From 701d593f7095db84459c76265349a83d30a4cae5 Mon Sep 17 00:00:00 2001
From: Andy Polyakov <appro@openssl.org>
Date: Sun, 15 Jul 2012 13:37:35 +0000
Subject: [PATCH] wp-x86_64.pl: ~10% performance improvement.

---
 crypto/whrlpool/asm/wp-x86_64.pl | 43 +++++++++++++++++---------------
 1 file changed, 23 insertions(+), 20 deletions(-)

diff --git a/crypto/whrlpool/asm/wp-x86_64.pl b/crypto/whrlpool/asm/wp-x86_64.pl
index 9be1cc2238..855aabc892 100644
--- a/crypto/whrlpool/asm/wp-x86_64.pl
+++ b/crypto/whrlpool/asm/wp-x86_64.pl
@@ -90,41 +90,44 @@ for($i=0;$i<8;$i++) { $code.="mov @mm[$i],64+$i*8(%rsp)\n"; }	# S=L
 $code.=<<___;
 	xor	%rsi,%rsi
 	mov	%rsi,24(%rbx)		# zero round counter
+	jmp	.Lround
 .align	16
 .Lround:
 	mov	4096(%rbp,%rsi,8),@mm[0]	# rc[r]
 	mov	0(%rsp),%eax
 	mov	4(%rsp),%ebx
+	movz	%al,%ecx
+	movz	%ah,%edx
 ___
 for($i=0;$i<8;$i++) {
     my $func = ($i==0)? "mov" : "xor";
     $code.=<<___;
-	mov	%al,%cl
-	mov	%ah,%dl
+	shr	\$16,%eax
 	lea	(%rcx,%rcx),%rsi
+	movz	%al,%ecx
 	lea	(%rdx,%rdx),%rdi
-	shr	\$16,%eax
+	movz	%ah,%edx
 	xor	0(%rbp,%rsi,8),@mm[0]
 	$func	7(%rbp,%rdi,8),@mm[1]
-	mov	%al,%cl
-	mov	%ah,%dl
 	mov	$i*8+8(%rsp),%eax		# ($i+1)*8
 	lea	(%rcx,%rcx),%rsi
+	movz	%bl,%ecx
 	lea	(%rdx,%rdx),%rdi
+	movz	%bh,%edx
 	$func	6(%rbp,%rsi,8),@mm[2]
 	$func	5(%rbp,%rdi,8),@mm[3]
-	mov	%bl,%cl
-	mov	%bh,%dl
+	shr	\$16,%ebx
 	lea	(%rcx,%rcx),%rsi
+	movz	%bl,%ecx
 	lea	(%rdx,%rdx),%rdi
-	shr	\$16,%ebx
+	movz	%bh,%edx
 	$func	4(%rbp,%rsi,8),@mm[4]
 	$func	3(%rbp,%rdi,8),@mm[5]
-	mov	%bl,%cl
-	mov	%bh,%dl
 	mov	$i*8+8+4(%rsp),%ebx		# ($i+1)*8+4
 	lea	(%rcx,%rcx),%rsi
+	movz	%al,%ecx
 	lea	(%rdx,%rdx),%rdi
+	movz	%ah,%edx
 	$func	2(%rbp,%rsi,8),@mm[6]
 	$func	1(%rbp,%rdi,8),@mm[7]
 ___
@@ -133,32 +136,32 @@ ___
 for($i=0;$i<8;$i++) { $code.="mov @mm[$i],$i*8(%rsp)\n"; }	# K=L
 for($i=0;$i<8;$i++) {
     $code.=<<___;
-	mov	%al,%cl
-	mov	%ah,%dl
+	shr	\$16,%eax
 	lea	(%rcx,%rcx),%rsi
+	movz	%al,%ecx
 	lea	(%rdx,%rdx),%rdi
-	shr	\$16,%eax
+	movz	%ah,%edx
 	xor	0(%rbp,%rsi,8),@mm[0]
 	xor	7(%rbp,%rdi,8),@mm[1]
-	mov	%al,%cl
-	mov	%ah,%dl
 	`"mov	64+$i*8+8(%rsp),%eax"	if($i<7);`	# 64+($i+1)*8
 	lea	(%rcx,%rcx),%rsi
+	movz	%bl,%ecx
 	lea	(%rdx,%rdx),%rdi
+	movz	%bh,%edx
 	xor	6(%rbp,%rsi,8),@mm[2]
 	xor	5(%rbp,%rdi,8),@mm[3]
-	mov	%bl,%cl
-	mov	%bh,%dl
+	shr	\$16,%ebx
 	lea	(%rcx,%rcx),%rsi
+	movz	%bl,%ecx
 	lea	(%rdx,%rdx),%rdi
-	shr	\$16,%ebx
+	movz	%bh,%edx
 	xor	4(%rbp,%rsi,8),@mm[4]
 	xor	3(%rbp,%rdi,8),@mm[5]
-	mov	%bl,%cl
-	mov	%bh,%dl
 	`"mov	64+$i*8+8+4(%rsp),%ebx"	if($i<7);`	# 64+($i+1)*8+4
 	lea	(%rcx,%rcx),%rsi
+	movz	%al,%ecx
 	lea	(%rdx,%rdx),%rdi
+	movz	%ah,%edx
 	xor	2(%rbp,%rsi,8),@mm[6]
 	xor	1(%rbp,%rdi,8),@mm[7]
 ___
-- 
2.40.0