aboutsummaryrefslogtreecommitdiff
path: root/openssl/crypto/bn/asm/x86/comba.pl
diff options
context:
space:
mode:
authormarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
committermarha <marha@users.sourceforge.net>2009-06-28 22:07:26 +0000
commit3562e78743202e43aec8727005182a2558117eca (patch)
tree8f9113a77d12470c5c851a2a8e4cb02e89df7d43 /openssl/crypto/bn/asm/x86/comba.pl
downloadvcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.gz
vcxsrv-3562e78743202e43aec8727005182a2558117eca.tar.bz2
vcxsrv-3562e78743202e43aec8727005182a2558117eca.zip
Checked in the following released items:
xkeyboard-config-1.4.tar.gz ttf-bitstream-vera-1.10.tar.gz font-alias-1.0.1.tar.gz font-sun-misc-1.0.0.tar.gz font-sun-misc-1.0.0.tar.gz font-sony-misc-1.0.0.tar.gz font-schumacher-misc-1.0.0.tar.gz font-mutt-misc-1.0.0.tar.gz font-misc-misc-1.0.0.tar.gz font-misc-meltho-1.0.0.tar.gz font-micro-misc-1.0.0.tar.gz font-jis-misc-1.0.0.tar.gz font-isas-misc-1.0.0.tar.gz font-dec-misc-1.0.0.tar.gz font-daewoo-misc-1.0.0.tar.gz font-cursor-misc-1.0.0.tar.gz font-arabic-misc-1.0.0.tar.gz font-winitzki-cyrillic-1.0.0.tar.gz font-misc-cyrillic-1.0.0.tar.gz font-cronyx-cyrillic-1.0.0.tar.gz font-screen-cyrillic-1.0.1.tar.gz font-xfree86-type1-1.0.1.tar.gz font-adobe-utopia-type1-1.0.1.tar.gz font-ibm-type1-1.0.0.tar.gz font-bitstream-type1-1.0.0.tar.gz font-bitstream-speedo-1.0.0.tar.gz font-bh-ttf-1.0.0.tar.gz font-bh-type1-1.0.0.tar.gz font-bitstream-100dpi-1.0.0.tar.gz font-bh-lucidatypewriter-100dpi-1.0.0.tar.gz font-bh-100dpi-1.0.0.tar.gz font-adobe-utopia-100dpi-1.0.1.tar.gz font-adobe-100dpi-1.0.0.tar.gz font-util-1.0.1.tar.gz font-bitstream-75dpi-1.0.0.tar.gz font-bh-lucidatypewriter-75dpi-1.0.0.tar.gz font-adobe-utopia-75dpi-1.0.1.tar.gz font-bh-75dpi-1.0.0.tar.gz bdftopcf-1.0.1.tar.gz font-adobe-75dpi-1.0.0.tar.gz mkfontscale-1.0.6.tar.gz openssl-0.9.8k.tar.gz bigreqsproto-1.0.2.tar.gz xtrans-1.2.2.tar.gz resourceproto-1.0.2.tar.gz inputproto-1.4.4.tar.gz compositeproto-0.4.tar.gz damageproto-1.1.0.tar.gz zlib-1.2.3.tar.gz xkbcomp-1.0.5.tar.gz freetype-2.3.9.tar.gz pthreads-w32-2-8-0-release.tar.gz pixman-0.12.0.tar.gz kbproto-1.0.3.tar.gz evieext-1.0.2.tar.gz fixesproto-4.0.tar.gz recordproto-1.13.2.tar.gz randrproto-1.2.2.tar.gz scrnsaverproto-1.1.0.tar.gz renderproto-0.9.3.tar.gz xcmiscproto-1.1.2.tar.gz fontsproto-2.0.2.tar.gz xextproto-7.0.3.tar.gz xproto-7.0.14.tar.gz libXdmcp-1.0.2.tar.gz libxkbfile-1.0.5.tar.gz libfontenc-1.0.4.tar.gz libXfont-1.3.4.tar.gz libX11-1.1.5.tar.gz libXau-1.0.4.tar.gz libxcb-1.1.tar.gz xorg-server-1.5.3.tar.gz
Diffstat (limited to 'openssl/crypto/bn/asm/x86/comba.pl')
-rw-r--r--openssl/crypto/bn/asm/x86/comba.pl277
1 files changed, 277 insertions, 0 deletions
diff --git a/openssl/crypto/bn/asm/x86/comba.pl b/openssl/crypto/bn/asm/x86/comba.pl
new file mode 100644
index 000000000..229125362
--- /dev/null
+++ b/openssl/crypto/bn/asm/x86/comba.pl
@@ -0,0 +1,277 @@
+#!/usr/local/bin/perl
+# x86 assember
+
+sub mul_add_c
+ {
+ local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("mul a[$ai]*b[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$b,"",0));
+
+ &mul("edx");
+ &add($c0,"eax");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # laod next a
+ &mov("eax",&wparam(0)) if $pos > 0; # load r[]
+ ###
+ &adc($c1,"edx");
+ &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # laod next b
+ &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # laod next b
+ ###
+ &adc($c2,0);
+ # is pos > 1, it means it is the last loop
+ &mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[];
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # laod next a
+ }
+
+sub sqr_add_c
+ {
+ local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("sqr a[$ai]*a[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$b,"",0));
+
+ if ($ai == $bi)
+ { &mul("eax");}
+ else
+ { &mul("edx");}
+ &add($c0,"eax");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
+ ###
+ &adc($c1,"edx");
+ &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb);
+ ###
+ &adc($c2,0);
+ # is pos > 1, it means it is the last loop
+ &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
+ }
+
+sub sqr_add_c2
+ {
+ local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("sqr a[$ai]*a[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$a,"",0));
+
+ if ($ai == $bi)
+ { &mul("eax");}
+ else
+ { &mul("edx");}
+ &add("eax","eax");
+ ###
+ &adc("edx","edx");
+ ###
+ &adc($c2,0);
+ &add($c0,"eax");
+ &adc($c1,"edx");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
+ &adc($c2,0);
+ &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
+ &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb);
+ ###
+ }
+
+sub bn_mul_comba
+ {
+ local($name,$num)=@_;
+ local($a,$b,$c0,$c1,$c2);
+ local($i,$as,$ae,$bs,$be,$ai,$bi);
+ local($tot,$end);
+
+ &function_begin_B($name,"");
+
+ $c0="ebx";
+ $c1="ecx";
+ $c2="ebp";
+ $a="esi";
+ $b="edi";
+
+ $as=0;
+ $ae=0;
+ $bs=0;
+ $be=0;
+ $tot=$num+$num-1;
+
+ &push("esi");
+ &mov($a,&wparam(1));
+ &push("edi");
+ &mov($b,&wparam(2));
+ &push("ebp");
+ &push("ebx");
+
+ &xor($c0,$c0);
+ &mov("eax",&DWP(0,$a,"",0)); # load the first word
+ &xor($c1,$c1);
+ &mov("edx",&DWP(0,$b,"",0)); # load the first second
+
+ for ($i=0; $i<$tot; $i++)
+ {
+ $ai=$as;
+ $bi=$bs;
+ $end=$be+1;
+
+ &comment("################## Calculate word $i");
+
+ for ($j=$bs; $j<$end; $j++)
+ {
+ &xor($c2,$c2) if ($j == $bs);
+ if (($j+1) == $end)
+ {
+ $v=1;
+ $v=2 if (($i+1) == $tot);
+ }
+ else
+ { $v=0; }
+ if (($j+1) != $end)
+ {
+ $na=($ai-1);
+ $nb=($bi+1);
+ }
+ else
+ {
+ $na=$as+($i < ($num-1));
+ $nb=$bs+($i >= ($num-1));
+ }
+#printf STDERR "[$ai,$bi] -> [$na,$nb]\n";
+ &mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb);
+ if ($v)
+ {
+ &comment("saved r[$i]");
+ # &mov("eax",&wparam(0));
+ # &mov(&DWP($i*4,"eax","",0),$c0);
+ ($c0,$c1,$c2)=($c1,$c2,$c0);
+ }
+ $ai--;
+ $bi++;
+ }
+ $as++ if ($i < ($num-1));
+ $ae++ if ($i >= ($num-1));
+
+ $bs++ if ($i >= ($num-1));
+ $be++ if ($i < ($num-1));
+ }
+ &comment("save r[$i]");
+ # &mov("eax",&wparam(0));
+ &mov(&DWP($i*4,"eax","",0),$c0);
+
+ &pop("ebx");
+ &pop("ebp");
+ &pop("edi");
+ &pop("esi");
+ &ret();
+ &function_end_B($name);
+ }
+
+sub bn_sqr_comba
+ {
+ local($name,$num)=@_;
+ local($r,$a,$c0,$c1,$c2)=@_;
+ local($i,$as,$ae,$bs,$be,$ai,$bi);
+ local($b,$tot,$end,$half);
+
+ &function_begin_B($name,"");
+
+ $c0="ebx";
+ $c1="ecx";
+ $c2="ebp";
+ $a="esi";
+ $r="edi";
+
+ &push("esi");
+ &push("edi");
+ &push("ebp");
+ &push("ebx");
+ &mov($r,&wparam(0));
+ &mov($a,&wparam(1));
+ &xor($c0,$c0);
+ &xor($c1,$c1);
+ &mov("eax",&DWP(0,$a,"",0)); # load the first word
+
+ $as=0;
+ $ae=0;
+ $bs=0;
+ $be=0;
+ $tot=$num+$num-1;
+
+ for ($i=0; $i<$tot; $i++)
+ {
+ $ai=$as;
+ $bi=$bs;
+ $end=$be+1;
+
+ &comment("############### Calculate word $i");
+ for ($j=$bs; $j<$end; $j++)
+ {
+ &xor($c2,$c2) if ($j == $bs);
+ if (($ai-1) < ($bi+1))
+ {
+ $v=1;
+ $v=2 if ($i+1) == $tot;
+ }
+ else
+ { $v=0; }
+ if (!$v)
+ {
+ $na=$ai-1;
+ $nb=$bi+1;
+ }
+ else
+ {
+ $na=$as+($i < ($num-1));
+ $nb=$bs+($i >= ($num-1));
+ }
+ if ($ai == $bi)
+ {
+ &sqr_add_c($r,$a,$ai,$bi,
+ $c0,$c1,$c2,$v,$i,$na,$nb);
+ }
+ else
+ {
+ &sqr_add_c2($r,$a,$ai,$bi,
+ $c0,$c1,$c2,$v,$i,$na,$nb);
+ }
+ if ($v)
+ {
+ &comment("saved r[$i]");
+ #&mov(&DWP($i*4,$r,"",0),$c0);
+ ($c0,$c1,$c2)=($c1,$c2,$c0);
+ last;
+ }
+ $ai--;
+ $bi++;
+ }
+ $as++ if ($i < ($num-1));
+ $ae++ if ($i >= ($num-1));
+
+ $bs++ if ($i >= ($num-1));
+ $be++ if ($i < ($num-1));
+ }
+ &mov(&DWP($i*4,$r,"",0),$c0);
+ &pop("ebx");
+ &pop("ebp");
+ &pop("edi");
+ &pop("esi");
+ &ret();
+ &function_end_B($name);
+ }
+
+1;