AES x86_64 assembler implementation.
authorAndy Polyakov <appro@openssl.org>
Tue, 12 Jul 2005 15:44:58 +0000 (15:44 +0000)
committerAndy Polyakov <appro@openssl.org>
Tue, 12 Jul 2005 15:44:58 +0000 (15:44 +0000)
Configure
crypto/aes/Makefile
crypto/aes/asm/aes-x86_64.pl [new file with mode: 0755]

index 933246a15709aa96a31e487734842364b3017e0b..a214db36a4d984b5365811a8f049431209a5bcab 100755 (executable)
--- a/Configure
+++ b/Configure
@@ -118,7 +118,7 @@ my $x86_elf_asm="x86cpuid-elf.o:bn86-elf.o co86-elf.o:dx86-elf.o yx86-elf.o:ax86
 my $x86_coff_asm="x86cpuid-cof.o:bn86-cof.o co86-cof.o:dx86-cof.o yx86-cof.o:ax86-cof.o:bx86-cof.o:mx86-cof.o:sx86-cof.o s512sse2-cof.o:cx86-cof.o:rx86-cof.o:rm86-cof.o:r586-cof.o";
 my $x86_out_asm="x86cpuid-out.o:bn86-out.o co86-out.o:dx86-out.o yx86-out.o:ax86-out.o:bx86-out.o:mx86-out.o:sx86-out.o s512sse2-out.o:cx86-out.o:rx86-out.o:rm86-out.o:r586-out.o";
 
-my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o::::md5-x86_64.o:::rc4-x86_64.o::";
+my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o::aes-x86_64.o::md5-x86_64.o:::rc4-x86_64.o::";
 my $ia64_asm=":bn-ia64.o::aes_core.o aes_cbc.o aes-ia64.o:::sha1-ia64.o sha256-ia64.o sha512-ia64.o::rc4-ia64.o::";
 
 my $no_asm="::::::::::";
index 057b70bdd1578dfa329f23bcbda0720297b59171..eed6542aaa66ac13dc88de5af9dcb6b066097812 100644 (file)
@@ -55,6 +55,9 @@ ax86-cof.s: asm/aes-586.pl ../perlasm/x86asm.pl
 ax86-out.s: asm/aes-586.pl ../perlasm/x86asm.pl
        (cd asm; $(PERL) aes-586.pl a.out $(CFLAGS) $(PROCESSOR) > ../$@)
 
+aes-x86_64.s: asm/aes-x86_64.pl
+       $(PERL) asm/aes-x86_64.pl $@
+
 files:
        $(PERL) $(TOP)/util/files.pl Makefile >> $(TOP)/MINFO
 
diff --git a/crypto/aes/asm/aes-x86_64.pl b/crypto/aes/asm/aes-x86_64.pl
new file mode 100755 (executable)
index 0000000..3113e35
--- /dev/null
@@ -0,0 +1,1258 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. Rights for redistribution and usage in source and binary
+# forms are granted according to the OpenSSL license.
+# ====================================================================
+#
+# Version 1.0.
+#
+# aes-*-cbc benchmarks are improved by 50% [compared to gcc 3.3.2 on
+# Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version
+# [you'll notice a lot of resemblance], such as compressed S-boxs
+# and prefetch of these tables in CBC mode, as well as avoiding L1
+# cache aliasing between stack frame and key schedule and already
+# mentioned tables.
+#
+# ECB performance is 15.6 cycles per processed byte for 128-bit key.
+
+$output=shift;
+open STDOUT,"| $^X ../perlasm/x86_64-xlate.pl $output";
+
+$code=".text\n";
+
+$s0="%eax";
+$s1="%ebx";
+$s2="%ecx";
+$s3="%edx";
+$inp="%rdi";
+$out="%rsi";
+$acc0="%ebp";
+$acc1="%r8d";
+$acc2="%r9d";
+$t0="%r10d";
+$t1="%r11d";
+$t2="%r12d";
+$cnt="%r13d";
+$tbl="%r14";
+$key="%r15";
+
+sub hi() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1h/; $r; }
+sub lo() { my $r=shift; $r =~ s/%[er]([a-d])x/%\1l/; $r; }
+sub _data_word()
+{ my $i;
+    while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
+}
+sub data_word()
+{ my $i;
+  my $last=pop(@_);
+    $code.=".long\t";
+    while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; }
+    $code.=sprintf"0x%08x\n",$last;
+}
+
+sub encstep()
+{ my ($i,@s) = @_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+       $code.="        mov     $s[0],$out\n"           if ($i!=3);
+                       $tmp1=$s[2]                     if ($i==3);
+       $code.="        mov     $s[2],$tmp1\n"          if ($i!=3);
+       $code.="        and     \$0xFF,$out\n";
+
+       $code.="        mov     0($tbl,$out,8),$out\n";
+       $code.="        shr     \$16,$tmp1\n";
+                       $tmp2=$s[3]                     if ($i==3);
+       $code.="        mov     $s[3],$tmp2\n"          if ($i!=3);
+
+                       $tmp0=$s[1]                     if ($i==3);
+       $code.="        movzb   ".&hi($s[1]).",$tmp0\n";
+       $code.="        and     \$0xFF,$tmp1\n";
+       $code.="        shr     \$24,$tmp2\n";
+
+       $code.="        xor     3($tbl,$tmp0,8),$out\n";
+       $code.="        xor     2($tbl,$tmp1,8),$out\n";
+       $code.="        xor     1($tbl,$tmp2,8),$out\n";
+
+       $code.="        mov     $t0,$s[1]\n"            if ($i==3);
+       $code.="        mov     $t1,$s[2]\n"            if ($i==3);
+       $code.="        mov     $t2,$s[3]\n"            if ($i==3);
+       $code.="\n";
+}
+
+sub enclast()
+{ my ($i,@s)=@_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+       $code.="        mov     $s[0],$out\n"           if ($i!=3);
+                       $tmp1=$s[2]                     if ($i==3);
+       $code.="        mov     $s[2],$tmp1\n"          if ($i!=3);
+       $code.="        and     \$0xFF,$out\n";
+
+       $code.="        mov     2($tbl,$out,8),$out\n";
+       $code.="        shr     \$16,$tmp1\n";
+                       $tmp2=$s[3]                     if ($i==3);
+       $code.="        mov     $s[3],$tmp2\n"          if ($i!=3);
+
+       $code.="        and     \$0x000000ff,$out\n";
+                       $tmp0=$s[1]                     if ($i==3);
+       $code.="        movzb   ".&hi($s[1]).",$tmp0\n";
+       $code.="        and     \$0xFF,$tmp1\n";
+       $code.="        shr     \$24,$tmp2\n";
+
+       $code.="        mov     0($tbl,$tmp0,8),$tmp0\n";
+       $code.="        mov     0($tbl,$tmp1,8),$tmp1\n";
+       $code.="        mov     2($tbl,$tmp2,8),$tmp2\n";
+
+       $code.="        and     \$0x0000ff00,$tmp0\n";
+       $code.="        and     \$0x00ff0000,$tmp1\n";
+       $code.="        and     \$0xff000000,$tmp2\n";
+
+       $code.="        xor     $tmp0,$out\n";
+       $code.="        mov     $t0,$s[1]\n"            if ($i==3);
+       $code.="        xor     $tmp1,$out\n";
+       $code.="        mov     $t1,$s[2]\n"            if ($i==3);
+       $code.="        xor     $tmp2,$out\n";
+       $code.="        mov     $t2,$s[3]\n"            if ($i==3);
+       $code.="\n";
+}
+
+$code.=<<___;
+.type  _x86_64_AES_encrypt,\@abi-omnipotent
+.align 16
+_x86_64_AES_encrypt:
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       mov     240($key),$cnt                  # load key->rounds
+       sub     \$1,$cnt
+.align 4
+.Lenc_loop:
+___
+       &encstep(0,$s0,$s1,$s2,$s3);
+       &encstep(1,$s1,$s2,$s3,$s0);
+       &encstep(2,$s2,$s3,$s0,$s1);
+       &encstep(3,$s3,$s0,$s1,$s2);
+$code.=<<___;
+       lea     16($key),$key
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       sub     \$1,$cnt
+       jnz     .Lenc_loop
+___
+       &enclast(0,$s0,$s1,$s2,$s3);
+       &enclast(1,$s1,$s2,$s3,$s0);
+       &enclast(2,$s2,$s3,$s0,$s1);
+       &enclast(3,$s3,$s0,$s1,$s2);
+$code.=<<___;
+       lea     16($key),$key
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       .byte   0xf3,0xc3                       # rep ret
+.size  _x86_64_AES_encrypt,.-_x86_64_AES_encrypt
+___
+
+# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
+$code.=<<___;
+.globl AES_encrypt
+.type  AES_encrypt,\@function,3
+.align 16
+AES_encrypt:
+       push    %rbx
+       push    %rbp
+       push    %r12
+       push    %r13
+       push    %r14
+       push    %r15
+
+       mov     %rdx,$key
+
+       .picmeup        $tbl
+       lea     AES_Te-.($tbl),$tbl
+
+       mov     0($inp),$s0
+       mov     4($inp),$s1
+       mov     8($inp),$s2
+       mov     12($inp),$s3
+
+       call    _x86_64_AES_encrypt
+
+       mov     $s0,0($out)
+       mov     $s1,4($out)
+       mov     $s2,8($out)
+       mov     $s3,12($out)
+
+       pop     %r15
+       pop     %r14
+       pop     %r13
+       pop     %r12
+       pop     %rbp
+       pop     %rbx
+       ret
+.size  AES_encrypt,.-AES_encrypt
+___
+
+#------------------------------------------------------------------#
+
+sub decstep()
+{ my ($i,@s) = @_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+       $code.="        mov     $s[0],$out\n"           if ($i!=3);
+                       $tmp1=$s[2]                     if ($i==3);
+       $code.="        mov     $s[2],$tmp1\n"          if ($i!=3);
+       $code.="        and     \$0xFF,$out\n";
+
+       $code.="        mov     0($tbl,$out,8),$out\n";
+       $code.="        shr     \$16,$tmp1\n";
+                       $tmp2=$s[3]                     if ($i==3);
+       $code.="        mov     $s[3],$tmp2\n"          if ($i!=3);
+
+                       $tmp0=$s[1]                     if ($i==3);
+       $code.="        movzb   ".&hi($s[1]).",$tmp0\n";
+       $code.="        and     \$0xFF,$tmp1\n";
+       $code.="        shr     \$24,$tmp2\n";
+
+       $code.="        xor     3($tbl,$tmp0,8),$out\n";
+       $code.="        xor     2($tbl,$tmp1,8),$out\n";
+       $code.="        xor     1($tbl,$tmp2,8),$out\n";
+
+       $code.="        mov     $t2,$s[1]\n"            if ($i==3);
+       $code.="        mov     $t1,$s[2]\n"            if ($i==3);
+       $code.="        mov     $t0,$s[3]\n"            if ($i==3);
+       $code.="\n";
+}
+
+sub declast()
+{ my ($i,@s)=@_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+       $code.="        mov     $s[0],$out\n"           if ($i!=3);
+                       $tmp1=$s[2]                     if ($i==3);
+       $code.="        mov     $s[2],$tmp1\n"          if ($i!=3);
+       $code.="        and     \$0xFF,$out\n";
+
+       $code.="        mov     2048($tbl,$out,4),$out\n";
+       $code.="        shr     \$16,$tmp1\n";
+                       $tmp2=$s[3]                     if ($i==3);
+       $code.="        mov     $s[3],$tmp2\n"          if ($i!=3);
+
+       $code.="        and     \$0x000000ff,$out\n";
+                       $tmp0=$s[1]                     if ($i==3);
+       $code.="        movzb   ".&hi($s[1]).",$tmp0\n";
+       $code.="        and     \$0xFF,$tmp1\n";
+       $code.="        shr     \$24,$tmp2\n";
+
+       $code.="        mov     2048($tbl,$tmp0,4),$tmp0\n";
+       $code.="        mov     2048($tbl,$tmp1,4),$tmp1\n";
+       $code.="        mov     2048($tbl,$tmp2,4),$tmp2\n";
+
+       $code.="        and     \$0x0000ff00,$tmp0\n";
+       $code.="        and     \$0x00ff0000,$tmp1\n";
+       $code.="        and     \$0xff000000,$tmp2\n";
+
+       $code.="        xor     $tmp0,$out\n";
+       $code.="        mov     $t2,$s[1]\n"            if ($i==3);
+       $code.="        xor     $tmp1,$out\n";
+       $code.="        mov     $t1,$s[2]\n"            if ($i==3);
+       $code.="        xor     $tmp2,$out\n";
+       $code.="        mov     $t0,$s[3]\n"            if ($i==3);
+       $code.="\n";
+}
+
+$code.=<<___;
+.type  _x86_64_AES_decrypt,\@abi-omnipotent
+.align 16
+_x86_64_AES_decrypt:
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       mov     240($key),$cnt                  # load key->rounds
+       sub     \$1,$cnt
+.align 4
+.Ldec_loop:
+___
+       &decstep(0,$s0,$s3,$s2,$s1);
+       &decstep(1,$s1,$s0,$s3,$s2);
+       &decstep(2,$s2,$s1,$s0,$s3);
+       &decstep(3,$s3,$s2,$s1,$s0);
+$code.=<<___;
+       lea     16($key),$key
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       sub     \$1,$cnt
+       jnz     .Ldec_loop
+___
+       &declast(0,$s0,$s3,$s2,$s1);
+       &declast(1,$s1,$s0,$s3,$s2);
+       &declast(2,$s2,$s1,$s0,$s3);
+       &declast(3,$s3,$s2,$s1,$s0);
+$code.=<<___;
+       lea     16($key),$key
+       xor     0($key),$s0                     # xor with key
+       xor     4($key),$s1
+       xor     8($key),$s2
+       xor     12($key),$s3
+
+       .byte   0xf3,0xc3                       # rep ret
+.size  _x86_64_AES_decrypt,.-_x86_64_AES_decrypt
+___
+
+# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
+$code.=<<___;
+.globl AES_decrypt
+.type  AES_decrypt,\@function,3
+.align 16
+AES_decrypt:
+       push    %rbx
+       push    %rbp
+       push    %r12
+       push    %r13
+       push    %r14
+       push    %r15
+
+       mov     %rdx,$key
+
+       .picmeup        $tbl
+       lea     AES_Td-.($tbl),$tbl
+
+       mov     0($inp),$s0
+       mov     4($inp),$s1
+       mov     8($inp),$s2
+       mov     12($inp),$s3
+
+       call    _x86_64_AES_decrypt
+
+       mov     $s0,0($out)
+       mov     $s1,4($out)
+       mov     $s2,8($out)
+       mov     $s3,12($out)
+
+       pop     %r15
+       pop     %r14
+       pop     %r13
+       pop     %r12
+       pop     %rbp
+       pop     %rbx
+       ret
+.size  AES_decrypt,.-AES_decrypt
+___
+#------------------------------------------------------------------#
+
+sub enckey()
+{
+$code.=<<___;
+       movz    %dl,%esi                # rk[i]>>0
+       mov     2(%rbp,%rsi,8),%ebx
+       movz    %dh,%esi                # rk[i]>>8
+       and     \$0xFF000000,%ebx
+       xor     %ebx,%eax
+
+       mov     2(%rbp,%rsi,8),%ebx
+       shr     \$16,%edx
+       and     \$0x000000FF,%ebx
+       movz    %dl,%esi                # rk[i]>>16
+       xor     %ebx,%eax
+
+       mov     0(%rbp,%rsi,8),%ebx
+       movz    %dh,%esi                # rk[i]>>24
+       and     \$0x0000FF00,%ebx
+       xor     %ebx,%eax
+
+       mov     0(%rbp,%rsi,8),%ebx
+       and     \$0x00FF0000,%ebx
+       xor     %ebx,%eax
+
+       xor     2048(%rbp,%rcx,4),%eax          # rcon
+___
+}
+
+# int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+$code.=<<___;
+.globl AES_set_encrypt_key
+.type  AES_set_encrypt_key,\@function,3
+.align 16
+AES_set_encrypt_key:
+       push    %rbx
+       push    %rbp
+
+       mov     %esi,%ecx                       # %ecx=bits
+       mov     %rdi,%rsi                       # %rsi=userKey
+       mov     %rdx,%rdi                       # %rdi=key
+
+       test    \$-1,%rsi
+       jz      .Lbadpointer
+       test    \$-1,%rdi
+       jz      .Lbadpointer
+
+       .picmeup %rbp
+       lea     AES_Te-.(%rbp),%rbp
+
+       cmp     \$128,%ecx
+       je      .L10rounds
+       cmp     \$192,%ecx
+       je      .L12rounds
+       cmp     \$256,%ecx
+       je      .L14rounds
+       mov     \$-2,%rax                       # invalid number of bits
+       jmp     .Lexit
+
+.L10rounds:
+       mov     0(%rsi),%eax                    # copy first 4 dwords
+       mov     4(%rsi),%ebx
+       mov     8(%rsi),%ecx
+       mov     12(%rsi),%edx
+       mov     %eax,0(%rdi)
+       mov     %ebx,4(%rdi)
+       mov     %ecx,8(%rdi)
+       mov     %edx,12(%rdi)
+
+       xor     %ecx,%ecx
+       jmp     .L10shortcut
+.align 4
+.L10loop:
+               mov     0(%rdi),%eax                    # rk[0]
+               mov     12(%rdi),%edx                   # rk[3]
+.L10shortcut:
+___
+               &enckey ();
+$code.=<<___;
+               mov     %eax,16(%rdi)                   # rk[4]
+               xor     4(%rdi),%eax
+               mov     %eax,20(%rdi)                   # rk[5]
+               xor     8(%rdi),%eax
+               mov     %eax,24(%rdi)                   # rk[6]
+               xor     12(%rdi),%eax
+               mov     %eax,28(%rdi)                   # rk[7]
+               add     \$1,%ecx
+               lea     16(%rdi),%rdi
+               cmp     \$10,%ecx
+       jl      .L10loop
+
+       movl    \$10,80(%rdi)                   # setup number of rounds
+       xor     %rax,%rax
+       jmp     .Lexit
+
+.L12rounds:
+       mov     0(%rsi),%eax                    # copy first 6 dwords
+       mov     4(%rsi),%ebx
+       mov     8(%rsi),%ecx
+       mov     12(%rsi),%edx
+       mov     %eax,0(%rdi)
+       mov     %ebx,4(%rdi)
+       mov     %ecx,8(%rdi)
+       mov     %edx,12(%rdi)
+       mov     16(%rsi),%ecx
+       mov     20(%rsi),%edx
+       mov     %ecx,16(%rdi)
+       mov     %edx,20(%rdi)
+
+       xor     %ecx,%ecx
+       jmp     .L12shortcut
+.align 4
+.L12loop:
+               mov     0(%rdi),%eax                    # rk[0]
+               mov     20(%rdi),%edx                   # rk[5]
+.L12shortcut:
+___
+               &enckey ();
+$code.=<<___;
+               mov     %eax,24(%rdi)                   # rk[6]
+               xor     4(%rdi),%eax
+               mov     %eax,28(%rdi)                   # rk[7]
+               xor     8(%rdi),%eax
+               mov     %eax,32(%rdi)                   # rk[8]
+               xor     12(%rdi),%eax
+               mov     %eax,36(%rdi)                   # rk[9]
+
+               cmp     \$7,%ecx
+               je      .L12break
+               add     \$1,%ecx
+
+               xor     16(%rdi),%eax
+               mov     %eax,40(%rdi)                   # rk[10]
+               xor     20(%rdi),%eax
+               mov     %eax,44(%rdi)                   # rk[11]
+
+               lea     24(%rdi),%rdi
+       jmp     .L12loop
+.L12break:
+       movl    \$12,72(%rdi)           # setup number of rounds
+       xor     %rax,%rax
+       jmp     .Lexit
+
+.L14rounds:            
+       mov     0(%rsi),%eax                    # copy first 8 dwords
+       mov     4(%rsi),%ebx
+       mov     8(%rsi),%ecx
+       mov     12(%rsi),%edx
+       mov     %eax,0(%rdi)
+       mov     %ebx,4(%rdi)
+       mov     %ecx,8(%rdi)
+       mov     %edx,12(%rdi)
+       mov     16(%rsi),%eax
+       mov     20(%rsi),%ebx
+       mov     24(%rsi),%ecx
+       mov     28(%rsi),%edx
+       mov     %eax,16(%rdi)
+       mov     %ebx,20(%rdi)
+       mov     %ecx,24(%rdi)
+       mov     %edx,28(%rdi)
+
+       xor     %ecx,%ecx
+       jmp     .L14shortcut
+.align 4
+.L14loop:
+               mov     28(%rdi),%edx                   # rk[4]
+.L14shortcut:
+               mov     0(%rdi),%eax                    # rk[0]
+___
+               &enckey ();
+$code.=<<___;
+               mov     %eax,32(%rdi)                   # rk[8]
+               xor     4(%rdi),%eax
+               mov     %eax,36(%rdi)                   # rk[9]
+               xor     8(%rdi),%eax
+               mov     %eax,40(%rdi)                   # rk[10]
+               xor     12(%rdi),%eax
+               mov     %eax,44(%rdi)                   # rk[11]
+
+               cmp     \$6,%ecx
+               je      .L14break
+               add     \$1,%ecx
+
+               mov     %eax,%edx
+               mov     16(%rdi),%eax                   # rk[4]
+               movz    %dl,%esi                        # rk[11]>>0
+               mov     2(%rbp,%rsi,8),%ebx
+               movz    %dh,%esi                        # rk[11]>>8
+               and     \$0x000000FF,%ebx
+               xor     %ebx,%eax
+
+               mov     0(%rbp,%rsi,8),%ebx
+               shr     \$16,%edx
+               and     \$0x0000FF00,%ebx
+               movz    %dl,%esi                        # rk[11]>>16
+               xor     %ebx,%eax
+
+               mov     0(%rbp,%rsi,8),%ebx
+               movz    %dh,%esi                        # rk[11]>>24
+               and     \$0x00FF0000,%ebx
+               xor     %ebx,%eax
+
+               mov     2(%rbp,%rsi,8),%ebx
+               and     \$0xFF000000,%ebx
+               xor     %ebx,%eax
+
+               mov     %eax,48(%rdi)                   # rk[12]
+               xor     20(%rdi),%eax
+               mov     %eax,52(%rdi)                   # rk[13]
+               xor     24(%rdi),%eax
+               mov     %eax,56(%rdi)                   # rk[14]
+               xor     28(%rdi),%eax
+               mov     %eax,60(%rdi)                   # rk[15]
+
+               lea     32(%rdi),%rdi
+       jmp     .L14loop
+.L14break:
+       movl    \$14,48(%rdi)           # setup number of rounds
+       xor     %rax,%rax
+       jmp     .Lexit
+
+.Lbadpointer:
+       mov     \$-1,%rax
+.Lexit:
+       pop     %rbp
+       pop     %rbx
+       ret
+.size  AES_set_encrypt_key,.-AES_set_encrypt_key
+___
+
+sub deckey()
+{ my ($i,$ptr,$te,$td) = @_;
+$code.=<<___;
+       mov     $i($ptr),%eax
+       mov     %eax,%edx
+       movz    %ah,%ebx
+       shr     \$16,%edx
+       and     \$0xFF,%eax
+       movzb   2($te,%rax,8),%rax
+       movzb   2($te,%rbx,8),%rbx
+       mov     0($td,%rax,8),%eax
+       xor     3($td,%rbx,8),%eax
+       movzb   %dh,%ebx
+       and     \$0xFF,%edx
+       movzb   2($te,%rdx,8),%rdx
+       movzb   2($te,%rbx,8),%rbx
+       xor     2($td,%rdx,8),%eax
+       xor     1($td,%rbx,8),%eax
+       mov     %eax,$i($ptr)
+___
+}
+
+# int AES_set_decrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+$code.=<<___;
+.globl AES_set_decrypt_key
+.type  AES_set_decrypt_key,\@function,3
+.align 16
+AES_set_decrypt_key:
+       push    %rdx
+       call    AES_set_encrypt_key
+       cmp     \$0,%eax
+       je      .Lproceed
+       lea     24(%rsp),%rsp
+       ret
+.Lproceed:
+       mov     (%rsp),%r8              # restore key schedule
+       mov     %rbx,(%rsp)
+
+       mov     240(%r8),%ecx           # pull number of rounds
+       xor     %rdi,%rdi
+       lea     (%rdi,%rcx,4),%rcx
+       mov     %r8,%rsi
+       lea     (%r8,%rcx,4),%rdi       # pointer to last chunk
+.align 4
+.Linvert:
+               mov     0(%rsi),%rax
+               mov     8(%rsi),%rbx
+               mov     0(%rdi),%rcx
+               mov     8(%rdi),%rdx
+               mov     %rax,0(%rdi)
+               mov     %rbx,8(%rdi)
+               mov     %rcx,0(%rsi)
+               mov     %rdx,8(%rsi)
+               lea     16(%rsi),%rsi
+               lea     -16(%rdi),%rdi
+               cmp     %rsi,%rdi
+       jne     .Linvert
+
+       .picmeup %r9
+       lea     AES_Td-.(%r9),%rdi
+       lea     AES_Te-AES_Td(%rdi),%r9
+
+       mov     %r8,%rsi
+       mov     240(%r8),%ecx           # pull number of rounds
+       sub     \$1,%ecx
+.align 4
+.Lpermute:
+               lea     16(%rsi),%rsi
+___
+               &deckey (0,"%rsi","%r9","%rdi");
+               &deckey (4,"%rsi","%r9","%rdi");
+               &deckey (8,"%rsi","%r9","%rdi");
+               &deckey (12,"%rsi","%r9","%rdi");
+$code.=<<___;
+               sub     \$1,%ecx
+       jnz     .Lpermute
+
+       xor     %rax,%rax
+       pop     %rbx
+       ret
+.size  AES_set_decrypt_key,.-AES_set_decrypt_key
+___
+
+# void AES_cbc_encrypt (const void char *inp, unsigned char *out,
+#                      size_t length, const AES_KEY *key,
+#                      unsigned char *ivp,const int enc);
+{
+# stack frame layout
+# -8(%rsp)             return address
+my $_rsp="0(%rsp)";            # saved %rsp
+my $_len="8(%rsp)";            # copy of 3rd parameter, length
+my $_key="16(%rsp)";           # copy of 4th parameter, key
+my $_ivp="24(%rsp)";           # copy of 5th parameter, ivp
+my $keyp="32(%rsp)";           # one to pass as $key
+my $ivec="40(%rsp)";           # ivec[16]
+my $aes_key="56(%rsp)";                # copy of aes_key
+my $mark="56+240(%rsp)";       # copy of aes_key->rounds
+
+$code.=<<___;
+.globl AES_cbc_encrypt
+.type  AES_cbc_encrypt,\@function,6
+.align 16
+AES_cbc_encrypt:
+       cmp     \$0,%rdx        # check length
+       je      .Lcbc_just_ret
+       push    %rbx
+       push    %rbp
+       push    %r12
+       push    %r13
+       push    %r14
+       push    %r15
+       pushfq
+       cld
+
+       .picmeup $tbl
+.Lcbc_pic_point:
+
+       cmp     \$0,%r9
+       je      .LDECRYPT
+
+       lea     AES_Te-.Lcbc_pic_point($tbl),$tbl
+
+       # allocate aligned stack frame...
+       lea     -64-248(%rsp),$key
+       and     \$-64,$key
+
+       # ... and make it doesn't alias with AES_Te modulo 4096
+       mov     $tbl,%r10
+       lea     2048($tbl),%r11
+       mov     $key,%r12
+       and     \$0xFFF,%r10    # s = $tbl&0xfff
+       and     \$0xFFF,%r11    # e = ($tbl+2048)&0xfff
+       and     \$0xFFF,%r12    # p = %rsp&0xfff
+
+       cmp     %r11,%r12       # if (p=>e) %rsp =- (p-e);
+       jb      .Lcbc_te_break_out
+       sub     %r11,%r12
+       sub     %r12,$key
+       jmp     .Lcbc_te_ok
+.Lcbc_te_break_out:            # else %rsp -= (p-s)&0xfff + framesz
+       sub     %r10,%r12
+       and     \$0xFFF,%r12
+       add     \$320,%r12
+       sub     %r12,$key
+.align 4
+.Lcbc_te_ok:
+
+       xchg    %rsp,$key
+       add     \$8,%rsp        # reserve for return address!
+       mov     $key,$_rsp      # save %rsp
+       mov     %rdx,$_len      # save copy of len
+       mov     %rcx,$_key      # save copy of key
+       mov     %r8,$_ivp       # save copy of ivp
+
+       movl    \$0,$mark       # copy of aes_key->rounds = 0;
+       mov     %rcx,$key
+
+       # do we copy key schedule to stack?
+       mov     $key,%r10
+       sub     $tbl,%r10
+       and     \$0xfff,%r10
+       cmp     \$2048,%r10
+       jb      .Lcbc_do_ecopy
+       cmp     \$4096-248,%r10
+       jb      .Lcbc_skip_ecopy
+.align 4
+.Lcbc_do_ecopy:
+               mov     %rsi,%r10       # backup $inp,$out
+               mov     %rdi,%r11
+               mov     $key,%rsi
+               lea     $aes_key,%rdi
+               lea     $aes_key,$key
+               mov     \$240/8,%ecx
+               .long   0x90A548F3      # rep movsq
+               mov     (%rsi),%eax     # copy aes_key->rounds
+               mov     %eax,(%rdi)
+               mov     %r10,%rsi       # restore $inp,$out
+               mov     %r11,%rdi
+.Lcbc_skip_ecopy:
+       mov     $key,$keyp      # save key pointer
+
+       mov     \$16,%ecx
+.align 4
+.Lcbc_prefetch_te:
+               mov     0($tbl),%r10
+               mov     32($tbl),%r11
+               mov     64($tbl),%r12
+               mov     96($tbl),%r13
+               lea     128($tbl),$tbl
+               sub     \$1,%ecx
+       jnz     .Lcbc_prefetch_te
+       sub     \$2048,$tbl
+
+       test    \$-16,%rdx              # check upon length
+       mov     %rdx,%r10
+       mov     0(%r8),$s0              # load iv
+       mov     4(%r8),$s1
+       mov     8(%r8),$s2
+       mov     12(%r8),$s3
+       jz      .Lcbc_enc_tail          # short input...
+
+.align 4
+.Lcbc_enc_loop:
+               xor     0($inp),$s0
+               xor     4($inp),$s1
+               xor     8($inp),$s2
+               xor     12($inp),$s3
+
+               mov     $keyp,$key      # restore key
+               call    _x86_64_AES_encrypt
+
+               mov     $s0,0($out)
+               mov     $s1,4($out)
+               mov     $s2,8($out)
+               mov     $s3,12($out)
+
+               mov     $_len,%r10
+               lea     16($inp),$inp
+               lea     16($out),$out
+               sub     \$16,%r10
+               test    \$-16,%r10
+               mov     %r10,$_len
+       jnz     .Lcbc_enc_loop
+       test    \$15,%r10
+       jnz     .Lcbc_enc_tail
+       mov     $_ivp,%r10      # restore ivp
+       mov     $s0,0(%r10)     # save ivec
+       mov     $s1,4(%r10)
+       mov     $s2,8(%r10)
+       mov     $s3,12(%r10)
+
+.align 4
+.Lcbc_cleanup:
+       cmpl    \$0,$mark       # was the key schedule copied?
+       lea     $aes_key,%rdi
+       mov     $_rsp,%rsp
+       je      .Lcbc_exit
+               mov     \$240/8,%ecx
+               xor     %rax,%rax
+               .long   0x90AB48F3      # rep stosq
+.Lcbc_exit:
+       popfq
+       pop     %r15
+       pop     %r14
+       pop     %r13
+       pop     %r12
+       pop     %rbp
+       pop     %rbx
+.Lcbc_just_ret:
+       ret
+.align 4
+.Lcbc_enc_tail:
+       cmp     $inp,$out
+       mov     $inp,%r11
+       mov     $out,%r12
+       je      .Lcbc_enc_in_place
+       mov     %r10,%rcx
+       xchg    %rsi,%rdi
+       .long   0xF689A4F3              # rep movsb
+.Lcbc_enc_in_place:
+       mov     \$16,%rcx               # zero tail
+       sub     %r10,%rcx
+       xor     %rax,%rax
+       .long   0xF689AAF3              # rep stosb
+       mov     %r12,$inp               # this is not a mistake!
+       mov     %r12,$out
+       movq    \$16,$_len              # len=16
+       jmp     .Lcbc_enc_loop          # one more spin...
+#----------------------------- DECRYPT -----------------------------#
+.align 16
+.LDECRYPT:
+       lea     AES_Td-.Lcbc_pic_point($tbl),$tbl
+
+       # allocate aligned stack frame...
+       lea     -64-248(%rsp),$key
+       and     \$-64,$key
+
+       # ... and make it doesn't alias with AES_Td modulo 4096
+       mov     $tbl,%r10
+       lea     3072($tbl),%r11
+       mov     $key,%r12
+       and     \$0xFFF,%r10    # s = $tbl&0xfff
+       and     \$0xFFF,%r11    # e = ($tbl+2048)&0xfff
+       and     \$0xFFF,%r12    # p = %rsp&0xfff
+
+       cmp     %r11,%r12       # if (p=>e) %rsp =- (p-e);
+       jb      .Lcbc_td_break_out
+       sub     %r11,%r12
+       sub     %r12,$key
+       jmp     .Lcbc_td_ok
+.Lcbc_td_break_out:            # else %rsp -= (p-s)&0xfff + framesz
+       sub     %r10,%r12
+       and     \$0xFFF,%r12
+       add     \$320,%r12
+       sub     %r12,$key
+.align 4
+.Lcbc_td_ok:
+
+       xchg    %rsp,$key
+       add     \$8,%rsp        # reserve for return address!
+       mov     $key,$_rsp      # save %rsp
+       mov     %rdx,$_len      # save copy of len
+       mov     %rcx,$_key      # save copy of key
+       mov     %r8,$_ivp       # save copy of ivp
+
+       movl    \$0,$mark       # copy of aes_key->rounds = 0;
+       mov     %rcx,$key
+
+       # do we copy key schedule to stack?
+       mov     $key,%r10
+       sub     $tbl,%r10
+       and     \$0xfff,%r10
+       cmp     \$3072,%r10
+       jb      .Lcbc_do_dcopy
+       cmp     \$4096-248,%r10
+       jb      .Lcbc_skip_dcopy
+.align 4
+.Lcbc_do_dcopy:
+               mov     %rsi,%r10       # backup $inp,$out
+               mov     %rdi,%r11
+               mov     $key,%rsi
+               lea     $aes_key,%rdi
+               lea     $aes_key,$key
+               mov     \$240/8,%ecx
+               .long   0x90A548F3      # rep movsq
+               mov     (%rsi),%eax     # copy aes_key->rounds
+               mov     %eax,(%rdi)
+               mov     %r10,%rsi       # restore $inp,$out
+               mov     %r11,%rdi
+.Lcbc_skip_dcopy:
+       mov     $key,$keyp      # save key pointer
+
+       mov     \$24,%ecx
+.align 4
+.Lcbc_prefetch_td:
+               mov     0($tbl),%r10
+               mov     32($tbl),%r11
+               mov     64($tbl),%r12
+               mov     96($tbl),%r13
+               lea     128($tbl),$tbl
+               sub     \$1,%ecx
+       jnz     .Lcbc_prefetch_td
+       sub     \$3072,$tbl
+
+       cmp     $inp,$out
+       je      .Lcbc_dec_in_place
+
+       mov     %r8,$ivec
+.align 4
+.Lcbc_dec_loop:
+               mov     0($inp),$s0             # read input
+               mov     4($inp),$s1
+               mov     8($inp),$s2
+               mov     12($inp),$s3
+
+               mov     $keyp,$key              # load key
+               call    _x86_64_AES_decrypt
+
+               mov     $ivec,%r8               # load ivp
+               xor     0(%r8),$s0              # xor iv
+               xor     4(%r8),$s1
+               xor     8(%r8),$s2
+               xor     12(%r8),$s3
+               mov     $inp,%r8                # current input, next iv
+
+               mov     $_len,%r10              # load len
+               sub     \$16,%r10
+               jc      .Lcbc_dec_partial
+               mov     %r10,$_len              # update len
+               mov     %r8,$ivec               # update ivp
+
+               mov     $s0,0($out)             # write output
+               mov     $s1,4($out)
+               mov     $s2,8($out)
+               mov     $s3,12($out)
+
+               lea     16($inp),$inp
+               lea     16($out),$out
+       jnz     .Lcbc_dec_loop
+.Lcbc_dec_end:
+       mov     $_ivp,%r9               # load user ivp
+       mov     0(%r8),%r10             # load iv
+       mov     8(%r8),%r11
+       mov     %r10,0(%r9)             # copy back to user
+       mov     %r11,8(%r9)
+       jmp     .Lcbc_cleanup
+
+.align 4
+.Lcbc_dec_partial:
+       mov     $s0,0+$ivec             # dump output to stack
+       mov     $s1,4+$ivec
+       mov     $s2,8+$ivec
+       mov     $s3,12+$ivec
+       mov     $out,%rdi
+       lea     $ivec,%rsi
+       mov     \$16,%rcx
+       add     %r10,%rcx               # number of bytes to copy
+       .long   0xF689A4F3              # rep movsb
+       jmp     .Lcbc_dec_end
+
+.align 16
+.Lcbc_dec_in_place:
+               mov     0($inp),$s0             # load input
+               mov     4($inp),$s1
+               mov     8($inp),$s2
+               mov     12($inp),$s3
+
+               mov     $keyp,$key
+               call    _x86_64_AES_decrypt
+
+               mov     $_ivp,%r8
+               xor     0(%r8),$s0
+               xor     4(%r8),$s1
+               xor     8(%r8),$s2
+               xor     12(%r8),$s3
+
+               mov     0($inp),%r10            # copy input to iv
+               mov     8($inp),%r11
+               mov     %r10,0(%r8)
+               mov     %r11,8(%r8)
+
+               mov     $s0,0($out)             # save output [zaps input]
+               mov     $s1,4($out)
+               mov     $s2,8($out)
+               mov     $s3,12($out)
+
+               mov     $_len,%rcx
+               lea     16($inp),$inp
+               lea     16($out),$out
+               sub     \$16,%rcx
+               jc      .Lcbc_dec_in_place_partial
+               mov     %rcx,$_len
+       jnz     .Lcbc_dec_in_place
+       jmp     .Lcbc_cleanup
+
+.align 4
+.Lcbc_dec_in_place_partial:
+       # one can argue if this is actually required
+       lea     ($out,%rcx),%rdi
+       lea     (%r8,%rcx),%rsi
+       neg     %rcx
+       .long   0xF689A4F3      # rep movsb     # restore tail
+       jmp     .Lcbc_cleanup
+.size  AES_cbc_encrypt,.-AES_cbc_encrypt
+___
+}
+
+$code.=<<___;
+.globl AES_Te
+.align 64
+AES_Te:
+___
+       &_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
+       &_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
+       &_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
+       &_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
+       &_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
+       &_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
+       &_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
+       &_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
+       &_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
+       &_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
+       &_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
+       &_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
+       &_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
+       &_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
+       &_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
+       &_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
+       &_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
+       &_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
+       &_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
+       &_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
+       &_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
+       &_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
+       &_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
+       &_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
+       &_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
+       &_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
+       &_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
+       &_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
+       &_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
+       &_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
+       &_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
+       &_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
+       &_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
+       &_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
+       &_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
+       &_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
+       &_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
+       &_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
+       &_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
+       &_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
+       &_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
+       &_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
+       &_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
+       &_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
+       &_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
+       &_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
+       &_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
+       &_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
+       &_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
+       &_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
+       &_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
+       &_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
+       &_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
+       &_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
+       &_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
+       &_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
+       &_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
+       &_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
+       &_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
+       &_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
+       &_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
+       &_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
+       &_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
+       &_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
+#rcon:
+$code.=<<___;
+       .long   0x00000001, 0x00000002, 0x00000004, 0x00000008
+       .long   0x00000010, 0x00000020, 0x00000040, 0x00000080
+       .long   0x0000001b, 0x00000036, 0, 0, 0, 0, 0, 0
+___
+$code.=<<___;
+.globl AES_Td
+.align 64
+AES_Td:
+___
+       &_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
+       &_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
+       &_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
+       &_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
+       &_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
+       &_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
+       &_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
+       &_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
+       &_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
+       &_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
+       &_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
+       &_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
+       &_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
+       &_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
+       &_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
+       &_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
+       &_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
+       &_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
+       &_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
+       &_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
+       &_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
+       &_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
+       &_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
+       &_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
+       &_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
+       &_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
+       &_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
+       &_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
+       &_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
+       &_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
+       &_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
+       &_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
+       &_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
+       &_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
+       &_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
+       &_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
+       &_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
+       &_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
+       &_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
+       &_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
+       &_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
+       &_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
+       &_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
+       &_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
+       &_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
+       &_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
+       &_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
+       &_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
+       &_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
+       &_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
+       &_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
+       &_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
+       &_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
+       &_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
+       &_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
+       &_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
+       &_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
+       &_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
+       &_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
+       &_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
+       &_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
+       &_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
+       &_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
+       &_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
+#Td4:
+       &data_word(0x52525252, 0x09090909, 0x6a6a6a6a, 0xd5d5d5d5);
+       &data_word(0x30303030, 0x36363636, 0xa5a5a5a5, 0x38383838);
+       &data_word(0xbfbfbfbf, 0x40404040, 0xa3a3a3a3, 0x9e9e9e9e);
+       &data_word(0x81818181, 0xf3f3f3f3, 0xd7d7d7d7, 0xfbfbfbfb);
+       &data_word(0x7c7c7c7c, 0xe3e3e3e3, 0x39393939, 0x82828282);
+       &data_word(0x9b9b9b9b, 0x2f2f2f2f, 0xffffffff, 0x87878787);
+       &data_word(0x34343434, 0x8e8e8e8e, 0x43434343, 0x44444444);
+       &data_word(0xc4c4c4c4, 0xdededede, 0xe9e9e9e9, 0xcbcbcbcb);
+       &data_word(0x54545454, 0x7b7b7b7b, 0x94949494, 0x32323232);
+       &data_word(0xa6a6a6a6, 0xc2c2c2c2, 0x23232323, 0x3d3d3d3d);
+       &data_word(0xeeeeeeee, 0x4c4c4c4c, 0x95959595, 0x0b0b0b0b);
+       &data_word(0x42424242, 0xfafafafa, 0xc3c3c3c3, 0x4e4e4e4e);
+       &data_word(0x08080808, 0x2e2e2e2e, 0xa1a1a1a1, 0x66666666);
+       &data_word(0x28282828, 0xd9d9d9d9, 0x24242424, 0xb2b2b2b2);
+       &data_word(0x76767676, 0x5b5b5b5b, 0xa2a2a2a2, 0x49494949);
+       &data_word(0x6d6d6d6d, 0x8b8b8b8b, 0xd1d1d1d1, 0x25252525);
+       &data_word(0x72727272, 0xf8f8f8f8, 0xf6f6f6f6, 0x64646464);
+       &data_word(0x86868686, 0x68686868, 0x98989898, 0x16161616);
+       &data_word(0xd4d4d4d4, 0xa4a4a4a4, 0x5c5c5c5c, 0xcccccccc);
+       &data_word(0x5d5d5d5d, 0x65656565, 0xb6b6b6b6, 0x92929292);
+       &data_word(0x6c6c6c6c, 0x70707070, 0x48484848, 0x50505050);
+       &data_word(0xfdfdfdfd, 0xedededed, 0xb9b9b9b9, 0xdadadada);
+       &data_word(0x5e5e5e5e, 0x15151515, 0x46464646, 0x57575757);
+       &data_word(0xa7a7a7a7, 0x8d8d8d8d, 0x9d9d9d9d, 0x84848484);
+       &data_word(0x90909090, 0xd8d8d8d8, 0xabababab, 0x00000000);
+       &data_word(0x8c8c8c8c, 0xbcbcbcbc, 0xd3d3d3d3, 0x0a0a0a0a);
+       &data_word(0xf7f7f7f7, 0xe4e4e4e4, 0x58585858, 0x05050505);
+       &data_word(0xb8b8b8b8, 0xb3b3b3b3, 0x45454545, 0x06060606);
+       &data_word(0xd0d0d0d0, 0x2c2c2c2c, 0x1e1e1e1e, 0x8f8f8f8f);
+       &data_word(0xcacacaca, 0x3f3f3f3f, 0x0f0f0f0f, 0x02020202);
+       &data_word(0xc1c1c1c1, 0xafafafaf, 0xbdbdbdbd, 0x03030303);
+       &data_word(0x01010101, 0x13131313, 0x8a8a8a8a, 0x6b6b6b6b);
+       &data_word(0x3a3a3a3a, 0x91919191, 0x11111111, 0x41414141);
+       &data_word(0x4f4f4f4f, 0x67676767, 0xdcdcdcdc, 0xeaeaeaea);
+       &data_word(0x97979797, 0xf2f2f2f2, 0xcfcfcfcf, 0xcececece);
+       &data_word(0xf0f0f0f0, 0xb4b4b4b4, 0xe6e6e6e6, 0x73737373);
+       &data_word(0x96969696, 0xacacacac, 0x74747474, 0x22222222);
+       &data_word(0xe7e7e7e7, 0xadadadad, 0x35353535, 0x85858585);
+       &data_word(0xe2e2e2e2, 0xf9f9f9f9, 0x37373737, 0xe8e8e8e8);
+       &data_word(0x1c1c1c1c, 0x75757575, 0xdfdfdfdf, 0x6e6e6e6e);
+       &data_word(0x47474747, 0xf1f1f1f1, 0x1a1a1a1a, 0x71717171);
+       &data_word(0x1d1d1d1d, 0x29292929, 0xc5c5c5c5, 0x89898989);
+       &data_word(0x6f6f6f6f, 0xb7b7b7b7, 0x62626262, 0x0e0e0e0e);
+       &data_word(0xaaaaaaaa, 0x18181818, 0xbebebebe, 0x1b1b1b1b);
+       &data_word(0xfcfcfcfc, 0x56565656, 0x3e3e3e3e, 0x4b4b4b4b);
+       &data_word(0xc6c6c6c6, 0xd2d2d2d2, 0x79797979, 0x20202020);
+       &data_word(0x9a9a9a9a, 0xdbdbdbdb, 0xc0c0c0c0, 0xfefefefe);
+       &data_word(0x78787878, 0xcdcdcdcd, 0x5a5a5a5a, 0xf4f4f4f4);
+       &data_word(0x1f1f1f1f, 0xdddddddd, 0xa8a8a8a8, 0x33333333);
+       &data_word(0x88888888, 0x07070707, 0xc7c7c7c7, 0x31313131);
+       &data_word(0xb1b1b1b1, 0x12121212, 0x10101010, 0x59595959);
+       &data_word(0x27272727, 0x80808080, 0xecececec, 0x5f5f5f5f);
+       &data_word(0x60606060, 0x51515151, 0x7f7f7f7f, 0xa9a9a9a9);
+       &data_word(0x19191919, 0xb5b5b5b5, 0x4a4a4a4a, 0x0d0d0d0d);
+       &data_word(0x2d2d2d2d, 0xe5e5e5e5, 0x7a7a7a7a, 0x9f9f9f9f);
+       &data_word(0x93939393, 0xc9c9c9c9, 0x9c9c9c9c, 0xefefefef);
+       &data_word(0xa0a0a0a0, 0xe0e0e0e0, 0x3b3b3b3b, 0x4d4d4d4d);
+       &data_word(0xaeaeaeae, 0x2a2a2a2a, 0xf5f5f5f5, 0xb0b0b0b0);
+       &data_word(0xc8c8c8c8, 0xebebebeb, 0xbbbbbbbb, 0x3c3c3c3c);
+       &data_word(0x83838383, 0x53535353, 0x99999999, 0x61616161);
+       &data_word(0x17171717, 0x2b2b2b2b, 0x04040404, 0x7e7e7e7e);
+       &data_word(0xbabababa, 0x77777777, 0xd6d6d6d6, 0x26262626);
+       &data_word(0xe1e1e1e1, 0x69696969, 0x14141414, 0x63636363);
+       &data_word(0x55555555, 0x21212121, 0x0c0c0c0c, 0x7d7d7d7d);
+
+print $code;
+
+close STDOUT;