spelling fixes, just comments and readme.

[openssl.git] / crypto / sha / asm / sha1-mb-x86_64.pl

diff --git a/crypto/sha/asm/sha1-mb-x86_64.pl b/crypto/sha/asm/sha1-mb-x86_64.pl
index 6a77ff4bf8190e2eec3541245f36671eee3d5e4d..51c73c05ace70ace4cebc3c2ce2e574fc27f7f89 100644 (file)
--- a/crypto/sha/asm/sha1-mb-x86_64.pl
+++ b/crypto/sha/asm/sha1-mb-x86_64.pl
@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2013-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
 
 # ====================================================================
 # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
@@ -19,6 +26,7 @@
 # Sandy Bridge (8.16   +5.15=13.3)/n   4.99    5.98            +80%
 # Ivy Bridge   (8.08   +5.14=13.2)/n   4.60    5.54            +68%
 # Haswell(iii) (8.96   +5.00=14.0)/n   3.57    4.55            +160%
+# Skylake      (8.70   +5.00=13.7)/n   3.64    4.20            +145%
 # Bulldozer    (9.76   +5.76=15.5)/n   5.95    6.37            +64%
 #
 # (i)  multi-block CBC encrypt with 128-bit key;
@@ -58,11 +66,11 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
        $avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /(^clang version|based on LLVM) ([3-9]\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([3-9]\.[0-9]+)/) {
        $avx = ($2>=3.0) + ($2>3.0);
 }
 
-open OUT,"| \"$^X\" $xlate $flavour $output";
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
 *STDOUT=*OUT;
 
 # void sha1_multi_block (
@@ -107,6 +115,19 @@ my ($i,$a,$b,$c,$d,$e)=@_;
 my $j=$i+1;
 my $k=$i+2;
 
+# Loads are performed 2+3/4 iterations in advance. 3/4 means that out
+# of 4 words you would expect to be loaded per given iteration one is
+# spilled to next iteration. In other words indices in four input
+# streams are distributed as following:
+#
+# $i==0:       0,0,0,0,1,1,1,1,2,2,2,
+# $i==1:       2,3,3,3,
+# $i==2:       3,4,4,4,
+# ...
+# $i==13:      14,15,15,15,
+# $i==14:      15
+# 
+# Then at $i==15 Xupdate is applied one iteration in advance...
 $code.=<<___ if ($i==0);
        movd            (@ptr[0]),@Xi[0]
         lea            `16*4`(@ptr[0]),@ptr[0]
@@ -153,7 +174,7 @@ $code.=<<___ if ($i<14);                    # just load input
        psrld   \$2,$b
        paddd   $t2,$e                          # e+=rol(a,5)
         pshufb $tx,@Xi[1]
-        movd           `4*$j-16*4`(@ptr[2]),$t2
+        movd           `4*$k-16*4`(@ptr[2]),$t2
        por     $t1,$b                          # b=rol(b,30)
 ___
 $code.=<<___ if ($i==14);                      # just load input
@@ -465,7 +486,7 @@ $code.=<<___;
        jnz     .Loop_grande
 
 .Ldone:
-       mov     `$REG_SZ*17`(%rsp),%rax         # orignal %rsp
+       mov     `$REG_SZ*17`(%rsp),%rax         # original %rsp
 ___
 $code.=<<___ if ($win64);
        movaps  -0xb8(%rax),%xmm6
@@ -525,7 +546,7 @@ $code.=<<___;
        movdqa  K_XX_XX+0x80(%rip),$BSWAP       # byte-n-word swap
 
 .Loop_grande_shaext:
-       mov     $num,`$REG_SZ*17+8`(%rsp)       # orignal $num
+       mov     $num,`$REG_SZ*17+8`(%rsp)       # original $num
        xor     $num,$num
 ___
 for($i=0;$i<2;$i++) {
@@ -1103,7 +1124,7 @@ $code.=<<___;
        jnz     .Loop_grande_avx
 
 .Ldone_avx:
-       mov     `$REG_SZ*17`(%rsp),%rax         # orignal %rsp
+       mov     `$REG_SZ*17`(%rsp),%rax         # original %rsp
        vzeroupper
 ___
 $code.=<<___ if ($win64);
@@ -1258,7 +1279,7 @@ $code.=<<___;
        #jnz    .Loop_grande_avx2
 
 .Ldone_avx2:
-       mov     `$REG_SZ*17`(%rsp),%rax         # orignal %rsp
+       mov     `$REG_SZ*17`(%rsp),%rax         # original %rsp
        vzeroupper
 ___
 $code.=<<___ if ($win64);