ppc64-mont.pl: adapt for 32-bit and engage for all builds.

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

diff --git a/crypto/bn/asm/ppc-mont.pl b/crypto/bn/asm/ppc-mont.pl
index 280d31244b72de6da467c35af9679172e40d2238..9257b2cd7153b21477e1e457aeac04239b399f35 100644 (file)
--- a/crypto/bn/asm/ppc-mont.pl
+++ b/crypto/bn/asm/ppc-mont.pl
@@ -2,8 +2,9 @@
 
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 
 # April 2006
@@ -23,9 +24,9 @@
 # 2048-bit     +18%
 # 4096-bit     +4%
 
-$output = shift;
+$flavour = shift;
 
-if ($output =~ /32\-mont\.s/) {
+if ($flavour =~ /32/) {
        $BITS=  32;
        $BNSZ=  $BITS/8;
        $SIZE_T=4;
@@ -42,9 +43,10 @@ if ($output =~ /32\-mont\.s/) {
        $UMULL= "mullw";        # unsigned multiply low
        $UMULH= "mulhwu";       # unsigned multiply high
        $UCMP=  "cmplw";        # unsigned compare
+       $SHRI=  "srwi";         # unsigned shift right by immediate     
        $PUSH=  $ST;
        $POP=   $LD;
-} elsif ($output =~ /64\-mont\.s/) {
+} elsif ($flavour =~ /64/) {
        $BITS=  64;
        $BNSZ=  $BITS/8;
        $SIZE_T=8;
@@ -62,12 +64,17 @@ if ($output =~ /32\-mont\.s/) {
        $UMULL= "mulld";        # unsigned multiply low
        $UMULH= "mulhdu";       # unsigned multiply high
        $UCMP=  "cmpld";        # unsigned compare
+       $SHRI=  "srdi";         # unsigned shift right by immediate     
        $PUSH=  $ST;
        $POP=   $LD;
-} else { die "nonsense $output"; }
+} else { die "nonsense $flavour"; }
 
-( defined shift || open STDOUT,"| $^X ../perlasm/ppc-xlate.pl $output" ) ||
-       die "can't call ../perlasm/ppc-xlate.pl: $!";
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $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";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
 
 $sp="r1";
 $toc="r2";
@@ -101,14 +108,19 @@ $code=<<___;
 .machine "any"
 .text
 
-.globl .bn_mul_mont
+.globl .bn_mul_mont_int
 .align 4
-.bn_mul_mont:
+.bn_mul_mont_int:
        cmpwi   $num,4
        mr      $rp,r3          ; $rp is reassigned
        li      r3,0
        bltlr
-
+___
+$code.=<<___ if ($BNSZ==4);
+       cmpwi   $num,32         ; longer key performance is not better
+       bgelr
+___
+$code.=<<___;
        slwi    $num,$num,`log($BNSZ)/log(2)`
        li      $tj,-4096
        addi    $ovf,$num,`$FRAME+$RZONE`
@@ -265,23 +277,33 @@ Linner:
        ble-    Louter
 \f
        addi    $num,$num,2     ; restore $num
+       subfc   $j,$j,$j        ; j=0 and "clear" XER[CA]
        addi    $tp,$sp,$FRAME
        mtctr   $num
+
+.align 4
+Lsub:  $LDX    $tj,$tp,$j
+       $LDX    $nj,$np,$j
+       subfe   $aj,$nj,$tj     ; tp[j]-np[j]
+       $STX    $aj,$rp,$j
+       addi    $j,$j,$BNSZ
+       bdnz-   Lsub
+
        li      $j,0
+       mtctr   $num
+       subfe   $ovf,$j,$ovf    ; handle upmost overflow bit
+       and     $ap,$tp,$ovf
+       andc    $np,$rp,$ovf
+       or      $ap,$ap,$np     ; ap=borrow?tp:rp
 
-       subfc.  $ovf,$j,$ovf    ; sets XER[CA]
-       bne     Lsub
-       $UCMP   $hi1,$nj
-       bge     Lsub
 .align 4
-Lcopy:
-       $LDX    $tj,$tp,$j
+Lcopy:                         ; copy or in-place refresh
+       $LDX    $tj,$ap,$j
        $STX    $tj,$rp,$j
        $STX    $j,$tp,$j       ; zap at once
        addi    $j,$j,$BNSZ
        bdnz-   Lcopy
 
-Lexit:
        $POP    r14,`4*$SIZE_T`($sp)
        $POP    r15,`5*$SIZE_T`($sp)
        $POP    r16,`6*$SIZE_T`($sp)
@@ -298,22 +320,7 @@ Lexit:
        li      r3,1
        blr
        .long   0
-.align 4
-Lsub:  $LDX    $tj,$tp,$j
-       $LDX    $nj,$np,$j
-       subfe   $tj,$nj,$tj     ; tp[j]-np[j]
-       $STX    $tj,$rp,$j
-       addi    $j,$j,$BNSZ
-       bdnz-   Lsub
-       li      $j,0
-       subfe.  $ovf,$j,$ovf
-       mtctr   $num
-       bne     Lcopy
-.align 4
-Lzap:  $STX    $j,$tp,$j
-       addi    $j,$j,$BNSZ
-       bdnz-   Lzap
-       b       Lexit
+.asciz  "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;