aboutsummaryrefslogtreecommitdiff
path: root/openssl/crypto/md5/asm
diff options
context:
space:
mode:
authormarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
committermarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
commit3562e78743202e43aec8727005182a2558117eca (patch)
tree8f9113a77d12470c5c851a2a8e4cb02e89df7d43 /openssl/crypto/md5/asm
downloadvcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.gz
vcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.bz2
vcxsrv-3562e78743202e43aec8727005182a2558117eca.zip
Checked in the following released items:
xkeyboard-config-1.4.tar.gz ttf-bitstream-vera-1.10.tar.gz font-alias-1.0.1.tar.gz font-sun-misc-1.0.0.tar.gz font-sun-misc-1.0.0.tar.gz font-sony-misc-1.0.0.tar.gz font-schumacher-misc-1.0.0.tar.gz font-mutt-misc-1.0.0.tar.gz font-misc-misc-1.0.0.tar.gz font-misc-meltho-1.0.0.tar.gz font-micro-misc-1.0.0.tar.gz font-jis-misc-1.0.0.tar.gz font-isas-misc-1.0.0.tar.gz font-dec-misc-1.0.0.tar.gz font-daewoo-misc-1.0.0.tar.gz font-cursor-misc-1.0.0.tar.gz font-arabic-misc-1.0.0.tar.gz font-winitzki-cyrillic-1.0.0.tar.gz font-misc-cyrillic-1.0.0.tar.gz font-cronyx-cyrillic-1.0.0.tar.gz font-screen-cyrillic-1.0.1.tar.gz font-xfree86-type1-1.0.1.tar.gz font-adobe-utopia-type1-1.0.1.tar.gz font-ibm-type1-1.0.0.tar.gz font-bitstream-type1-1.0.0.tar.gz font-bitstream-speedo-1.0.0.tar.gz font-bh-ttf-1.0.0.tar.gz font-bh-type1-1.0.0.tar.gz font-bitstream-100dpi-1.0.0.tar.gz font-bh-lucidatypewriter-100dpi-1.0.0.tar.gz font-bh-100dpi-1.0.0.tar.gz font-adobe-utopia-100dpi-1.0.1.tar.gz font-adobe-100dpi-1.0.0.tar.gz font-util-1.0.1.tar.gz font-bitstream-75dpi-1.0.0.tar.gz font-bh-lucidatypewriter-75dpi-1.0.0.tar.gz font-adobe-utopia-75dpi-1.0.1.tar.gz font-bh-75dpi-1.0.0.tar.gz bdftopcf-1.0.1.tar.gz font-adobe-75dpi-1.0.0.tar.gz mkfontscale-1.0.6.tar.gz openssl-0.9.8k.tar.gz bigreqsproto-1.0.2.tar.gz xtrans-1.2.2.tar.gz resourceproto-1.0.2.tar.gz inputproto-1.4.4.tar.gz compositeproto-0.4.tar.gz damageproto-1.1.0.tar.gz zlib-1.2.3.tar.gz xkbcomp-1.0.5.tar.gz freetype-2.3.9.tar.gz pthreads-w32-2-8-0-release.tar.gz pixman-0.12.0.tar.gz kbproto-1.0.3.tar.gz evieext-1.0.2.tar.gz fixesproto-4.0.tar.gz recordproto-1.13.2.tar.gz randrproto-1.2.2.tar.gz scrnsaverproto-1.1.0.tar.gz renderproto-0.9.3.tar.gz xcmiscproto-1.1.2.tar.gz fontsproto-2.0.2.tar.gz xextproto-7.0.3.tar.gz xproto-7.0.14.tar.gz libXdmcp-1.0.2.tar.gz libxkbfile-1.0.5.tar.gz libfontenc-1.0.4.tar.gz libXfont-1.3.4.tar.gz libX11-1.1.5.tar.gz libXau-1.0.4.tar.gz libxcb-1.1.tar.gz xorg-server-1.5.3.tar.gz
Diffstat (limited to 'openssl/crypto/md5/asm')
-rw-r--r--openssl/crypto/md5/asm/md5-586.pl306
-rw-r--r--openssl/crypto/md5/asm/md5-x86_64.pl245
2 files changed, 551 insertions, 0 deletions
diff --git a/openssl/crypto/md5/asm/md5-586.pl b/openssl/crypto/md5/asm/md5-586.pl
new file mode 100644
index 000000000..76ac235f7
--- /dev/null
+++ b/openssl/crypto/md5/asm/md5-586.pl
@@ -0,0 +1,306 @@
+#!/usr/local/bin/perl
+
+# Normal is the
+# md5_block_x86(MD5_CTX *c, ULONG *X);
+# version, non-normal is the
+# md5_block_x86(MD5_CTX *c, ULONG *X,int blocks);
+
+$normal=0;
+
+push(@INC,"perlasm","../../perlasm");
+require "x86asm.pl";
+
+&asm_init($ARGV[0],$0);
+
+$A="eax";
+$B="ebx";
+$C="ecx";
+$D="edx";
+$tmp1="edi";
+$tmp2="ebp";
+$X="esi";
+
+# What we need to load into $tmp for the next round
+%Ltmp1=("R0",&Np($C), "R1",&Np($C), "R2",&Np($C), "R3",&Np($D));
+@xo=(
+ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, # R0
+ 1, 6, 11, 0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12, # R1
+ 5, 8, 11, 14, 1, 4, 7, 10, 13, 0, 3, 6, 9, 12, 15, 2, # R2
+ 0, 7, 14, 5, 12, 3, 10, 1, 8, 15, 6, 13, 4, 11, 2, 9, # R3
+ );
+
+&md5_block("md5_block_asm_data_order");
+&asm_finish();
+
+sub Np
+ {
+ local($p)=@_;
+ local(%n)=($A,$D,$B,$A,$C,$B,$D,$C);
+ return($n{$p});
+ }
+
+sub R0
+ {
+ local($pos,$a,$b,$c,$d,$K,$ki,$s,$t)=@_;
+
+ &mov($tmp1,$C) if $pos < 0;
+ &mov($tmp2,&DWP($xo[$ki]*4,$K,"",0)) if $pos < 0; # very first one
+
+ # body proper
+
+ &comment("R0 $ki");
+ &xor($tmp1,$d); # F function - part 2
+
+ &and($tmp1,$b); # F function - part 3
+ &lea($a,&DWP($t,$a,$tmp2,1));
+
+ &xor($tmp1,$d); # F function - part 4
+
+ &add($a,$tmp1);
+ &mov($tmp1,&Np($c)) if $pos < 1; # next tmp1 for R0
+ &mov($tmp1,&Np($c)) if $pos == 1; # next tmp1 for R1
+
+ &rotl($a,$s);
+
+ &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
+
+ &add($a,$b);
+ }
+
+sub R1
+ {
+ local($pos,$a,$b,$c,$d,$K,$ki,$s,$t)=@_;
+
+ &comment("R1 $ki");
+
+ &lea($a,&DWP($t,$a,$tmp2,1));
+
+ &xor($tmp1,$b); # G function - part 2
+ &and($tmp1,$d); # G function - part 3
+
+ &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
+ &xor($tmp1,$c); # G function - part 4
+
+ &add($a,$tmp1);
+ &mov($tmp1,&Np($c)) if $pos < 1; # G function - part 1
+ &mov($tmp1,&Np($c)) if $pos == 1; # G function - part 1
+
+ &rotl($a,$s);
+
+ &add($a,$b);
+ }
+
+sub R2
+ {
+ local($n,$pos,$a,$b,$c,$d,$K,$ki,$s,$t)=@_;
+ # This one is different, only 3 logical operations
+
+if (($n & 1) == 0)
+ {
+ &comment("R2 $ki");
+ # make sure to do 'D' first, not 'B', else we clash with
+ # the last add from the previous round.
+
+ &xor($tmp1,$d); # H function - part 2
+
+ &xor($tmp1,$b); # H function - part 3
+ &lea($a,&DWP($t,$a,$tmp2,1));
+
+ &add($a,$tmp1);
+
+ &rotl($a,$s);
+
+ &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0));
+ &mov($tmp1,&Np($c));
+ }
+else
+ {
+ &comment("R2 $ki");
+ # make sure to do 'D' first, not 'B', else we clash with
+ # the last add from the previous round.
+
+ &lea($a,&DWP($t,$a,$tmp2,1));
+
+ &add($b,$c); # MOVED FORWARD
+ &xor($tmp1,$d); # H function - part 2
+
+ &xor($tmp1,$b); # H function - part 3
+ &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
+
+ &add($a,$tmp1);
+ &mov($tmp1,&Np($c)) if $pos < 1; # H function - part 1
+ &mov($tmp1,-1) if $pos == 1; # I function - part 1
+
+ &rotl($a,$s);
+
+ &add($a,$b);
+ }
+ }
+
+sub R3
+ {
+ local($pos,$a,$b,$c,$d,$K,$ki,$s,$t)=@_;
+
+ &comment("R3 $ki");
+
+ # &not($tmp1)
+ &xor($tmp1,$d) if $pos < 0; # I function - part 2
+
+ &or($tmp1,$b); # I function - part 3
+ &lea($a,&DWP($t,$a,$tmp2,1));
+
+ &xor($tmp1,$c); # I function - part 4
+ &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if $pos != 2; # load X/k value
+ &mov($tmp2,&wparam(0)) if $pos == 2;
+
+ &add($a,$tmp1);
+ &mov($tmp1,-1) if $pos < 1; # H function - part 1
+ &add($K,64) if $pos >=1 && !$normal;
+
+ &rotl($a,$s);
+
+ &xor($tmp1,&Np($d)) if $pos <= 0; # I function - part = first time
+ &mov($tmp1,&DWP( 0,$tmp2,"",0)) if $pos > 0;
+ &add($a,$b);
+ }
+
+
+sub md5_block
+ {
+ local($name)=@_;
+
+ &function_begin_B($name,"",3);
+
+ # parameter 1 is the MD5_CTX structure.
+ # A 0
+ # B 4
+ # C 8
+ # D 12
+
+ &push("esi");
+ &push("edi");
+ &mov($tmp1, &wparam(0)); # edi
+ &mov($X, &wparam(1)); # esi
+ &mov($C, &wparam(2));
+ &push("ebp");
+ &shl($C, 6);
+ &push("ebx");
+ &add($C, $X); # offset we end at
+ &sub($C, 64);
+ &mov($A, &DWP( 0,$tmp1,"",0));
+ &push($C); # Put on the TOS
+ &mov($B, &DWP( 4,$tmp1,"",0));
+ &mov($C, &DWP( 8,$tmp1,"",0));
+ &mov($D, &DWP(12,$tmp1,"",0));
+
+ &set_label("start") unless $normal;
+ &comment("");
+ &comment("R0 section");
+
+ &R0(-2,$A,$B,$C,$D,$X, 0, 7,0xd76aa478);
+ &R0( 0,$D,$A,$B,$C,$X, 1,12,0xe8c7b756);
+ &R0( 0,$C,$D,$A,$B,$X, 2,17,0x242070db);
+ &R0( 0,$B,$C,$D,$A,$X, 3,22,0xc1bdceee);
+ &R0( 0,$A,$B,$C,$D,$X, 4, 7,0xf57c0faf);
+ &R0( 0,$D,$A,$B,$C,$X, 5,12,0x4787c62a);
+ &R0( 0,$C,$D,$A,$B,$X, 6,17,0xa8304613);
+ &R0( 0,$B,$C,$D,$A,$X, 7,22,0xfd469501);
+ &R0( 0,$A,$B,$C,$D,$X, 8, 7,0x698098d8);
+ &R0( 0,$D,$A,$B,$C,$X, 9,12,0x8b44f7af);
+ &R0( 0,$C,$D,$A,$B,$X,10,17,0xffff5bb1);
+ &R0( 0,$B,$C,$D,$A,$X,11,22,0x895cd7be);
+ &R0( 0,$A,$B,$C,$D,$X,12, 7,0x6b901122);
+ &R0( 0,$D,$A,$B,$C,$X,13,12,0xfd987193);
+ &R0( 0,$C,$D,$A,$B,$X,14,17,0xa679438e);
+ &R0( 1,$B,$C,$D,$A,$X,15,22,0x49b40821);
+
+ &comment("");
+ &comment("R1 section");
+ &R1(-1,$A,$B,$C,$D,$X,16, 5,0xf61e2562);
+ &R1( 0,$D,$A,$B,$C,$X,17, 9,0xc040b340);
+ &R1( 0,$C,$D,$A,$B,$X,18,14,0x265e5a51);
+ &R1( 0,$B,$C,$D,$A,$X,19,20,0xe9b6c7aa);
+ &R1( 0,$A,$B,$C,$D,$X,20, 5,0xd62f105d);
+ &R1( 0,$D,$A,$B,$C,$X,21, 9,0x02441453);
+ &R1( 0,$C,$D,$A,$B,$X,22,14,0xd8a1e681);
+ &R1( 0,$B,$C,$D,$A,$X,23,20,0xe7d3fbc8);
+ &R1( 0,$A,$B,$C,$D,$X,24, 5,0x21e1cde6);
+ &R1( 0,$D,$A,$B,$C,$X,25, 9,0xc33707d6);
+ &R1( 0,$C,$D,$A,$B,$X,26,14,0xf4d50d87);
+ &R1( 0,$B,$C,$D,$A,$X,27,20,0x455a14ed);
+ &R1( 0,$A,$B,$C,$D,$X,28, 5,0xa9e3e905);
+ &R1( 0,$D,$A,$B,$C,$X,29, 9,0xfcefa3f8);
+ &R1( 0,$C,$D,$A,$B,$X,30,14,0x676f02d9);
+ &R1( 1,$B,$C,$D,$A,$X,31,20,0x8d2a4c8a);
+
+ &comment("");
+ &comment("R2 section");
+ &R2( 0,-1,$A,$B,$C,$D,$X,32, 4,0xfffa3942);
+ &R2( 1, 0,$D,$A,$B,$C,$X,33,11,0x8771f681);
+ &R2( 2, 0,$C,$D,$A,$B,$X,34,16,0x6d9d6122);
+ &R2( 3, 0,$B,$C,$D,$A,$X,35,23,0xfde5380c);
+ &R2( 4, 0,$A,$B,$C,$D,$X,36, 4,0xa4beea44);
+ &R2( 5, 0,$D,$A,$B,$C,$X,37,11,0x4bdecfa9);
+ &R2( 6, 0,$C,$D,$A,$B,$X,38,16,0xf6bb4b60);
+ &R2( 7, 0,$B,$C,$D,$A,$X,39,23,0xbebfbc70);
+ &R2( 8, 0,$A,$B,$C,$D,$X,40, 4,0x289b7ec6);
+ &R2( 9, 0,$D,$A,$B,$C,$X,41,11,0xeaa127fa);
+ &R2(10, 0,$C,$D,$A,$B,$X,42,16,0xd4ef3085);
+ &R2(11, 0,$B,$C,$D,$A,$X,43,23,0x04881d05);
+ &R2(12, 0,$A,$B,$C,$D,$X,44, 4,0xd9d4d039);
+ &R2(13, 0,$D,$A,$B,$C,$X,45,11,0xe6db99e5);
+ &R2(14, 0,$C,$D,$A,$B,$X,46,16,0x1fa27cf8);
+ &R2(15, 1,$B,$C,$D,$A,$X,47,23,0xc4ac5665);
+
+ &comment("");
+ &comment("R3 section");
+ &R3(-1,$A,$B,$C,$D,$X,48, 6,0xf4292244);
+ &R3( 0,$D,$A,$B,$C,$X,49,10,0x432aff97);
+ &R3( 0,$C,$D,$A,$B,$X,50,15,0xab9423a7);
+ &R3( 0,$B,$C,$D,$A,$X,51,21,0xfc93a039);
+ &R3( 0,$A,$B,$C,$D,$X,52, 6,0x655b59c3);
+ &R3( 0,$D,$A,$B,$C,$X,53,10,0x8f0ccc92);
+ &R3( 0,$C,$D,$A,$B,$X,54,15,0xffeff47d);
+ &R3( 0,$B,$C,$D,$A,$X,55,21,0x85845dd1);
+ &R3( 0,$A,$B,$C,$D,$X,56, 6,0x6fa87e4f);
+ &R3( 0,$D,$A,$B,$C,$X,57,10,0xfe2ce6e0);
+ &R3( 0,$C,$D,$A,$B,$X,58,15,0xa3014314);
+ &R3( 0,$B,$C,$D,$A,$X,59,21,0x4e0811a1);
+ &R3( 0,$A,$B,$C,$D,$X,60, 6,0xf7537e82);
+ &R3( 0,$D,$A,$B,$C,$X,61,10,0xbd3af235);
+ &R3( 0,$C,$D,$A,$B,$X,62,15,0x2ad7d2bb);
+ &R3( 2,$B,$C,$D,$A,$X,63,21,0xeb86d391);
+
+ # &mov($tmp2,&wparam(0)); # done in the last R3
+ # &mov($tmp1, &DWP( 0,$tmp2,"",0)); # done is the last R3
+
+ &add($A,$tmp1);
+ &mov($tmp1, &DWP( 4,$tmp2,"",0));
+
+ &add($B,$tmp1);
+ &mov($tmp1, &DWP( 8,$tmp2,"",0));
+
+ &add($C,$tmp1);
+ &mov($tmp1, &DWP(12,$tmp2,"",0));
+
+ &add($D,$tmp1);
+ &mov(&DWP( 0,$tmp2,"",0),$A);
+
+ &mov(&DWP( 4,$tmp2,"",0),$B);
+ &mov($tmp1,&swtmp(0)) unless $normal;
+
+ &mov(&DWP( 8,$tmp2,"",0),$C);
+ &mov(&DWP(12,$tmp2,"",0),$D);
+
+ &cmp($tmp1,$X) unless $normal; # check count
+ &jae(&label("start")) unless $normal;
+
+ &pop("eax"); # pop the temp variable off the stack
+ &pop("ebx");
+ &pop("ebp");
+ &pop("edi");
+ &pop("esi");
+ &ret();
+ &function_end_B($name);
+ }
+
diff --git a/openssl/crypto/md5/asm/md5-x86_64.pl b/openssl/crypto/md5/asm/md5-x86_64.pl
new file mode 100644
index 000000000..9a6fa6722
--- /dev/null
+++ b/openssl/crypto/md5/asm/md5-x86_64.pl
@@ -0,0 +1,245 @@
+#!/usr/bin/perl -w
+#
+# MD5 optimized for AMD64.
+#
+# Author: Marc Bevand <bevand_m (at) epita.fr>
+# Licence: I hereby disclaim the copyright on this code and place it
+# in the public domain.
+#
+
+use strict;
+
+my $code;
+
+# round1_step() does:
+# dst = x + ((dst + F(x,y,z) + X[k] + T_i) <<< s)
+# %r10d = X[k_next]
+# %r11d = z' (copy of z for the next step)
+# Each round1_step() takes about 5.71 clocks (9 instructions, 1.58 IPC)
+sub round1_step
+{
+ my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
+ $code .= " mov 0*4(%rsi), %r10d /* (NEXT STEP) X[0] */\n" if ($pos == -1);
+ $code .= " mov %edx, %r11d /* (NEXT STEP) z' = %edx */\n" if ($pos == -1);
+ $code .= <<EOF;
+ xor $y, %r11d /* y ^ ... */
+ lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
+ and $x, %r11d /* x & ... */
+ xor $z, %r11d /* z ^ ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
+ add %r11d, $dst /* dst += ... */
+ rol \$$s, $dst /* dst <<< s */
+ mov $y, %r11d /* (NEXT STEP) z' = $y */
+ add $x, $dst /* dst += x */
+EOF
+}
+
+# round2_step() does:
+# dst = x + ((dst + G(x,y,z) + X[k] + T_i) <<< s)
+# %r10d = X[k_next]
+# %r11d = y' (copy of y for the next step)
+# Each round2_step() takes about 6.22 clocks (9 instructions, 1.45 IPC)
+sub round2_step
+{
+ my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
+ $code .= " mov 1*4(%rsi), %r10d /* (NEXT STEP) X[1] */\n" if ($pos == -1);
+ $code .= " mov %ecx, %r11d /* (NEXT STEP) y' = %ecx */\n" if ($pos == -1);
+ $code .= <<EOF;
+ xor $x, %r11d /* x ^ ... */
+ lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
+ and $z, %r11d /* z & ... */
+ xor $y, %r11d /* y ^ ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
+ add %r11d, $dst /* dst += ... */
+ rol \$$s, $dst /* dst <<< s */
+ mov $x, %r11d /* (NEXT STEP) y' = $x */
+ add $x, $dst /* dst += x */
+EOF
+}
+
+# round3_step() does:
+# dst = x + ((dst + H(x,y,z) + X[k] + T_i) <<< s)
+# %r10d = X[k_next]
+# %r11d = y' (copy of y for the next step)
+# Each round3_step() takes about 4.26 clocks (8 instructions, 1.88 IPC)
+sub round3_step
+{
+ my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
+ $code .= " mov 5*4(%rsi), %r10d /* (NEXT STEP) X[5] */\n" if ($pos == -1);
+ $code .= " mov %ecx, %r11d /* (NEXT STEP) y' = %ecx */\n" if ($pos == -1);
+ $code .= <<EOF;
+ lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
+ xor $z, %r11d /* z ^ ... */
+ xor $x, %r11d /* x ^ ... */
+ add %r11d, $dst /* dst += ... */
+ rol \$$s, $dst /* dst <<< s */
+ mov $x, %r11d /* (NEXT STEP) y' = $x */
+ add $x, $dst /* dst += x */
+EOF
+}
+
+# round4_step() does:
+# dst = x + ((dst + I(x,y,z) + X[k] + T_i) <<< s)
+# %r10d = X[k_next]
+# %r11d = not z' (copy of not z for the next step)
+# Each round4_step() takes about 5.27 clocks (9 instructions, 1.71 IPC)
+sub round4_step
+{
+ my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
+ $code .= " mov 0*4(%rsi), %r10d /* (NEXT STEP) X[0] */\n" if ($pos == -1);
+ $code .= " mov \$0xffffffff, %r11d\n" if ($pos == -1);
+ $code .= " xor %edx, %r11d /* (NEXT STEP) not z' = not %edx*/\n"
+ if ($pos == -1);
+ $code .= <<EOF;
+ lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
+ or $x, %r11d /* x | ... */
+ xor $y, %r11d /* y ^ ... */
+ add %r11d, $dst /* dst += ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
+ mov \$0xffffffff, %r11d
+ rol \$$s, $dst /* dst <<< s */
+ xor $y, %r11d /* (NEXT STEP) not z' = not $y */
+ add $x, $dst /* dst += x */
+EOF
+}
+
+my $output = shift;
+open STDOUT,"| $^X ../perlasm/x86_64-xlate.pl $output";
+
+$code .= <<EOF;
+.text
+.align 16
+
+.globl md5_block_asm_data_order
+.type md5_block_asm_data_order,\@function,3
+md5_block_asm_data_order:
+ push %rbp
+ push %rbx
+ push %r14
+ push %r15
+
+ # rdi = arg #1 (ctx, MD5_CTX pointer)
+ # rsi = arg #2 (ptr, data pointer)
+ # rdx = arg #3 (nbr, number of 16-word blocks to process)
+ mov %rdi, %rbp # rbp = ctx
+ shl \$6, %rdx # rdx = nbr in bytes
+ lea (%rsi,%rdx), %rdi # rdi = end
+ mov 0*4(%rbp), %eax # eax = ctx->A
+ mov 1*4(%rbp), %ebx # ebx = ctx->B
+ mov 2*4(%rbp), %ecx # ecx = ctx->C
+ mov 3*4(%rbp), %edx # edx = ctx->D
+ # end is 'rdi'
+ # ptr is 'rsi'
+ # A is 'eax'
+ # B is 'ebx'
+ # C is 'ecx'
+ # D is 'edx'
+
+ cmp %rdi, %rsi # cmp end with ptr
+ je .Lend # jmp if ptr == end
+
+ # BEGIN of loop over 16-word blocks
+.Lloop: # save old values of A, B, C, D
+ mov %eax, %r8d
+ mov %ebx, %r9d
+ mov %ecx, %r14d
+ mov %edx, %r15d
+EOF
+round1_step(-1,'%eax','%ebx','%ecx','%edx', '1','0xd76aa478', '7');
+round1_step( 0,'%edx','%eax','%ebx','%ecx', '2','0xe8c7b756','12');
+round1_step( 0,'%ecx','%edx','%eax','%ebx', '3','0x242070db','17');
+round1_step( 0,'%ebx','%ecx','%edx','%eax', '4','0xc1bdceee','22');
+round1_step( 0,'%eax','%ebx','%ecx','%edx', '5','0xf57c0faf', '7');
+round1_step( 0,'%edx','%eax','%ebx','%ecx', '6','0x4787c62a','12');
+round1_step( 0,'%ecx','%edx','%eax','%ebx', '7','0xa8304613','17');
+round1_step( 0,'%ebx','%ecx','%edx','%eax', '8','0xfd469501','22');
+round1_step( 0,'%eax','%ebx','%ecx','%edx', '9','0x698098d8', '7');
+round1_step( 0,'%edx','%eax','%ebx','%ecx','10','0x8b44f7af','12');
+round1_step( 0,'%ecx','%edx','%eax','%ebx','11','0xffff5bb1','17');
+round1_step( 0,'%ebx','%ecx','%edx','%eax','12','0x895cd7be','22');
+round1_step( 0,'%eax','%ebx','%ecx','%edx','13','0x6b901122', '7');
+round1_step( 0,'%edx','%eax','%ebx','%ecx','14','0xfd987193','12');
+round1_step( 0,'%ecx','%edx','%eax','%ebx','15','0xa679438e','17');
+round1_step( 1,'%ebx','%ecx','%edx','%eax', '0','0x49b40821','22');
+
+round2_step(-1,'%eax','%ebx','%ecx','%edx', '6','0xf61e2562', '5');
+round2_step( 0,'%edx','%eax','%ebx','%ecx','11','0xc040b340', '9');
+round2_step( 0,'%ecx','%edx','%eax','%ebx', '0','0x265e5a51','14');
+round2_step( 0,'%ebx','%ecx','%edx','%eax', '5','0xe9b6c7aa','20');
+round2_step( 0,'%eax','%ebx','%ecx','%edx','10','0xd62f105d', '5');
+round2_step( 0,'%edx','%eax','%ebx','%ecx','15', '0x2441453', '9');
+round2_step( 0,'%ecx','%edx','%eax','%ebx', '4','0xd8a1e681','14');
+round2_step( 0,'%ebx','%ecx','%edx','%eax', '9','0xe7d3fbc8','20');
+round2_step( 0,'%eax','%ebx','%ecx','%edx','14','0x21e1cde6', '5');
+round2_step( 0,'%edx','%eax','%ebx','%ecx', '3','0xc33707d6', '9');
+round2_step( 0,'%ecx','%edx','%eax','%ebx', '8','0xf4d50d87','14');
+round2_step( 0,'%ebx','%ecx','%edx','%eax','13','0x455a14ed','20');
+round2_step( 0,'%eax','%ebx','%ecx','%edx', '2','0xa9e3e905', '5');
+round2_step( 0,'%edx','%eax','%ebx','%ecx', '7','0xfcefa3f8', '9');
+round2_step( 0,'%ecx','%edx','%eax','%ebx','12','0x676f02d9','14');
+round2_step( 1,'%ebx','%ecx','%edx','%eax', '0','0x8d2a4c8a','20');
+
+round3_step(-1,'%eax','%ebx','%ecx','%edx', '8','0xfffa3942', '4');
+round3_step( 0,'%edx','%eax','%ebx','%ecx','11','0x8771f681','11');
+round3_step( 0,'%ecx','%edx','%eax','%ebx','14','0x6d9d6122','16');
+round3_step( 0,'%ebx','%ecx','%edx','%eax', '1','0xfde5380c','23');
+round3_step( 0,'%eax','%ebx','%ecx','%edx', '4','0xa4beea44', '4');
+round3_step( 0,'%edx','%eax','%ebx','%ecx', '7','0x4bdecfa9','11');
+round3_step( 0,'%ecx','%edx','%eax','%ebx','10','0xf6bb4b60','16');
+round3_step( 0,'%ebx','%ecx','%edx','%eax','13','0xbebfbc70','23');
+round3_step( 0,'%eax','%ebx','%ecx','%edx', '0','0x289b7ec6', '4');
+round3_step( 0,'%edx','%eax','%ebx','%ecx', '3','0xeaa127fa','11');
+round3_step( 0,'%ecx','%edx','%eax','%ebx', '6','0xd4ef3085','16');
+round3_step( 0,'%ebx','%ecx','%edx','%eax', '9', '0x4881d05','23');
+round3_step( 0,'%eax','%ebx','%ecx','%edx','12','0xd9d4d039', '4');
+round3_step( 0,'%edx','%eax','%ebx','%ecx','15','0xe6db99e5','11');
+round3_step( 0,'%ecx','%edx','%eax','%ebx', '2','0x1fa27cf8','16');
+round3_step( 1,'%ebx','%ecx','%edx','%eax', '0','0xc4ac5665','23');
+
+round4_step(-1,'%eax','%ebx','%ecx','%edx', '7','0xf4292244', '6');
+round4_step( 0,'%edx','%eax','%ebx','%ecx','14','0x432aff97','10');
+round4_step( 0,'%ecx','%edx','%eax','%ebx', '5','0xab9423a7','15');
+round4_step( 0,'%ebx','%ecx','%edx','%eax','12','0xfc93a039','21');
+round4_step( 0,'%eax','%ebx','%ecx','%edx', '3','0x655b59c3', '6');
+round4_step( 0,'%edx','%eax','%ebx','%ecx','10','0x8f0ccc92','10');
+round4_step( 0,'%ecx','%edx','%eax','%ebx', '1','0xffeff47d','15');
+round4_step( 0,'%ebx','%ecx','%edx','%eax', '8','0x85845dd1','21');
+round4_step( 0,'%eax','%ebx','%ecx','%edx','15','0x6fa87e4f', '6');
+round4_step( 0,'%edx','%eax','%ebx','%ecx', '6','0xfe2ce6e0','10');
+round4_step( 0,'%ecx','%edx','%eax','%ebx','13','0xa3014314','15');
+round4_step( 0,'%ebx','%ecx','%edx','%eax', '4','0x4e0811a1','21');
+round4_step( 0,'%eax','%ebx','%ecx','%edx','11','0xf7537e82', '6');
+round4_step( 0,'%edx','%eax','%ebx','%ecx', '2','0xbd3af235','10');
+round4_step( 0,'%ecx','%edx','%eax','%ebx', '9','0x2ad7d2bb','15');
+round4_step( 1,'%ebx','%ecx','%edx','%eax', '0','0xeb86d391','21');
+$code .= <<EOF;
+ # add old values of A, B, C, D
+ add %r8d, %eax
+ add %r9d, %ebx
+ add %r14d, %ecx
+ add %r15d, %edx
+
+ # loop control
+ add \$64, %rsi # ptr += 64
+ cmp %rdi, %rsi # cmp end with ptr
+ jb .Lloop # jmp if ptr < end
+ # END of loop over 16-word blocks
+
+.Lend:
+ mov %eax, 0*4(%rbp) # ctx->A = A
+ mov %ebx, 1*4(%rbp) # ctx->B = B
+ mov %ecx, 2*4(%rbp) # ctx->C = C
+ mov %edx, 3*4(%rbp) # ctx->D = D
+
+ pop %r15
+ pop %r14
+ pop %rbx
+ pop %rbp
+ ret
+.size md5_block_asm_data_order,.-md5_block_asm_data_order
+EOF
+
+print $code;
+
+close STDOUT;