This commit was generated by cvs2git to track changes on a CVS vendor

branch.

5 files changed, 218 insertions, 189 deletions

diff --git a/src/lib/libcrypto/bn/asm/alpha-mont.pl b/src/lib/libcrypto/bn/asm/alpha-mont.pl
index 7a2cc3173b..f7e0ca1646 100644
--- a/src/lib/libcrypto/bn/asm/alpha-mont.pl
+++ b/src/lib/libcrypto/bn/asm/alpha-mont.pl
@@ -53,15 +53,15 @@ $code=<<___;
 .align  5
 .ent    bn_mul_mont
 bn_mul_mont:
-        lda     sp,-40(sp)
+        lda     sp,-48(sp)
         stq     ra,0(sp)
         stq     s3,8(sp)
         stq     s4,16(sp)
         stq     s5,24(sp)
         stq     fp,32(sp)
         mov     sp,fp
-        .mask   0x0400f000,-40
-        .frame  fp,40,ra
+        .mask   0x0400f000,-48
+        .frame  fp,48,ra
         .prologue 0
 
         .align  4
@@ -306,7 +306,7 @@ bn_mul_mont:
         ldq     s4,16(sp)
         ldq     s5,24(sp)
         ldq     fp,32(sp)
-        lda     sp,40(sp)
+        lda     sp,48(sp)
         ret     (ra)
 .end    bn_mul_mont
 .rdata
diff --git a/src/lib/libcrypto/bn/asm/armv4-mont.pl b/src/lib/libcrypto/bn/asm/armv4-mont.pl
index 05d5dc1a48..14e0d2d1dd 100644
--- a/src/lib/libcrypto/bn/asm/armv4-mont.pl
+++ b/src/lib/libcrypto/bn/asm/armv4-mont.pl
@@ -193,6 +193,7 @@ bn_mul_mont:
         bx      lr                      @ interoperable with Thumb ISA:-)
 .size   bn_mul_mont,.-bn_mul_mont
 .asciz  "Montgomery multiplication for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
+.align  2
 ___
 
 $code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm;    # make it possible to compile with -march=armv4
diff --git a/src/lib/libcrypto/bn/asm/ppc.pl b/src/lib/libcrypto/bn/asm/ppc.pl
index 08e0053473..37c65d3511 100644
--- a/src/lib/libcrypto/bn/asm/ppc.pl
+++ b/src/lib/libcrypto/bn/asm/ppc.pl
@@ -100,9 +100,9 @@
 #       me a note at schari@us.ibm.com
 #
 
-$opf = shift;
+$flavour = shift;
 
-if ($opf =~ /32\.s/) {
+if ($flavour =~ /32/) {
         $BITS=  32;
         $BNSZ=  $BITS/8;
         $ISA=   "\"ppc\"";
@@ -125,7 +125,7 @@ if ($opf =~ /32\.s/) {
         $INSR=  "insrwi";       # insert right
         $ROTL=  "rotlwi";       # rotate left by immediate
         $TR=    "tw";           # conditional trap
-} elsif ($opf =~ /64\.s/) {
+} elsif ($flavour =~ /64/) {
         $BITS=  64;
         $BNSZ=  $BITS/8;
         $ISA=   "\"ppc64\"";
@@ -149,93 +149,16 @@ if ($opf =~ /32\.s/) {
         $INSR=  "insrdi";       # insert right 
         $ROTL=  "rotldi";       # rotate left by immediate
         $TR=    "td";           # conditional trap
-} else { die "nonsense $opf"; }
+} else { die "nonsense $flavour"; }
 
-( defined shift || open STDOUT,">$opf" ) || die "can't open $opf: $!";
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
 
-# function entry points from the AIX code
-#
-# There are other, more elegant, ways to handle this. We (IBM) chose
-# this approach as it plays well with scripts we run to 'namespace'
-# OpenSSL .i.e. we add a prefix to all the public symbols so we can
-# co-exist in the same process with other implementations of OpenSSL.
-# 'cleverer' ways of doing these substitutions tend to hide data we
-# need to be obvious.
-#
-my @items = ("bn_sqr_comba4",
-             "bn_sqr_comba8",
-             "bn_mul_comba4",
-             "bn_mul_comba8",
-             "bn_sub_words",
-             "bn_add_words",
-             "bn_div_words",
-             "bn_sqr_words",
-             "bn_mul_words",
-             "bn_mul_add_words");
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
 
-if    ($opf =~ /linux/) {  do_linux();  }
-elsif ($opf =~ /aix/)   {  do_aix();    }
-elsif ($opf =~ /osx/)   {  do_osx();    }
-else                    {  do_bsd();    }
-
-sub do_linux {
-    $d=&data();
-
-    if ($BITS==64) {
-      foreach $t (@items) {
-        $d =~ s/\.$t:/\
-\t.section\t".opd","aw"\
-\t.align\t3\
-\t.globl\t$t\
-$t:\
-\t.quad\t.$t,.TOC.\@tocbase,0\
-\t.size\t$t,24\
-\t.previous\n\
-\t.type\t.$t,\@function\
-\t.globl\t.$t\
-.$t:/g;
-      }
-    }
-    else {
-      foreach $t (@items) {
-        $d=~s/\.$t/$t/g;
-      }
-    }
-    # hide internal labels to avoid pollution of name table...
-    $d=~s/Lppcasm_/.Lppcasm_/gm;
-    print $d;
-}
-
-sub do_aix {
-    # AIX assembler is smart enough to please the linker without
-    # making us do something special...
-    print &data();
-}
-
-# MacOSX 32 bit
-sub do_osx {
-    $d=&data();
-    # Change the bn symbol prefix from '.' to '_'
-    foreach $t (@items) {
-      $d=~s/\.$t/_$t/g;
-    }
-    # Change .machine to something OS X asm will accept
-    $d=~s/\.machine.*/.text/g;
-    $d=~s/\#/;/g; # change comment from '#' to ';'
-    print $d;
-}
-
-# BSD (Untested)
-sub do_bsd {
-    $d=&data();
-    foreach $t (@items) {
-      $d=~s/\.$t/_$t/g;
-    }
-    print $d;
-}
-
-sub data {
-        local($data)=<<EOF;
+$data=<<EOF;
 #--------------------------------------------------------------------
 #
 #
@@ -297,33 +220,20 @@ sub data {
 #
 #       Defines to be used in the assembly code.
 #       
-.set r0,0       # we use it as storage for value of 0
-.set SP,1       # preserved
-.set RTOC,2     # preserved 
-.set r3,3       # 1st argument/return value
-.set r4,4       # 2nd argument/volatile register
-.set r5,5       # 3rd argument/volatile register
-.set r6,6       # ...
-.set r7,7
-.set r8,8
-.set r9,9
-.set r10,10
-.set r11,11
-.set r12,12
-.set r13,13     # not used, nor any other "below" it...
-
-.set BO_IF_NOT,4
-.set BO_IF,12
-.set BO_dCTR_NZERO,16
-.set BO_dCTR_ZERO,18
-.set BO_ALWAYS,20
-.set CR0_LT,0;
-.set CR0_GT,1;
-.set CR0_EQ,2
-.set CR1_FX,4;
-.set CR1_FEX,5;
-.set CR1_VX,6
-.set LR,8
+#.set r0,0      # we use it as storage for value of 0
+#.set SP,1      # preserved
+#.set RTOC,2    # preserved 
+#.set r3,3      # 1st argument/return value
+#.set r4,4      # 2nd argument/volatile register
+#.set r5,5      # 3rd argument/volatile register
+#.set r6,6      # ...
+#.set r7,7
+#.set r8,8
+#.set r9,9
+#.set r10,10
+#.set r11,11
+#.set r12,12
+#.set r13,13    # not used, nor any other "below" it...
 
 #       Declare function names to be global
 #       NOTE:   For gcc these names MUST be changed to remove
@@ -344,7 +254,7 @@ sub data {
         
 # .text section
         
-        .machine        $ISA
+        .machine        "any"
 
 #
 #       NOTE:   The following label name should be changed to
@@ -478,7 +388,7 @@ sub data {
 
         $ST             r9,`6*$BNSZ`(r3)        #r[6]=c1
         $ST             r10,`7*$BNSZ`(r3)       #r[7]=c2
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -903,7 +813,7 @@ sub data {
         $ST             r9, `15*$BNSZ`(r3)      #r[15]=c1;
 
 
-        bclr    BO_ALWAYS,CR0_LT
+        blr
 
         .long   0x00000000
 
@@ -1055,7 +965,7 @@ sub data {
 
         $ST     r10,`6*$BNSZ`(r3)       #r[6]=c1
         $ST     r11,`7*$BNSZ`(r3)       #r[7]=c2
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -1591,7 +1501,7 @@ sub data {
         adde    r10,r10,r9
         $ST     r12,`14*$BNSZ`(r3)      #r[14]=c3;
         $ST     r10,`15*$BNSZ`(r3)      #r[15]=c1;
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -1623,7 +1533,7 @@ sub data {
         subfc.  r7,r0,r6        # If r6 is 0 then result is 0.
                                 # if r6 > 0 then result !=0
                                 # In either case carry bit is set.
-        bc      BO_IF,CR0_EQ,Lppcasm_sub_adios
+        beq     Lppcasm_sub_adios
         addi    r4,r4,-$BNSZ
         addi    r3,r3,-$BNSZ
         addi    r5,r5,-$BNSZ
@@ -1635,11 +1545,11 @@ Lppcasm_sub_mainloop:
                                 # if carry = 1 this is r7-r8. Else it
                                 # is r7-r8 -1 as we need.
         $STU    r6,$BNSZ(r3)
-        bc      BO_dCTR_NZERO,CR0_EQ,Lppcasm_sub_mainloop
+        bdnz-   Lppcasm_sub_mainloop
 Lppcasm_sub_adios:      
         subfze  r3,r0           # if carry bit is set then r3 = 0 else -1
         andi.   r3,r3,1         # keep only last bit.
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 
@@ -1670,7 +1580,7 @@ Lppcasm_sub_adios:
 #       check for r6 = 0. Is this needed?
 #
         addic.  r6,r6,0         #test r6 and clear carry bit.
-        bc      BO_IF,CR0_EQ,Lppcasm_add_adios
+        beq     Lppcasm_add_adios
         addi    r4,r4,-$BNSZ
         addi    r3,r3,-$BNSZ
         addi    r5,r5,-$BNSZ
@@ -1680,10 +1590,10 @@ Lppcasm_add_mainloop:
         $LDU    r8,$BNSZ(r5)
         adde    r8,r7,r8
         $STU    r8,$BNSZ(r3)
-        bc      BO_dCTR_NZERO,CR0_EQ,Lppcasm_add_mainloop
+        bdnz-   Lppcasm_add_mainloop
 Lppcasm_add_adios:      
         addze   r3,r0                   #return carry bit.
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -1707,24 +1617,24 @@ Lppcasm_add_adios:
 #       r5 = d
         
         $UCMPI  0,r5,0                  # compare r5 and 0
-        bc      BO_IF_NOT,CR0_EQ,Lppcasm_div1   # proceed if d!=0
+        bne     Lppcasm_div1            # proceed if d!=0
         li      r3,-1                   # d=0 return -1
-        bclr    BO_ALWAYS,CR0_LT        
+        blr
 Lppcasm_div1:
         xor     r0,r0,r0                #r0=0
         li      r8,$BITS
         $CNTLZ. r7,r5                   #r7 = num leading 0s in d.
-        bc      BO_IF,CR0_EQ,Lppcasm_div2       #proceed if no leading zeros
+        beq     Lppcasm_div2            #proceed if no leading zeros
         subf    r8,r7,r8                #r8 = BN_num_bits_word(d)
         $SHR.   r9,r3,r8                #are there any bits above r8'th?
         $TR     16,r9,r0                #if there're, signal to dump core...
 Lppcasm_div2:
         $UCMP   0,r3,r5                 #h>=d?
-        bc      BO_IF,CR0_LT,Lppcasm_div3       #goto Lppcasm_div3 if not
+        blt     Lppcasm_div3            #goto Lppcasm_div3 if not
         subf    r3,r5,r3                #h-=d ; 
 Lppcasm_div3:                           #r7 = BN_BITS2-i. so r7=i
         cmpi    0,0,r7,0                # is (i == 0)?
-        bc      BO_IF,CR0_EQ,Lppcasm_div4
+        beq     Lppcasm_div4
         $SHL    r3,r3,r7                # h = (h<< i)
         $SHR    r8,r4,r8                # r8 = (l >> BN_BITS2 -i)
         $SHL    r5,r5,r7                # d<<=i
@@ -1741,7 +1651,7 @@ Lppcasm_divouterloop:
         $SHRI   r11,r4,`$BITS/2`        #r11= (l&BN_MASK2h)>>BN_BITS4
                                         # compute here for innerloop.
         $UCMP   0,r8,r9                 # is (h>>BN_BITS4)==dh
-        bc      BO_IF_NOT,CR0_EQ,Lppcasm_div5   # goto Lppcasm_div5 if not
+        bne     Lppcasm_div5            # goto Lppcasm_div5 if not
 
         li      r8,-1
         $CLRU   r8,r8,`$BITS/2`         #q = BN_MASK2l 
@@ -1762,9 +1672,9 @@ Lppcasm_divinnerloop:
                                         # the following 2 instructions do that
         $SHLI   r7,r10,`$BITS/2`        # r7 = (t<<BN_BITS4)
         or      r7,r7,r11               # r7|=((l&BN_MASK2h)>>BN_BITS4)
-        $UCMP   1,r6,r7                 # compare (tl <= r7)
-        bc      BO_IF_NOT,CR0_EQ,Lppcasm_divinnerexit
-        bc      BO_IF_NOT,CR1_FEX,Lppcasm_divinnerexit
+        $UCMP   cr1,r6,r7               # compare (tl <= r7)
+        bne     Lppcasm_divinnerexit
+        ble     cr1,Lppcasm_divinnerexit
         addi    r8,r8,-1                #q--
         subf    r12,r9,r12              #th -=dh
         $CLRU   r10,r5,`$BITS/2`        #r10=dl. t is no longer needed in loop.
@@ -1773,14 +1683,14 @@ Lppcasm_divinnerloop:
 Lppcasm_divinnerexit:
         $SHRI   r10,r6,`$BITS/2`        #t=(tl>>BN_BITS4)
         $SHLI   r11,r6,`$BITS/2`        #tl=(tl<<BN_BITS4)&BN_MASK2h;
-        $UCMP   1,r4,r11                # compare l and tl
+        $UCMP   cr1,r4,r11              # compare l and tl
         add     r12,r12,r10             # th+=t
-        bc      BO_IF_NOT,CR1_FX,Lppcasm_div7  # if (l>=tl) goto Lppcasm_div7
+        bge     cr1,Lppcasm_div7        # if (l>=tl) goto Lppcasm_div7
         addi    r12,r12,1               # th++
 Lppcasm_div7:
         subf    r11,r11,r4              #r11=l-tl
-        $UCMP   1,r3,r12                #compare h and th
-        bc      BO_IF_NOT,CR1_FX,Lppcasm_div8   #if (h>=th) goto Lppcasm_div8
+        $UCMP   cr1,r3,r12              #compare h and th
+        bge     cr1,Lppcasm_div8        #if (h>=th) goto Lppcasm_div8
         addi    r8,r8,-1                # q--
         add     r3,r5,r3                # h+=d
 Lppcasm_div8:
@@ -1791,12 +1701,12 @@ Lppcasm_div8:
                                         # the following 2 instructions will do this.
         $INSR   r11,r12,`$BITS/2`,`$BITS/2`     # r11 is the value we want rotated $BITS/2.
         $ROTL   r3,r11,`$BITS/2`        # rotate by $BITS/2 and store in r3
-        bc      BO_dCTR_ZERO,CR0_EQ,Lppcasm_div9#if (count==0) break ;
+        bdz     Lppcasm_div9            #if (count==0) break ;
         $SHLI   r0,r8,`$BITS/2`         #ret =q<<BN_BITS4
         b       Lppcasm_divouterloop
 Lppcasm_div9:
         or      r3,r8,r0
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -1822,7 +1732,7 @@ Lppcasm_div9:
 #       No unrolling done here. Not performance critical.
 
         addic.  r5,r5,0                 #test r5.
-        bc      BO_IF,CR0_EQ,Lppcasm_sqr_adios
+        beq     Lppcasm_sqr_adios
         addi    r4,r4,-$BNSZ
         addi    r3,r3,-$BNSZ
         mtctr   r5
@@ -1833,9 +1743,9 @@ Lppcasm_sqr_mainloop:
         $UMULH  r8,r6,r6
         $STU    r7,$BNSZ(r3)
         $STU    r8,$BNSZ(r3)
-        bc      BO_dCTR_NZERO,CR0_EQ,Lppcasm_sqr_mainloop
+        bdnz-   Lppcasm_sqr_mainloop
 Lppcasm_sqr_adios:      
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 
@@ -1858,7 +1768,7 @@ Lppcasm_sqr_adios:
         xor     r0,r0,r0
         xor     r12,r12,r12             # used for carry
         rlwinm. r7,r5,30,2,31           # num >> 2
-        bc      BO_IF,CR0_EQ,Lppcasm_mw_REM
+        beq     Lppcasm_mw_REM
         mtctr   r7
 Lppcasm_mw_LOOP:        
                                         #mul(rp[0],ap[0],w,c1);
@@ -1896,11 +1806,11 @@ Lppcasm_mw_LOOP:
         
         addi    r3,r3,`4*$BNSZ`
         addi    r4,r4,`4*$BNSZ`
-        bc      BO_dCTR_NZERO,CR0_EQ,Lppcasm_mw_LOOP
+        bdnz-   Lppcasm_mw_LOOP
 
 Lppcasm_mw_REM:
         andi.   r5,r5,0x3
-        bc      BO_IF,CR0_EQ,Lppcasm_mw_OVER
+        beq     Lppcasm_mw_OVER
                                         #mul(rp[0],ap[0],w,c1);
         $LD     r8,`0*$BNSZ`(r4)
         $UMULL  r9,r6,r8
@@ -1912,7 +1822,7 @@ Lppcasm_mw_REM:
         
         addi    r5,r5,-1
         cmpli   0,0,r5,0
-        bc      BO_IF,CR0_EQ,Lppcasm_mw_OVER
+        beq     Lppcasm_mw_OVER
 
         
                                         #mul(rp[1],ap[1],w,c1);
@@ -1926,7 +1836,7 @@ Lppcasm_mw_REM:
         
         addi    r5,r5,-1
         cmpli   0,0,r5,0
-        bc      BO_IF,CR0_EQ,Lppcasm_mw_OVER
+        beq     Lppcasm_mw_OVER
         
                                         #mul_add(rp[2],ap[2],w,c1);
         $LD     r8,`2*$BNSZ`(r4)
@@ -1939,7 +1849,7 @@ Lppcasm_mw_REM:
                 
 Lppcasm_mw_OVER:        
         addi    r3,r12,0
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
 
 #
@@ -1964,7 +1874,7 @@ Lppcasm_mw_OVER:
         xor     r0,r0,r0                #r0 = 0
         xor     r12,r12,r12             #r12 = 0 . used for carry               
         rlwinm. r7,r5,30,2,31           # num >> 2
-        bc      BO_IF,CR0_EQ,Lppcasm_maw_leftover       # if (num < 4) go LPPCASM_maw_leftover
+        beq     Lppcasm_maw_leftover    # if (num < 4) go LPPCASM_maw_leftover
         mtctr   r7
 Lppcasm_maw_mainloop:   
                                         #mul_add(rp[0],ap[0],w,c1);
@@ -2017,11 +1927,11 @@ Lppcasm_maw_mainloop:
         $ST     r11,`3*$BNSZ`(r3)
         addi    r3,r3,`4*$BNSZ`
         addi    r4,r4,`4*$BNSZ`
-        bc      BO_dCTR_NZERO,CR0_EQ,Lppcasm_maw_mainloop
+        bdnz-   Lppcasm_maw_mainloop
         
 Lppcasm_maw_leftover:
         andi.   r5,r5,0x3
-        bc      BO_IF,CR0_EQ,Lppcasm_maw_adios
+        beq     Lppcasm_maw_adios
         addi    r3,r3,-$BNSZ
         addi    r4,r4,-$BNSZ
                                         #mul_add(rp[0],ap[0],w,c1);
@@ -2036,7 +1946,7 @@ Lppcasm_maw_leftover:
         addze   r12,r10
         $ST     r9,0(r3)
         
-        bc      BO_dCTR_ZERO,CR0_EQ,Lppcasm_maw_adios
+        bdz     Lppcasm_maw_adios
                                         #mul_add(rp[1],ap[1],w,c1);
         $LDU    r8,$BNSZ(r4)    
         $UMULL  r9,r6,r8
@@ -2048,7 +1958,7 @@ Lppcasm_maw_leftover:
         addze   r12,r10
         $ST     r9,0(r3)
         
-        bc      BO_dCTR_ZERO,CR0_EQ,Lppcasm_maw_adios
+        bdz     Lppcasm_maw_adios
                                         #mul_add(rp[2],ap[2],w,c1);
         $LDU    r8,$BNSZ(r4)
         $UMULL  r9,r6,r8
@@ -2062,17 +1972,10 @@ Lppcasm_maw_leftover:
                 
 Lppcasm_maw_adios:      
         addi    r3,r12,0
-        bclr    BO_ALWAYS,CR0_LT
+        blr
         .long   0x00000000
         .align  4
 EOF
-        $data =~ s/\`([^\`]*)\`/eval $1/gem;
-
-        # if some assembler chokes on some simplified mnemonic,
-        # this is the spot to fix it up, e.g.:
-        # GNU as doesn't seem to accept cmplw, 32-bit unsigned compare
-        $data =~ s/^(\s*)cmplw(\s+)([^,]+),(.*)/$1cmpl$2$3,0,$4/gm;
-        # assembler X doesn't accept li, load immediate value
-        #$data =~ s/^(\s*)li(\s+)([^,]+),(.*)/$1addi$2$3,0,$4/gm;
-        return($data);
-}
+$data =~ s/\`([^\`]*)\`/eval $1/gem;
+print $data;
+close STDOUT;
diff --git a/src/lib/libcrypto/bn/asm/x86_64-gcc.c b/src/lib/libcrypto/bn/asm/x86_64-gcc.c
index f13f52dd85..acb0b40118 100644
--- a/src/lib/libcrypto/bn/asm/x86_64-gcc.c
+++ b/src/lib/libcrypto/bn/asm/x86_64-gcc.c
@@ -1,4 +1,5 @@
-#ifdef __SUNPRO_C
+#include "../bn_lcl.h"
+#if !(defined(__GNUC__) && __GNUC__>=2)
 # include "../bn_asm.c" /* kind of dirty hack for Sun Studio */
 #else
 /*
@@ -54,7 +55,15 @@
  *    machine.
  */
 
+#ifdef _WIN64
+#define BN_ULONG unsigned long long
+#else
 #define BN_ULONG unsigned long
+#endif
+
+#undef mul
+#undef mul_add
+#undef sqr
 
 /*
  * "m"(a), "+m"(r)      is the way to favor DirectPath µ-code;
@@ -97,7 +106,7 @@
                 : "a"(a)                \
                 : "cc");
 
-BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
         {
         BN_ULONG c1=0;
 
@@ -121,7 +130,7 @@ BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
         return(c1);
         } 
 
-BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
         {
         BN_ULONG c1=0;
 
@@ -144,7 +153,7 @@ BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
         return(c1);
         } 
 
-void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
         {
         if (n <= 0) return;
 
@@ -175,14 +184,14 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
         return ret;
 }
 
-BN_ULONG bn_add_words (BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int n)
+BN_ULONG bn_add_words (BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,int n)
 { BN_ULONG ret=0,i=0;
 
         if (n <= 0) return 0;
 
         asm (
         "       subq    %2,%2           \n"
-        ".align 16                      \n"
+        ".p2align 4                     \n"
         "1:     movq    (%4,%2,8),%0    \n"
         "       adcq    (%5,%2,8),%0    \n"
         "       movq    %0,(%3,%2,8)    \n"
@@ -198,14 +207,14 @@ BN_ULONG bn_add_words (BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int n)
 }
 
 #ifndef SIMICS
-BN_ULONG bn_sub_words (BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int n)
+BN_ULONG bn_sub_words (BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,int n)
 { BN_ULONG ret=0,i=0;
 
         if (n <= 0) return 0;
 
         asm (
         "       subq    %2,%2           \n"
-        ".align 16                      \n"
+        ".p2align 4                     \n"
         "1:     movq    (%4,%2,8),%0    \n"
         "       sbbq    (%5,%2,8),%0    \n"
         "       movq    %0,(%3,%2,8)    \n"
@@ -485,7 +494,7 @@ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
         r[7]=c2;
         }
 
-void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
         {
         BN_ULONG t1,t2;
         BN_ULONG c1,c2,c3;
@@ -561,7 +570,7 @@ void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
         r[15]=c1;
         }
 
-void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
         {
         BN_ULONG t1,t2;
         BN_ULONG c1,c2,c3;
diff --git a/src/lib/libcrypto/bn/asm/x86_64-mont.pl b/src/lib/libcrypto/bn/asm/x86_64-mont.pl
index c43b69592a..3b7a6f243f 100755
--- a/src/lib/libcrypto/bn/asm/x86_64-mont.pl
+++ b/src/lib/libcrypto/bn/asm/x86_64-mont.pl
@@ -15,14 +15,18 @@
 # respectful 50%. It remains to be seen if loop unrolling and
 # dedicated squaring routine can provide further improvement...
 
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
 ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
 die "can't locate x86_64-xlate.pl";
 
-open STDOUT,"| $^X $xlate $output";
+open STDOUT,"| $^X $xlate $flavour $output";
 
 # int bn_mul_mont(
 $rp="%rdi";     # BN_ULONG *rp,
@@ -55,13 +59,14 @@ bn_mul_mont:
         push    %r15
 
         mov     ${num}d,${num}d
-        lea     2($num),%rax
-        mov     %rsp,%rbp
-        neg     %rax
-        lea     (%rsp,%rax,8),%rsp      # tp=alloca(8*(num+2))
+        lea     2($num),%r10
+        mov     %rsp,%r11
+        neg     %r10
+        lea     (%rsp,%r10,8),%rsp      # tp=alloca(8*(num+2))
         and     \$-1024,%rsp            # minimize TLB usage
 
-        mov     %rbp,8(%rsp,$num,8)     # tp[num+1]=%rsp
+        mov     %r11,8(%rsp,$num,8)     # tp[num+1]=%rsp
+.Lprologue:
         mov     %rdx,$bp                # $bp reassigned, remember?
 
         mov     ($n0),$n0               # pull n0[0] value
@@ -197,18 +202,129 @@ bn_mul_mont:
         dec     $j
         jge     .Lcopy
 
-        mov     8(%rsp,$num,8),%rsp     # restore %rsp
+        mov     8(%rsp,$num,8),%rsi     # restore %rsp
         mov     \$1,%rax
+        mov     (%rsi),%r15
+        mov     8(%rsi),%r14
+        mov     16(%rsi),%r13
+        mov     24(%rsi),%r12
+        mov     32(%rsi),%rbp
+        mov     40(%rsi),%rbx
+        lea     48(%rsi),%rsp
+.Lepilogue:
+        ret
+.size   bn_mul_mont,.-bn_mul_mont
+.asciz  "Montgomery Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align  16
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#               CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern __imp_RtlVirtualUnwind
+.type   se_handler,\@abi-omnipotent
+.align  16
+se_handler:
+        push    %rsi
+        push    %rdi
+        push    %rbx
+        push    %rbp
+        push    %r12
+        push    %r13
+        push    %r14
+        push    %r15
+        pushfq
+        sub     \$64,%rsp
+
+        mov     120($context),%rax      # pull context->Rax
+        mov     248($context),%rbx      # pull context->Rip
+
+        lea     .Lprologue(%rip),%r10
+        cmp     %r10,%rbx               # context->Rip<.Lprologue
+        jb      .Lin_prologue
+
+        mov     152($context),%rax      # pull context->Rsp
+
+        lea     .Lepilogue(%rip),%r10
+        cmp     %r10,%rbx               # context->Rip>=.Lepilogue
+        jae     .Lin_prologue
+
+        mov     192($context),%r10      # pull $num
+        mov     8(%rax,%r10,8),%rax     # pull saved stack pointer
+        lea     48(%rax),%rax
+
+        mov     -8(%rax),%rbx
+        mov     -16(%rax),%rbp
+        mov     -24(%rax),%r12
+        mov     -32(%rax),%r13
+        mov     -40(%rax),%r14
+        mov     -48(%rax),%r15
+        mov     %rbx,144($context)      # restore context->Rbx
+        mov     %rbp,160($context)      # restore context->Rbp
+        mov     %r12,216($context)      # restore context->R12
+        mov     %r13,224($context)      # restore context->R13
+        mov     %r14,232($context)      # restore context->R14
+        mov     %r15,240($context)      # restore context->R15
+
+.Lin_prologue:
+        mov     8(%rax),%rdi
+        mov     16(%rax),%rsi
+        mov     %rax,152($context)      # restore context->Rsp
+        mov     %rsi,168($context)      # restore context->Rsi
+        mov     %rdi,176($context)      # restore context->Rdi
+
+        mov     40($disp),%rdi          # disp->ContextRecord
+        mov     $context,%rsi           # context
+        mov     \$154,%ecx              # sizeof(CONTEXT)
+        .long   0xa548f3fc              # cld; rep movsq
+
+        mov     $disp,%rsi
+        xor     %rcx,%rcx               # arg1, UNW_FLAG_NHANDLER
+        mov     8(%rsi),%rdx            # arg2, disp->ImageBase
+        mov     0(%rsi),%r8             # arg3, disp->ControlPc
+        mov     16(%rsi),%r9            # arg4, disp->FunctionEntry
+        mov     40(%rsi),%r10           # disp->ContextRecord
+        lea     56(%rsi),%r11           # &disp->HandlerData
+        lea     24(%rsi),%r12           # &disp->EstablisherFrame
+        mov     %r10,32(%rsp)           # arg5
+        mov     %r11,40(%rsp)           # arg6
+        mov     %r12,48(%rsp)           # arg7
+        mov     %rcx,56(%rsp)           # arg8, (NULL)
+        call    *__imp_RtlVirtualUnwind(%rip)
+
+        mov     \$1,%eax                # ExceptionContinueSearch
+        add     \$64,%rsp
+        popfq
         pop     %r15
         pop     %r14
         pop     %r13
         pop     %r12
         pop     %rbp
         pop     %rbx
+        pop     %rdi
+        pop     %rsi
         ret
-.size   bn_mul_mont,.-bn_mul_mont
-.asciz  "Montgomery Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.size   se_handler,.-se_handler
+
+.section        .pdata
+.align  4
+        .rva    .LSEH_begin_bn_mul_mont
+        .rva    .LSEH_end_bn_mul_mont
+        .rva    .LSEH_info_bn_mul_mont
+
+.section        .xdata
+.align  8
+.LSEH_info_bn_mul_mont:
+        .byte   9,0,0,0
+        .rva    se_handler
 ___
+}
 
 print $code;
 close STDOUT;