X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=blobdiff_plain;f=crypto%2Fcast%2Fasm%2Fcast-586.pl;h=ec2eab179d536d2cf3c5b8417377201340393de6;hp=6be0bfe57245e75fd0f90fafaffffcf4de51bf03;hb=ae4af7a067531245c20f032afe61a68d60d39d11;hpb=2613c1fa2f08814cd276800008b5a20c159ea7ff

diff --git a/crypto/cast/asm/cast-586.pl b/crypto/cast/asm/cast-586.pl
index 6be0bfe572..ec2eab179d 100644
--- a/crypto/cast/asm/cast-586.pl
+++ b/crypto/cast/asm/cast-586.pl
@@ -1,9 +1,13 @@
 #!/usr/local/bin/perl
 
-# define for pentium pro friendly version
+# This flag makes the inner loop one cycle longer, but generates 
+# code that runs %30 faster on the pentium pro/II, 44% faster
+# of PIII, while only %7 slower on the pentium.
+# By default, this flag is on.
 $ppro=1;
 
-push(@INC,"perlasm","../../perlasm");
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
 require "x86asm.pl";
 require "cbc.pl";
 
@@ -139,11 +143,11 @@ sub E_CAST {
     &rotl(	$tmp4,		&LB($tmp1));
 
     if ($ppro) {
-	&mov(	$tmp2,		$tmp4);		# B
 	&xor(	$tmp1,		$tmp1);
+	&mov(	$tmp2,		0xff);
 	
 	&movb(	&LB($tmp1),	&HB($tmp4));	# A
-	&and(	$tmp2,		0xff);
+	&and(	$tmp2,		$tmp4);
 
 	&shr(	$tmp4,		16); 		#
 	&xor(	$tmp3,		$tmp3);