make update

[openssl.git] / crypto / bn / asm / alpha-mont.pl

diff --git a/crypto/bn/asm/alpha-mont.pl b/crypto/bn/asm/alpha-mont.pl
index d84093754804d82484aca34cabab8952816e8835..03596e2014d4035d02440316cc1cf025f726d772 100644 (file)
--- a/crypto/bn/asm/alpha-mont.pl
+++ b/crypto/bn/asm/alpha-mont.pl
@@ -41,8 +41,12 @@ $j="s4";
 $m1="s5";
 
 $code=<<___;
+#ifdef __linux__
+#include <asm/regdef.h>
+#else
 #include <asm.h>
 #include <regdef.h>
+#endif
 
 .text
 
@@ -53,15 +57,15 @@ $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
@@ -76,7 +80,7 @@ bn_mul_mont:
        ldq     $aj,8($ap)
        subq    sp,AT,sp
        ldq     $bi,0($bp)      # bp[0]
-       mov     -4096,AT
+       lda     AT,-4096(zero)  # mov   -4096,AT
        ldq     $n0,0($n0)
        and     sp,AT,sp
 
@@ -106,9 +110,9 @@ bn_mul_mont:
 .align 4
 .L1st:
        .set    noreorder
-       ldq     $aj,($aj)
+       ldq     $aj,0($aj)
        addl    $j,1,$j
-       ldq     $nj,($nj)
+       ldq     $nj,0($nj)
        lda     $tp,8($tp)
 
        addq    $alo,$hi0,$lo0
@@ -159,12 +163,12 @@ bn_mul_mont:
 .align 4
 .Louter:
        s8addq  $i,$bp,$bi
-       ldq     $hi0,($ap)
+       ldq     $hi0,0($ap)
        ldq     $aj,8($ap)
-       ldq     $bi,($bi)
-       ldq     $hi1,($np)
+       ldq     $bi,0($bi)
+       ldq     $hi1,0($np)
        ldq     $nj,8($np)
-       ldq     $tj,(sp)
+       ldq     $tj,0(sp)
 
        mulq    $hi0,$bi,$lo0
        umulh   $hi0,$bi,$hi0
@@ -195,10 +199,10 @@ bn_mul_mont:
        .set    noreorder
        ldq     $tj,8($tp)      #L0
        nop                     #U1
-       ldq     $aj,($aj)       #L1
+       ldq     $aj,0($aj)      #L1
        s8addq  $j,$np,$nj      #U0
 
-       ldq     $nj,($nj)       #L0
+       ldq     $nj,0($nj)      #L0
        nop                     #U1
        addq    $alo,$hi0,$lo0  #L1
        lda     $tp,8($tp)
@@ -247,7 +251,7 @@ bn_mul_mont:
        addq    $hi1,v0,$hi1
 
        addq    $hi1,$hi0,$lo1
-       stq     $j,($tp)
+       stq     $j,0($tp)
        cmpult  $lo1,$hi0,$hi1
        addq    $lo1,$tj,$lo1
        cmpult  $lo1,$tj,AT
@@ -258,56 +262,46 @@ bn_mul_mont:
        stq     $hi1,16($tp)
        bne     $tj,.Louter
 \f
-       s8addq  $num,sp,$ap
-       mov     $rp,$bp
+       s8addq  $num,sp,$tj     # &tp[num]
+       mov     $rp,$bp         # put rp aside
        mov     sp,$tp
-       mov     0,$hi0
-
-       bne     $hi1,.Lsub
-       cmpult  $nj,$lo1,AT
-       bne     AT,.Lsub
+       mov     sp,$ap
+       mov     0,$hi0          # clear borrow bit
 
 .align 4
-.Lcopy:        ldq     AT,($tp)
+.Lsub: ldq     $lo0,0($tp)
+       ldq     $lo1,0($np)
        lda     $tp,8($tp)
-       stq     AT,($rp)
-       cmpult  $tp,$ap,AT
-       stq     zero,-8($tp)
-       nop
-       lda     $rp,8($rp)
-       bne     AT,.Lcopy
-       mov     1,v0
-       br      .Lexit
-
-.align 4
-.Lsub: ldq     $lo0,($tp)
-       ldq     $lo1,($np)
-       subq    $lo0,$lo1,$lo1
+       lda     $np,8($np)
+       subq    $lo0,$lo1,$lo1  # tp[i]-np[i]
        cmpult  $lo0,$lo1,AT
        subq    $lo1,$hi0,$lo0
        cmpult  $lo1,$lo0,$hi0
-       lda     $tp,8($tp)
        or      $hi0,AT,$hi0
-       lda     $np,8($np)
-       stq     $lo0,($rp)
-       cmpult  $tp,$ap,v0
+       stq     $lo0,0($rp)
+       cmpult  $tp,$tj,v0
        lda     $rp,8($rp)
        bne     v0,.Lsub
 
-       subq    $hi1,$hi0,$hi0
+       subq    $hi1,$hi0,$hi0  # handle upmost overflow bit
        mov     sp,$tp
-       cmpule  $hi1,$hi0,AT
-       mov     $bp,$rp
-       bne     AT,.Lcopy
+       mov     $bp,$rp         # restore rp
+
+       and     sp,$hi0,$ap
+       bic     $bp,$hi0,$bp
+       bis     $bp,$ap,$ap     # ap=borrow?tp:rp
 
 .align 4
-.Lzap: stq     zero,($tp)
-       cmpult  $tp,$ap,AT
+.Lcopy:        ldq     $aj,0($ap)      # copy or in-place refresh
        lda     $tp,8($tp)
-       bne     AT,.Lzap
+       lda     $rp,8($rp)
+       lda     $ap,8($ap)
+       stq     zero,-8($tp)    # zap tp
+       cmpult  $tp,$tj,AT
+       stq     $aj,-8($rp)
+       bne     AT,.Lcopy
        mov     1,v0
 
-.align 4
 .Lexit:
        .set    noreorder
        mov     fp,sp
@@ -316,11 +310,11 @@ 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
-.asciiz        "Montgomery Multiplication for Alpha, CRYPTOGAMS by <appro\@openssl.org>"
+.ascii "Montgomery Multiplication for Alpha, CRYPTOGAMS by <appro\@openssl.org>"
+.align 2
 ___
 
 print $code;