Pedantic polish to aes-586.pl:-)
[openssl.git] / crypto / aes / asm / aes-586.pl
index c61ed6b..e279420 100755 (executable)
@@ -6,15 +6,21 @@
 # forms are granted according to the OpenSSL license.
 # ====================================================================
 #
+# Version 3.5.
+#
 # You might fail to appreciate this module performance from the first
-# try. If compared to "vanilla" linux-ia32-icc target, i.e. Intel C
-# without -KPIC, performance appears to be virtually identical... But
-# try to configure with shared library support... Aha! Intel compiler
-# "suddenly" lags behind by 30% [on P4]:-) And if compared to
-# position-independent code generated by GNU C, this code performs
-# more than *twice* as fast! Yes, all this buzz about PIC means that
-# [unlike other implementations] this module was explicitly designed
-# to be safe to use even in shared library context...
+# try. If compared to "vanilla" linux-ia32-icc target, i.e. considered
+# to be *the* best Intel C compiler without -KPIC, performance appears
+# to be virtually identical... But try to re-configure with shared
+# library support... Aha! Intel compiler "suddenly" lags behind by 30%
+# [on P4, more on others]:-) And if compared to position-independent
+# code generated by GNU C, this code performs *more* than *twice* as
+# fast! Yes, all this buzz about PIC means that unlike other hand-
+# coded implementations, this one was explicitly designed to be safe
+# to use even in shared library context... This also means that this
+# code isn't necessarily absolutely fastest "ever," because in order
+# to achieve position independence an extra register has to be
+# off-loaded to stack, which affects the benchmark result.
 #
 # Special note about instruction choice. Do you recall RC4_INT code
 # performing poorly on P4? It might be the time to figure out why.
 # intermediate implementation, which was spilling yet another register
 # to stack... Final offset*4 code below runs just a tad faster on P4,
 # but exhibits up to 10% improvement on other cores.
+#
+# Second version is "monolithic" replacement for aes_core.c, which in
+# addition to AES_[de|en]crypt implements AES_set_[de|en]cryption_key.
+# This made it possible to implement little-endian variant of the
+# algorithm without modifying the base C code. Motivating factor for
+# the undertaken effort was that it appeared that in tight IA-32
+# register window little-endian flavor could achieve slightly higher
+# Instruction Level Parallelism, and it indeed resulted in up to 15%
+# better performance on most recent µ-archs...
+#
+# Third version adds AES_cbc_encrypt implementation, which resulted in
+# up to 40% performance imrovement of CBC benchmark results. 40% was
+# observed on P4 core, where "overall" imrovement coefficient, i.e. if
+# compared to PIC generated by GCC and in CBC mode, was observed to be
+# as large as 4x:-) CBC performance is virtually identical to ECB now
+# and on some platforms even better, e.g. 17.6 "small" cycles/byte on
+# Opteron, because certain function prologues and epilogues are
+# effectively taken out of the loop...
+#
+# Version 3.2 implements compressed tables and prefetch of these tables
+# in CBC[!] mode. Former means that 3/4 of table references are now
+# misaligned, which unfortunately has negative impact on elder IA-32
+# implementations, Pentium suffered 30% penalty, PIII - 10%.
+#
+# Version 3.3 avoids L1 cache aliasing between stack frame and
+# S-boxes, and 3.4 - L1 cache aliasing even between key schedule. The
+# latter is achieved by copying the key schedule to controlled place in
+# stack. This unfortunately has rather strong impact on small block CBC
+# performance, ~2x deterioration on 16-byte block if compared to 3.3.
+#
+# Version 3.5 checks if there is L1 cache aliasing between user-supplied
+# key schedule and S-boxes and abstains from copying the former if
+# there is no. This allows end-user to consciously retain small block
+# performance by aligning key schedule in specific manner.
+#
+# Current ECB performance numbers for 128-bit key in CPU cycles per
+# processed byte [measure commonly used by AES benchmarkers] are:
+#
+#              small footprint         fully unrolled
+# P4           24                      22
+# AMD K8       20                      19
+# PIII         25                      23
+# Pentium      81                      78
 
 push(@INC,"perlasm","../../perlasm");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"aes-586.pl",$ARGV[$#ARGV] eq "386");
 
-$small_footprint=1;    # $small_footprint=1 code is 4-6% slower, but
-                       # 5 times smaller! I default to compact code.
 $s0="eax";
 $s1="ebx";
 $s2="ecx";
 $s3="edx";
+$key="edi";
+$acc="esi";
+
+$compromise=0;         # $compromise=128 abstains from copying key
+                       # schedule to stack when encrypting inputs
+                       # shorter than 128 bytes at the cost of
+                       # risksing aliasing with S-boxes. In return
+                       # you get way better, up to +70%, small block
+                       # performance.
+$small_footprint=1;    # $small_footprint=1 code is ~5% slower [on
+                       # recent µ-archs], but ~5 times smaller!
+                       # I favor compact code to minimize cache
+                       # contention and in hope to "collect" 5% back
+                       # in real-life applications...
+$vertical_spin=0;      # shift "verticaly" defaults to 0, because of
+                       # its proof-of-concept status...
+
+# Note that there is no decvert(), as well as last encryption round is
+# performed with "horizontal" shifts. This is because this "vertical"
+# implementation [one which groups shifts on a given $s[i] to form a
+# "column," unlike "horizontal" one, which groups shifts on different
+# $s[i] to form a "row"] is work in progress. It was observed to run
+# few percents faster on Intel cores, but not AMD. On AMD K8 core it's
+# whole 12% slower:-( So we face a trade-off... Shall it be resolved
+# some day? Till then the code is considered experimental and by
+# default remains dormant...
+
+sub encvert()
+{ my ($te,@s) = @_;
+  my $v0 = $acc, $v1 = $key;
+
+       &mov    ($v0,$s[3]);                            # copy s3
+       &mov    (&DWP(4,"esp"),$s[2]);                  # save s2
+       &mov    ($v1,$s[0]);                            # copy s0
+       &mov    (&DWP(8,"esp"),$s[1]);                  # save s1
+
+       &movz   ($s[2],&HB($s[0]));
+       &and    ($s[0],0xFF);
+       &mov    ($s[0],&DWP(0,$te,$s[0],8));            # s0>>0
+       &shr    ($v1,16);
+       &mov    ($s[3],&DWP(3,$te,$s[2],8));            # s0>>8
+       &movz   ($s[1],&HB($v1));
+       &and    ($v1,0xFF);
+       &mov    ($s[2],&DWP(2,$te,$v1,8));              # s0>>16
+        &mov   ($v1,$v0);
+       &mov    ($s[1],&DWP(1,$te,$s[1],8));            # s0>>24
+
+       &and    ($v0,0xFF);
+       &xor    ($s[3],&DWP(0,$te,$v0,8));              # s3>>0
+       &movz   ($v0,&HB($v1));
+       &shr    ($v1,16);
+       &xor    ($s[2],&DWP(3,$te,$v0,8));              # s3>>8
+       &movz   ($v0,&HB($v1));
+       &and    ($v1,0xFF);
+       &xor    ($s[1],&DWP(2,$te,$v1,8));              # s3>>16
+        &mov   ($v1,&DWP(4,"esp"));                    # restore s2
+       &xor    ($s[0],&DWP(1,$te,$v0,8));              # s3>>24
+
+       &mov    ($v0,$v1);
+       &and    ($v1,0xFF);
+       &xor    ($s[2],&DWP(0,$te,$v1,8));              # s2>>0
+       &movz   ($v1,&HB($v0));
+       &shr    ($v0,16);
+       &xor    ($s[1],&DWP(3,$te,$v1,8));              # s2>>8
+       &movz   ($v1,&HB($v0));
+       &and    ($v0,0xFF);
+       &xor    ($s[0],&DWP(2,$te,$v0,8));              # s2>>16
+        &mov   ($v0,&DWP(8,"esp"));                    # restore s1
+       &xor    ($s[3],&DWP(1,$te,$v1,8));              # s2>>24
+
+       &mov    ($v1,$v0);
+       &and    ($v0,0xFF);
+       &xor    ($s[1],&DWP(0,$te,$v0,8));              # s1>>0
+       &movz   ($v0,&HB($v1));
+       &shr    ($v1,16);
+       &xor    ($s[0],&DWP(3,$te,$v0,8));              # s1>>8
+       &movz   ($v0,&HB($v1));
+       &and    ($v1,0xFF);
+       &xor    ($s[3],&DWP(2,$te,$v1,8));              # s1>>16
+        &mov   ($key,&DWP(12,"esp"));                  # reincarnate v1 as key
+       &xor    ($s[2],&DWP(1,$te,$v0,8));              # s1>>24
+}
 
 sub encstep()
 { my ($i,$te,@s) = @_;
-  my $tmp,$out;
-
-       # lines marked with ## denote same $sN...
-       if ($i==3)  {   &mov    ("edi",&DWP(12,"esp"));
-                       &movz   ($out=$s[0],&HB($s[0]));        }       ##
-       else        {   &mov    ($out="esi",$s[0]);
-                       &shr    ($out,24);                      }
-                       &mov    ($out,&DWP(1024*0,$te,$out,4));
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
 
-       if ($i==2)  {   &movz   ($tmp="edi",&LB($s[1]));        }       ##
-       else        {   $i==3?$tmp=$s[1]:&mov($tmp="edi",$s[1]);
-                       &shr    ($tmp,16);
-                       &and    ($tmp,0xFF);                    }
-                       &xor    ($out,&DWP(1024*1,$te,$tmp,4));
+       # lines marked with #%e?x[i] denote "reordered" instructions...
+       if ($i==3)  {   &mov    ($key,&DWP(12,"esp"));          }##%edx
+       else        {   &mov    ($out,$s[0]);
+                       &and    ($out,0xFF);                    }
+       if ($i==1)  {   &shr    ($s[0],16);                     }#%ebx[1]
+       if ($i==2)  {   &shr    ($s[0],24);                     }#%ecx[2]
+                       &mov    ($out,&DWP(0,$te,$out,8));
 
-       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],&DWP(0,"esp")); }
-       else        {   $tmp="edi";                             }
-                       &movz   ($tmp,&HB($s[2]));
-                       &xor    ($out,&DWP(1024*2,$te,$tmp,4));
-       if ($i==1)  {   &shr    ($s[2],16);                     }       ##
+       if ($i==3)  {   $tmp=$s[1];                             }##%eax
+                       &movz   ($tmp,&HB($s[1]));
+                       &xor    ($out,&DWP(3,$te,$tmp,8));
 
-       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(4,"esp")); }
-       else        {   &mov    ($tmp="edi",$s[3]);             } 
+       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],&DWP(4,"esp")); }##%ebx
+       else        {   &mov    ($tmp,$s[2]);
+                       &shr    ($tmp,16);                      }
+       if ($i==2)  {   &and    ($s[1],0xFF);                   }#%edx[2]
                        &and    ($tmp,0xFF);
-                       &xor    ($out,&DWP(1024*3,$te,$tmp,4));
-       if ($i<2)   {   &mov    (&DWP(4*$i,"esp"),$out);        }
-       if ($i==3)  {   &mov    ($s[3],"esi");                  }
+                       &xor    ($out,&DWP(2,$te,$tmp,8));
+
+       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(8,"esp")); }##%ecx
+       elsif($i==2){   &movz   ($tmp,&HB($s[3]));              }#%ebx[2]
+       else        {   &mov    ($tmp,$s[3]); 
+                       &shr    ($tmp,24)                       }
+                       &xor    ($out,&DWP(1,$te,$tmp,8));
+       if ($i<2)   {   &mov    (&DWP(4+4*$i,"esp"),$out);      }
+       if ($i==3)  {   &mov    ($s[3],$acc);                   }
+                       &comment();
 }
 
 sub enclast()
 { my ($i,$te,@s)=@_;
-  my $tmp,$out;
-
-       if ($i==3)  {   &mov    ("edi",&DWP(12,"esp"));
-                       &movz   ($out=$s[0],&HB($s[0]));        }       ##
-       else        {   &mov    ($out="esi",$s[0]);
-                       &shr    ($out,24);                      }
-                       &mov    ($out,&DWP(0,$te,$out,4));
-                       &and    ($out,0xff000000);
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
 
-       if ($i==2)  {   &movz   ($tmp="edi",&LB($s[1]));        }       ##
-       else        {   $i==3?$tmp=$s[1]:&mov($tmp="edi",$s[1]);
-                       &shr    ($tmp,16);
-                       &and    ($tmp,0xFF);                    }
-                       &mov    ($tmp,&DWP(0,$te,$tmp,4));
-                       &and    ($tmp,0x00ff0000);
-                       &xor    ($out,$tmp);
+       if ($i==3)  {   &mov    ($key,&DWP(12,"esp"));          }##%edx
+       else        {   &mov    ($out,$s[0]);                   }
+                       &and    ($out,0xFF);
+       if ($i==1)  {   &shr    ($s[0],16);                     }#%ebx[1]
+       if ($i==2)  {   &shr    ($s[0],24);                     }#%ecx[2]
+                       &mov    ($out,&DWP(2,$te,$out,8));
+                       &and    ($out,0x000000ff);
 
-       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],&DWP(0,"esp")); }
-       else        {   $tmp="edi";                             }
-                       &movz   ($tmp,&HB($s[2]));
-                       &mov    ($tmp,&DWP(0,$te,$tmp,4));
+       if ($i==3)  {   $tmp=$s[1];                             }##%eax
+                       &movz   ($tmp,&HB($s[1]));
+                       &mov    ($tmp,&DWP(0,$te,$tmp,8));
                        &and    ($tmp,0x0000ff00);
-       if ($i==1)  {   &shr    ($s[2],16);                     }       ##
                        &xor    ($out,$tmp);
 
-       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(4,"esp")); }
-       else        {   &mov    ($tmp="edi",$s[3]);             } 
+       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],&DWP(4,"esp")); }##%ebx
+       else        {   mov     ($tmp,$s[2]);
+                       &shr    ($tmp,16);                      }
+       if ($i==2)  {   &and    ($s[1],0xFF);                   }#%edx[2]
                        &and    ($tmp,0xFF);
-                       &mov    ($tmp,&DWP(0,$te,$tmp,4));
-                       &and    ($tmp,0x000000ff);
+                       &mov    ($tmp,&DWP(0,$te,$tmp,8));
+                       &and    ($tmp,0x00ff0000);
+                       &xor    ($out,$tmp);
+
+       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(8,"esp")); }##%ecx
+       elsif($i==2){   &movz   ($tmp,&HB($s[3]));              }#%ebx[2]
+       else        {   &mov    ($tmp,$s[3]);
+                       &shr    ($tmp,24);                      }
+                       &mov    ($tmp,&DWP(2,$te,$tmp,8));
+                       &and    ($tmp,0xff000000);
                        &xor    ($out,$tmp);
-       if ($i<2)   {   &mov    (&DWP(4*$i,"esp"),$out);        }
-       if ($i==3)  {   &mov    ($s[3],"esi");                  }
+       if ($i<2)   {   &mov    (&DWP(4+4*$i,"esp"),$out);      }
+       if ($i==3)  {   &mov    ($s[3],$acc);                   }
 }
 
-# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
+sub _data_word() { my $i; while(defined($i=shift)) { &data_word($i,$i); } }
+
 &public_label("AES_Te");
-&function_begin("AES_encrypt");
-       &mov    ("esi",&wparam(0));             # load inp
-       &mov    ("edi",&wparam(2));             # load key
-
-        &call   (&label("pic_point"));          # make it PIC!
-&set_label("pic_point");
-        &blindpop("ebp");
-        &lea    ("ebp",&DWP(&label("AES_Te")."-".&label("pic_point"),"ebp"));
-
-       # allocate aligned stack frame
-       &mov    ("eax","esp");
-       &sub    ("esp",20);
-       &and    ("esp",-16);
-
-       &mov    (&DWP(12,"esp"),"edi");         # save key
-       &mov    (&DWP(16,"esp"),"eax");         # save %esp
-
-       &mov    ($s0,&DWP(0,"esi"));            # load input data
-       &mov    ($s1,&DWP(4,"esi"));
-       &mov    ($s2,&DWP(8,"esi"));
-       &mov    ($s3,&DWP(12,"esi"));
-       #
-       # It's perfectly possible to implement algorithm as
-       # little-endian and get rid of bswaps... It would give
-       # less than 1% performance improvement, so I judge it
-       # doesn't worth the trouble...
-       #
-       &bswap  ($s0);
-       &bswap  ($s1);
-       &bswap  ($s2);
-       &bswap  ($s3);
-       &xor    ($s0,&DWP(0,"edi"));
-       &xor    ($s1,&DWP(4,"edi"));
-       &xor    ($s2,&DWP(8,"edi"));
-       &xor    ($s3,&DWP(12,"edi"));
-
-       &mov    ("esi",&DWP(240,"edi"));        # load key->rounds
+&function_begin_B("_x86_AES_encrypt");
+       if ($vertical_spin) {
+               # I need high parts of volatile registers to be accessible...
+               &exch   ($s1="edi",$key="ebx");
+               &mov    ($s2="esi",$acc="ecx");
+       }
+
+       # note that caller is expected to allocate stack frame for me!
+       &mov    (&DWP(12,"esp"),$key);          # save key
+
+       &xor    ($s0,&DWP(0,$key));             # xor with key
+       &xor    ($s1,&DWP(4,$key));
+       &xor    ($s2,&DWP(8,$key));
+       &xor    ($s3,&DWP(12,$key));
+
+       &mov    ($acc,&DWP(240,$key));          # load key->rounds
 
        if ($small_footprint) {
-           &lea        ("esi",&DWP(-2,"esi","esi"));
-           &lea        ("esi",&DWP(0,"edi","esi",8));
-           &mov        (&DWP(8,"esp"),"esi");  # end of key schedule
+           &lea        ($acc,&DWP(-2,$acc,$acc));
+           &lea        ($acc,&DWP(0,$key,$acc,8));
+           &mov        (&DWP(16,"esp"),$acc);  # end of key schedule
            &align      (4);
            &set_label("loop");
-               &encstep(0,"ebp",$s0,$s1,$s2,$s3);
-               &encstep(1,"ebp",$s1,$s2,$s3,$s0);
-               &encstep(2,"ebp",$s2,$s3,$s0,$s1);
-               &encstep(3,"ebp",$s3,$s0,$s1,$s2);
-               &add    ("edi",16);                     # advance rd_key
-               &xor    ($s0,&DWP(0,"edi"));
-               &xor    ($s1,&DWP(4,"edi"));
-               &xor    ($s2,&DWP(8,"edi"));
-               &xor    ($s3,&DWP(12,"edi"));
-           &cmp        ("edi",&DWP(8,"esp"));
-           &mov        (&DWP(12,"esp"),"edi");
+               if ($vertical_spin) {
+                   &encvert("ebp",$s0,$s1,$s2,$s3);
+               } else {
+                   &encstep(0,"ebp",$s0,$s1,$s2,$s3);
+                   &encstep(1,"ebp",$s1,$s2,$s3,$s0);
+                   &encstep(2,"ebp",$s2,$s3,$s0,$s1);
+                   &encstep(3,"ebp",$s3,$s0,$s1,$s2);
+               }
+               &add    ($key,16);              # advance rd_key
+               &xor    ($s0,&DWP(0,$key));
+               &xor    ($s1,&DWP(4,$key));
+               &xor    ($s2,&DWP(8,$key));
+               &xor    ($s3,&DWP(12,$key));
+           &cmp        ($key,&DWP(16,"esp"));
+           &mov        (&DWP(12,"esp"),$key);
            &jb         (&label("loop"));
        }
        else {
-           &cmp        ("esi",10);
+           &cmp        ($acc,10);
            &jle        (&label("10rounds"));
-           &cmp        ("esi",12);
+           &cmp        ($acc,12);
            &jle        (&label("12rounds"));
 
        &set_label("14rounds");
            for ($i=1;$i<3;$i++) {
-               &encstep(0,"ebp",$s0,$s1,$s2,$s3);
-               &encstep(1,"ebp",$s1,$s2,$s3,$s0);
-               &encstep(2,"ebp",$s2,$s3,$s0,$s1);
-               &encstep(3,"ebp",$s3,$s0,$s1,$s2);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               if ($vertical_spin) {
+                   &encvert("ebp",$s0,$s1,$s2,$s3);
+               } else {
+                   &encstep(0,"ebp",$s0,$s1,$s2,$s3);
+                   &encstep(1,"ebp",$s1,$s2,$s3,$s0);
+                   &encstep(2,"ebp",$s2,$s3,$s0,$s1);
+                   &encstep(3,"ebp",$s3,$s0,$s1,$s2);
+               }
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
-           &add        ("edi",32);
-           &mov        (&DWP(12,"esp"),"edi");         # advance rd_key
+           &add        ($key,32);
+           &mov        (&DWP(12,"esp"),$key);  # advance rd_key
        &set_label("12rounds");
            for ($i=1;$i<3;$i++) {
-               &encstep(0,"ebp",$s0,$s1,$s2,$s3);
-               &encstep(1,"ebp",$s1,$s2,$s3,$s0);
-               &encstep(2,"ebp",$s2,$s3,$s0,$s1);
-               &encstep(3,"ebp",$s3,$s0,$s1,$s2);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               if ($vertical_spin) {
+                   &encvert("ebp",$s0,$s1,$s2,$s3);
+               } else {
+                   &encstep(0,"ebp",$s0,$s1,$s2,$s3);
+                   &encstep(1,"ebp",$s1,$s2,$s3,$s0);
+                   &encstep(2,"ebp",$s2,$s3,$s0,$s1);
+                   &encstep(3,"ebp",$s3,$s0,$s1,$s2);
+               }
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
-           &add        ("edi",32);
-           &mov        (&DWP(12,"esp"),"edi");         # advance rd_key
+           &add        ($key,32);
+           &mov        (&DWP(12,"esp"),$key);  # advance rd_key
        &set_label("10rounds");
            for ($i=1;$i<10;$i++) {
-               &encstep(0,"ebp",$s0,$s1,$s2,$s3);
-               &encstep(1,"ebp",$s1,$s2,$s3,$s0);
-               &encstep(2,"ebp",$s2,$s3,$s0,$s1);
-               &encstep(3,"ebp",$s3,$s0,$s1,$s2);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               if ($vertical_spin) {
+                   &encvert("ebp",$s0,$s1,$s2,$s3);
+               } else {
+                   &encstep(0,"ebp",$s0,$s1,$s2,$s3);
+                   &encstep(1,"ebp",$s1,$s2,$s3,$s0);
+                   &encstep(2,"ebp",$s2,$s3,$s0,$s1);
+                   &encstep(3,"ebp",$s3,$s0,$s1,$s2);
+               }
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
        }
 
-       &add    ("ebp",4*1024);                 # skip to Te4
+       if ($vertical_spin) {
+           # "reincarnate" some registers for "horizontal" spin...
+           &mov        ($s1="ebx",$key="edi");
+           &mov        ($s2="ecx",$acc="esi");
+       }
        &enclast(0,"ebp",$s0,$s1,$s2,$s3);
        &enclast(1,"ebp",$s1,$s2,$s3,$s0);
        &enclast(2,"ebp",$s2,$s3,$s0,$s1);
        &enclast(3,"ebp",$s3,$s0,$s1,$s2);
 
-       &mov    ("esp",&DWP(16,"esp"));         # restore %esp
-       &add    ("edi",$small_footprint?16:160);
-       &xor    ($s0,&DWP(0,"edi"));
-       &xor    ($s1,&DWP(4,"edi"));
-       &xor    ($s2,&DWP(8,"edi"));
-       &xor    ($s3,&DWP(12,"edi"));
-       &bswap  ($s0);
-       &bswap  ($s1);
-       &bswap  ($s2);
-       &bswap  ($s3);
-       &mov    ("edi",&wparam(1));             # load out
-       &mov    (&DWP(0,"edi"),$s0);            # write output data
-       &mov    (&DWP(4,"edi"),$s1);
-       &mov    (&DWP(8,"edi"),$s2);
-       &mov    (&DWP(12,"edi"),$s3);
-
-       &pop    ("edi");
-       &pop    ("esi");
-       &pop    ("ebx");
-       &pop    ("ebp");
+       &add    ($key,$small_footprint?16:160);
+       &xor    ($s0,&DWP(0,$key));
+       &xor    ($s1,&DWP(4,$key));
+       &xor    ($s2,&DWP(8,$key));
+       &xor    ($s3,&DWP(12,$key));
+
        &ret    ();
 
 &set_label("AES_Te",64);       # Yes! I keep it in the code segment!
-       &data_word(0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d);
-       &data_word(0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554);
-       &data_word(0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d);
-       &data_word(0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a);
-       &data_word(0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87);
-       &data_word(0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b);
-       &data_word(0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea);
-       &data_word(0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b);
-       &data_word(0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a);
-       &data_word(0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f);
-       &data_word(0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108);
-       &data_word(0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f);
-       &data_word(0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e);
-       &data_word(0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5);
-       &data_word(0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d);
-       &data_word(0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f);
-       &data_word(0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e);
-       &data_word(0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb);
-       &data_word(0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce);
-       &data_word(0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497);
-       &data_word(0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c);
-       &data_word(0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed);
-       &data_word(0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b);
-       &data_word(0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a);
-       &data_word(0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16);
-       &data_word(0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594);
-       &data_word(0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81);
-       &data_word(0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3);
-       &data_word(0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a);
-       &data_word(0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504);
-       &data_word(0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163);
-       &data_word(0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d);
-       &data_word(0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f);
-       &data_word(0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739);
-       &data_word(0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47);
-       &data_word(0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395);
-       &data_word(0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f);
-       &data_word(0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883);
-       &data_word(0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c);
-       &data_word(0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76);
-       &data_word(0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e);
-       &data_word(0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4);
-       &data_word(0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6);
-       &data_word(0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b);
-       &data_word(0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7);
-       &data_word(0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0);
-       &data_word(0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25);
-       &data_word(0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818);
-       &data_word(0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72);
-       &data_word(0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651);
-       &data_word(0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21);
-       &data_word(0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85);
-       &data_word(0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa);
-       &data_word(0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12);
-       &data_word(0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0);
-       &data_word(0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9);
-       &data_word(0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133);
-       &data_word(0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7);
-       &data_word(0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920);
-       &data_word(0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a);
-       &data_word(0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17);
-       &data_word(0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8);
-       &data_word(0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11);
-       &data_word(0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a);
-#Te1:
-       &data_word(0xa5c66363, 0x84f87c7c, 0x99ee7777, 0x8df67b7b);
-       &data_word(0x0dfff2f2, 0xbdd66b6b, 0xb1de6f6f, 0x5491c5c5);
-       &data_word(0x50603030, 0x03020101, 0xa9ce6767, 0x7d562b2b);
-       &data_word(0x19e7fefe, 0x62b5d7d7, 0xe64dabab, 0x9aec7676);
-       &data_word(0x458fcaca, 0x9d1f8282, 0x4089c9c9, 0x87fa7d7d);
-       &data_word(0x15effafa, 0xebb25959, 0xc98e4747, 0x0bfbf0f0);
-       &data_word(0xec41adad, 0x67b3d4d4, 0xfd5fa2a2, 0xea45afaf);
-       &data_word(0xbf239c9c, 0xf753a4a4, 0x96e47272, 0x5b9bc0c0);
-       &data_word(0xc275b7b7, 0x1ce1fdfd, 0xae3d9393, 0x6a4c2626);
-       &data_word(0x5a6c3636, 0x417e3f3f, 0x02f5f7f7, 0x4f83cccc);
-       &data_word(0x5c683434, 0xf451a5a5, 0x34d1e5e5, 0x08f9f1f1);
-       &data_word(0x93e27171, 0x73abd8d8, 0x53623131, 0x3f2a1515);
-       &data_word(0x0c080404, 0x5295c7c7, 0x65462323, 0x5e9dc3c3);
-       &data_word(0x28301818, 0xa1379696, 0x0f0a0505, 0xb52f9a9a);
-       &data_word(0x090e0707, 0x36241212, 0x9b1b8080, 0x3ddfe2e2);
-       &data_word(0x26cdebeb, 0x694e2727, 0xcd7fb2b2, 0x9fea7575);
-       &data_word(0x1b120909, 0x9e1d8383, 0x74582c2c, 0x2e341a1a);
-       &data_word(0x2d361b1b, 0xb2dc6e6e, 0xeeb45a5a, 0xfb5ba0a0);
-       &data_word(0xf6a45252, 0x4d763b3b, 0x61b7d6d6, 0xce7db3b3);
-       &data_word(0x7b522929, 0x3edde3e3, 0x715e2f2f, 0x97138484);
-       &data_word(0xf5a65353, 0x68b9d1d1, 0x00000000, 0x2cc1eded);
-       &data_word(0x60402020, 0x1fe3fcfc, 0xc879b1b1, 0xedb65b5b);
-       &data_word(0xbed46a6a, 0x468dcbcb, 0xd967bebe, 0x4b723939);
-       &data_word(0xde944a4a, 0xd4984c4c, 0xe8b05858, 0x4a85cfcf);
-       &data_word(0x6bbbd0d0, 0x2ac5efef, 0xe54faaaa, 0x16edfbfb);
-       &data_word(0xc5864343, 0xd79a4d4d, 0x55663333, 0x94118585);
-       &data_word(0xcf8a4545, 0x10e9f9f9, 0x06040202, 0x81fe7f7f);
-       &data_word(0xf0a05050, 0x44783c3c, 0xba259f9f, 0xe34ba8a8);
-       &data_word(0xf3a25151, 0xfe5da3a3, 0xc0804040, 0x8a058f8f);
-       &data_word(0xad3f9292, 0xbc219d9d, 0x48703838, 0x04f1f5f5);
-       &data_word(0xdf63bcbc, 0xc177b6b6, 0x75afdada, 0x63422121);
-       &data_word(0x30201010, 0x1ae5ffff, 0x0efdf3f3, 0x6dbfd2d2);
-       &data_word(0x4c81cdcd, 0x14180c0c, 0x35261313, 0x2fc3ecec);
-       &data_word(0xe1be5f5f, 0xa2359797, 0xcc884444, 0x392e1717);
-       &data_word(0x5793c4c4, 0xf255a7a7, 0x82fc7e7e, 0x477a3d3d);
-       &data_word(0xacc86464, 0xe7ba5d5d, 0x2b321919, 0x95e67373);
-       &data_word(0xa0c06060, 0x98198181, 0xd19e4f4f, 0x7fa3dcdc);
-       &data_word(0x66442222, 0x7e542a2a, 0xab3b9090, 0x830b8888);
-       &data_word(0xca8c4646, 0x29c7eeee, 0xd36bb8b8, 0x3c281414);
-       &data_word(0x79a7dede, 0xe2bc5e5e, 0x1d160b0b, 0x76addbdb);
-       &data_word(0x3bdbe0e0, 0x56643232, 0x4e743a3a, 0x1e140a0a);
-       &data_word(0xdb924949, 0x0a0c0606, 0x6c482424, 0xe4b85c5c);
-       &data_word(0x5d9fc2c2, 0x6ebdd3d3, 0xef43acac, 0xa6c46262);
-       &data_word(0xa8399191, 0xa4319595, 0x37d3e4e4, 0x8bf27979);
-       &data_word(0x32d5e7e7, 0x438bc8c8, 0x596e3737, 0xb7da6d6d);
-       &data_word(0x8c018d8d, 0x64b1d5d5, 0xd29c4e4e, 0xe049a9a9);
-       &data_word(0xb4d86c6c, 0xfaac5656, 0x07f3f4f4, 0x25cfeaea);
-       &data_word(0xafca6565, 0x8ef47a7a, 0xe947aeae, 0x18100808);
-       &data_word(0xd56fbaba, 0x88f07878, 0x6f4a2525, 0x725c2e2e);
-       &data_word(0x24381c1c, 0xf157a6a6, 0xc773b4b4, 0x5197c6c6);
-       &data_word(0x23cbe8e8, 0x7ca1dddd, 0x9ce87474, 0x213e1f1f);
-       &data_word(0xdd964b4b, 0xdc61bdbd, 0x860d8b8b, 0x850f8a8a);
-       &data_word(0x90e07070, 0x427c3e3e, 0xc471b5b5, 0xaacc6666);
-       &data_word(0xd8904848, 0x05060303, 0x01f7f6f6, 0x121c0e0e);
-       &data_word(0xa3c26161, 0x5f6a3535, 0xf9ae5757, 0xd069b9b9);
-       &data_word(0x91178686, 0x5899c1c1, 0x273a1d1d, 0xb9279e9e);
-       &data_word(0x38d9e1e1, 0x13ebf8f8, 0xb32b9898, 0x33221111);
-       &data_word(0xbbd26969, 0x70a9d9d9, 0x89078e8e, 0xa7339494);
-       &data_word(0xb62d9b9b, 0x223c1e1e, 0x92158787, 0x20c9e9e9);
-       &data_word(0x4987cece, 0xffaa5555, 0x78502828, 0x7aa5dfdf);
-       &data_word(0x8f038c8c, 0xf859a1a1, 0x80098989, 0x171a0d0d);
-       &data_word(0xda65bfbf, 0x31d7e6e6, 0xc6844242, 0xb8d06868);
-       &data_word(0xc3824141, 0xb0299999, 0x775a2d2d, 0x111e0f0f);
-       &data_word(0xcb7bb0b0, 0xfca85454, 0xd66dbbbb, 0x3a2c1616);
-#Te2:
-       &data_word(0x63a5c663, 0x7c84f87c, 0x7799ee77, 0x7b8df67b);
-       &data_word(0xf20dfff2, 0x6bbdd66b, 0x6fb1de6f, 0xc55491c5);
-       &data_word(0x30506030, 0x01030201, 0x67a9ce67, 0x2b7d562b);
-       &data_word(0xfe19e7fe, 0xd762b5d7, 0xabe64dab, 0x769aec76);
-       &data_word(0xca458fca, 0x829d1f82, 0xc94089c9, 0x7d87fa7d);
-       &data_word(0xfa15effa, 0x59ebb259, 0x47c98e47, 0xf00bfbf0);
-       &data_word(0xadec41ad, 0xd467b3d4, 0xa2fd5fa2, 0xafea45af);
-       &data_word(0x9cbf239c, 0xa4f753a4, 0x7296e472, 0xc05b9bc0);
-       &data_word(0xb7c275b7, 0xfd1ce1fd, 0x93ae3d93, 0x266a4c26);
-       &data_word(0x365a6c36, 0x3f417e3f, 0xf702f5f7, 0xcc4f83cc);
-       &data_word(0x345c6834, 0xa5f451a5, 0xe534d1e5, 0xf108f9f1);
-       &data_word(0x7193e271, 0xd873abd8, 0x31536231, 0x153f2a15);
-       &data_word(0x040c0804, 0xc75295c7, 0x23654623, 0xc35e9dc3);
-       &data_word(0x18283018, 0x96a13796, 0x050f0a05, 0x9ab52f9a);
-       &data_word(0x07090e07, 0x12362412, 0x809b1b80, 0xe23ddfe2);
-       &data_word(0xeb26cdeb, 0x27694e27, 0xb2cd7fb2, 0x759fea75);
-       &data_word(0x091b1209, 0x839e1d83, 0x2c74582c, 0x1a2e341a);
-       &data_word(0x1b2d361b, 0x6eb2dc6e, 0x5aeeb45a, 0xa0fb5ba0);
-       &data_word(0x52f6a452, 0x3b4d763b, 0xd661b7d6, 0xb3ce7db3);
-       &data_word(0x297b5229, 0xe33edde3, 0x2f715e2f, 0x84971384);
-       &data_word(0x53f5a653, 0xd168b9d1, 0x00000000, 0xed2cc1ed);
-       &data_word(0x20604020, 0xfc1fe3fc, 0xb1c879b1, 0x5bedb65b);
-       &data_word(0x6abed46a, 0xcb468dcb, 0xbed967be, 0x394b7239);
-       &data_word(0x4ade944a, 0x4cd4984c, 0x58e8b058, 0xcf4a85cf);
-       &data_word(0xd06bbbd0, 0xef2ac5ef, 0xaae54faa, 0xfb16edfb);
-       &data_word(0x43c58643, 0x4dd79a4d, 0x33556633, 0x85941185);
-       &data_word(0x45cf8a45, 0xf910e9f9, 0x02060402, 0x7f81fe7f);
-       &data_word(0x50f0a050, 0x3c44783c, 0x9fba259f, 0xa8e34ba8);
-       &data_word(0x51f3a251, 0xa3fe5da3, 0x40c08040, 0x8f8a058f);
-       &data_word(0x92ad3f92, 0x9dbc219d, 0x38487038, 0xf504f1f5);
-       &data_word(0xbcdf63bc, 0xb6c177b6, 0xda75afda, 0x21634221);
-       &data_word(0x10302010, 0xff1ae5ff, 0xf30efdf3, 0xd26dbfd2);
-       &data_word(0xcd4c81cd, 0x0c14180c, 0x13352613, 0xec2fc3ec);
-       &data_word(0x5fe1be5f, 0x97a23597, 0x44cc8844, 0x17392e17);
-       &data_word(0xc45793c4, 0xa7f255a7, 0x7e82fc7e, 0x3d477a3d);
-       &data_word(0x64acc864, 0x5de7ba5d, 0x192b3219, 0x7395e673);
-       &data_word(0x60a0c060, 0x81981981, 0x4fd19e4f, 0xdc7fa3dc);
-       &data_word(0x22664422, 0x2a7e542a, 0x90ab3b90, 0x88830b88);
-       &data_word(0x46ca8c46, 0xee29c7ee, 0xb8d36bb8, 0x143c2814);
-       &data_word(0xde79a7de, 0x5ee2bc5e, 0x0b1d160b, 0xdb76addb);
-       &data_word(0xe03bdbe0, 0x32566432, 0x3a4e743a, 0x0a1e140a);
-       &data_word(0x49db9249, 0x060a0c06, 0x246c4824, 0x5ce4b85c);
-       &data_word(0xc25d9fc2, 0xd36ebdd3, 0xacef43ac, 0x62a6c462);
-       &data_word(0x91a83991, 0x95a43195, 0xe437d3e4, 0x798bf279);
-       &data_word(0xe732d5e7, 0xc8438bc8, 0x37596e37, 0x6db7da6d);
-       &data_word(0x8d8c018d, 0xd564b1d5, 0x4ed29c4e, 0xa9e049a9);
-       &data_word(0x6cb4d86c, 0x56faac56, 0xf407f3f4, 0xea25cfea);
-       &data_word(0x65afca65, 0x7a8ef47a, 0xaee947ae, 0x08181008);
-       &data_word(0xbad56fba, 0x7888f078, 0x256f4a25, 0x2e725c2e);
-       &data_word(0x1c24381c, 0xa6f157a6, 0xb4c773b4, 0xc65197c6);
-       &data_word(0xe823cbe8, 0xdd7ca1dd, 0x749ce874, 0x1f213e1f);
-       &data_word(0x4bdd964b, 0xbddc61bd, 0x8b860d8b, 0x8a850f8a);
-       &data_word(0x7090e070, 0x3e427c3e, 0xb5c471b5, 0x66aacc66);
-       &data_word(0x48d89048, 0x03050603, 0xf601f7f6, 0x0e121c0e);
-       &data_word(0x61a3c261, 0x355f6a35, 0x57f9ae57, 0xb9d069b9);
-       &data_word(0x86911786, 0xc15899c1, 0x1d273a1d, 0x9eb9279e);
-       &data_word(0xe138d9e1, 0xf813ebf8, 0x98b32b98, 0x11332211);
-       &data_word(0x69bbd269, 0xd970a9d9, 0x8e89078e, 0x94a73394);
-       &data_word(0x9bb62d9b, 0x1e223c1e, 0x87921587, 0xe920c9e9);
-       &data_word(0xce4987ce, 0x55ffaa55, 0x28785028, 0xdf7aa5df);
-       &data_word(0x8c8f038c, 0xa1f859a1, 0x89800989, 0x0d171a0d);
-       &data_word(0xbfda65bf, 0xe631d7e6, 0x42c68442, 0x68b8d068);
-       &data_word(0x41c38241, 0x99b02999, 0x2d775a2d, 0x0f111e0f);
-       &data_word(0xb0cb7bb0, 0x54fca854, 0xbbd66dbb, 0x163a2c16);
-#Te3:
-       &data_word(0x6363a5c6, 0x7c7c84f8, 0x777799ee, 0x7b7b8df6);
-       &data_word(0xf2f20dff, 0x6b6bbdd6, 0x6f6fb1de, 0xc5c55491);
-       &data_word(0x30305060, 0x01010302, 0x6767a9ce, 0x2b2b7d56);
-       &data_word(0xfefe19e7, 0xd7d762b5, 0xababe64d, 0x76769aec);
-       &data_word(0xcaca458f, 0x82829d1f, 0xc9c94089, 0x7d7d87fa);
-       &data_word(0xfafa15ef, 0x5959ebb2, 0x4747c98e, 0xf0f00bfb);
-       &data_word(0xadadec41, 0xd4d467b3, 0xa2a2fd5f, 0xafafea45);
-       &data_word(0x9c9cbf23, 0xa4a4f753, 0x727296e4, 0xc0c05b9b);
-       &data_word(0xb7b7c275, 0xfdfd1ce1, 0x9393ae3d, 0x26266a4c);
-       &data_word(0x36365a6c, 0x3f3f417e, 0xf7f702f5, 0xcccc4f83);
-       &data_word(0x34345c68, 0xa5a5f451, 0xe5e534d1, 0xf1f108f9);
-       &data_word(0x717193e2, 0xd8d873ab, 0x31315362, 0x15153f2a);
-       &data_word(0x04040c08, 0xc7c75295, 0x23236546, 0xc3c35e9d);
-       &data_word(0x18182830, 0x9696a137, 0x05050f0a, 0x9a9ab52f);
-       &data_word(0x0707090e, 0x12123624, 0x80809b1b, 0xe2e23ddf);
-       &data_word(0xebeb26cd, 0x2727694e, 0xb2b2cd7f, 0x75759fea);
-       &data_word(0x09091b12, 0x83839e1d, 0x2c2c7458, 0x1a1a2e34);
-       &data_word(0x1b1b2d36, 0x6e6eb2dc, 0x5a5aeeb4, 0xa0a0fb5b);
-       &data_word(0x5252f6a4, 0x3b3b4d76, 0xd6d661b7, 0xb3b3ce7d);
-       &data_word(0x29297b52, 0xe3e33edd, 0x2f2f715e, 0x84849713);
-       &data_word(0x5353f5a6, 0xd1d168b9, 0x00000000, 0xeded2cc1);
-       &data_word(0x20206040, 0xfcfc1fe3, 0xb1b1c879, 0x5b5bedb6);
-       &data_word(0x6a6abed4, 0xcbcb468d, 0xbebed967, 0x39394b72);
-       &data_word(0x4a4ade94, 0x4c4cd498, 0x5858e8b0, 0xcfcf4a85);
-       &data_word(0xd0d06bbb, 0xefef2ac5, 0xaaaae54f, 0xfbfb16ed);
-       &data_word(0x4343c586, 0x4d4dd79a, 0x33335566, 0x85859411);
-       &data_word(0x4545cf8a, 0xf9f910e9, 0x02020604, 0x7f7f81fe);
-       &data_word(0x5050f0a0, 0x3c3c4478, 0x9f9fba25, 0xa8a8e34b);
-       &data_word(0x5151f3a2, 0xa3a3fe5d, 0x4040c080, 0x8f8f8a05);
-       &data_word(0x9292ad3f, 0x9d9dbc21, 0x38384870, 0xf5f504f1);
-       &data_word(0xbcbcdf63, 0xb6b6c177, 0xdada75af, 0x21216342);
-       &data_word(0x10103020, 0xffff1ae5, 0xf3f30efd, 0xd2d26dbf);
-       &data_word(0xcdcd4c81, 0x0c0c1418, 0x13133526, 0xecec2fc3);
-       &data_word(0x5f5fe1be, 0x9797a235, 0x4444cc88, 0x1717392e);
-       &data_word(0xc4c45793, 0xa7a7f255, 0x7e7e82fc, 0x3d3d477a);
-       &data_word(0x6464acc8, 0x5d5de7ba, 0x19192b32, 0x737395e6);
-       &data_word(0x6060a0c0, 0x81819819, 0x4f4fd19e, 0xdcdc7fa3);
-       &data_word(0x22226644, 0x2a2a7e54, 0x9090ab3b, 0x8888830b);
-       &data_word(0x4646ca8c, 0xeeee29c7, 0xb8b8d36b, 0x14143c28);
-       &data_word(0xdede79a7, 0x5e5ee2bc, 0x0b0b1d16, 0xdbdb76ad);
-       &data_word(0xe0e03bdb, 0x32325664, 0x3a3a4e74, 0x0a0a1e14);
-       &data_word(0x4949db92, 0x06060a0c, 0x24246c48, 0x5c5ce4b8);
-       &data_word(0xc2c25d9f, 0xd3d36ebd, 0xacacef43, 0x6262a6c4);
-       &data_word(0x9191a839, 0x9595a431, 0xe4e437d3, 0x79798bf2);
-       &data_word(0xe7e732d5, 0xc8c8438b, 0x3737596e, 0x6d6db7da);
-       &data_word(0x8d8d8c01, 0xd5d564b1, 0x4e4ed29c, 0xa9a9e049);
-       &data_word(0x6c6cb4d8, 0x5656faac, 0xf4f407f3, 0xeaea25cf);
-       &data_word(0x6565afca, 0x7a7a8ef4, 0xaeaee947, 0x08081810);
-       &data_word(0xbabad56f, 0x787888f0, 0x25256f4a, 0x2e2e725c);
-       &data_word(0x1c1c2438, 0xa6a6f157, 0xb4b4c773, 0xc6c65197);
-       &data_word(0xe8e823cb, 0xdddd7ca1, 0x74749ce8, 0x1f1f213e);
-       &data_word(0x4b4bdd96, 0xbdbddc61, 0x8b8b860d, 0x8a8a850f);
-       &data_word(0x707090e0, 0x3e3e427c, 0xb5b5c471, 0x6666aacc);
-       &data_word(0x4848d890, 0x03030506, 0xf6f601f7, 0x0e0e121c);
-       &data_word(0x6161a3c2, 0x35355f6a, 0x5757f9ae, 0xb9b9d069);
-       &data_word(0x86869117, 0xc1c15899, 0x1d1d273a, 0x9e9eb927);
-       &data_word(0xe1e138d9, 0xf8f813eb, 0x9898b32b, 0x11113322);
-       &data_word(0x6969bbd2, 0xd9d970a9, 0x8e8e8907, 0x9494a733);
-       &data_word(0x9b9bb62d, 0x1e1e223c, 0x87879215, 0xe9e920c9);
-       &data_word(0xcece4987, 0x5555ffaa, 0x28287850, 0xdfdf7aa5);
-       &data_word(0x8c8c8f03, 0xa1a1f859, 0x89898009, 0x0d0d171a);
-       &data_word(0xbfbfda65, 0xe6e631d7, 0x4242c684, 0x6868b8d0);
-       &data_word(0x4141c382, 0x9999b029, 0x2d2d775a, 0x0f0f111e);
-       &data_word(0xb0b0cb7b, 0x5454fca8, 0xbbbbd66d, 0x16163a2c);
-#Te4:
-       &data_word(0x63636363, 0x7c7c7c7c, 0x77777777, 0x7b7b7b7b);
-       &data_word(0xf2f2f2f2, 0x6b6b6b6b, 0x6f6f6f6f, 0xc5c5c5c5);
-       &data_word(0x30303030, 0x01010101, 0x67676767, 0x2b2b2b2b);
-       &data_word(0xfefefefe, 0xd7d7d7d7, 0xabababab, 0x76767676);
-       &data_word(0xcacacaca, 0x82828282, 0xc9c9c9c9, 0x7d7d7d7d);
-       &data_word(0xfafafafa, 0x59595959, 0x47474747, 0xf0f0f0f0);
-       &data_word(0xadadadad, 0xd4d4d4d4, 0xa2a2a2a2, 0xafafafaf);
-       &data_word(0x9c9c9c9c, 0xa4a4a4a4, 0x72727272, 0xc0c0c0c0);
-       &data_word(0xb7b7b7b7, 0xfdfdfdfd, 0x93939393, 0x26262626);
-       &data_word(0x36363636, 0x3f3f3f3f, 0xf7f7f7f7, 0xcccccccc);
-       &data_word(0x34343434, 0xa5a5a5a5, 0xe5e5e5e5, 0xf1f1f1f1);
-       &data_word(0x71717171, 0xd8d8d8d8, 0x31313131, 0x15151515);
-       &data_word(0x04040404, 0xc7c7c7c7, 0x23232323, 0xc3c3c3c3);
-       &data_word(0x18181818, 0x96969696, 0x05050505, 0x9a9a9a9a);
-       &data_word(0x07070707, 0x12121212, 0x80808080, 0xe2e2e2e2);
-       &data_word(0xebebebeb, 0x27272727, 0xb2b2b2b2, 0x75757575);
-       &data_word(0x09090909, 0x83838383, 0x2c2c2c2c, 0x1a1a1a1a);
-       &data_word(0x1b1b1b1b, 0x6e6e6e6e, 0x5a5a5a5a, 0xa0a0a0a0);
-       &data_word(0x52525252, 0x3b3b3b3b, 0xd6d6d6d6, 0xb3b3b3b3);
-       &data_word(0x29292929, 0xe3e3e3e3, 0x2f2f2f2f, 0x84848484);
-       &data_word(0x53535353, 0xd1d1d1d1, 0x00000000, 0xedededed);
-       &data_word(0x20202020, 0xfcfcfcfc, 0xb1b1b1b1, 0x5b5b5b5b);
-       &data_word(0x6a6a6a6a, 0xcbcbcbcb, 0xbebebebe, 0x39393939);
-       &data_word(0x4a4a4a4a, 0x4c4c4c4c, 0x58585858, 0xcfcfcfcf);
-       &data_word(0xd0d0d0d0, 0xefefefef, 0xaaaaaaaa, 0xfbfbfbfb);
-       &data_word(0x43434343, 0x4d4d4d4d, 0x33333333, 0x85858585);
-       &data_word(0x45454545, 0xf9f9f9f9, 0x02020202, 0x7f7f7f7f);
-       &data_word(0x50505050, 0x3c3c3c3c, 0x9f9f9f9f, 0xa8a8a8a8);
-       &data_word(0x51515151, 0xa3a3a3a3, 0x40404040, 0x8f8f8f8f);
-       &data_word(0x92929292, 0x9d9d9d9d, 0x38383838, 0xf5f5f5f5);
-       &data_word(0xbcbcbcbc, 0xb6b6b6b6, 0xdadadada, 0x21212121);
-       &data_word(0x10101010, 0xffffffff, 0xf3f3f3f3, 0xd2d2d2d2);
-       &data_word(0xcdcdcdcd, 0x0c0c0c0c, 0x13131313, 0xecececec);
-       &data_word(0x5f5f5f5f, 0x97979797, 0x44444444, 0x17171717);
-       &data_word(0xc4c4c4c4, 0xa7a7a7a7, 0x7e7e7e7e, 0x3d3d3d3d);
-       &data_word(0x64646464, 0x5d5d5d5d, 0x19191919, 0x73737373);
-       &data_word(0x60606060, 0x81818181, 0x4f4f4f4f, 0xdcdcdcdc);
-       &data_word(0x22222222, 0x2a2a2a2a, 0x90909090, 0x88888888);
-       &data_word(0x46464646, 0xeeeeeeee, 0xb8b8b8b8, 0x14141414);
-       &data_word(0xdededede, 0x5e5e5e5e, 0x0b0b0b0b, 0xdbdbdbdb);
-       &data_word(0xe0e0e0e0, 0x32323232, 0x3a3a3a3a, 0x0a0a0a0a);
-       &data_word(0x49494949, 0x06060606, 0x24242424, 0x5c5c5c5c);
-       &data_word(0xc2c2c2c2, 0xd3d3d3d3, 0xacacacac, 0x62626262);
-       &data_word(0x91919191, 0x95959595, 0xe4e4e4e4, 0x79797979);
-       &data_word(0xe7e7e7e7, 0xc8c8c8c8, 0x37373737, 0x6d6d6d6d);
-       &data_word(0x8d8d8d8d, 0xd5d5d5d5, 0x4e4e4e4e, 0xa9a9a9a9);
-       &data_word(0x6c6c6c6c, 0x56565656, 0xf4f4f4f4, 0xeaeaeaea);
-       &data_word(0x65656565, 0x7a7a7a7a, 0xaeaeaeae, 0x08080808);
-       &data_word(0xbabababa, 0x78787878, 0x25252525, 0x2e2e2e2e);
-       &data_word(0x1c1c1c1c, 0xa6a6a6a6, 0xb4b4b4b4, 0xc6c6c6c6);
-       &data_word(0xe8e8e8e8, 0xdddddddd, 0x74747474, 0x1f1f1f1f);
-       &data_word(0x4b4b4b4b, 0xbdbdbdbd, 0x8b8b8b8b, 0x8a8a8a8a);
-       &data_word(0x70707070, 0x3e3e3e3e, 0xb5b5b5b5, 0x66666666);
-       &data_word(0x48484848, 0x03030303, 0xf6f6f6f6, 0x0e0e0e0e);
-       &data_word(0x61616161, 0x35353535, 0x57575757, 0xb9b9b9b9);
-       &data_word(0x86868686, 0xc1c1c1c1, 0x1d1d1d1d, 0x9e9e9e9e);
-       &data_word(0xe1e1e1e1, 0xf8f8f8f8, 0x98989898, 0x11111111);
-       &data_word(0x69696969, 0xd9d9d9d9, 0x8e8e8e8e, 0x94949494);
-       &data_word(0x9b9b9b9b, 0x1e1e1e1e, 0x87878787, 0xe9e9e9e9);
-       &data_word(0xcececece, 0x55555555, 0x28282828, 0xdfdfdfdf);
-       &data_word(0x8c8c8c8c, 0xa1a1a1a1, 0x89898989, 0x0d0d0d0d);
-       &data_word(0xbfbfbfbf, 0xe6e6e6e6, 0x42424242, 0x68686868);
-       &data_word(0x41414141, 0x99999999, 0x2d2d2d2d, 0x0f0f0f0f);
-       &data_word(0xb0b0b0b0, 0x54545454, 0xbbbbbbbb, 0x16161616);
-&function_end_B("AES_encrypt");
+       &_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:
+       &data_word(0x00000001, 0x00000002, 0x00000004, 0x00000008);
+       &data_word(0x00000010, 0x00000020, 0x00000040, 0x00000080);
+       &data_word(0x0000001b, 0x00000036, 0, 0, 0, 0, 0, 0);
+&function_end_B("_x86_AES_encrypt");
+
+# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
+&public_label("AES_Te");
+&function_begin("AES_encrypt");
+       &mov    ($acc,&wparam(0));              # load inp
+       &mov    ($key,&wparam(2));              # load key
+
+       &mov    ($s0,"esp");
+       &sub    ("esp",24);
+       &and    ("esp",-64);
+       &add    ("esp",4);
+       &mov    (&DWP(16,"esp"),$s0);
+
+       &call   (&label("pic_point"));          # make it PIC!
+       &set_label("pic_point");
+       &blindpop("ebp");
+       &lea    ("ebp",&DWP(&label("AES_Te")."-".&label("pic_point"),"ebp"));
+
+       &mov    ($s0,&DWP(0,$acc));             # load input data
+       &mov    ($s1,&DWP(4,$acc));
+       &mov    ($s2,&DWP(8,$acc));
+       &mov    ($s3,&DWP(12,$acc));
+
+       &call   ("_x86_AES_encrypt");
+
+       &mov    ("esp",&DWP(16,"esp"));
+
+       &mov    ($acc,&wparam(1));              # load out
+       &mov    (&DWP(0,$acc),$s0);             # write output data
+       &mov    (&DWP(4,$acc),$s1);
+       &mov    (&DWP(8,$acc),$s2);
+       &mov    (&DWP(12,$acc),$s3);
+&function_end("AES_encrypt");
+
+#------------------------------------------------------------------#
 
 sub decstep()
 { my ($i,$td,@s) = @_;
-  my $tmp,$out;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
 
-       if ($i==2)  {   &movz   ($out="esi",&HB($s[0]));        }       ##
-       elsif($i==3){   &mov    ("edi",&DWP(12,"esp"));
-                       &shr    ($out=$s[0],24);                }
-       else        {   &mov    ($out="esi",$s[0]);
-                       &shr    ($out,24);                      }
-                       &mov    ($out,&DWP(1024*0,$td,$out,4));
+       # no instructions are reordered, as performance appears
+       # optimal... or rather that all attempts to reorder didn't
+       # result in better performance [which by the way is not a
+       # bit lower than ecryption].
+       if($i==3)   {   &mov    ($key,&DWP(12,"esp"));          }
+       else        {   &mov    ($out,$s[0]);                   }
+                       &and    ($out,0xFF);
+                       &mov    ($out,&DWP(0,$td,$out,8));
 
-       if ($i==3)  {   $tmp=$s[1];                             }       ##
-       else        {   &mov    ($tmp="edi",$s[1]);
-                       &shr    ($tmp,16);                      }
-                       &and    ($tmp,0xFF);
-                       &xor    ($out,&DWP(1024*1,$td,$tmp,4));
+       if ($i==3)  {   $tmp=$s[1];                             }
+                       &movz   ($tmp,&HB($s[1]));
+                       &xor    ($out,&DWP(3,$td,$tmp,8));
 
-       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],"esi");         }
-       else        {   $tmp="edi";                             }
-                       &movz   ($tmp,&HB($s[2]));
-                       &xor    ($out,&DWP(1024*2,$td,$tmp,4));
-
-       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(4,"esp")); }
-       else        {   &mov    ($tmp="edi",$s[3]);             }
-       if ($i==1)  {   &shr    ($s[3],16);                     }       ##
+       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],$acc);          }
+       else        {   &mov    ($tmp,$s[2]);                   }
+                       &shr    ($tmp,16);
                        &and    ($tmp,0xFF);
-                       &xor    ($out,&DWP(1024*3,$td,$tmp,4));
-       if ($i<2)   {   &mov    (&DWP(4*$i,"esp"),$out);        }
-       if ($i==3)  {   &mov    ($s[3],&DWP(0,"esp"));          }
+                       &xor    ($out,&DWP(2,$td,$tmp,8));
+
+       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(8,"esp")); }
+       else        {   &mov    ($tmp,$s[3]);                   }
+                       &shr    ($tmp,24);
+                       &xor    ($out,&DWP(1,$td,$tmp,8));
+       if ($i<2)   {   &mov    (&DWP(4+4*$i,"esp"),$out);      }
+       if ($i==3)  {   &mov    ($s[3],&DWP(4,"esp"));          }
+                       &comment();
 }
 
 sub declast()
 { my ($i,$td,@s)=@_;
-  my $tmp,$out;
-
-       if ($i==2)  {   &movz   ($out="esi",&HB($s[0]));        }       ##
-       elsif($i==3){   &mov    ("edi",&DWP(12,"esp"));
-                       &shr    ($out=$s[0],24);                }
-       else        {   &mov    ($out="esi",$s[0]);
-                       &shr    ($out,24);                      }
-                       &mov    ($out,&DWP(0,$td,$out,4));
-                       &and    ($out,0xff000000);
-
-       if ($i==3)  {   $tmp=$s[1];                             }       ##
-       else        {   &mov    ($tmp="edi",$s[1]);
-                       &shr    ($tmp,16);                      }
-                       &and    ($tmp,0xFF);
-                       &mov    ($tmp,&DWP(0,$td,$tmp,4));
-                       &and    ($tmp,0x00ff0000);
-                       &xor    ($out,$tmp);
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+       if($i==3)   {   &mov    ($key,&DWP(12,"esp"));          }
+       else        {   &mov    ($out,$s[0]);                   }
+                       &and    ($out,0xFF);
+                       &mov    ($out,&DWP(2048,$td,$out,4));
+                       &and    ($out,0x000000ff);
 
-       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],"esi");         }
-       else        {   $tmp="edi";                             }
-                       &movz   ($tmp,&HB($s[2]));
-                       &mov    ($tmp,&DWP(0,$td,$tmp,4));
+       if ($i==3)  {   $tmp=$s[1];                             }
+                       &movz   ($tmp,&HB($s[1]));
+                       &mov    ($tmp,&DWP(2048,$td,$tmp,4));
                        &and    ($tmp,0x0000ff00);
                        &xor    ($out,$tmp);
 
-       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(4,"esp")); }
-       else        {   &mov    ($tmp="edi",$s[3]);             }
-       if ($i==1)  {   &shr    ($s[3],16);                     }       ##
+       if ($i==3)  {   $tmp=$s[2]; &mov ($s[1],$acc);          }
+       else        {   mov     ($tmp,$s[2]);                   }
+                       &shr    ($tmp,16);
                        &and    ($tmp,0xFF);
-                       &mov    ($tmp,&DWP(0,$td,$tmp,4));
-                       &and    ($tmp,0x000000ff);
+                       &mov    ($tmp,&DWP(2048,$td,$tmp,4));
+                       &and    ($tmp,0x00ff0000);
                        &xor    ($out,$tmp);
-       if ($i<2)   {   &mov    (&DWP(4*$i,"esp"),$out);        }
-       if ($i==3)  {   &mov    ($s[3],&DWP(0,"esp"));          }
+
+       if ($i==3)  {   $tmp=$s[3]; &mov ($s[2],&DWP(8,"esp")); }
+       else        {   &mov    ($tmp,$s[3]);                   }
+                       &shr    ($tmp,24);
+                       &mov    ($tmp,&DWP(2048,$td,$tmp,4));
+                       &and    ($tmp,0xff000000);
+                       &xor    ($out,$tmp);
+       if ($i<2)   {   &mov    (&DWP(4+4*$i,"esp"),$out);      }
+       if ($i==3)  {   &mov    ($s[3],&DWP(4,"esp"));          }
 }
 
-# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
 &public_label("AES_Td");
-&function_begin("AES_decrypt");
-       &mov    ("esi",&wparam(0));             # load inp
-       &mov    ("edi",&wparam(2));             # load key
-
-        &call   (&label("pic_point"));          # make it PIC!
-&set_label("pic_point");
-        &blindpop("ebp");
-        &lea    ("ebp",&DWP(&label("AES_Td")."-".&label("pic_point"),"ebp"));
-
-       # allocate aligned stack frame
-       &mov    ("eax","esp");
-       &sub    ("esp",20);
-       &and    ("esp",-16);
-
-       &mov    (&DWP(12,"esp"),"edi");         # save key
-       &mov    (&DWP(16,"esp"),"eax");         # save %esp
-
-       &mov    ($s0,&DWP(0,"esi"));            # load input data
-       &mov    ($s1,&DWP(4,"esi"));
-       &mov    ($s2,&DWP(8,"esi"));
-       &mov    ($s3,&DWP(12,"esi"));
-       &bswap  ($s0);
-       &bswap  ($s1);
-       &bswap  ($s2);
-       &bswap  ($s3);
-       &xor    ($s0,&DWP(0,"edi"));
-       &xor    ($s1,&DWP(4,"edi"));
-       &xor    ($s2,&DWP(8,"edi"));
-       &xor    ($s3,&DWP(12,"edi"));
-
-       &mov    ("esi",&DWP(240,"edi"));        # load key->rounds
+&function_begin_B("_x86_AES_decrypt");
+       # note that caller is expected to allocate stack frame for me!
+       &mov    (&DWP(12,"esp"),$key);          # save key
+
+       &xor    ($s0,&DWP(0,$key));             # xor with key
+       &xor    ($s1,&DWP(4,$key));
+       &xor    ($s2,&DWP(8,$key));
+       &xor    ($s3,&DWP(12,$key));
+
+       &mov    ($acc,&DWP(240,$key));          # load key->rounds
 
        if ($small_footprint) {
-           &lea        ("esi",&DWP(-2,"esi","esi"));
-           &lea        ("esi",&DWP(0,"edi","esi",8));
-           &mov        (&DWP(8,"esp"),"esi");  # end of key schedule
+           &lea        ($acc,&DWP(-2,$acc,$acc));
+           &lea        ($acc,&DWP(0,$key,$acc,8));
+           &mov        (&DWP(16,"esp"),$acc);  # end of key schedule
            &align      (4);
            &set_label("loop");
                &decstep(0,"ebp",$s0,$s3,$s2,$s1);
                &decstep(1,"ebp",$s1,$s0,$s3,$s2);
                &decstep(2,"ebp",$s2,$s1,$s0,$s3);
                &decstep(3,"ebp",$s3,$s2,$s1,$s0);
-               &add    ("edi",16);                     # advance rd_key
-               &xor    ($s0,&DWP(0,"edi"));
-               &xor    ($s1,&DWP(4,"edi"));
-               &xor    ($s2,&DWP(8,"edi"));
-               &xor    ($s3,&DWP(12,"edi"));
-           &cmp        ("edi",&DWP(8,"esp"));
-           &mov        (&DWP(12,"esp"),"edi");
+               &add    ($key,16);              # advance rd_key
+               &xor    ($s0,&DWP(0,$key));
+               &xor    ($s1,&DWP(4,$key));
+               &xor    ($s2,&DWP(8,$key));
+               &xor    ($s3,&DWP(12,$key));
+           &cmp        ($key,&DWP(16,"esp"));
+           &mov        (&DWP(12,"esp"),$key);
            &jb         (&label("loop"));
        }
        else {
-           &cmp        ("esi",10);
+           &cmp        ($acc,10);
            &jle        (&label("10rounds"));
-           &cmp        ("esi",12);
+           &cmp        ($acc,12);
            &jle        (&label("12rounds"));
 
        &set_label("14rounds");
@@ -704,327 +580,117 @@ sub declast()
                &decstep(1,"ebp",$s1,$s0,$s3,$s2);
                &decstep(2,"ebp",$s2,$s1,$s0,$s3);
                &decstep(3,"ebp",$s3,$s2,$s1,$s0);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
-           &add        ("edi",32);
-           &mov        (&DWP(12,"esp"),"edi");         # advance rd_key
+           &add        ($key,32);
+           &mov        (&DWP(12,"esp"),$key);  # advance rd_key
        &set_label("12rounds");
            for ($i=1;$i<3;$i++) {
                &decstep(0,"ebp",$s0,$s3,$s2,$s1);
                &decstep(1,"ebp",$s1,$s0,$s3,$s2);
                &decstep(2,"ebp",$s2,$s1,$s0,$s3);
                &decstep(3,"ebp",$s3,$s2,$s1,$s0);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
-           &add        ("edi",32);
-           &mov        (&DWP(12,"esp"),"edi");         # advance rd_key
+           &add        ($key,32);
+           &mov        (&DWP(12,"esp"),$key);  # advance rd_key
        &set_label("10rounds");
            for ($i=1;$i<10;$i++) {
                &decstep(0,"ebp",$s0,$s3,$s2,$s1);
                &decstep(1,"ebp",$s1,$s0,$s3,$s2);
                &decstep(2,"ebp",$s2,$s1,$s0,$s3);
                &decstep(3,"ebp",$s3,$s2,$s1,$s0);
-               &xor    ($s0,&DWP(16*$i+0,"edi"));
-               &xor    ($s1,&DWP(16*$i+4,"edi"));
-               &xor    ($s2,&DWP(16*$i+8,"edi"));
-               &xor    ($s3,&DWP(16*$i+12,"edi"));
+               &xor    ($s0,&DWP(16*$i+0,$key));
+               &xor    ($s1,&DWP(16*$i+4,$key));
+               &xor    ($s2,&DWP(16*$i+8,$key));
+               &xor    ($s3,&DWP(16*$i+12,$key));
            }
        }
 
-       &add    ("ebp",4*1024);                 # skip to Te4
        &declast(0,"ebp",$s0,$s3,$s2,$s1);
        &declast(1,"ebp",$s1,$s0,$s3,$s2);
        &declast(2,"ebp",$s2,$s1,$s0,$s3);
        &declast(3,"ebp",$s3,$s2,$s1,$s0);
 
-       &mov    ("esp",&DWP(16,"esp"));         # restore %esp
-       &add    ("edi",$small_footprint?16:160);
-       &xor    ($s0,&DWP(0,"edi"));
-       &xor    ($s1,&DWP(4,"edi"));
-       &xor    ($s2,&DWP(8,"edi"));
-       &xor    ($s3,&DWP(12,"edi"));
-       &bswap  ($s0);
-       &bswap  ($s1);
-       &bswap  ($s2);
-       &bswap  ($s3);
-       &mov    ("edi",&wparam(1));             # load out
-       &mov    (&DWP(0,"edi"),$s0);            # write output data
-       &mov    (&DWP(4,"edi"),$s1);
-       &mov    (&DWP(8,"edi"),$s2);
-       &mov    (&DWP(12,"edi"),$s3);
-
-       &pop    ("edi");
-       &pop    ("esi");
-       &pop    ("ebx");
-       &pop    ("ebp");
+       &add    ($key,$small_footprint?16:160);
+       &xor    ($s0,&DWP(0,$key));
+       &xor    ($s1,&DWP(4,$key));
+       &xor    ($s2,&DWP(8,$key));
+       &xor    ($s3,&DWP(12,$key));
+
        &ret    ();
 
 &set_label("AES_Td",64);       # Yes! I keep it in the code segment!
-       &data_word(0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96);
-       &data_word(0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393);
-       &data_word(0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25);
-       &data_word(0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f);
-       &data_word(0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1);
-       &data_word(0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6);
-       &data_word(0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da);
-       &data_word(0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844);
-       &data_word(0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd);
-       &data_word(0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4);
-       &data_word(0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45);
-       &data_word(0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94);
-       &data_word(0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7);
-       &data_word(0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a);
-       &data_word(0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5);
-       &data_word(0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c);
-       &data_word(0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1);
-       &data_word(0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a);
-       &data_word(0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75);
-       &data_word(0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051);
-       &data_word(0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46);
-       &data_word(0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff);
-       &data_word(0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77);
-       &data_word(0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb);
-       &data_word(0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000);
-       &data_word(0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e);
-       &data_word(0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927);
-       &data_word(0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a);
-       &data_word(0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e);
-       &data_word(0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16);
-       &data_word(0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d);
-       &data_word(0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8);
-       &data_word(0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd);
-       &data_word(0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34);
-       &data_word(0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163);
-       &data_word(0xd731dcca, 0x42638510, 0x13972240, 0x84c61120);
-       &data_word(0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d);
-       &data_word(0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0);
-       &data_word(0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422);
-       &data_word(0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef);
-       &data_word(0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36);
-       &data_word(0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4);
-       &data_word(0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662);
-       &data_word(0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5);
-       &data_word(0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3);
-       &data_word(0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b);
-       &data_word(0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8);
-       &data_word(0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6);
-       &data_word(0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6);
-       &data_word(0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0);
-       &data_word(0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815);
-       &data_word(0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f);
-       &data_word(0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df);
-       &data_word(0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f);
-       &data_word(0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e);
-       &data_word(0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713);
-       &data_word(0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89);
-       &data_word(0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c);
-       &data_word(0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf);
-       &data_word(0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86);
-       &data_word(0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f);
-       &data_word(0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541);
-       &data_word(0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190);
-       &data_word(0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742);
-#Td1:
-       &data_word(0x5051f4a7, 0x537e4165, 0xc31a17a4, 0x963a275e);
-       &data_word(0xcb3bab6b, 0xf11f9d45, 0xabacfa58, 0x934be303);
-       &data_word(0x552030fa, 0xf6ad766d, 0x9188cc76, 0x25f5024c);
-       &data_word(0xfc4fe5d7, 0xd7c52acb, 0x80263544, 0x8fb562a3);
-       &data_word(0x49deb15a, 0x6725ba1b, 0x9845ea0e, 0xe15dfec0);
-       &data_word(0x02c32f75, 0x12814cf0, 0xa38d4697, 0xc66bd3f9);
-       &data_word(0xe7038f5f, 0x9515929c, 0xebbf6d7a, 0xda955259);
-       &data_word(0x2dd4be83, 0xd3587421, 0x2949e069, 0x448ec9c8);
-       &data_word(0x6a75c289, 0x78f48e79, 0x6b99583e, 0xdd27b971);
-       &data_word(0xb6bee14f, 0x17f088ad, 0x66c920ac, 0xb47dce3a);
-       &data_word(0x1863df4a, 0x82e51a31, 0x60975133, 0x4562537f);
-       &data_word(0xe0b16477, 0x84bb6bae, 0x1cfe81a0, 0x94f9082b);
-       &data_word(0x58704868, 0x198f45fd, 0x8794de6c, 0xb7527bf8);
-       &data_word(0x23ab73d3, 0xe2724b02, 0x57e31f8f, 0x2a6655ab);
-       &data_word(0x07b2eb28, 0x032fb5c2, 0x9a86c57b, 0xa5d33708);
-       &data_word(0xf2302887, 0xb223bfa5, 0xba02036a, 0x5ced1682);
-       &data_word(0x2b8acf1c, 0x92a779b4, 0xf0f307f2, 0xa14e69e2);
-       &data_word(0xcd65daf4, 0xd50605be, 0x1fd13462, 0x8ac4a6fe);
-       &data_word(0x9d342e53, 0xa0a2f355, 0x32058ae1, 0x75a4f6eb);
-       &data_word(0x390b83ec, 0xaa4060ef, 0x065e719f, 0x51bd6e10);
-       &data_word(0xf93e218a, 0x3d96dd06, 0xaedd3e05, 0x464de6bd);
-       &data_word(0xb591548d, 0x0571c45d, 0x6f0406d4, 0xff605015);
-       &data_word(0x241998fb, 0x97d6bde9, 0xcc894043, 0x7767d99e);
-       &data_word(0xbdb0e842, 0x8807898b, 0x38e7195b, 0xdb79c8ee);
-       &data_word(0x47a17c0a, 0xe97c420f, 0xc9f8841e, 0x00000000);
-       &data_word(0x83098086, 0x48322bed, 0xac1e1170, 0x4e6c5a72);
-       &data_word(0xfbfd0eff, 0x560f8538, 0x1e3daed5, 0x27362d39);
-       &data_word(0x640a0fd9, 0x21685ca6, 0xd19b5b54, 0x3a24362e);
-       &data_word(0xb10c0a67, 0x0f9357e7, 0xd2b4ee96, 0x9e1b9b91);
-       &data_word(0x4f80c0c5, 0xa261dc20, 0x695a774b, 0x161c121a);
-       &data_word(0x0ae293ba, 0xe5c0a02a, 0x433c22e0, 0x1d121b17);
-       &data_word(0x0b0e090d, 0xadf28bc7, 0xb92db6a8, 0xc8141ea9);
-       &data_word(0x8557f119, 0x4caf7507, 0xbbee99dd, 0xfda37f60);
-       &data_word(0x9ff70126, 0xbc5c72f5, 0xc544663b, 0x345bfb7e);
-       &data_word(0x768b4329, 0xdccb23c6, 0x68b6edfc, 0x63b8e4f1);
-       &data_word(0xcad731dc, 0x10426385, 0x40139722, 0x2084c611);
-       &data_word(0x7d854a24, 0xf8d2bb3d, 0x11aef932, 0x6dc729a1);
-       &data_word(0x4b1d9e2f, 0xf3dcb230, 0xec0d8652, 0xd077c1e3);
-       &data_word(0x6c2bb316, 0x99a970b9, 0xfa119448, 0x2247e964);
-       &data_word(0xc4a8fc8c, 0x1aa0f03f, 0xd8567d2c, 0xef223390);
-       &data_word(0xc787494e, 0xc1d938d1, 0xfe8ccaa2, 0x3698d40b);
-       &data_word(0xcfa6f581, 0x28a57ade, 0x26dab78e, 0xa43fadbf);
-       &data_word(0xe42c3a9d, 0x0d507892, 0x9b6a5fcc, 0x62547e46);
-       &data_word(0xc2f68d13, 0xe890d8b8, 0x5e2e39f7, 0xf582c3af);
-       &data_word(0xbe9f5d80, 0x7c69d093, 0xa96fd52d, 0xb3cf2512);
-       &data_word(0x3bc8ac99, 0xa710187d, 0x6ee89c63, 0x7bdb3bbb);
-       &data_word(0x09cd2678, 0xf46e5918, 0x01ec9ab7, 0xa8834f9a);
-       &data_word(0x65e6956e, 0x7eaaffe6, 0x0821bccf, 0xe6ef15e8);
-       &data_word(0xd9bae79b, 0xce4a6f36, 0xd4ea9f09, 0xd629b07c);
-       &data_word(0xaf31a4b2, 0x312a3f23, 0x30c6a594, 0xc035a266);
-       &data_word(0x37744ebc, 0xa6fc82ca, 0xb0e090d0, 0x1533a7d8);
-       &data_word(0x4af10498, 0xf741ecda, 0x0e7fcd50, 0x2f1791f6);
-       &data_word(0x8d764dd6, 0x4d43efb0, 0x54ccaa4d, 0xdfe49604);
-       &data_word(0xe39ed1b5, 0x1b4c6a88, 0xb8c12c1f, 0x7f466551);
-       &data_word(0x049d5eea, 0x5d018c35, 0x73fa8774, 0x2efb0b41);
-       &data_word(0x5ab3671d, 0x5292dbd2, 0x33e91056, 0x136dd647);
-       &data_word(0x8c9ad761, 0x7a37a10c, 0x8e59f814, 0x89eb133c);
-       &data_word(0xeecea927, 0x35b761c9, 0xede11ce5, 0x3c7a47b1);
-       &data_word(0x599cd2df, 0x3f55f273, 0x791814ce, 0xbf73c737);
-       &data_word(0xea53f7cd, 0x5b5ffdaa, 0x14df3d6f, 0x867844db);
-       &data_word(0x81caaff3, 0x3eb968c4, 0x2c382434, 0x5fc2a340);
-       &data_word(0x72161dc3, 0x0cbce225, 0x8b283c49, 0x41ff0d95);
-       &data_word(0x7139a801, 0xde080cb3, 0x9cd8b4e4, 0x906456c1);
-       &data_word(0x617bcb84, 0x70d532b6, 0x74486c5c, 0x42d0b857);
-#Td2:
-       &data_word(0xa75051f4, 0x65537e41, 0xa4c31a17, 0x5e963a27);
-       &data_word(0x6bcb3bab, 0x45f11f9d, 0x58abacfa, 0x03934be3);
-       &data_word(0xfa552030, 0x6df6ad76, 0x769188cc, 0x4c25f502);
-       &data_word(0xd7fc4fe5, 0xcbd7c52a, 0x44802635, 0xa38fb562);
-       &data_word(0x5a49deb1, 0x1b6725ba, 0x0e9845ea, 0xc0e15dfe);
-       &data_word(0x7502c32f, 0xf012814c, 0x97a38d46, 0xf9c66bd3);
-       &data_word(0x5fe7038f, 0x9c951592, 0x7aebbf6d, 0x59da9552);
-       &data_word(0x832dd4be, 0x21d35874, 0x692949e0, 0xc8448ec9);
-       &data_word(0x896a75c2, 0x7978f48e, 0x3e6b9958, 0x71dd27b9);
-       &data_word(0x4fb6bee1, 0xad17f088, 0xac66c920, 0x3ab47dce);
-       &data_word(0x4a1863df, 0x3182e51a, 0x33609751, 0x7f456253);
-       &data_word(0x77e0b164, 0xae84bb6b, 0xa01cfe81, 0x2b94f908);
-       &data_word(0x68587048, 0xfd198f45, 0x6c8794de, 0xf8b7527b);
-       &data_word(0xd323ab73, 0x02e2724b, 0x8f57e31f, 0xab2a6655);
-       &data_word(0x2807b2eb, 0xc2032fb5, 0x7b9a86c5, 0x08a5d337);
-       &data_word(0x87f23028, 0xa5b223bf, 0x6aba0203, 0x825ced16);
-       &data_word(0x1c2b8acf, 0xb492a779, 0xf2f0f307, 0xe2a14e69);
-       &data_word(0xf4cd65da, 0xbed50605, 0x621fd134, 0xfe8ac4a6);
-       &data_word(0x539d342e, 0x55a0a2f3, 0xe132058a, 0xeb75a4f6);
-       &data_word(0xec390b83, 0xefaa4060, 0x9f065e71, 0x1051bd6e);
-       &data_word(0x8af93e21, 0x063d96dd, 0x05aedd3e, 0xbd464de6);
-       &data_word(0x8db59154, 0x5d0571c4, 0xd46f0406, 0x15ff6050);
-       &data_word(0xfb241998, 0xe997d6bd, 0x43cc8940, 0x9e7767d9);
-       &data_word(0x42bdb0e8, 0x8b880789, 0x5b38e719, 0xeedb79c8);
-       &data_word(0x0a47a17c, 0x0fe97c42, 0x1ec9f884, 0x00000000);
-       &data_word(0x86830980, 0xed48322b, 0x70ac1e11, 0x724e6c5a);
-       &data_word(0xfffbfd0e, 0x38560f85, 0xd51e3dae, 0x3927362d);
-       &data_word(0xd9640a0f, 0xa621685c, 0x54d19b5b, 0x2e3a2436);
-       &data_word(0x67b10c0a, 0xe70f9357, 0x96d2b4ee, 0x919e1b9b);
-       &data_word(0xc54f80c0, 0x20a261dc, 0x4b695a77, 0x1a161c12);
-       &data_word(0xba0ae293, 0x2ae5c0a0, 0xe0433c22, 0x171d121b);
-       &data_word(0x0d0b0e09, 0xc7adf28b, 0xa8b92db6, 0xa9c8141e);
-       &data_word(0x198557f1, 0x074caf75, 0xddbbee99, 0x60fda37f);
-       &data_word(0x269ff701, 0xf5bc5c72, 0x3bc54466, 0x7e345bfb);
-       &data_word(0x29768b43, 0xc6dccb23, 0xfc68b6ed, 0xf163b8e4);
-       &data_word(0xdccad731, 0x85104263, 0x22401397, 0x112084c6);
-       &data_word(0x247d854a, 0x3df8d2bb, 0x3211aef9, 0xa16dc729);
-       &data_word(0x2f4b1d9e, 0x30f3dcb2, 0x52ec0d86, 0xe3d077c1);
-       &data_word(0x166c2bb3, 0xb999a970, 0x48fa1194, 0x642247e9);
-       &data_word(0x8cc4a8fc, 0x3f1aa0f0, 0x2cd8567d, 0x90ef2233);
-       &data_word(0x4ec78749, 0xd1c1d938, 0xa2fe8cca, 0x0b3698d4);
-       &data_word(0x81cfa6f5, 0xde28a57a, 0x8e26dab7, 0xbfa43fad);
-       &data_word(0x9de42c3a, 0x920d5078, 0xcc9b6a5f, 0x4662547e);
-       &data_word(0x13c2f68d, 0xb8e890d8, 0xf75e2e39, 0xaff582c3);
-       &data_word(0x80be9f5d, 0x937c69d0, 0x2da96fd5, 0x12b3cf25);
-       &data_word(0x993bc8ac, 0x7da71018, 0x636ee89c, 0xbb7bdb3b);
-       &data_word(0x7809cd26, 0x18f46e59, 0xb701ec9a, 0x9aa8834f);
-       &data_word(0x6e65e695, 0xe67eaaff, 0xcf0821bc, 0xe8e6ef15);
-       &data_word(0x9bd9bae7, 0x36ce4a6f, 0x09d4ea9f, 0x7cd629b0);
-       &data_word(0xb2af31a4, 0x23312a3f, 0x9430c6a5, 0x66c035a2);
-       &data_word(0xbc37744e, 0xcaa6fc82, 0xd0b0e090, 0xd81533a7);
-       &data_word(0x984af104, 0xdaf741ec, 0x500e7fcd, 0xf62f1791);
-       &data_word(0xd68d764d, 0xb04d43ef, 0x4d54ccaa, 0x04dfe496);
-       &data_word(0xb5e39ed1, 0x881b4c6a, 0x1fb8c12c, 0x517f4665);
-       &data_word(0xea049d5e, 0x355d018c, 0x7473fa87, 0x412efb0b);
-       &data_word(0x1d5ab367, 0xd25292db, 0x5633e910, 0x47136dd6);
-       &data_word(0x618c9ad7, 0x0c7a37a1, 0x148e59f8, 0x3c89eb13);
-       &data_word(0x27eecea9, 0xc935b761, 0xe5ede11c, 0xb13c7a47);
-       &data_word(0xdf599cd2, 0x733f55f2, 0xce791814, 0x37bf73c7);
-       &data_word(0xcdea53f7, 0xaa5b5ffd, 0x6f14df3d, 0xdb867844);
-       &data_word(0xf381caaf, 0xc43eb968, 0x342c3824, 0x405fc2a3);
-       &data_word(0xc372161d, 0x250cbce2, 0x498b283c, 0x9541ff0d);
-       &data_word(0x017139a8, 0xb3de080c, 0xe49cd8b4, 0xc1906456);
-       &data_word(0x84617bcb, 0xb670d532, 0x5c74486c, 0x5742d0b8);
-#Td3:
-       &data_word(0xf4a75051, 0x4165537e, 0x17a4c31a, 0x275e963a);
-       &data_word(0xab6bcb3b, 0x9d45f11f, 0xfa58abac, 0xe303934b);
-       &data_word(0x30fa5520, 0x766df6ad, 0xcc769188, 0x024c25f5);
-       &data_word(0xe5d7fc4f, 0x2acbd7c5, 0x35448026, 0x62a38fb5);
-       &data_word(0xb15a49de, 0xba1b6725, 0xea0e9845, 0xfec0e15d);
-       &data_word(0x2f7502c3, 0x4cf01281, 0x4697a38d, 0xd3f9c66b);
-       &data_word(0x8f5fe703, 0x929c9515, 0x6d7aebbf, 0x5259da95);
-       &data_word(0xbe832dd4, 0x7421d358, 0xe0692949, 0xc9c8448e);
-       &data_word(0xc2896a75, 0x8e7978f4, 0x583e6b99, 0xb971dd27);
-       &data_word(0xe14fb6be, 0x88ad17f0, 0x20ac66c9, 0xce3ab47d);
-       &data_word(0xdf4a1863, 0x1a3182e5, 0x51336097, 0x537f4562);
-       &data_word(0x6477e0b1, 0x6bae84bb, 0x81a01cfe, 0x082b94f9);
-       &data_word(0x48685870, 0x45fd198f, 0xde6c8794, 0x7bf8b752);
-       &data_word(0x73d323ab, 0x4b02e272, 0x1f8f57e3, 0x55ab2a66);
-       &data_word(0xeb2807b2, 0xb5c2032f, 0xc57b9a86, 0x3708a5d3);
-       &data_word(0x2887f230, 0xbfa5b223, 0x036aba02, 0x16825ced);
-       &data_word(0xcf1c2b8a, 0x79b492a7, 0x07f2f0f3, 0x69e2a14e);
-       &data_word(0xdaf4cd65, 0x05bed506, 0x34621fd1, 0xa6fe8ac4);
-       &data_word(0x2e539d34, 0xf355a0a2, 0x8ae13205, 0xf6eb75a4);
-       &data_word(0x83ec390b, 0x60efaa40, 0x719f065e, 0x6e1051bd);
-       &data_word(0x218af93e, 0xdd063d96, 0x3e05aedd, 0xe6bd464d);
-       &data_word(0x548db591, 0xc45d0571, 0x06d46f04, 0x5015ff60);
-       &data_word(0x98fb2419, 0xbde997d6, 0x4043cc89, 0xd99e7767);
-       &data_word(0xe842bdb0, 0x898b8807, 0x195b38e7, 0xc8eedb79);
-       &data_word(0x7c0a47a1, 0x420fe97c, 0x841ec9f8, 0x00000000);
-       &data_word(0x80868309, 0x2bed4832, 0x1170ac1e, 0x5a724e6c);
-       &data_word(0x0efffbfd, 0x8538560f, 0xaed51e3d, 0x2d392736);
-       &data_word(0x0fd9640a, 0x5ca62168, 0x5b54d19b, 0x362e3a24);
-       &data_word(0x0a67b10c, 0x57e70f93, 0xee96d2b4, 0x9b919e1b);
-       &data_word(0xc0c54f80, 0xdc20a261, 0x774b695a, 0x121a161c);
-       &data_word(0x93ba0ae2, 0xa02ae5c0, 0x22e0433c, 0x1b171d12);
-       &data_word(0x090d0b0e, 0x8bc7adf2, 0xb6a8b92d, 0x1ea9c814);
-       &data_word(0xf1198557, 0x75074caf, 0x99ddbbee, 0x7f60fda3);
-       &data_word(0x01269ff7, 0x72f5bc5c, 0x663bc544, 0xfb7e345b);
-       &data_word(0x4329768b, 0x23c6dccb, 0xedfc68b6, 0xe4f163b8);
-       &data_word(0x31dccad7, 0x63851042, 0x97224013, 0xc6112084);
-       &data_word(0x4a247d85, 0xbb3df8d2, 0xf93211ae, 0x29a16dc7);
-       &data_word(0x9e2f4b1d, 0xb230f3dc, 0x8652ec0d, 0xc1e3d077);
-       &data_word(0xb3166c2b, 0x70b999a9, 0x9448fa11, 0xe9642247);
-       &data_word(0xfc8cc4a8, 0xf03f1aa0, 0x7d2cd856, 0x3390ef22);
-       &data_word(0x494ec787, 0x38d1c1d9, 0xcaa2fe8c, 0xd40b3698);
-       &data_word(0xf581cfa6, 0x7ade28a5, 0xb78e26da, 0xadbfa43f);
-       &data_word(0x3a9de42c, 0x78920d50, 0x5fcc9b6a, 0x7e466254);
-       &data_word(0x8d13c2f6, 0xd8b8e890, 0x39f75e2e, 0xc3aff582);
-       &data_word(0x5d80be9f, 0xd0937c69, 0xd52da96f, 0x2512b3cf);
-       &data_word(0xac993bc8, 0x187da710, 0x9c636ee8, 0x3bbb7bdb);
-       &data_word(0x267809cd, 0x5918f46e, 0x9ab701ec, 0x4f9aa883);
-       &data_word(0x956e65e6, 0xffe67eaa, 0xbccf0821, 0x15e8e6ef);
-       &data_word(0xe79bd9ba, 0x6f36ce4a, 0x9f09d4ea, 0xb07cd629);
-       &data_word(0xa4b2af31, 0x3f23312a, 0xa59430c6, 0xa266c035);
-       &data_word(0x4ebc3774, 0x82caa6fc, 0x90d0b0e0, 0xa7d81533);
-       &data_word(0x04984af1, 0xecdaf741, 0xcd500e7f, 0x91f62f17);
-       &data_word(0x4dd68d76, 0xefb04d43, 0xaa4d54cc, 0x9604dfe4);
-       &data_word(0xd1b5e39e, 0x6a881b4c, 0x2c1fb8c1, 0x65517f46);
-       &data_word(0x5eea049d, 0x8c355d01, 0x877473fa, 0x0b412efb);
-       &data_word(0x671d5ab3, 0xdbd25292, 0x105633e9, 0xd647136d);
-       &data_word(0xd7618c9a, 0xa10c7a37, 0xf8148e59, 0x133c89eb);
-       &data_word(0xa927eece, 0x61c935b7, 0x1ce5ede1, 0x47b13c7a);
-       &data_word(0xd2df599c, 0xf2733f55, 0x14ce7918, 0xc737bf73);
-       &data_word(0xf7cdea53, 0xfdaa5b5f, 0x3d6f14df, 0x44db8678);
-       &data_word(0xaff381ca, 0x68c43eb9, 0x24342c38, 0xa3405fc2);
-       &data_word(0x1dc37216, 0xe2250cbc, 0x3c498b28, 0x0d9541ff);
-       &data_word(0xa8017139, 0x0cb3de08, 0xb4e49cd8, 0x56c19064);
-       &data_word(0xcb84617b, 0x32b670d5, 0x6c5c7448, 0xb85742d0);
+       &_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);
@@ -1090,6 +756,796 @@ sub declast()
        &data_word(0xbabababa, 0x77777777, 0xd6d6d6d6, 0x26262626);
        &data_word(0xe1e1e1e1, 0x69696969, 0x14141414, 0x63636363);
        &data_word(0x55555555, 0x21212121, 0x0c0c0c0c, 0x7d7d7d7d);
-&function_end_B("AES_decrypt");
+&function_end_B("_x86_AES_decrypt");
+
+# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
+&public_label("AES_Td");
+&function_begin("AES_decrypt");
+       &mov    ($acc,&wparam(0));              # load inp
+       &mov    ($key,&wparam(2));              # load key
+
+       &mov    ($s0,"esp");
+       &sub    ("esp",24);
+       &and    ("esp",-64);
+       &add    ("esp",4);
+       &mov    (&DWP(16,"esp"),$s0);
+
+       &call   (&label("pic_point"));          # make it PIC!
+       &set_label("pic_point");
+       &blindpop("ebp");
+       &lea    ("ebp",&DWP(&label("AES_Td")."-".&label("pic_point"),"ebp"));
+
+       &mov    ($s0,&DWP(0,$acc));             # load input data
+       &mov    ($s1,&DWP(4,$acc));
+       &mov    ($s2,&DWP(8,$acc));
+       &mov    ($s3,&DWP(12,$acc));
+
+       &call   ("_x86_AES_decrypt");
+
+       &mov    ("esp",&DWP(16,"esp"));
+
+       &mov    ($acc,&wparam(1));              # load out
+       &mov    (&DWP(0,$acc),$s0);             # write output data
+       &mov    (&DWP(4,$acc),$s1);
+       &mov    (&DWP(8,$acc),$s2);
+       &mov    (&DWP(12,$acc),$s3);
+&function_end("AES_decrypt");
+
+# 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
+# -4(%esp)     0(%esp)         return address
+# 0(%esp)      4(%esp)         tmp1
+# 4(%esp)      8(%esp)         tmp2
+# 8(%esp)      12(%esp)        key
+# 12(%esp)     16(%esp)        end of key schedule
+my $_esp=&DWP(16,"esp");       #saved %esp
+my $_inp=&DWP(20,"esp");       #copy of wparam(0)
+my $_out=&DWP(24,"esp");       #copy of wparam(1)
+my $_len=&DWP(28,"esp");       #copy of wparam(2)
+my $_key=&DWP(32,"esp");       #copy of wparam(3)
+my $_ivp=&DWP(36,"esp");       #copy of wparam(4)
+my $_tmp=&DWP(40,"esp");       #volatile variable
+my $ivec=&DWP(44,"esp");       #ivec[16]
+my $aes_key=&DWP(60,"esp");    #copy of aes_key
+my $mark=&DWP(60+240,"esp");   #copy of aes_key->rounds
+
+&public_label("AES_Te");
+&public_label("AES_Td");
+&function_begin("AES_cbc_encrypt");
+       &mov    ($s2 eq "ecx"? $s2 : "",&wparam(2));    # load len
+       &cmp    ($s2,0);
+       &je     (&label("enc_out"));
+
+       &call   (&label("pic_point"));          # make it PIC!
+       &set_label("pic_point");
+       &blindpop("ebp");
+
+       &pushf  ();
+       &cld    ();
+
+       &cmp    (&wparam(5),0);
+       &je     (&label("DECRYPT"));
+
+       &lea    ("ebp",&DWP(&label("AES_Te")."-".&label("pic_point"),"ebp"));
+
+       # allocate aligned stack frame...
+       &lea    ($key,&DWP(-64-244,"esp"));
+       &and    ($key,-64);
+
+       # ... and make sure it doesn't alias with AES_Te modulo 4096
+       &mov    ($s0,"ebp");
+       &lea    ($s1,&DWP(2048,"ebp"));
+       &mov    ($s3,$key);
+       &and    ($s0,0xfff);            # s = %ebp&0xfff
+       &and    ($s1,0xfff);            # e = (%ebp+2048)&0xfff
+       &and    ($s3,0xfff);            # p = %esp&0xfff
+
+       &cmp    ($s3,$s1);              # if (p>=e) %esp =- (p-e);
+       &jb     (&label("te_break_out"));
+       &sub    ($s3,$s1);
+       &sub    ($key,$s3);
+       &jmp    (&label("te_ok"));
+       &set_label("te_break_out");     # else %esp -= (p-s)&0xfff + framesz;
+       &sub    ($s3,$s0);
+       &and    ($s3,0xfff);
+       &add    ($s3,64+256);
+       &sub    ($key,$s3);
+       &align  (4);
+       &set_label("te_ok");
+
+       &mov    ($s0,&wparam(0));       # load inp
+       &mov    ($s1,&wparam(1));       # load out
+       &mov    ($s3,&wparam(3));       # load key
+       &mov    ($acc,&wparam(4));      # load ivp
+
+       &exch   ("esp",$key);
+       &add    ("esp",4);              # reserve for return address!
+       &mov    ($_esp,$key);           # save %esp
+
+       &mov    ($_inp,$s0);            # save copy of inp
+       &mov    ($_out,$s1);            # save copy of out
+       &mov    ($_len,$s2);            # save copy of len
+       &mov    ($_key,$s3);            # save copy of key
+       &mov    ($_ivp,$acc);           # save copy of ivp
+
+       &mov    ($mark,0);              # copy of aes_key->rounds = 0;
+       if ($compromise) {
+               &cmp    ($s2,$compromise);
+               &jb     (&label("skip_ecopy"));
+       }
+       # do we copy key schedule to stack?
+       &mov    ($s1 eq "ebx" ? $s1 : "",$s3);
+       &mov    ($s2 eq "ecx" ? $s2 : "",244/4);
+       &sub    ($s1,"ebp");
+       &mov    ("esi",$s3);
+       &and    ($s1,0xfff);
+       &lea    ("edi",$aes_key);
+       &cmp    ($s1,2048);
+       &jb     (&label("do_ecopy"));
+       &cmp    ($s1,4096-244);
+       &jb     (&label("skip_ecopy"));
+       &align  (4);
+       &set_label("do_ecopy");
+               &mov    ($_key,"edi");
+               &data_word(0xA5F3F689); # rep movsd
+       &set_label("skip_ecopy");
+
+       &mov    ($acc,$s0);
+       &mov    ($key,16);
+       &align  (4);
+       &set_label("prefetch_te");
+               &mov    ($s0,&DWP(0,"ebp"));
+               &mov    ($s1,&DWP(32,"ebp"));
+               &mov    ($s2,&DWP(64,"ebp"));
+               &mov    ($s3,&DWP(96,"ebp"));
+               &lea    ("ebp",&DWP(128,"ebp"));
+               &dec    ($key);
+       &jnz    (&label("prefetch_te"));
+       &sub    ("ebp",2048);
+
+       &mov    ($s2,$_len);
+       &mov    ($key,$_ivp);
+       &test   ($s2,0xFFFFFFF0);
+       &jz     (&label("enc_tail"));           # short input...
+
+       &mov    ($s0,&DWP(0,$key));             # load iv
+       &mov    ($s1,&DWP(4,$key));
+
+       &align  (4);
+       &set_label("enc_loop");
+               &mov    ($s2,&DWP(8,$key));
+               &mov    ($s3,&DWP(12,$key));
+
+               &xor    ($s0,&DWP(0,$acc));     # xor input data
+               &xor    ($s1,&DWP(4,$acc));
+               &xor    ($s2,&DWP(8,$acc));
+               &xor    ($s3,&DWP(12,$acc));
+
+               &mov    ($key,$_key);           # load key
+               &call   ("_x86_AES_encrypt");
+
+               &mov    ($acc,$_inp);           # load inp
+               &mov    ($key,$_out);           # load out
+
+               &mov    (&DWP(0,$key),$s0);     # save output data
+               &mov    (&DWP(4,$key),$s1);
+               &mov    (&DWP(8,$key),$s2);
+               &mov    (&DWP(12,$key),$s3);
+
+               &mov    ($s2,$_len);            # load len
+
+               &lea    ($acc,&DWP(16,$acc));
+               &mov    ($_inp,$acc);           # save inp
+
+               &lea    ($s3,&DWP(16,$key));
+               &mov    ($_out,$s3);            # save out
+
+               &sub    ($s2,16);
+               &test   ($s2,0xFFFFFFF0);
+               &mov    ($_len,$s2);            # save len
+       &jnz    (&label("enc_loop"));
+       &test   ($s2,15);
+       &jnz    (&label("enc_tail"));
+       &mov    ($acc,$_ivp);           # load ivp
+       &mov    ($s2,&DWP(8,$key));     # restore last dwords
+       &mov    ($s3,&DWP(12,$key));
+       &mov    (&DWP(0,$acc),$s0);     # save ivec
+       &mov    (&DWP(4,$acc),$s1);
+       &mov    (&DWP(8,$acc),$s2);
+       &mov    (&DWP(12,$acc),$s3);
+
+       &cmp    ($mark,0);              # was the key schedule copied?
+       &mov    ("edi",$_key);
+       &mov    ("esp",$_esp);
+       &je     (&label("skip_ezero"));
+       # zero copy of key schedule
+       &mov    ("ecx",240/4);
+       &xor    ("eax","eax");
+       &align  (4);
+       &data_word(0xABF3F689); # rep stosd
+       &set_label("skip_ezero")
+       &popf   ();
+    &set_label("enc_out");
+       &function_end_A();
+       &pushf  ();                     # kludge, never executed
+
+    &align     (4);
+    &set_label("enc_tail");
+       &push   ($key eq "edi" ? $key : "");    # push ivp
+       &mov    ($key,$_out);                   # load out
+       &mov    ($s1,16);
+       &sub    ($s1,$s2);
+       &cmp    ($key,$acc);                    # compare with inp
+       &je     (&label("enc_in_place"));
+       &align  (4);
+       &data_word(0xA4F3F689); # rep movsb     # copy input
+       &jmp    (&label("enc_skip_in_place"));
+    &set_label("enc_in_place");
+       &lea    ($key,&DWP(0,$key,$s2));
+    &set_label("enc_skip_in_place");
+       &mov    ($s2,$s1);
+       &xor    ($s0,$s0);
+       &align  (4);
+       &data_word(0xAAF3F689); # rep stosb     # zero tail
+       &pop    ($key);                         # pop ivp
+
+       &mov    ($acc,$_out);                   # output as input
+       &mov    ($s0,&DWP(0,$key));
+       &mov    ($s1,&DWP(4,$key));
+       &mov    ($_len,16);                     # len=16
+       &jmp    (&label("enc_loop"));           # one more spin...
+
+#----------------------------- DECRYPT -----------------------------#
+&align (4);
+&set_label("DECRYPT");
+       &lea    ("ebp",&DWP(&label("AES_Td")."-".&label("pic_point"),"ebp"));
+
+       # allocate aligned stack frame...
+       &lea    ($key,&DWP(-64-244,"esp"));
+       &and    ($key,-64);
+
+       # ... and make sure it doesn't alias with AES_Td modulo 4096
+       &mov    ($s0,"ebp");
+       &lea    ($s1,&DWP(3072,"ebp"));
+       &mov    ($s3,$key);
+       &and    ($s0,0xfff);            # s = %ebp&0xfff
+       &and    ($s1,0xfff);            # e = (%ebp+3072)&0xfff
+       &and    ($s3,0xfff);            # p = %esp&0xfff
+
+       &cmp    ($s3,$s1);              # if (p>=e) %esp =- (p-e);
+       &jb     (&label("td_break_out"));
+       &sub    ($s3,$s1);
+       &sub    ($key,$s3);
+       &jmp    (&label("td_ok"));
+       &set_label("td_break_out");     # else %esp -= (p-s)&0xfff + framesz;
+       &sub    ($s3,$s0);
+       &and    ($s3,0xfff);
+       &add    ($s3,64+256);
+       &sub    ($key,$s3);
+       &align  (4);
+       &set_label("td_ok");
+
+       &mov    ($s0,&wparam(0));       # load inp
+       &mov    ($s1,&wparam(1));       # load out
+       &mov    ($s3,&wparam(3));       # load key
+       &mov    ($acc,&wparam(4));      # load ivp
+
+       &exch   ("esp",$key);
+       &add    ("esp",4);              # reserve for return address!
+       &mov    ($_esp,$key);           # save %esp
+
+       &mov    ($_inp,$s0);            # save copy of inp
+       &mov    ($_out,$s1);            # save copy of out
+       &mov    ($_len,$s2);            # save copy of len
+       &mov    ($_key,$s3);            # save copy of key
+       &mov    ($_ivp,$acc);           # save copy of ivp
+
+       &mov    ($mark,0);              # copy of aes_key->rounds = 0;
+       if ($compromise) {
+               &cmp    ($s2,$compromise);
+               &jb     (&label("skip_dcopy"));
+       }
+       # do we copy key schedule to stack?
+       &mov    ($s1 eq "ebx" ? $s1 : "",$s3);
+       &mov    ($s2 eq "ecx" ? $s2 : "",244/4);
+       &sub    ($s1,"ebp");
+       &mov    ("esi",$s3);
+       &and    ($s1,0xfff);
+       &lea    ("edi",$aes_key);
+       &cmp    ($s1,3072);
+       &jb     (&label("do_dcopy"));
+       &cmp    ($s1,4096-244);
+       &jb     (&label("skip_dcopy"));
+       &align  (4);
+       &set_label("do_dcopy");
+               &mov    ($_key,"edi");
+               &data_word(0xA5F3F689); # rep movsd
+       &set_label("skip_dcopy");
+
+       &mov    ($acc,$s0);
+       &mov    ($key,24);
+       &align  (4);
+       &set_label("prefetch_td");
+               &mov    ($s0,&DWP(0,"ebp"));
+               &mov    ($s1,&DWP(32,"ebp"));
+               &mov    ($s2,&DWP(64,"ebp"));
+               &mov    ($s3,&DWP(96,"ebp"));
+               &lea    ("ebp",&DWP(128,"ebp"));
+               &dec    ($key);
+       &jnz    (&label("prefetch_td"));
+       &sub    ("ebp",3072);
+
+       &cmp    ($acc,$_out);
+       &je     (&label("dec_in_place"));       # in-place processing...
+
+       &mov    ($key,$_ivp);           # load ivp
+       &mov    ($_tmp,$key);
+
+       &align  (4);
+       &set_label("dec_loop");
+               &mov    ($s0,&DWP(0,$acc));     # read input
+               &mov    ($s1,&DWP(4,$acc));
+               &mov    ($s2,&DWP(8,$acc));
+               &mov    ($s3,&DWP(12,$acc));
+
+               &mov    ($key,$_key);           # load key
+               &call   ("_x86_AES_decrypt");
+
+               &mov    ($key,$_tmp);           # load ivp
+               &mov    ($acc,$_len);           # load len
+               &xor    ($s0,&DWP(0,$key));     # xor iv
+               &xor    ($s1,&DWP(4,$key));
+               &xor    ($s2,&DWP(8,$key));
+               &xor    ($s3,&DWP(12,$key));
+
+               &sub    ($acc,16);
+               &jc     (&label("dec_partial"));
+               &mov    ($_len,$acc);           # save len
+               &mov    ($acc,$_inp);           # load inp
+               &mov    ($key,$_out);           # load out
+
+               &mov    (&DWP(0,$key),$s0);     # write output
+               &mov    (&DWP(4,$key),$s1);
+               &mov    (&DWP(8,$key),$s2);
+               &mov    (&DWP(12,$key),$s3);
+
+               &mov    ($_tmp,$acc);           # save ivp
+               &lea    ($acc,&DWP(16,$acc));
+               &mov    ($_inp,$acc);           # save inp
+
+               &lea    ($key,&DWP(16,$key));
+               &mov    ($_out,$key);           # save out
+
+       &jnz    (&label("dec_loop"));
+       &mov    ($key,$_tmp);           # load temp ivp
+    &set_label("dec_end");
+       &mov    ($acc,$_ivp);           # load user ivp
+       &mov    ($s0,&DWP(0,$key));     # load iv
+       &mov    ($s1,&DWP(4,$key));
+       &mov    ($s2,&DWP(8,$key));
+       &mov    ($s3,&DWP(12,$key));
+       &mov    (&DWP(0,$acc),$s0);     # copy back to user
+       &mov    (&DWP(4,$acc),$s1);
+       &mov    (&DWP(8,$acc),$s2);
+       &mov    (&DWP(12,$acc),$s3);
+       &jmp    (&label("dec_out"));
+
+    &align     (4);
+    &set_label("dec_partial");
+       &lea    ($key,$ivec);
+       &mov    (&DWP(0,$key),$s0);     # dump output to stack
+       &mov    (&DWP(4,$key),$s1);
+       &mov    (&DWP(8,$key),$s2);
+       &mov    (&DWP(12,$key),$s3);
+       &lea    ($s2 eq "ecx" ? $s2 : "",&DWP(16,$acc));
+       &mov    ($acc eq "esi" ? $acc : "",$key);
+       &mov    ($key eq "edi" ? $key : "",$_out);      # load out
+       &data_word(0xA4F3F689); # rep movsb             # copy output
+       &mov    ($key,$_inp);                           # use inp as temp ivp
+       &jmp    (&label("dec_end"));
+
+    &align     (4);
+    &set_label("dec_in_place");
+       &set_label("dec_in_place_loop");
+               &lea    ($key,$ivec);
+               &mov    ($s0,&DWP(0,$acc));     # read input
+               &mov    ($s1,&DWP(4,$acc));
+               &mov    ($s2,&DWP(8,$acc));
+               &mov    ($s3,&DWP(12,$acc));
+
+               &mov    (&DWP(0,$key),$s0);     # copy to temp
+               &mov    (&DWP(4,$key),$s1);
+               &mov    (&DWP(8,$key),$s2);
+               &mov    (&DWP(12,$key),$s3);
+
+               &mov    ($key,$_key);           # load key
+               &call   ("_x86_AES_decrypt");
+
+               &mov    ($key,$_ivp);           # load ivp
+               &mov    ($acc,$_out);           # load out
+               &xor    ($s0,&DWP(0,$key));     # xor iv
+               &xor    ($s1,&DWP(4,$key));
+               &xor    ($s2,&DWP(8,$key));
+               &xor    ($s3,&DWP(12,$key));
+
+               &mov    (&DWP(0,$acc),$s0);     # write output
+               &mov    (&DWP(4,$acc),$s1);
+               &mov    (&DWP(8,$acc),$s2);
+               &mov    (&DWP(12,$acc),$s3);
+
+               &lea    ($acc,&DWP(16,$acc));
+               &mov    ($_out,$acc);           # save out
+
+               &lea    ($acc,$ivec);
+               &mov    ($s0,&DWP(0,$acc));     # read temp
+               &mov    ($s1,&DWP(4,$acc));
+               &mov    ($s2,&DWP(8,$acc));
+               &mov    ($s3,&DWP(12,$acc));
+
+               &mov    (&DWP(0,$key),$s0);     # copy iv
+               &mov    (&DWP(4,$key),$s1);
+               &mov    (&DWP(8,$key),$s2);
+               &mov    (&DWP(12,$key),$s3);
+
+               &mov    ($acc,$_inp);           # load inp
+
+               &lea    ($acc,&DWP(16,$acc));
+               &mov    ($_inp,$acc);           # save inp
+
+               &mov    ($s2,$_len);            # load len
+               &sub    ($s2,16);
+               &jc     (&label("dec_in_place_partial"));
+               &mov    ($_len,$s2);            # save len
+       &jnz    (&label("dec_in_place_loop"));
+       &jmp    (&label("dec_out"));
+
+    &align     (4);
+    &set_label("dec_in_place_partial");
+       # one can argue if this is actually required...
+       &mov    ($key eq "edi" ? $key : "",$_out);
+       &lea    ($acc eq "esi" ? $acc : "",$ivec);
+       &lea    ($key,&DWP(0,$key,$s2));
+       &lea    ($acc,&DWP(16,$acc,$s2));
+       &neg    ($s2 eq "ecx" ? $s2 : "");
+       &data_word(0xA4F3F689); # rep movsb     # restore tail
+
+    &align     (4);
+    &set_label("dec_out");
+    &cmp       ($mark,0);              # was the key schedule copied?
+    &mov       ("edi",$_key);
+    &mov       ("esp",$_esp);
+    &je                (&label("skip_dzero"));
+    # zero copy of key schedule
+    &mov       ("ecx",240/4);
+    &xor       ("eax","eax");
+    &align     (4);
+    &data_word(0xABF3F689);    # rep stosd
+    &set_label("skip_dzero")
+    &popf      ();
+&function_end("AES_cbc_encrypt");
+}
+
+#------------------------------------------------------------------#
+
+sub enckey()
+{
+       &movz   ("esi",&LB("edx"));             # rk[i]>>0
+       &mov    ("ebx",&DWP(2,"ebp","esi",8));
+       &movz   ("esi",&HB("edx"));             # rk[i]>>8
+       &and    ("ebx",0xFF000000);
+       &xor    ("eax","ebx");
+
+       &mov    ("ebx",&DWP(2,"ebp","esi",8));
+       &shr    ("edx",16);
+       &and    ("ebx",0x000000FF);
+       &movz   ("esi",&LB("edx"));             # rk[i]>>16
+       &xor    ("eax","ebx");
+
+       &mov    ("ebx",&DWP(0,"ebp","esi",8));
+       &movz   ("esi",&HB("edx"));             # rk[i]>>24
+       &and    ("ebx",0x0000FF00);
+       &xor    ("eax","ebx");
+
+       &mov    ("ebx",&DWP(0,"ebp","esi",8));
+       &and    ("ebx",0x00FF0000);
+       &xor    ("eax","ebx");
+
+       &xor    ("eax",&DWP(2048,"ebp","ecx",4));       # rcon
+}
+
+# int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+&public_label("AES_Te");
+&function_begin("AES_set_encrypt_key");
+       &mov    ("esi",&wparam(0));             # user supplied key
+       &mov    ("edi",&wparam(2));             # private key schedule
+
+       &test   ("esi",-1);
+       &jz     (&label("badpointer"));
+       &test   ("edi",-1);
+       &jz     (&label("badpointer"));
+
+       &call   (&label("pic_point"));
+       &set_label("pic_point");
+       &blindpop("ebp");
+       &lea    ("ebp",&DWP(&label("AES_Te")."-".&label("pic_point"),"ebp"));
+
+       &mov    ("ecx",&wparam(1));             # number of bits in key
+       &cmp    ("ecx",128);
+       &je     (&label("10rounds"));
+       &cmp    ("ecx",192);
+       &je     (&label("12rounds"));
+       &cmp    ("ecx",256);
+       &je     (&label("14rounds"));
+       &mov    ("eax",-2);                     # invalid number of bits
+       &jmp    (&label("exit"));
+
+    &set_label("10rounds");
+       &mov    ("eax",&DWP(0,"esi"));          # copy first 4 dwords
+       &mov    ("ebx",&DWP(4,"esi"));
+       &mov    ("ecx",&DWP(8,"esi"));
+       &mov    ("edx",&DWP(12,"esi"));
+       &mov    (&DWP(0,"edi"),"eax");
+       &mov    (&DWP(4,"edi"),"ebx");
+       &mov    (&DWP(8,"edi"),"ecx");
+       &mov    (&DWP(12,"edi"),"edx");
+
+       &xor    ("ecx","ecx");
+       &jmp    (&label("10shortcut"));
+
+       &align  (4);
+       &set_label("10loop");
+               &mov    ("eax",&DWP(0,"edi"));          # rk[0]
+               &mov    ("edx",&DWP(12,"edi"));         # rk[3]
+       &set_label("10shortcut");
+               &enckey ();
+
+               &mov    (&DWP(16,"edi"),"eax");         # rk[4]
+               &xor    ("eax",&DWP(4,"edi"));
+               &mov    (&DWP(20,"edi"),"eax");         # rk[5]
+               &xor    ("eax",&DWP(8,"edi"));
+               &mov    (&DWP(24,"edi"),"eax");         # rk[6]
+               &xor    ("eax",&DWP(12,"edi"));
+               &mov    (&DWP(28,"edi"),"eax");         # rk[7]
+               &inc    ("ecx");
+               &add    ("edi",16);
+               &cmp    ("ecx",10);
+       &jl     (&label("10loop"));
+
+       &mov    (&DWP(80,"edi"),10);            # setup number of rounds
+       &xor    ("eax","eax");
+       &jmp    (&label("exit"));
+               
+    &set_label("12rounds");
+       &mov    ("eax",&DWP(0,"esi"));          # copy first 6 dwords
+       &mov    ("ebx",&DWP(4,"esi"));
+       &mov    ("ecx",&DWP(8,"esi"));
+       &mov    ("edx",&DWP(12,"esi"));
+       &mov    (&DWP(0,"edi"),"eax");
+       &mov    (&DWP(4,"edi"),"ebx");
+       &mov    (&DWP(8,"edi"),"ecx");
+       &mov    (&DWP(12,"edi"),"edx");
+       &mov    ("ecx",&DWP(16,"esi"));
+       &mov    ("edx",&DWP(20,"esi"));
+       &mov    (&DWP(16,"edi"),"ecx");
+       &mov    (&DWP(20,"edi"),"edx");
+
+       &xor    ("ecx","ecx");
+       &jmp    (&label("12shortcut"));
+
+       &align  (4);
+       &set_label("12loop");
+               &mov    ("eax",&DWP(0,"edi"));          # rk[0]
+               &mov    ("edx",&DWP(20,"edi"));         # rk[5]
+       &set_label("12shortcut");
+               &enckey ();
+
+               &mov    (&DWP(24,"edi"),"eax");         # rk[6]
+               &xor    ("eax",&DWP(4,"edi"));
+               &mov    (&DWP(28,"edi"),"eax");         # rk[7]
+               &xor    ("eax",&DWP(8,"edi"));
+               &mov    (&DWP(32,"edi"),"eax");         # rk[8]
+               &xor    ("eax",&DWP(12,"edi"));
+               &mov    (&DWP(36,"edi"),"eax");         # rk[9]
+
+               &cmp    ("ecx",7);
+               &je     (&label("12break"));
+               &inc    ("ecx");
+
+               &xor    ("eax",&DWP(16,"edi"));
+               &mov    (&DWP(40,"edi"),"eax");         # rk[10]
+               &xor    ("eax",&DWP(20,"edi"));
+               &mov    (&DWP(44,"edi"),"eax");         # rk[11]
+
+               &add    ("edi",24);
+       &jmp    (&label("12loop"));
+
+       &set_label("12break");
+       &mov    (&DWP(72,"edi"),12);            # setup number of rounds
+       &xor    ("eax","eax");
+       &jmp    (&label("exit"));
+
+    &set_label("14rounds");
+       &mov    ("eax",&DWP(0,"esi"));          # copy first 8 dwords
+       &mov    ("ebx",&DWP(4,"esi"));
+       &mov    ("ecx",&DWP(8,"esi"));
+       &mov    ("edx",&DWP(12,"esi"));
+       &mov    (&DWP(0,"edi"),"eax");
+       &mov    (&DWP(4,"edi"),"ebx");
+       &mov    (&DWP(8,"edi"),"ecx");
+       &mov    (&DWP(12,"edi"),"edx");
+       &mov    ("eax",&DWP(16,"esi"));
+       &mov    ("ebx",&DWP(20,"esi"));
+       &mov    ("ecx",&DWP(24,"esi"));
+       &mov    ("edx",&DWP(28,"esi"));
+       &mov    (&DWP(16,"edi"),"eax");
+       &mov    (&DWP(20,"edi"),"ebx");
+       &mov    (&DWP(24,"edi"),"ecx");
+       &mov    (&DWP(28,"edi"),"edx");
+
+       &xor    ("ecx","ecx");
+       &jmp    (&label("14shortcut"));
+
+       &align  (4);
+       &set_label("14loop");
+               &mov    ("edx",&DWP(28,"edi"));         # rk[7]
+       &set_label("14shortcut");
+               &mov    ("eax",&DWP(0,"edi"));          # rk[0]
+
+               &enckey ();
+
+               &mov    (&DWP(32,"edi"),"eax");         # rk[8]
+               &xor    ("eax",&DWP(4,"edi"));
+               &mov    (&DWP(36,"edi"),"eax");         # rk[9]
+               &xor    ("eax",&DWP(8,"edi"));
+               &mov    (&DWP(40,"edi"),"eax");         # rk[10]
+               &xor    ("eax",&DWP(12,"edi"));
+               &mov    (&DWP(44,"edi"),"eax");         # rk[11]
+
+               &cmp    ("ecx",6);
+               &je     (&label("14break"));
+               &inc    ("ecx");
+
+               &mov    ("edx","eax");
+               &mov    ("eax",&DWP(16,"edi"));         # rk[4]
+               &movz   ("esi",&LB("edx"));             # rk[11]>>0
+               &mov    ("ebx",&DWP(2,"ebp","esi",8));
+               &movz   ("esi",&HB("edx"));             # rk[11]>>8
+               &and    ("ebx",0x000000FF);
+               &xor    ("eax","ebx");
+
+               &mov    ("ebx",&DWP(0,"ebp","esi",8));
+               &shr    ("edx",16);
+               &and    ("ebx",0x0000FF00);
+               &movz   ("esi",&LB("edx"));             # rk[11]>>16
+               &xor    ("eax","ebx");
+
+               &mov    ("ebx",&DWP(0,"ebp","esi",8));
+               &movz   ("esi",&HB("edx"));             # rk[11]>>24
+               &and    ("ebx",0x00FF0000);
+               &xor    ("eax","ebx");
+
+               &mov    ("ebx",&DWP(2,"ebp","esi",8));
+               &and    ("ebx",0xFF000000);
+               &xor    ("eax","ebx");
+
+               &mov    (&DWP(48,"edi"),"eax");         # rk[12]
+               &xor    ("eax",&DWP(20,"edi"));
+               &mov    (&DWP(52,"edi"),"eax");         # rk[13]
+               &xor    ("eax",&DWP(24,"edi"));
+               &mov    (&DWP(56,"edi"),"eax");         # rk[14]
+               &xor    ("eax",&DWP(28,"edi"));
+               &mov    (&DWP(60,"edi"),"eax");         # rk[15]
+
+               &add    ("edi",32);
+       &jmp    (&label("14loop"));
+
+       &set_label("14break");
+       &mov    (&DWP(48,"edi"),14);            # setup number of rounds
+       &xor    ("eax","eax");
+       &jmp    (&label("exit"));
+
+    &set_label("badpointer");
+       &mov    ("eax",-1);
+    &set_label("exit");
+&function_end("AES_set_encrypt_key");
+
+sub deckey()
+{ my ($i,$ptr,$te,$td) = @_;
+
+       &mov    ("eax",&DWP($i,$ptr));
+       &mov    ("edx","eax");
+       &movz   ("ebx",&HB("eax"));
+       &shr    ("edx",16);
+       &and    ("eax",0xFF);
+       &movz   ("eax",&BP(2,$te,"eax",8));
+       &movz   ("ebx",&BP(2,$te,"ebx",8));
+       &mov    ("eax",&DWP(0,$td,"eax",8));
+       &xor    ("eax",&DWP(3,$td,"ebx",8));
+       &movz   ("ebx",&HB("edx"));
+       &and    ("edx",0xFF);
+       &movz   ("edx",&BP(2,$te,"edx",8));
+       &movz   ("ebx",&BP(2,$te,"ebx",8));
+       &xor    ("eax",&DWP(2,$td,"edx",8));
+       &xor    ("eax",&DWP(1,$td,"ebx",8));
+       &mov    (&DWP($i,$ptr),"eax");
+}
+
+# int AES_set_decrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+&public_label("AES_Td");
+&public_label("AES_Te");
+&function_begin_B("AES_set_decrypt_key");
+       &mov    ("eax",&wparam(0));
+       &mov    ("ecx",&wparam(1));
+       &mov    ("edx",&wparam(2));
+       &sub    ("esp",12);
+       &mov    (&DWP(0,"esp"),"eax");
+       &mov    (&DWP(4,"esp"),"ecx");
+       &mov    (&DWP(8,"esp"),"edx");
+       &call   ("AES_set_encrypt_key");
+       &add    ("esp",12);
+       &cmp    ("eax",0);
+       &je     (&label("proceed"));
+       &ret    ();
+
+    &set_label("proceed");
+       &push   ("ebp");
+       &push   ("ebx");
+       &push   ("esi");
+       &push   ("edi");
+
+       &mov    ("esi",&wparam(2));
+       &mov    ("ecx",&DWP(240,"esi"));        # pull number of rounds
+       &lea    ("ecx",&DWP(0,"","ecx",4));
+       &lea    ("edi",&DWP(0,"esi","ecx",4));  # pointer to last chunk
+
+       &align  (4);
+       &set_label("invert");                   # invert order of chunks
+               &mov    ("eax",&DWP(0,"esi"));
+               &mov    ("ebx",&DWP(4,"esi"));
+               &mov    ("ecx",&DWP(0,"edi"));
+               &mov    ("edx",&DWP(4,"edi"));
+               &mov    (&DWP(0,"edi"),"eax");
+               &mov    (&DWP(4,"edi"),"ebx");
+               &mov    (&DWP(0,"esi"),"ecx");
+               &mov    (&DWP(4,"esi"),"edx");
+               &mov    ("eax",&DWP(8,"esi"));
+               &mov    ("ebx",&DWP(12,"esi"));
+               &mov    ("ecx",&DWP(8,"edi"));
+               &mov    ("edx",&DWP(12,"edi"));
+               &mov    (&DWP(8,"edi"),"eax");
+               &mov    (&DWP(12,"edi"),"ebx");
+               &mov    (&DWP(8,"esi"),"ecx");
+               &mov    (&DWP(12,"esi"),"edx");
+               &add    ("esi",16);
+               &sub    ("edi",16);
+               &cmp    ("esi","edi");
+       &jne    (&label("invert"));
+
+       &call   (&label("pic_point"));
+       &set_label("pic_point");
+       blindpop("ebp");
+       &lea    ("edi",&DWP(&label("AES_Td")."-".&label("pic_point"),"ebp"));
+       &lea    ("ebp",&DWP(&label("AES_Te")."-".&label("pic_point"),"ebp"));
+
+       &mov    ("esi",&wparam(2));
+       &mov    ("ecx",&DWP(240,"esi"));        # pull number of rounds
+       &dec    ("ecx");
+       &align  (4);
+       &set_label("permute");                  # permute the key schedule
+               &add    ("esi",16);
+               &deckey (0,"esi","ebp","edi");
+               &deckey (4,"esi","ebp","edi");
+               &deckey (8,"esi","ebp","edi");
+               &deckey (12,"esi","ebp","edi");
+               &dec    ("ecx");
+       &jnz    (&label("permute"));
+
+       &xor    ("eax","eax");                  # return success
+&function_end("AES_set_decrypt_key");
 
 &asm_finish();