h>
@@ -233,7 +233,7 @@ typedef struct bignum_st
 	BN_ULONG *d;	/* Pointer to an array of 'BN_BITS2' bit chunks. */
 	int top;	/* Index of last used d +1. */
 	/* The next are internal book keeping for bn_expand. */
-	int max;	/* Size of the d array. */
+	int dmax;	/* Size of the d array. */
 	int neg;	/* one if the number is negative */
 	int flags;
 	} BIGNUM;
@@ -364,6 +364,8 @@ int	BN_mod_exp(BIGNUM *r, BIGNUM *a, const BIGNUM *p,
 		   const BIGNUM *m,BN_CTX *ctx);
 int	BN_mod_exp_mont(BIGNUM *r, BIGNUM *a, const BIGNUM *p,
 			const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *m_ctx);
+int	BN_mod_exp_mont_word(BIGNUM *r, BN_ULONG a, const BIGNUM *p,
+			const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *m_ctx);
 int	BN_mod_exp2_mont(BIGNUM *r, BIGNUM *a1, BIGNUM *p1,BIGNUM *a2,
 		BIGNUM *p2,BIGNUM *m,BN_CTX *ctx,BN_MONT_CTX *m_ctx);
 int	BN_mod_exp_simple(BIGNUM *r, BIGNUM *a, BIGNUM *p,
@@ -433,9 +435,9 @@ int	BN_div_recp(BIGNUM *dv, BIGNUM *rem, BIGNUM *m,
 
 /* library internal functions */
 
-#define bn_expand(a,bits) ((((((bits+BN_BITS2-1))/BN_BITS2)) <= (a)->max)?\
+#define bn_expand(a,bits) ((((((bits+BN_BITS2-1))/BN_BITS2)) <= (a)->dmax)?\
 	(a):bn_expand2((a),(bits)/BN_BITS2+1))
-#define bn_wexpand(a,words) (((words) <= (a)->max)?(a):bn_expand2((a),(words)))
+#define bn_wexpand(a,words) (((words) <= (a)->dmax)?(a):bn_expand2((a),(words)))
 BIGNUM *bn_expand2(BIGNUM *a, int words);
 
 #define bn_fix_top(a) \
@@ -483,7 +485,9 @@ BN_ULONG bn_sub_words(BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int num);
 #define BN_F_BN_CTX_NEW					 106
 #define BN_F_BN_DIV					 107
 #define BN_F_BN_EXPAND2					 108
+#define BN_F_BN_MOD_EXP2_MONT				 118
 #define BN_F_BN_MOD_EXP_MONT				 109
+#define BN_F_BN_MOD_EXP_MONT_WORD			 117
 #define BN_F_BN_MOD_INVERSE				 110
 #define BN_F_BN_MOD_MUL_RECIPROCAL			 111
 #define BN_F_BN_MPI2BN					 112
diff --git a/lib/libcrypto/bn/bn_asm.c b/lib/libcrypto/bn/bn_asm.c
index 3329cc18e6c..44e52a40db8 100644
--- a/lib/libcrypto/bn/bn_asm.c
+++ b/lib/libcrypto/bn/bn_asm.c
@@ -227,7 +227,7 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 
 #else
 
-/* Divide h-l by d and return the result. */
+/* Divide h,l by d and return the result. */
 /* I need to test this some more :-( */
 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 	{
@@ -237,13 +237,8 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 	if (d == 0) return(BN_MASK2);
 
 	i=BN_num_bits_word(d);
-	if ((i != BN_BITS2) && (h > (BN_ULONG)1<<i))
-		{
-#if !defined(NO_STDIO) && !defined(WIN16)
-		fprintf(stderr,"Division would overflow (%d)\n",i);
-#endif
-		abort();
-		}
+	assert((i == BN_BITS2) || (h > (BN_ULONG)1<<i));
+
 	i=BN_BITS2-i;
 	if (h >= d) h-=d;
 
diff --git a/lib/libcrypto/bn/bn_blind.c b/lib/libcrypto/bn/bn_blind.c
index 1b1bb060463..2d287e6d1bb 100644
--- a/lib/libcrypto/bn/bn_blind.c
+++ b/lib/libcrypto/bn/bn_blind.c
@@ -67,7 +67,7 @@ BN_BLINDING *BN_BLINDING_new(BIGNUM *A, BIGNUM *Ai, BIGNUM *mod)
 	bn_check_top(Ai);
 	bn_check_top(mod);
 
-	if ((ret=(BN_BLINDING *)Malloc(sizeof(BN_BLINDING))) == NULL)
+	if ((ret=(BN_BLINDING *)OPENSSL_malloc(sizeof(BN_BLINDING))) == NULL)
 		{
 		BNerr(BN_F_BN_BLINDING_NEW,ERR_R_MALLOC_FAILURE);
 		return(NULL);
@@ -91,7 +91,7 @@ void BN_BLINDING_free(BN_BLINDING *r)
 
 	if (r->A  != NULL) BN_free(r->A );
 	if (r->Ai != NULL) BN_free(r->Ai);
-	Free(r);
+	OPENSSL_free(r);
 	}
 
 int BN_BLINDING_update(BN_BLINDING *b, BN_CTX *ctx)
diff --git a/lib/libcrypto/bn/bn_ctx.c b/lib/libcrypto/bn/bn_ctx.c
index 46132fd1806..b1a8d7571e8 100644
--- a/lib/libcrypto/bn/bn_ctx.c
+++ b/lib/libcrypto/bn/bn_ctx.c
@@ -69,7 +69,7 @@ BN_CTX *BN_CTX_new(void)
 	{
 	BN_CTX *ret;
 
-	ret=(BN_CTX *)Malloc(sizeof(BN_CTX));
+	ret=(BN_CTX *)OPENSSL_malloc(sizeof(BN_CTX));
 	if (ret == NULL)
 		{
 		BNerr(BN_F_BN_CTX_NEW,ERR_R_MALLOC_FAILURE);
@@ -102,7 +102,7 @@ void BN_CTX_free(BN_CTX *ctx)
 	for (i=0; i < BN_CTX_NUM; i++)
 		BN_clear_free(&(ctx->bn[i]));
 	if (ctx->flags & BN_FLG_MALLOCED)
-		Free(ctx);
+		OPENSSL_free(ctx);
 	}
 
 void BN_CTX_start(BN_CTX *ctx)
diff --git a/lib/libcrypto/bn/bn_div.c b/lib/libcrypto/bn/bn_div.c
index 07af1d3b449..c3772c243be 100644
--- a/lib/libcrypto/bn/bn_div.c
+++ b/lib/libcrypto/bn/bn_div.c
@@ -205,7 +205,7 @@ int BN_div(BIGNUM *dv, BIGNUM *rm, const BIGNUM *num, const BIGNUM *divisor,
 	BN_init(&wnum);
 	wnum.d=	 &(snum->d[loop]);
 	wnum.top= div_n;
-	wnum.max= snum->max+1; /* a bit of a lie */
+	wnum.dmax= snum->dmax+1; /* a bit of a lie */
 
 	/* Get the top 2 words of sdiv */
 	/* i=sdiv->top; */
diff --git a/lib/libcrypto/bn/bn_err.c b/lib/libcrypto/bn/bn_err.c
index 988270bcf4f..86550c4c21d 100644
--- a/lib/libcrypto/bn/bn_err.c
+++ b/lib/libcrypto/bn/bn_err.c
@@ -76,7 +76,9 @@ static ERR_STRING_DATA BN_str_functs[]=
 {ERR_PACK(0,BN_F_BN_CTX_NEW,0),	"BN_CTX_new"},
 {ERR_PACK(0,BN_F_BN_DIV,0),	"BN_div"},
 {ERR_PACK(0,BN_F_BN_EXPAND2,0),	"bn_expand2"},
+{ERR_PACK(0,BN_F_BN_MOD_EXP2_MONT,0),	"BN_mod_exp2_mont"},
 {ERR_PACK(0,BN_F_BN_MOD_EXP_MONT,0),	"BN_mod_exp_mont"},
+{ERR_PACK(0,BN_F_BN_MOD_EXP_MONT_WORD,0),	"BN_mod_exp_mont_word"},
 {ERR_PACK(0,BN_F_BN_MOD_INVERSE,0),	"BN_mod_inverse"},
 {ERR_PACK(0,BN_F_BN_MOD_MUL_RECIPROCAL,0),	"BN_mod_mul_reciprocal"},
 {ERR_PACK(0,BN_F_BN_MPI2BN,0),	"BN_mpi2bn"},
diff --git a/lib/libcrypto/bn/bn_exp.c b/lib/libcrypto/bn/bn_exp.c
index 0c116016754..d2c91628acb 100644
--- a/lib/libcrypto/bn/bn_exp.c
+++ b/lib/libcrypto/bn/bn_exp.c
@@ -55,18 +55,66 @@
  * copied and put under another distribution licence
  * [including the GNU Public Licence.]
  */
+/* ====================================================================
+ * Copyright (c) 1998-2000 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
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer. 
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ *    software must display the following acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ *    endorse or promote products derived from this software without
+ *    prior written permission. For written permission, please contact
+ *    openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ *    nor may "OpenSSL" appear in their names without prior written
+ *    permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ *    acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * 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).
+ *
+ */
+
 
 #include <stdio.h>
 #include "cryptlib.h"
 #include "bn_lcl.h"
-#ifdef ATALLA
-# include <alloca.h>
-# include <atasi.h>
-# include <assert.h>
-# include <dlfcn.h>
-#endif
 
-#define TABLE_SIZE	16
+#define TABLE_SIZE	32
 
 /* slow but works */
 int BN_mod_mul(BIGNUM *ret, BIGNUM *a, BIGNUM *b, const BIGNUM *m, BN_CTX *ctx)
@@ -91,42 +139,6 @@ err:
 	return(r);
 	}
 
-#if 0
-/* this one works - simple but works */
-int BN_mod_exp(BIGNUM *r, BIGNUM *a, BIGNUM *p, BIGNUM *m, BN_CTX *ctx)
-	{
-	int i,bits,ret=0;
-	BIGNUM *v,*tmp;
-
-	BN_CTX_start(ctx);
-	v = BN_CTX_get(ctx);
-	tmp = BN_CTX_get(ctx);
-	if (v == NULL || tmp == NULL) goto err;
-
-	if (BN_copy(v,a) == NULL) goto err;
-	bits=BN_num_bits(p);
-
-	if (BN_is_odd(p))
-		{ if (BN_copy(r,a) == NULL) goto err; }
-	else	{ if (!BN_one(r)) goto err; }
-
-	for (i=1; i<bits; i++)
-		{
-		if (!BN_sqr(tmp,v,ctx)) goto err;
-		if (!BN_mod(v,tmp,m,ctx)) goto err;
-		if (BN_is_bit_set(p,i))
-			{
-			if (!BN_mul(tmp,r,v,ctx)) goto err;
-			if (!BN_mod(r,tmp,m,ctx)) goto err;
-			}
-		}
-	ret=1;
-err:
-	BN_CTX_end(ctx);
-	return(ret);
-	}
-
-#endif
 
 /* this one works - simple but works */
 int BN_exp(BIGNUM *r, BIGNUM *a, BIGNUM *p, BN_CTX *ctx)
@@ -163,172 +175,6 @@ err:
 	return(ret);
 	}
 
-#ifdef ATALLA
-
-/*
- * This routine will dynamically check for the existance of an Atalla AXL-200
- * SSL accelerator module.  If one is found, the variable
- * asi_accelerator_present is set to 1 and the function pointers
- * ptr_ASI_xxxxxx above will be initialized to corresponding ASI API calls.
- */
-typedef int tfnASI_GetPerformanceStatistics(int reset_flag,
-					    unsigned int *ret_buf);
-typedef int tfnASI_GetHardwareConfig(long card_num, unsigned int *ret_buf);
-typedef int tfnASI_RSAPrivateKeyOpFn(RSAPrivateKey * rsaKey,
-				     unsigned char *output,
-				     unsigned char *input,
-				     unsigned int modulus_len);
-
-static tfnASI_GetHardwareConfig *ptr_ASI_GetHardwareConfig;
-static tfnASI_RSAPrivateKeyOpFn *ptr_ASI_RSAPrivateKeyOpFn;
-static tfnASI_GetPerformanceStatistics *ptr_ASI_GetPerformanceStatistics;
-static int asi_accelerator_present;
-static int tried_atalla;
-
-void atalla_initialize_accelerator_handle(void)
-	{
-	void *dl_handle;
-	int status;
-	unsigned int config_buf[1024]; 
-	static int tested;
-
-	if(tested)
-		return;
-
-	tested=1;
-
-	bzero((void *)config_buf, 1024);
-
-	/*
-	 * Check to see if the library is present on the system
-	 */
-	dl_handle = dlopen("atasi.so", RTLD_NOW);
-	if (dl_handle == (void *) NULL)
-		{
-/*		printf("atasi.so library is not present on the system\n");
-		printf("No HW acceleration available\n");*/
-		return;
-	        }
-
-	/*
-	 * The library is present.  Now we'll check to insure that the
-	 * LDM is up and running. First we'll get the address of the
-	 * function in the atasi library that we need to see if the
-	 * LDM is operating.
-	 */
-
-	ptr_ASI_GetHardwareConfig =
-	  (tfnASI_GetHardwareConfig *)dlsym(dl_handle,"ASI_GetHardwareConfig");
-
-	if (ptr_ASI_GetHardwareConfig)
-		{
-		/*
-		 * We found the call, now we'll get our config
-		 * status.  If we get a non 0 result, the LDM is not
-		 * running and we cannot use the Atalla ASI *
-		 * library.
-		 */
-		status = (*ptr_ASI_GetHardwareConfig)(0L, config_buf);
-		if (status != 0)
-			{
-			printf("atasi.so library is present but not initialized\n");
-			printf("No HW acceleration available\n");
-			return;
-			}    
-	        }
-	else
-		{
-/*		printf("We found the library, but not the function. Very Strange!\n");*/
-		return ;
-	      	}
-
-	/* 
-	 * It looks like we have acceleration capabilities.  Load up the
-	 * pointers to our ASI API calls.
-	 */
-	ptr_ASI_RSAPrivateKeyOpFn=
-	  (tfnASI_RSAPrivateKeyOpFn *)dlsym(dl_handle, "ASI_RSAPrivateKeyOpFn");
-	if (ptr_ASI_RSAPrivateKeyOpFn == NULL)
-		{
-/*		printf("We found the library, but no RSA function. Very Strange!\n");*/
-		return;
-	        }
-
-	ptr_ASI_GetPerformanceStatistics =
-	  (tfnASI_GetPerformanceStatistics *)dlsym(dl_handle, "ASI_GetPerformanceStatistics");
-	if (ptr_ASI_GetPerformanceStatistics == NULL)
-		{
-/*		printf("We found the library, but no stat function. Very Strange!\n");*/
-		return;
-	      }
-
-	/*
-	 * Indicate that acceleration is available
-	 */
-	asi_accelerator_present = 1;
-
-/*	printf("This system has acceleration!\n");*/
-
-	return;
-	}
-
-/* make sure this only gets called once when bn_mod_exp calls bn_mod_exp_mont */
-int BN_mod_exp_atalla(BIGNUM *r, BIGNUM *a, const BIGNUM *p, const BIGNUM *m)
-	{
-	unsigned char *abin;
-	unsigned char *pbin;
-	unsigned char *mbin;
-	unsigned char *rbin;
-	int an,pn,mn,ret;
-	RSAPrivateKey keydata;
-
-	atalla_initialize_accelerator_handle();
-	if(!asi_accelerator_present)
-		return 0;
-
-
-/* We should be able to run without size testing */
-# define ASIZE	128
-	an=BN_num_bytes(a);
-	pn=BN_num_bytes(p);
-	mn=BN_num_bytes(m);
-
-	if(an <= ASIZE && pn <= ASIZE && mn <= ASIZE)
-	    {
-	    int size=mn;
-
-	    assert(an <= mn);
-	    abin=alloca(size);
-	    memset(abin,'\0',mn);
-	    BN_bn2bin(a,abin+size-an);
-
-	    pbin=alloca(pn);
-	    BN_bn2bin(p,pbin);
-
-	    mbin=alloca(size);
-	    memset(mbin,'\0',mn);
-	    BN_bn2bin(m,mbin+size-mn);
-
-	    rbin=alloca(size);
-
-	    memset(&keydata,'\0',sizeof keydata);
-	    keydata.privateExponent.data=pbin;
-	    keydata.privateExponent.len=pn;
-	    keydata.modulus.data=mbin;
-	    keydata.modulus.len=size;
-
-	    ret=(*ptr_ASI_RSAPrivateKeyOpFn)(&keydata,rbin,abin,keydata.modulus.len);
-/*fprintf(stderr,"!%s\n",BN_bn2hex(a));*/
-	    if(!ret)
-	        {
-		BN_bin2bn(rbin,keydata.modulus.len,r);
-/*fprintf(stderr,"?%s\n",BN_bn2hex(r));*/
-		return 1;
-	        }
-	    }
-	return 0;
-        }
-#endif /* def ATALLA */
 
 int BN_mod_exp(BIGNUM *r, BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
 	       BN_CTX *ctx)
@@ -339,13 +185,6 @@ int BN_mod_exp(BIGNUM *r, BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
 	bn_check_top(p);
 	bn_check_top(m);
 
-#ifdef ATALLA
-	if(BN_mod_exp_atalla(r,a,p,m))
-	    return 1;
-/* If it fails, try the other methods (but don't try atalla again) */
-	tried_atalla=1;
-#endif
-
 #ifdef MONT_MUL_MOD
 	/* I have finally been able to take out this pre-condition of
 	 * the top bit being set.  It was caused by an error in BN_div
@@ -354,7 +193,15 @@ int BN_mod_exp(BIGNUM *r, BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
 /*	if ((m->d[m->top-1]&BN_TBIT) && BN_is_odd(m)) */
 
 	if (BN_is_odd(m))
-		{ ret=BN_mod_exp_mont(r,a,p,m,ctx,NULL); }
+		{
+		if (a->top == 1)
+			{
+			BN_ULONG A = a->d[0];
+			ret=BN_mod_exp_mont_word(r,A,p,m,ctx,NULL);
+			}
+		else
+			ret=BN_mod_exp_mont(r,a,p,m,ctx,NULL);
+		}
 	else
 #endif
 #ifdef RECP_MUL_MOD
@@ -363,14 +210,10 @@ int BN_mod_exp(BIGNUM *r, BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
 		{ ret=BN_mod_exp_simple(r,a,p,m,ctx); }
 #endif
 
-#ifdef ATALLA
-	tried_atalla=0;
-#endif
-
 	return(ret);
 	}
 
-/* #ifdef RECP_MUL_MOD */
+
 int BN_mod_exp_recp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
 		    const BIGNUM *m, BN_CTX *ctx)
 	{
@@ -398,27 +241,22 @@ int BN_mod_exp_recp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
 	ts=1;
 
 	if (!BN_mod(&(val[0]),a,m,ctx)) goto err;		/* 1 */
-	if (!BN_mod_mul_reciprocal(aa,&(val[0]),&(val[0]),&recp,ctx))
-		goto err;				/* 2 */
-
-	if (bits <= 17) /* This is probably 3 or 0x10001, so just do singles */
-		window=1;
-	else if (bits >= 256)
-		window=5;	/* max size of window */
-	else if (bits >= 128)
-		window=4;
-	else
-		window=3;
 
-	j=1<<(window-1);
-	for (i=1; i<j; i++)
+	window = BN_window_bits_for_exponent_size(bits);
+	if (window > 1)
 		{
-		BN_init(&val[i]);
-		if (!BN_mod_mul_reciprocal(&(val[i]),&(val[i-1]),aa,&recp,ctx))
-			goto err;
+		if (!BN_mod_mul_reciprocal(aa,&(val[0]),&(val[0]),&recp,ctx))
+			goto err;				/* 2 */
+		j=1<<(window-1);
+		for (i=1; i<j; i++)
+			{
+			BN_init(&val[i]);
+			if (!BN_mod_mul_reciprocal(&(val[i]),&(val[i-1]),aa,&recp,ctx))
+				goto err;
+			}
+		ts=i;
 		}
-	ts=i;
-
+		
 	start=1;	/* This is used to avoid multiplication etc
 			 * when there is only the value '1' in the
 			 * buffer. */
@@ -485,9 +323,8 @@ err:
 	BN_RECP_CTX_free(&recp);
 	return(ret);
 	}
-/* #endif */
 
-/* #ifdef MONT_MUL_MOD */
+
 int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 		    const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
 	{
@@ -502,12 +339,6 @@ int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 	bn_check_top(p);
 	bn_check_top(m);
 
-#ifdef ATALLA
-	if(!tried_atalla && BN_mod_exp_atalla(rr,a,p,m))
-	    return 1;
-/* If it fails, try the other methods */
-#endif
-
 	if (!(m->d[0] & 1))
 		{
 		BNerr(BN_F_BN_MOD_EXP_MONT,BN_R_CALLED_WITH_EVEN_MODULUS);
@@ -527,11 +358,9 @@ int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 	/* If this is not done, things will break in the montgomery
 	 * part */
 
-#if 1
 	if (in_mont != NULL)
 		mont=in_mont;
 	else
-#endif
 		{
 		if ((mont=BN_MONT_CTX_new()) == NULL) goto err;
 		if (!BN_MONT_CTX_set(mont,m,ctx)) goto err;
@@ -541,31 +370,27 @@ int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 	ts=1;
 	if (BN_ucmp(a,m) >= 0)
 		{
-		BN_mod(&(val[0]),a,m,ctx);
+		if (!BN_mod(&(val[0]),a,m,ctx))
+			goto err;
 		aa= &(val[0]);
 		}
 	else
 		aa=a;
 	if (!BN_to_montgomery(&(val[0]),aa,mont,ctx)) goto err; /* 1 */
-	if (!BN_mod_mul_montgomery(d,&(val[0]),&(val[0]),mont,ctx)) goto err; /* 2 */
-
-	if (bits <= 20) /* This is probably 3 or 0x10001, so just do singles */
-		window=1;
-	else if (bits >= 256)
-		window=5;	/* max size of window */
-	else if (bits >= 128)
-		window=4;
-	else
-		window=3;
 
-	j=1<<(window-1);
-	for (i=1; i<j; i++)
+	window = BN_window_bits_for_exponent_size(bits);
+	if (window > 1)
 		{
-		BN_init(&(val[i]));
-		if (!BN_mod_mul_montgomery(&(val[i]),&(val[i-1]),d,mont,ctx))
-			goto err;
+		if (!BN_mod_mul_montgomery(d,&(val[0]),&(val[0]),mont,ctx)) goto err; /* 2 */
+		j=1<<(window-1);
+		for (i=1; i<j; i++)
+			{
+			BN_init(&(val[i]));
+			if (!BN_mod_mul_montgomery(&(val[i]),&(val[i-1]),d,mont,ctx))
+				goto err;
+			}
+		ts=i;
 		}
-	ts=i;
 
 	start=1;	/* This is used to avoid multiplication etc
 			 * when there is only the value '1' in the
@@ -574,7 +399,7 @@ int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 	wstart=bits-1;	/* The top bit of the window */
 	wend=0;		/* The bottom bit of the window */
 
-        if (!BN_to_montgomery(r,BN_value_one(),mont,ctx)) goto err;
+	if (!BN_to_montgomery(r,BN_value_one(),mont,ctx)) goto err;
 	for (;;)
 		{
 		if (BN_is_bit_set(p,wstart) == 0)
@@ -626,7 +451,7 @@ int BN_mod_exp_mont(BIGNUM *rr, BIGNUM *a, const BIGNUM *p,
 		start=0;
 		if (wstart < 0) break;
 		}
-	BN_from_montgomery(rr,r,mont,ctx);
+	if (!BN_from_montgomery(rr,r,mont,ctx)) goto err;
 	ret=1;
 err:
 	if ((in_mont == NULL) && (mont != NULL)) BN_MONT_CTX_free(mont);
@@ -635,7 +460,134 @@ err:
 		BN_clear_free(&(val[i]));
 	return(ret);
 	}
-/* #endif */
+
+int BN_mod_exp_mont_word(BIGNUM *rr, BN_ULONG a, const BIGNUM *p,
+                         const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
+	{
+	BN_MONT_CTX *mont = NULL;
+	int b, bits, ret=0;
+	int r_is_one;
+	BN_ULONG w, next_w;
+	BIGNUM *d, *r, *t;
+	BIGNUM *swap_tmp;
+#define BN_MOD_MUL_WORD(r, w, m) \
+		(BN_mul_word(r, (w)) && \
+		(/* BN_ucmp(r, (m)) < 0 ? 1 :*/  \
+			(BN_mod(t, r, m, ctx) && (swap_tmp = r, r = t, t = swap_tmp, 1))))
+		/* BN_MOD_MUL_WORD is only used with 'w' large,
+		  * so the BN_ucmp test is probably more overhead
+		  * than always using BN_mod (which uses BN_copy if
+		  * a similar test returns true). */
+#define BN_TO_MONTGOMERY_WORD(r, w, mont) \
+		(BN_set_word(r, (w)) && BN_to_montgomery(r, r, (mont), ctx))
+
+	bn_check_top(p);
+	bn_check_top(m);
+
+	if (!(m->d[0] & 1))
+		{
+		BNerr(BN_F_BN_MOD_EXP_MONT_WORD,BN_R_CALLED_WITH_EVEN_MODULUS);
+		return(0);
+		}
+	bits = BN_num_bits(p);
+	if (bits == 0)
+		{
+		BN_one(rr);
+		return(1);
+		}
+	BN_CTX_start(ctx);
+	d = BN_CTX_get(ctx);
+	r = BN_CTX_get(ctx);
+	t = BN_CTX_get(ctx);
+	if (d == NULL || r == NULL || t == NULL) goto err;
+
+	if (in_mont != NULL)
+		mont=in_mont;
+	else
+		{
+		if ((mont = BN_MONT_CTX_new()) == NULL) goto err;
+		if (!BN_MONT_CTX_set(mont, m, ctx)) goto err;
+		}
+
+	r_is_one = 1; /* except for Montgomery factor */
+
+	/* bits-1 >= 0 */
+
+	/* The result is accumulated in the product r*w. */
+	w = a; /* bit 'bits-1' of 'p' is always set */
+	for (b = bits-2; b >= 0; b--)
+		{
+		/* First, square r*w. */
+		next_w = w*w;
+		if ((next_w/w) != w) /* overflow */
+			{
+			if (r_is_one)
+				{
+				if (!BN_TO_MONTGOMERY_WORD(r, w, mont)) goto err;
+				r_is_one = 0;
+				}
+			else
+				{
+				if (!BN_MOD_MUL_WORD(r, w, m)) goto err;
+				}
+			next_w = 1;
+			}
+		w = next_w;
+		if (!r_is_one)
+			{
+			if (!BN_mod_mul_montgomery(r, r, r, mont, ctx)) goto err;
+			}
+
+		/* Second, multiply r*w by 'a' if exponent bit is set. */
+		if (BN_is_bit_set(p, b))
+			{
+			next_w = w*a;
+			if ((next_w/a) != w) /* overflow */
+				{
+				if (r_is_one)
+					{
+					if (!BN_TO_MONTGOMERY_WORD(r, w, mont)) goto err;
+					r_is_one = 0;
+					}
+				else
+					{
+					if (!BN_MOD_MUL_WORD(r, w, m)) goto err;
+					}
+				next_w = a;
+				}
+			w = next_w;
+			}
+		}
+
+	/* Finally, set r:=r*w. */
+	if (w != 1)
+		{
+		if (r_is_one)
+			{
+			if (!BN_TO_MONTGOMERY_WORD(r, w, mont)) goto err;
+			r_is_one = 0;
+			}
+		else
+			{
+			if (!BN_MOD_MUL_WORD(r, w, m)) goto err;
+			}
+		}
+
+	if (r_is_one) /* can happen only if a == 1*/
+		{
+		if (!BN_one(rr)) goto err;
+		}
+	else
+		{
+		if (!BN_from_montgomery(rr, r, mont, ctx)) goto err;
+		}
+	ret = 1;
+err:
+	if ((in_mont == NULL) && (mont != NULL)) BN_MONT_CTX_free(mont);
+	BN_CTX_end(ctx);
+	return(ret);
+	}
+
 
 /* The old fallback, simple version :-) */
 int BN_mod_exp_simple(BIGNUM *r, BIGNUM *a, BIGNUM *p, BIGNUM *m,
@@ -660,26 +612,21 @@ int BN_mod_exp_simple(BIGNUM *r, BIGNUM *a, BIGNUM *p, BIGNUM *m,
 	BN_init(&(val[0]));
 	ts=1;
 	if (!BN_mod(&(val[0]),a,m,ctx)) goto err;		/* 1 */
-	if (!BN_mod_mul(d,&(val[0]),&(val[0]),m,ctx))
-		goto err;				/* 2 */
-
-	if (bits <= 17) /* This is probably 3 or 0x10001, so just do singles */
-		window=1;
-	else if (bits >= 256)
-		window=5;	/* max size of window */
-	else if (bits >= 128)
-		window=4;
-	else
-		window=3;
 
-	j=1<<(window-1);
-	for (i=1; i<j; i++)
+	window = BN_window_bits_for_exponent_size(bits);
+	if (window > 1)
 		{
-		BN_init(&(val[i]));
-		if (!BN_mod_mul(&(val[i]),&(val[i-1]),d,m,ctx))
-			goto err;
+		if (!BN_mod_mul(d,&(val[0]),&(val[0]),m,ctx))
+			goto err;				/* 2 */
+		j=1<<(window-1);
+		for (i=1; i<j; i++)
+			{
+			BN_init(&(val[i]));
+			if (!BN_mod_mul(&(val[i]),&(val[i-1]),d,m,ctx))
+				goto err;
+			}
+		ts=i;
 		}
-	ts=i;
 
 	start=1;	/* This is used to avoid multiplication etc
 			 * when there is only the value '1' in the
diff --git a/lib/libcrypto/bn/bn_exp2.c b/lib/libcrypto/bn/bn_exp2.c
index 4f4e9e32998..29029f4c724 100644
--- a/lib/libcrypto/bn/bn_exp2.c
+++ b/lib/libcrypto/bn/bn_exp2.c
@@ -1,27 +1,128 @@
+/* crypto/bn/bn_exp2.c */
+/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
+ * All rights reserved.
+ *
+ * This package is an SSL implementation written
+ * by Eric Young (eay@cryptsoft.com).
+ * The implementation was written so as to conform with Netscapes SSL.
+ * 
+ * This library is free for commercial and non-commercial use as long as
+ * the following conditions are aheared to.  The following conditions
+ * apply to all code found in this distribution, be it the RC4, RSA,
+ * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
+ * included with this distribution is covered by the same copyright terms
+ * except that the holder is Tim Hudson (tjh@cryptsoft.com).
+ * 
+ * Copyright remains Eric Young's, and as such any Copyright notices in
+ * the code are not to be removed.
+ * If this package is used in a product, Eric Young should be given attribution
+ * as the author of the parts of the library used.
+ * This can be in the form of a textual message at program startup or
+ * in documentation (online or textual) provided with the package.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. All advertising materials mentioning features or use of this software
+ *    must display the following acknowledgement:
+ *    "This product includes cryptographic software written by
+ *     Eric Young (eay@cryptsoft.com)"
+ *    The word 'cryptographic' can be left out if the rouines from the library
+ *    being used are not cryptographic related :-).
+ * 4. If you include any Windows specific code (or a derivative thereof) from 
+ *    the apps directory (application code) you must include an acknowledgement:
+ *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
+ * 
+ * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ * 
+ * The licence and distribution terms for any publically available version or
+ * derivative of this code cannot be changed.  i.e. this code cannot simply be
+ * copied and put under another distribution licence
+ * [including the GNU Public Licence.]
+ */
+/* ====================================================================
+ * Copyright (c) 1998-2000 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
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer. 
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ *    software must display the following acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ *    endorse or promote products derived from this software without
+ *    prior written permission. For written permission, please contact
+ *    openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ *    nor may "OpenSSL" appear in their names without prior written
+ *    permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ *    acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * 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).
+ *
+ */
+
 #include <stdio.h>
 #include "cryptlib.h"
 #include "bn_lcl.h"
 
-/* I've done some timing with different table sizes.
- * The main hassle is that even with bits set at 3, this requires
- * 63 BIGNUMs to store the pre-calculated values.
- *          512   1024 
- * bits=1  75.4%  79.4%
- * bits=2  61.2%  62.4%
- * bits=3  61.3%  59.3%
- * The lack of speed improvement is also a function of the pre-calculation
- * which could be removed.
- */
-#define EXP2_TABLE_BITS	2 /* 1  2  3  4  5  */
-#define EXP2_TABLE_SIZE	4 /* 2  4  8 16 32  */
+#define TABLE_SIZE	32
 
 int BN_mod_exp2_mont(BIGNUM *rr, BIGNUM *a1, BIGNUM *p1, BIGNUM *a2,
 	     BIGNUM *p2, BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
 	{
-	int i,j,k,bits,bits1,bits2,ret=0,wstart,wend,window,xvalue,yvalue;
-	int start=1,ts=0,x,y;
-	BIGNUM *d,*aa1,*aa2,*r;
-	BIGNUM val[EXP2_TABLE_SIZE][EXP2_TABLE_SIZE];
+	int i,j,bits,b,bits1,bits2,ret=0,wpos1,wpos2,window1,window2,wvalue1,wvalue2;
+	int r_is_one=1,ts1=0,ts2=0;
+	BIGNUM *d,*r;
+	BIGNUM *a_mod_m;
+	BIGNUM val1[TABLE_SIZE], val2[TABLE_SIZE];
 	BN_MONT_CTX *mont=NULL;
 
 	bn_check_top(a1);
@@ -32,7 +133,7 @@ int BN_mod_exp2_mont(BIGNUM *rr, BIGNUM *a1, BIGNUM *p1, BIGNUM *a2,
 
 	if (!(m->d[0] & 1))
 		{
-		BNerr(BN_F_BN_MOD_EXP_MONT,BN_R_CALLED_WITH_EVEN_MODULUS);
+		BNerr(BN_F_BN_MOD_EXP2_MONT,BN_R_CALLED_WITH_EVEN_MODULUS);
 		return(0);
 		}
 	bits1=BN_num_bits(p1);
@@ -42,17 +143,13 @@ int BN_mod_exp2_mont(BIGNUM *rr, BIGNUM *a1, BIGNUM *p1, BIGNUM *a2,
 		BN_one(rr);
 		return(1);
 		}
+	bits=(bits1 > bits2)?bits1:bits2;
 
 	BN_CTX_start(ctx);
 	d = BN_CTX_get(ctx);
 	r = BN_CTX_get(ctx);
 	if (d == NULL || r == NULL) goto err;
 
-	bits=(bits1 > bits2)?bits1:bits2;
-
-	/* If this is not done, things will break in the montgomery
-	 * part */
-
 	if (in_mont != NULL)
 		mont=in_mont;
 	else
@@ -61,139 +158,143 @@ int BN_mod_exp2_mont(BIGNUM *rr, BIGNUM *a1, BIGNUM *p1, BIGNUM *a2,
 		if (!BN_MONT_CTX_set(mont,m,ctx)) goto err;
 		}
 
-	BN_init(&(val[0][0]));
-	BN_init(&(val[1][1]));
-	BN_init(&(val[0][1]));
-	BN_init(&(val[1][0]));
-	ts=1;
+	window1 = BN_window_bits_for_exponent_size(bits1);
+	window2 = BN_window_bits_for_exponent_size(bits2);
+
+	/*
+	 * Build table for a1:   val1[i] := a1^(2*i + 1) mod m  for i = 0 .. 2^(window1-1)
+	 */
+	BN_init(&val1[0]);
+	ts1=1;
 	if (BN_ucmp(a1,m) >= 0)
 		{
-		BN_mod(&(val[1][0]),a1,m,ctx);
-		aa1= &(val[1][0]);
+		if (!BN_mod(&(val1[0]),a1,m,ctx))
+			goto err;
+		a_mod_m = &(val1[0]);
 		}
 	else
-		aa1=a1;
+		a_mod_m = a1;
+	if (!BN_to_montgomery(&(val1[0]),a_mod_m,mont,ctx)) goto err;
+	if (window1 > 1)
+		{
+		if (!BN_mod_mul_montgomery(d,&(val1[0]),&(val1[0]),mont,ctx)) goto err;
+
+		j=1<<(window1-1);
+		for (i=1; i<j; i++)
+			{
+			BN_init(&(val1[i]));
+			if (!BN_mod_mul_montgomery(&(val1[i]),&(val1[i-1]),d,mont,ctx))
+				goto err;
+			}
+		ts1=i;
+		}
+
+
+	/*
+	 * Build table for a2:   val2[i] := a2^(2*i + 1) mod m  for i = 0 .. 2^(window2-1)
+	 */
+	BN_init(&val2[0]);
+	ts2=1;
 	if (BN_ucmp(a2,m) >= 0)
 		{
-		BN_mod(&(val[0][1]),a2,m,ctx);
-		aa2= &(val[0][1]);
+		if (!BN_mod(&(val2[0]),a2,m,ctx))
+			goto err;
+		a_mod_m = &(val2[0]);
 		}
 	else
-		aa2=a2;
-	if (!BN_to_montgomery(&(val[1][0]),aa1,mont,ctx)) goto err;
-	if (!BN_to_montgomery(&(val[0][1]),aa2,mont,ctx)) goto err;
-	if (!BN_mod_mul_montgomery(&(val[1][1]),
-		&(val[1][0]),&(val[0][1]),mont,ctx))
-		goto err;
-
-#if 0
-	if (bits <= 20) /* This is probably 3 or 0x10001, so just do singles */
-		window=1;
-	else if (bits > 250)
-		window=5;	/* max size of window */
-	else if (bits >= 120)
-		window=4;
-	else
-		window=3;
-#else
-	window=EXP2_TABLE_BITS;
-#endif
-
-	k=1<<window;
-	for (x=0; x<k; x++)
+		a_mod_m = a2;
+	if (!BN_to_montgomery(&(val2[0]),a_mod_m,mont,ctx)) goto err;
+	if (window2 > 1)
 		{
-		if (x >= 2)
-			{
-			BN_init(&(val[x][0]));
-			BN_init(&(val[x][1]));
-			if (!BN_mod_mul_montgomery(&(val[x][0]),
-				&(val[1][0]),&(val[x-1][0]),mont,ctx)) goto err;
-			if (!BN_mod_mul_montgomery(&(val[x][1]),
-				&(val[1][0]),&(val[x-1][1]),mont,ctx)) goto err;
-			}
-		for (y=2; y<k; y++)
+		if (!BN_mod_mul_montgomery(d,&(val2[0]),&(val2[0]),mont,ctx)) goto err;
+
+		j=1<<(window2-1);
+		for (i=1; i<j; i++)
 			{
-			BN_init(&(val[x][y]));
-			if (!BN_mod_mul_montgomery(&(val[x][y]),
-				&(val[x][y-1]),&(val[0][1]),mont,ctx))
+			BN_init(&(val2[i]));
+			if (!BN_mod_mul_montgomery(&(val2[i]),&(val2[i-1]),d,mont,ctx))
 				goto err;
 			}
+		ts2=i;
 		}
-	ts=k;
-
-	start=1;	/* This is used to avoid multiplication etc
-			 * when there is only the value '1' in the
-			 * buffer. */
-	xvalue=0;	/* The 'x value' of the window */
-	yvalue=0;	/* The 'y value' of the window */
-	wstart=bits-1;	/* The top bit of the window */
-	wend=0;		/* The bottom bit of the window */
-
-        if (!BN_to_montgomery(r,BN_value_one(),mont,ctx)) goto err;
-	for (;;)
+
+
+	/* Now compute the power product, using independent windows. */
+	r_is_one=1;
+	wvalue1=0;  /* The 'value' of the first window */
+	wvalue2=0;  /* The 'value' of the second window */
+	wpos1=0;    /* If wvalue1 > 0, the bottom bit of the first window */
+	wpos2=0;    /* If wvalue2 > 0, the bottom bit of the second window */
+
+	if (!BN_to_montgomery(r,BN_value_one(),mont,ctx)) goto err;
+	for (b=bits-1; b>=0; b--)
 		{
-		xvalue=BN_is_bit_set(p1,wstart);
-		yvalue=BN_is_bit_set(p2,wstart);
-		if (!(xvalue || yvalue))
+		if (!r_is_one)
 			{
-			if (!start)
+			if (!BN_mod_mul_montgomery(r,r,r,mont,ctx))
+				goto err;
+			}
+		
+		if (!wvalue1)
+			if (BN_is_bit_set(p1, b))
 				{
-				if (!BN_mod_mul_montgomery(r,r,r,mont,ctx))
-					goto err;
+				/* consider bits b-window1+1 .. b for this window */
+				i = b-window1+1;
+				while (!BN_is_bit_set(p1, i)) /* works for i<0 */
+					i++;
+				wpos1 = i;
+				wvalue1 = 1;
+				for (i = b-1; i >= wpos1; i--)
+					{
+					wvalue1 <<= 1;
+					if (BN_is_bit_set(p1, i))
+						wvalue1++;
+					}
 				}
-			wstart--;
-			if (wstart < 0) break;
-			continue;
-			}
-		/* We now have wstart on a 'set' bit, we now need to work out
-		 * how bit a window to do.  To do this we need to scan
-		 * forward until the last set bit before the end of the
-		 * window */
-		j=wstart;
-		/* xvalue=BN_is_bit_set(p1,wstart); already set */
-		/* yvalue=BN_is_bit_set(p1,wstart); already set */
-		wend=0;
-		for (i=1; i<window; i++)
-			{
-			if (wstart-i < 0) break;
-			xvalue+=xvalue;
-			xvalue|=BN_is_bit_set(p1,wstart-i);
-			yvalue+=yvalue;
-			yvalue|=BN_is_bit_set(p2,wstart-i);
-			}
-
-		/* i is the size of the current window */
-		/* add the 'bytes above' */
-		if (!start)
-			for (j=0; j<i; j++)
+		
+		if (!wvalue2)
+			if (BN_is_bit_set(p2, b))
 				{
-				if (!BN_mod_mul_montgomery(r,r,r,mont,ctx))
-					goto err;
+				/* consider bits b-window2+1 .. b for this window */
+				i = b-window2+1;
+				while (!BN_is_bit_set(p2, i))
+					i++;
+				wpos2 = i;
+				wvalue2 = 1;
+				for (i = b-1; i >= wpos2; i--)
+					{
+					wvalue2 <<= 1;
+					if (BN_is_bit_set(p2, i))
+						wvalue2++;
+					}
 				}
+
+		if (wvalue1 && b == wpos1)
+			{
+			/* wvalue1 is odd and < 2^window1 */
+			if (!BN_mod_mul_montgomery(r,r,&(val1[wvalue1>>1]),mont,ctx))
+				goto err;
+			wvalue1 = 0;
+			r_is_one = 0;
+			}
 		
-		/* wvalue will be an odd number < 2^window */
-		if (xvalue || yvalue)
+		if (wvalue2 && b == wpos2)
 			{
-			if (!BN_mod_mul_montgomery(r,r,&(val[xvalue][yvalue]),
-				mont,ctx)) goto err;
+			/* wvalue2 is odd and < 2^window2 */
+			if (!BN_mod_mul_montgomery(r,r,&(val2[wvalue2>>1]),mont,ctx))
+				goto err;
+			wvalue2 = 0;
+			r_is_one = 0;
 			}
-
-		/* move the 'window' down further */
-		wstart-=i;
-		start=0;
-		if (wstart < 0) break;
 		}
 	BN_from_montgomery(rr,r,mont,ctx);
 	ret=1;
 err:
 	if ((in_mont == NULL) && (mont != NULL)) BN_MONT_CTX_free(mont);
 	BN_CTX_end(ctx);
-	for (i=0; i<ts; i++)
-		{
-		for (j=0; j<ts; j++)
-			{
-			BN_clear_free(&(val[i][j]));
-			}
-		}
+	for (i=0; i<ts1; i++)
+		BN_clear_free(&(val1[i]));
+	for (i=0; i<ts2; i++)
+		BN_clear_free(&(val2[i]));
 	return(ret);
 	}
diff --git a/lib/libcrypto/bn/bn_lcl.h b/lib/libcrypto/bn/bn_lcl.h
index e36ccbc4c2d..9c959921b49 100644
--- a/lib/libcrypto/bn/bn_lcl.h
+++ b/lib/libcrypto/bn/bn_lcl.h
@@ -55,6 +55,59 @@
  * copied and put under another distribution licence
  * [including the GNU Public Licence.]
  */
+/* ====================================================================
+ * Copyright (c) 1998-2000 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
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer. 
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ *    software must display the following acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ *    endorse or promote products derived from this software without
+ *    prior written permission. For written permission, please contact
+ *    openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ *    nor may "OpenSSL" appear in their names without prior written
+ *    permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ *    acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * 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).
+ *
+ */
 
 #ifndef HEADER_BN_LCL_H
 #define HEADER_BN_LCL_H
@@ -65,6 +118,51 @@
 extern "C" {
 #endif
 
+
+/*
+ * BN_window_bits_for_exponent_size -- macro for sliding window mod_exp functions
+ *
+ *
+ * For window size 'w' (w >= 2) and a random 'b' bits exponent,
+ * the number of multiplications is a constant plus on average
+ *
+ *    2^(w-1) + (b-w)/(w+1);
+ *
+ * here  2^(w-1)  is for precomputing the table (we actually need
+ * entries only for windows that have the lowest bit set), and
+ * (b-w)/(w+1)  is an approximation for the expected number of
+ * w-bit windows, not counting the first one.
+ *
+ * Thus we should use
+ *
+ *    w >= 6  if        b > 671
+ *     w = 5  if  671 > b > 239
+ *     w = 4  if  239 > b >  79
+ *     w = 3  if   79 > b >  23
+ *    w <= 2  if   23 > b
+ *
+ * (with draws in between).  Very small exponents are often selected
+ * with low Hamming weight, so we use  w = 1  for b <= 23.
+ */
+#if 1
+#define BN_window_bits_for_exponent_size(b) \
+		((b) > 671 ? 6 : \
+		 (b) > 239 ? 5 : \
+		 (b) >  79 ? 4 : \
+		 (b) >  23 ? 3 : 1)
+#else
+/* Old SSLeay/OpenSSL table.
+ * Maximum window size was 5, so this table differs for b==1024;
+ * but it coincides for other interesting values (b==160, b==512).
+ */
+#define BN_window_bits_for_exponent_size(b) \
+		((b) > 255 ? 5 : \
+		 (b) > 127 ? 4 : \
+		 (b) >  17 ? 3 : 1)
+#endif	 
+
+
+
 /* Pentium pro 16,16,16,32,64 */
 /* Alpha       16,16,16,16.64 */
 #define BN_MULL_SIZE_NORMAL			(16) /* 32 */
@@ -130,7 +228,7 @@ extern "C" {
 /* This is used for internal error checking and is not normally used */
 #ifdef BN_DEBUG
 # include <assert.h>
-# define bn_check_top(a) assert ((a)->top >= 0 && (a)->top <= (a)->max);
+# define bn_check_top(a) assert ((a)->top >= 0 && (a)->top <= (a)->dmax);
 #else
 # define bn_check_top(a)
 #endif
diff --git a/lib/libcrypto/bn/bn_lib.c b/lib/libcrypto/bn/bn_lib.c
index 0e6b12d9c36..b6b0ce4b3c9 100644
--- a/lib/libcrypto/bn/bn_lib.c
+++ b/lib/libcrypto/bn/bn_lib.c
@@ -56,6 +56,12 @@
  * [including the GNU Public Licence.]
  */
 
+#ifndef BN_DEBUG
+# undef NDEBUG /* avoid conflicting definitions */
+# define NDEBUG
+#endif
+
+#include <assert.h>
 #include <stdio.h>
 #include "cryptlib.h"
 #include "bn_lcl.h"
@@ -244,14 +250,8 @@ int BN_num_bits(const BIGNUM *a)
 
 	if (a->top == 0) return(0);
 	l=a->d[a->top-1];
+	assert(l != 0);
 	i=(a->top-1)*BN_BITS2;
-	if (l == 0)
-		{
-#if !defined(NO_STDIO) && !defined(WIN16)
-		fprintf(stderr,"BAD TOP VALUE\n");
-#endif
-		abort();
-		}
 	return(i+BN_num_bits_word(l));
 	}
 
@@ -262,24 +262,24 @@ void BN_clear_free(BIGNUM *a)
 	if (a == NULL) return;
 	if (a->d != NULL)
 		{
-		memset(a->d,0,a->max*sizeof(a->d[0]));
+		memset(a->d,0,a->dmax*sizeof(a->d[0]));
 		if (!(BN_get_flags(a,BN_FLG_STATIC_DATA)))
-			Free(a->d);
+			OPENSSL_free(a->d);
 		}
 	i=BN_get_flags(a,BN_FLG_MALLOCED);
 	memset(a,0,sizeof(BIGNUM));
 	if (i)
-		Free(a);
+		OPENSSL_free(a);
 	}
 
 void BN_free(BIGNUM *a)
 	{
 	if (a == NULL) return;
 	if ((a->d != NULL) && !(BN_get_flags(a,BN_FLG_STATIC_DATA)))
-		Free(a->d);
+		OPENSSL_free(a->d);
 	a->flags|=BN_FLG_FREE; /* REMOVE? */
 	if (a->flags & BN_FLG_MALLOCED)
-		Free(a);
+		OPENSSL_free(a);
 	}
 
 void BN_init(BIGNUM *a)
@@ -291,7 +291,7 @@ BIGNUM *BN_new(void)
 	{
 	BIGNUM *ret;
 
-	if ((ret=(BIGNUM *)Malloc(sizeof(BIGNUM))) == NULL)
+	if ((ret=(BIGNUM *)OPENSSL_malloc(sizeof(BIGNUM))) == NULL)
 		{
 		BNerr(BN_F_BN_NEW,ERR_R_MALLOC_FAILURE);
 		return(NULL);
@@ -299,7 +299,7 @@ BIGNUM *BN_new(void)
 	ret->flags=BN_FLG_MALLOCED;
 	ret->top=0;
 	ret->neg=0;
-	ret->max=0;
+	ret->dmax=0;
 	ret->d=NULL;
 	return(ret);
 	}
@@ -317,7 +317,7 @@ BIGNUM *bn_expand2(BIGNUM *b, int words)
 
 	bn_check_top(b);
 
-	if (words > b->max)
+	if (words > b->dmax)
 		{
 		bn_check_top(b);	
 		if (BN_get_flags(b,BN_FLG_STATIC_DATA))
@@ -325,7 +325,7 @@ BIGNUM *bn_expand2(BIGNUM *b, int words)
 			BNerr(BN_F_BN_EXPAND2,BN_R_EXPAND_ON_STATIC_BIGNUM_DATA);
 			return(NULL);
 			}
-		a=A=(BN_ULONG *)Malloc(sizeof(BN_ULONG)*(words+1));
+		a=A=(BN_ULONG *)OPENSSL_malloc(sizeof(BN_ULONG)*(words+1));
 		if (A == NULL)
 			{
 			BNerr(BN_F_BN_EXPAND2,ERR_R_MALLOC_FAILURE);
@@ -423,21 +423,21 @@ BIGNUM *bn_expand2(BIGNUM *b, int words)
 				case 0:	; /* ultrix cc workaround, see above */
 				}
 #endif
-			Free(b->d);
+			OPENSSL_free(b->d);
 			}
 
 		b->d=a;
-		b->max=words;
+		b->dmax=words;
 
 		/* Now need to zero any data between b->top and b->max */
 
 		A= &(b->d[b->top]);
-		for (i=(b->max - b->top)>>3; i>0; i--,A+=8)
+		for (i=(b->dmax - b->top)>>3; i>0; i--,A+=8)
 			{
 			A[0]=0; A[1]=0; A[2]=0; A[3]=0;
 			A[4]=0; A[5]=0; A[6]=0; A[7]=0;
 			}
-		for (i=(b->max - b->top)&7; i>0; i--,A++)
+		for (i=(b->dmax - b->top)&7; i>0; i--,A++)
 			A[0]=0;
 #else
 			memset(A,0,sizeof(BN_ULONG)*(words+1));
@@ -508,7 +508,7 @@ BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
 void BN_clear(BIGNUM *a)
 	{
 	if (a->d != NULL)
-		memset(a->d,0,a->max*sizeof(a->d[0]));
+		memset(a->d,0,a->dmax*sizeof(a->d[0]));
 	a->top=0;
 	a->neg=0;
 	}
diff --git a/lib/libcrypto/bn/bn_mont.c b/lib/libcrypto/bn/bn_mont.c
index 598fecbf0c8..8cf1febacca 100644
--- a/lib/libcrypto/bn/bn_mont.c
+++ b/lib/libcrypto/bn/bn_mont.c
@@ -85,16 +85,7 @@ int BN_mod_mul_montgomery(BIGNUM *r, BIGNUM *a, BIGNUM *b,
 
 	if (a == b)
 		{
-#if 0
-		bn_wexpand(tmp,a->top*2);
-		bn_wexpand(tmp2,a->top*4);
-		bn_sqr_recursive(tmp->d,a->d,a->top,tmp2->d);
-		tmp->top=a->top*2;
-		if (tmp->d[tmp->top-1] == 0)
-			tmp->top--;
-#else
 		if (!BN_sqr(tmp,a,ctx)) goto err;
-#endif
 		}
 	else
 		{
@@ -157,7 +148,22 @@ int BN_from_montgomery(BIGNUM *ret, BIGNUM *a, BN_MONT_CTX *mont,
 #endif
 	for (i=0; i<nl; i++)
 		{
+#ifdef __TANDEM
+                {
+                   long long t1;
+                   long long t2;
+                   long long t3;
+                   t1 = rp[0] * (n0 & 0177777);
+                   t2 = 037777600000l;
+                   t2 = n0 & t2;
+                   t3 = rp[0] & 0177777;
+                   t2 = (t3 * t2) & BN_MASK2;
+                   t1 = t1 + t2;
+                   v=bn_mul_add_words(rp,np,nl,(BN_ULONG) t1);
+                }
+#else
 		v=bn_mul_add_words(rp,np,nl,(rp[0]*n0)&BN_MASK2);
+#endif
 		nrp++;
 		rp++;
 		if (((nrp[-1]+=v)&BN_MASK2) >= v)
@@ -175,6 +181,7 @@ int BN_from_montgomery(BIGNUM *ret, BIGNUM *a, BN_MONT_CTX *mont,
 #if 0
 	BN_rshift(ret,r,mont->ri);
 #else
+	ret->neg = r->neg;
 	x=ri;
 	rp=ret->d;
 	ap= &(r->d[x]);
@@ -234,7 +241,7 @@ BN_MONT_CTX *BN_MONT_CTX_new(void)
 	{
 	BN_MONT_CTX *ret;
 
-	if ((ret=(BN_MONT_CTX *)Malloc(sizeof(BN_MONT_CTX))) == NULL)
+	if ((ret=(BN_MONT_CTX *)OPENSSL_malloc(sizeof(BN_MONT_CTX))) == NULL)
 		return(NULL);
 
 	BN_MONT_CTX_init(ret);
@@ -260,7 +267,7 @@ void BN_MONT_CTX_free(BN_MONT_CTX *mont)
 	BN_free(&(mont->N));
 	BN_free(&(mont->Ni));
 	if (mont->flags & BN_FLG_MALLOCED)
-		Free(mont);
+		OPENSSL_free(mont);
 	}
 
 int BN_MONT_CTX_set(BN_MONT_CTX *mont, const BIGNUM *mod, BN_CTX *ctx)
@@ -284,7 +291,7 @@ int BN_MONT_CTX_set(BN_MONT_CTX *mont, const BIGNUM *mod, BN_CTX *ctx)
 		buf[1]=0;
 		tmod.d=buf;
 		tmod.top=1;
-		tmod.max=2;
+		tmod.dmax=2;
 		tmod.neg=mod->neg;
 							/* Ri = R^-1 mod N*/
 		if ((BN_mod_inverse(&Ri,R,&tmod,ctx)) == NULL)
diff --git a/lib/libcrypto/bn/bn_mul.c b/lib/libcrypto/bn/bn_mul.c
index 3e8baaad9a0..3e8d8b9567a 100644
--- a/lib/libcrypto/bn/bn_mul.c
+++ b/lib/libcrypto/bn/bn_mul.c
@@ -631,7 +631,6 @@ int BN_mul(BIGNUM *r, BIGNUM *a, BIGNUM *b, BN_CTX *ctx)
 
 	al=a->top;
 	bl=b->top;
-	r->neg=a->neg^b->neg;
 
 	if ((al == 0) || (bl == 0))
 		{
@@ -647,6 +646,7 @@ int BN_mul(BIGNUM *r, BIGNUM *a, BIGNUM *b, BN_CTX *ctx)
 		}
 	else
 		rr = r;
+	rr->neg=a->neg^b->neg;
 
 #if defined(BN_MUL_COMBA) || defined(BN_RECURSION)
 	i = al-bl;
diff --git a/lib/libcrypto/bn/bn_print.c b/lib/libcrypto/bn/bn_print.c
index 782a96e7e0a..532e66bcc39 100644
--- a/lib/libcrypto/bn/bn_print.c
+++ b/lib/libcrypto/bn/bn_print.c
@@ -64,14 +64,14 @@
 
 static const char *Hex="0123456789ABCDEF";
 
-/* Must 'Free' the returned data */
+/* Must 'OPENSSL_free' the returned data */
 char *BN_bn2hex(const BIGNUM *a)
 	{
 	int i,j,v,z=0;
 	char *buf;
 	char *p;
 
-	buf=(char *)Malloc(a->top*BN_BYTES*2+2);
+	buf=(char *)OPENSSL_malloc(a->top*BN_BYTES*2+2);
 	if (buf == NULL)
 		{
 		BNerr(BN_F_BN_BN2HEX,ERR_R_MALLOC_FAILURE);
@@ -99,7 +99,7 @@ err:
 	return(buf);
 	}
 
-/* Must 'Free' the returned data */
+/* Must 'OPENSSL_free' the returned data */
 char *BN_bn2dec(const BIGNUM *a)
 	{
 	int i=0,num;
@@ -110,8 +110,8 @@ char *BN_bn2dec(const BIGNUM *a)
 
 	i=BN_num_bits(a)*3;
 	num=(i/10+i/1000+3)+1;
-	bn_data=(BN_ULONG *)Malloc((num/BN_DEC_NUM+1)*sizeof(BN_ULONG));
-	buf=(char *)Malloc(num+3);
+	bn_data=(BN_ULONG *)OPENSSL_malloc((num/BN_DEC_NUM+1)*sizeof(BN_ULONG));
+	buf=(char *)OPENSSL_malloc(num+3);
 	if ((buf == NULL) || (bn_data == NULL))
 		{
 		BNerr(BN_F_BN_BN2DEC,ERR_R_MALLOC_FAILURE);
@@ -149,7 +149,7 @@ char *BN_bn2dec(const BIGNUM *a)
 			}
 		}
 err:
-	if (bn_data != NULL) Free(bn_data);
+	if (bn_data != NULL) OPENSSL_free(bn_data);
 	if (t != NULL) BN_free(t);
 	return(buf);
 	}
diff --git a/lib/libcrypto/bn/bn_rand.c b/lib/libcrypto/bn/bn_rand.c
index 943712c15b8..21ecbc04ed0 100644
--- a/lib/libcrypto/bn/bn_rand.c
+++ b/lib/libcrypto/bn/bn_rand.c
@@ -68,11 +68,17 @@ static int bnrand(int pseudorand, BIGNUM *rnd, int bits, int top, int bottom)
 	int ret=0,bit,bytes,mask;
 	time_t tim;
 
+	if (bits == 0)
+		{
+		BN_zero(rnd);
+		return 1;
+		}
+
 	bytes=(bits+7)/8;
 	bit=(bits-1)%8;
 	mask=0xff<<bit;
 
-	buf=(unsigned char *)Malloc(bytes);
+	buf=(unsigned char *)OPENSSL_malloc(bytes);
 	if (buf == NULL)
 		{
 		BNerr(BN_F_BN_RAND,ERR_R_MALLOC_FAILURE);
@@ -120,7 +126,7 @@ err:
 	if (buf != NULL)
 		{
 		memset(buf,0,bytes);
-		Free(buf);
+		OPENSSL_free(buf);
 		}
 	return(ret);
 	}
diff --git a/lib/libcrypto/bn/bn_recp.c b/lib/libcrypto/bn/bn_recp.c
index a8796bd0aac..d019941d6be 100644
--- a/lib/libcrypto/bn/bn_recp.c
+++ b/lib/libcrypto/bn/bn_recp.c
@@ -72,7 +72,7 @@ BN_RECP_CTX *BN_RECP_CTX_new(void)
 	{
 	BN_RECP_CTX *ret;
 
-	if ((ret=(BN_RECP_CTX *)Malloc(sizeof(BN_RECP_CTX))) == NULL)
+	if ((ret=(BN_RECP_CTX *)OPENSSL_malloc(sizeof(BN_RECP_CTX))) == NULL)
 		return(NULL);
 
 	BN_RECP_CTX_init(ret);
@@ -88,7 +88,7 @@ void BN_RECP_CTX_free(BN_RECP_CTX *recp)
 	BN_free(&(recp->N));
 	BN_free(&(recp->Nr));
 	if (recp->flags & BN_FLG_MALLOCED)
-		Free(recp);
+		OPENSSL_free(recp);
 	}
 
 int BN_RECP_CTX_set(BN_RECP_CTX *recp, const BIGNUM *d, BN_CTX *ctx)
diff --git a/lib/libcrypto/bn/bn_shift.c b/lib/libcrypto/bn/bn_shift.c
index 61aae65a6bf..0883247384e 100644
--- a/lib/libcrypto/bn/bn_shift.c
+++ b/lib/libcrypto/bn/bn_shift.c
@@ -162,7 +162,7 @@ int BN_rshift(BIGNUM *r, BIGNUM *a, int n)
 	nw=n/BN_BITS2;
 	rb=n%BN_BITS2;
 	lb=BN_BITS2-rb;
-	if (nw > a->top)
+	if (nw > a->top || a->top == 0)
 		{
 		BN_zero(r);
 		return(1);
diff --git a/lib/libcrypto/bn/bn_sqr.c b/lib/libcrypto/bn/bn_sqr.c
index fe00c5f69a0..75f4f38392d 100644
--- a/lib/libcrypto/bn/bn_sqr.c
+++ b/lib/libcrypto/bn/bn_sqr.c
@@ -188,7 +188,7 @@ void bn_sqr_normal(BN_ULONG *r, BN_ULONG *a, int n, BN_ULONG *tmp)
 
 #ifdef BN_RECURSION
 /* r is 2*n words in size,
- * a and b are both n words in size.
+ * a and b are both n words in size.    (There's not actually a 'b' here ...)
  * n must be a power of 2.
  * We multiply and return the result.
  * t must be 2*n words in size
diff --git a/lib/libcrypto/bn/bn_word.c b/lib/libcrypto/bn/bn_word.c
index 73157a7d43f..cd59baa2c49 100644
--- a/lib/libcrypto/bn/bn_word.c
+++ b/lib/libcrypto/bn/bn_word.c
@@ -115,7 +115,7 @@ int BN_add_word(BIGNUM *a, BN_ULONG w)
 		a->neg=0;
 		i=BN_sub_word(a,w);
 		if (!BN_is_zero(a))
-			a->neg=1;
+			a->neg=!(a->neg);
 		return(i);
 		}
 	w&=BN_MASK2;
@@ -140,7 +140,7 @@ int BN_sub_word(BIGNUM *a, BN_ULONG w)
 	{
 	int i;
 
-	if (a->neg)
+	if (BN_is_zero(a) || a->neg)
 		{
 		a->neg=0;
 		i=BN_add_word(a,w);
@@ -182,11 +182,16 @@ int BN_mul_word(BIGNUM *a, BN_ULONG w)
 	w&=BN_MASK2;
 	if (a->top)
 		{
-		ll=bn_mul_words(a->d,a->d,a->top,w);
-		if (ll)
+		if (w == 0)
+			BN_zero(a);
+		else
 			{
-			if (bn_wexpand(a,a->top+1) == NULL) return(0);
-			a->d[a->top++]=ll;
+			ll=bn_mul_words(a->d,a->d,a->top,w);
+			if (ll)
+				{
+				if (bn_wexpand(a,a->top+1) == NULL) return(0);
+				a->d[a->top++]=ll;
+				}
 			}
 		}
 	return(1);
diff --git a/lib/libcrypto/bn/vms-helper.c b/lib/libcrypto/bn/vms-helper.c
index 73af3370695..0fa79c4edb5 100644
--- a/lib/libcrypto/bn/vms-helper.c
+++ b/lib/libcrypto/bn/vms-helper.c
@@ -59,8 +59,10 @@
 
 bn_div_words_abort(int i)
 {
+#ifdef BN_DEBUG
 #if !defined(NO_STDIO) && !defined(WIN16)
 	fprintf(stderr,"Division would overflow (%d)\n",i);
 #endif
 	abort();
+#endif
 }