poly1305/asm/poly1305-x86_64.pl: switch to vpermdd in table expansion.
[openssl.git] / crypto / des / asm / des-586.pl
index 72ca9705118383335aa94c523f0d3bb2cce03714..d45102c1da5aadfb30623c3c4ecb54d8df9e665e 100644 (file)
@@ -1,10 +1,16 @@
-#!/usr/local/bin/perl
+#! /usr/bin/env perl
+# Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
 #
+# Licensed under the OpenSSL license (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
 # The inner loop instruction sequence and the IP/FP modifications are from
 # Svend Olaf Mikkelsen <svolaf@inet.uni-c.dk>
-#
 
-push(@INC,"perlasm","../../perlasm");
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
 require "x86asm.pl";
 require "cbc.pl";
 require "desboth.pl";
@@ -14,12 +20,23 @@ require "desboth.pl";
 # format.
 #
 
+$output=pop;
+open STDOUT,">$output";
+
 &asm_init($ARGV[0],"des-586.pl");
 
 $L="edi";
 $R="esi";
+$trans="ebp";
+$small_footprint=1 if (grep(/\-DOPENSSL_SMALL_FOOTPRINT/,@ARGV));
+# one can discuss setting this variable to 1 unconditionally, as
+# the folded loop is only 3% slower than unrolled, but >7 times smaller
+
+&public_label("DES_SPtrans");
+&static_label("des_sptrans");
 
-&external_label("DES_SPtrans");
+&DES_encrypt_internal();
+&DES_decrypt_internal();
 &DES_encrypt("DES_encrypt1",1);
 &DES_encrypt("DES_encrypt2",0);
 &DES_encrypt3("DES_encrypt3",1);
@@ -30,18 +47,93 @@ $R="esi";
 
 &asm_finish();
 
+close STDOUT;
+
+sub DES_encrypt_internal()
+       {
+       &function_begin_B("_x86_DES_encrypt");
+
+       if ($small_footprint)
+           {
+           &lea("edx",&DWP(128,"ecx"));
+           &push("edx");
+           &push("ecx");
+           &set_label("eloop");
+               &D_ENCRYPT(0,$L,$R,0,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("");
+               &D_ENCRYPT(1,$R,$L,2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("");
+               &add("ecx",16);
+               &cmp("ecx",&swtmp(1));
+               &mov(&swtmp(0),"ecx");
+               &jb(&label("eloop"));
+           &add("esp",8);
+           }
+       else
+           {
+           &push("ecx");
+           for ($i=0; $i<16; $i+=2)
+               {
+               &comment("Round $i");
+               &D_ENCRYPT($i,$L,$R,$i*2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("Round ".sprintf("%d",$i+1));
+               &D_ENCRYPT($i+1,$R,$L,($i+1)*2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               }
+           &add("esp",4);
+       }
+       &ret();
+
+       &function_end_B("_x86_DES_encrypt");
+       }
+
+sub DES_decrypt_internal()
+       {
+       &function_begin_B("_x86_DES_decrypt");
+
+       if ($small_footprint)
+           {
+           &push("ecx");
+           &lea("ecx",&DWP(128,"ecx"));
+           &push("ecx");
+           &set_label("dloop");
+               &D_ENCRYPT(0,$L,$R,-2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("");
+               &D_ENCRYPT(1,$R,$L,-4,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("");
+               &sub("ecx",16);
+               &cmp("ecx",&swtmp(1));
+               &mov(&swtmp(0),"ecx");
+               &ja(&label("dloop"));
+           &add("esp",8);
+           }
+       else
+           {
+           &push("ecx");
+           for ($i=15; $i>0; $i-=2)
+               {
+               &comment("Round $i");
+               &D_ENCRYPT(15-$i,$L,$R,$i*2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               &comment("Round ".sprintf("%d",$i-1));
+               &D_ENCRYPT(15-$i+1,$R,$L,($i-1)*2,$trans,"eax","ebx","ecx","edx",&swtmp(0));
+               }
+           &add("esp",4);
+           }
+       &ret();
+
+       &function_end_B("_x86_DES_decrypt");
+       }
+
 sub DES_encrypt
        {
        local($name,$do_ip)=@_;
 
-       &function_begin_B($name,"EXTRN   _DES_SPtrans:DWORD");
+       &function_begin_B($name);
 
        &push("esi");
        &push("edi");
 
        &comment("");
        &comment("Load the 2 words");
-       $trans="ebp";
 
        if ($do_ip)
                {
@@ -74,39 +166,20 @@ sub DES_encrypt
                }
 
        # PIC-ification:-)
-       &picmeup($trans,"DES_SPtrans");
-       #if ($cpp)      { &picmeup($trans,"DES_SPtrans");   }
-       #else           { &lea($trans,&DWP("DES_SPtrans")); }
+       &call   (&label("pic_point"));
+       &set_label("pic_point");
+       &blindpop($trans);
+       &lea    ($trans,&DWP(&label("des_sptrans")."-".&label("pic_point"),$trans));
 
        &mov(   "ecx",  &wparam(1)      );
-       &cmp("ebx","0");
-       &je(&label("start_decrypt"));
-
-       for ($i=0; $i<16; $i+=2)
-               {
-               &comment("");
-               &comment("Round $i");
-               &D_ENCRYPT($i,$L,$R,$i*2,$trans,"eax","ebx","ecx","edx");
-
-               &comment("");
-               &comment("Round ".sprintf("%d",$i+1));
-               &D_ENCRYPT($i+1,$R,$L,($i+1)*2,$trans,"eax","ebx","ecx","edx");
-               }
-       &jmp(&label("end"));
-
-       &set_label("start_decrypt");
-
-       for ($i=15; $i>0; $i-=2)
-               {
-               &comment("");
-               &comment("Round $i");
-               &D_ENCRYPT(15-$i,$L,$R,$i*2,$trans,"eax","ebx","ecx","edx");
-               &comment("");
-               &comment("Round ".sprintf("%d",$i-1));
-               &D_ENCRYPT(15-$i+1,$R,$L,($i-1)*2,$trans,"eax","ebx","ecx","edx");
-               }
 
-       &set_label("end");
+       &cmp("ebx","0");
+       &je(&label("decrypt"));
+       &call("_x86_DES_encrypt");
+       &jmp(&label("done"));
+       &set_label("decrypt");
+       &call("_x86_DES_decrypt");
+       &set_label("done");
 
        if ($do_ip)
                {
@@ -140,7 +213,7 @@ sub DES_encrypt
 
 sub D_ENCRYPT
        {
-       local($r,$L,$R,$S,$trans,$u,$tmp1,$tmp2,$t)=@_;
+       local($r,$L,$R,$S,$trans,$u,$tmp1,$tmp2,$t,$wp1)=@_;
 
         &mov(  $u,             &DWP(&n2a($S*4),$tmp2,"",0));
        &xor(   $tmp1,          $tmp1);
@@ -167,7 +240,7 @@ sub D_ENCRYPT
        &and(   $t,             "0xff"  );
         &xor(  $L,             &DWP("0x600",$trans,$tmp1,0));
         &xor(  $L,             &DWP("0x700",$trans,$tmp2,0));
-       &mov(   $tmp2,          &wparam(1)      );
+       &mov(   $tmp2,          $wp1    );
         &xor(  $L,             &DWP("0x400",$trans,$u,0));
         &xor(  $L,             &DWP("0x500",$trans,$t,0));
        }
@@ -210,7 +283,7 @@ sub IP_new
        &R_PERM_OP($l,$tt,$r,14,"0x33333333",$r);
        &R_PERM_OP($tt,$r,$l,22,"0x03fc03fc",$r);
        &R_PERM_OP($l,$r,$tt, 9,"0xaaaaaaaa",$r);
-       
+
        if ($lr != 3)
                {
                if (($lr-3) < 0)
@@ -252,8 +325,8 @@ sub FP_new
 
 sub DES_SPtrans
        {
-       &public_label("DES_SPtrans");
        &set_label("DES_SPtrans",64);
+       &set_label("des_sptrans");
        &data_word(0x02080800, 0x00080000, 0x02000002, 0x02080802);
        &data_word(0x02000000, 0x00080802, 0x00080002, 0x02000002);
        &data_word(0x00080802, 0x02080800, 0x02080000, 0x00000802);