I've introduced a bug to i386 RC4 assembler, which would emerge with

certain mix of calls to RC4 routine not covered by rc4test.c.
It's fixed now. In addition this patch inadvertently fixes minor
performance problem: in 0.9.7 context P4 was performing 12% slower
than the original implementation...

diff --git a/crypto/rc4/asm/rc4-586.pl b/crypto/rc4/asm/rc4-586.pl
index 07b2bc6fcdb9777ceef4a631ff846e97e4266426..d6e98f08117f1b580751d2c96db583e1eec85423 100644 (file)
--- a/crypto/rc4/asm/rc4-586.pl
+++ b/crypto/rc4/asm/rc4-586.pl
@@ -68,13 +68,13 @@ sub RC4_loop
        # &mov( $tx,            &DWP(0,$d,$x,4)) if $p < 0;
 
        &add(   &LB($y),        &LB($tx));
        # &mov( $tx,            &DWP(0,$d,$x,4)) if $p < 0;
 
        &add(   &LB($y),        &LB($tx));

-        &inc(  &LB($x));                       # NEXT ROUND

        &mov(   $ty,            &DWP(0,$d,$y,4));
         # XXX
        &mov(   $ty,            &DWP(0,$d,$y,4));
         # XXX

-       &mov(   &DWP(-4,$d,$x,4),$ty);                  # AGI
+       &mov(   &DWP(0,$d,$x,4),$ty);

         &add(  $ty,            $tx);
        &mov(   &DWP(0,$d,$y,4),$tx);
         &and(  $ty,            0xff);
         &add(  $ty,            $tx);
        &mov(   &DWP(0,$d,$y,4),$tx);
         &and(  $ty,            0xff);

+        &inc(  &LB($x));                       # NEXT ROUND

        &mov(   $tx,            &DWP(0,$d,$x,4)) if $p < 1; # NEXT ROUND
         &mov(  $ty,            &DWP(0,$d,$ty,4));
 
        &mov(   $tx,            &DWP(0,$d,$x,4)) if $p < 1; # NEXT ROUND
         &mov(  $ty,            &DWP(0,$d,$ty,4));