my $bits1="THIRTY_TWO_BIT ";
my $bits2="SIXTY_FOUR_BIT ";
-my $x86_elf_asm="x86cpuid-elf.o:bn86-elf.o co86-elf.o:dx86-elf.o yx86-elf.o:ax86-elf.o:bx86-elf.o:mx86-elf.o:sx86-elf.o s512sse2-elf.o:cx86-elf.o:rx86-elf.o:rm86-elf.o:r586-elf.o";
-my $x86_coff_asm="x86cpuid-cof.o:bn86-cof.o co86-cof.o:dx86-cof.o yx86-cof.o:ax86-cof.o:bx86-cof.o:mx86-cof.o:sx86-cof.o s512sse2-cof.o:cx86-cof.o:rx86-cof.o:rm86-cof.o:r586-cof.o";
-my $x86_out_asm="x86cpuid-out.o:bn86-out.o co86-out.o:dx86-out.o yx86-out.o:ax86-out.o:bx86-out.o:mx86-out.o:sx86-out.o s512sse2-out.o:cx86-out.o:rx86-out.o:rm86-out.o:r586-out.o";
+my $x86_elf_asm="x86cpuid-elf.o:bn86-elf.o co86-elf.o:dx86-elf.o yx86-elf.o:ax86-elf.o:bx86-elf.o:mx86-elf.o:sx86-elf.o s512sse2-elf.o:cx86-elf.o:rx86-elf.o:rm86-elf.o:r586-elf.o rc4_skey.o";
+my $x86_coff_asm="x86cpuid-cof.o:bn86-cof.o co86-cof.o:dx86-cof.o yx86-cof.o:ax86-cof.o:bx86-cof.o:mx86-cof.o:sx86-cof.o s512sse2-cof.o:cx86-cof.o:rx86-cof.o:rm86-cof.o:r586-cof.o rc4_skey.o";
+my $x86_out_asm="x86cpuid-out.o:bn86-out.o co86-out.o:dx86-out.o yx86-out.o:ax86-out.o:bx86-out.o:mx86-out.o:sx86-out.o s512sse2-out.o:cx86-out.o:rx86-out.o:rm86-out.o:r586-out.o rc4_skey.o";
my $x86_64_asm="x86_64cpuid.o:x86_64-gcc.o::aes-x86_64.o::md5-x86_64.o:sha1-x86_64.o sha256-x86_64.o sha512-x86_64.o::rc4-x86_64.o::";
-my $ia64_asm=":bn-ia64.o::aes_core.o aes_cbc.o aes-ia64.o:::sha1-ia64.o sha256-ia64.o sha512-ia64.o::rc4-ia64.o::";
+my $ia64_asm=":bn-ia64.o::aes_core.o aes_cbc.o aes-ia64.o:::sha1-ia64.o sha256-ia64.o sha512-ia64.o::rc4-ia64.o rc4_skey.o::";
my $no_asm="::::::::::";
my $aes_enc="aes_core.o aes_cbc.o";
my $bf_enc ="bf_enc.o";
my $cast_enc="c_enc.o";
-my $rc4_enc="rc4_enc.o";
+my $rc4_enc="rc4_enc.o rc4_skey.o";
my $rc5_enc="rc5_enc.o";
my $md5_obj="";
my $sha1_obj="";
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$cast_obj = cx86-out.o
$rc4_obj = rx86-out.o
$rmd160_obj = rm86-out.o
-$rc5_obj = r586-out.o
+$rc5_obj = r586-out.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= bsd-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= bsd-shared
$shared_cflag = -fPIC
$cast_obj = cx86-cof.o
$rc4_obj = rx86-cof.o
$rmd160_obj = rm86-cof.o
-$rc5_obj = r586-cof.o
+$rc5_obj = r586-cof.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= cygwin-shared
$shared_cflag = -D_WINDLL
$cast_obj = cx86-out.o
$rc4_obj = rx86-out.o
$rmd160_obj = rm86-out.o
-$rc5_obj = r586-out.o
+$rc5_obj = r586-out.o rc4_skey.o
$dso_scheme =
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= bsd-gcc-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= bsd-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme =
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme =
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme =
$shared_target=
$shared_cflag =
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-out.o
$rc4_obj = rx86-out.o
$rmd160_obj = rm86-out.o
-$rc5_obj = r586-out.o
+$rc5_obj = r586-out.o rc4_skey.o
$dso_scheme =
$shared_target=
$shared_cflag =
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= linux-shared
$shared_cflag = -KPIC
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$md5_obj =
$sha1_obj = sha1-ia64.o sha256-ia64.o sha512-ia64.o
$cast_obj =
-$rc4_obj = rc4-ia64.o
+$rc4_obj = rc4-ia64.o rc4_skey.o
$rmd160_obj =
$rc5_obj =
$dso_scheme = dlfcn
$cast_obj = cx86-cof.o
$rc4_obj = rx86-cof.o
$rmd160_obj = rm86-cof.o
-$rc5_obj = r586-cof.o
+$rc5_obj = r586-cof.o rc4_skey.o
$dso_scheme = win32
$shared_target= cygwin-shared
$shared_cflag = -D_WINDLL -DOPENSSL_USE_APPLINK
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= svr3-shared
$shared_cflag = -Kpic
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= svr3-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= solaris-shared
$shared_cflag = -fPIC
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= svr5-shared
$shared_cflag = -Kpic
$cast_obj = cx86-elf.o
$rc4_obj = rx86-elf.o
$rmd160_obj = rm86-elf.o
-$rc5_obj = r586-elf.o
+$rc5_obj = r586-elf.o rc4_skey.o
$dso_scheme = dlfcn
$shared_target= gnu-shared
$shared_cflag = -fPIC
CFLAG=-g
AR= ar r
-RC4_ENC=rc4_enc.o
+RC4_ENC=rc4_enc.o rc4_skey.o
CFLAGS= $(INCLUDES) $(CFLAG)
ASFLAGS= $(INCLUDES) $(ASFLAG)
LIB=$(TOP)/libcrypto.a
LIBSRC=rc4_skey.c rc4_enc.c
-LIBOBJ=rc4_skey.o $(RC4_ENC)
+LIBOBJ=$(RC4_ENC)
SRC= $(LIBSRC)
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. Rights for redistribution and usage in source and binary
-# forms are granted according to the OpenSSL license.
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# 2.22x RC4 tune-up:-) It should be noted though that my hand [as in
# is not implemented, then this final RC4_CHAR code-path should be
# preferred, as it provides better *all-round* performance].
+# Intel Core2 was observed to perform poorly on both code paths:-( It
+# apparently suffers from some kind of partial register stall, which
+# occurs in 64-bit mode only [as virtually identical 32-bit loop was
+# observed to outperform 64-bit one by almost 50%]. Adding two movzb to
+# cloop1 boosts its performance by 80%! This loop appears to be optimal
+# fit for Core2 and therefore the code was modified to skip cloop8 on
+# this CPU.
+
$output=shift;
-open STDOUT,"| $^X ../perlasm/x86_64-xlate.pl $output";
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open STDOUT,"| $^X $xlate $output";
$dat="%rdi"; # arg1
$len="%rsi"; # arg2
movzb ($dat,$XX[0]),$TX[0]#d
test \$-8,$len
jz .Lcloop1
+ cmp \$0,260($dat)
+ jnz .Lcloop1
push %rbx
jmp .Lcloop8
.align 16
movb $TY#b,($dat,$XX[0])
add $TX[0]#b,$TY#b
add \$1,$XX[0]#b
+ movzb $TY#b,$TY#d
+ movzb $XX[0]#b,$XX[0]#d
movzb ($dat,$TY),$TY#d
movzb ($dat,$XX[0]),$TX[0]#d
xorb ($inp),$TY#b
.size RC4,.-RC4
___
+$idx="%r8";
+$ido="%r9";
+
+$code.=<<___;
+.extern OPENSSL_ia32cap_P
+.globl RC4_set_key
+.type RC4_set_key,\@function,3
+.align 16
+RC4_set_key:
+ lea 8($dat),$dat
+ lea ($inp,$len),$inp
+ neg $len
+ mov $len,%rcx
+ xor %eax,%eax
+ xor $ido,$ido
+ xor %r10,%r10
+ xor %r11,%r11
+
+ mov OPENSSL_ia32cap_P(%rip),$idx#d
+ bt \$20,$idx#d
+ jnc .Lw1stloop
+ bt \$30,$idx#d
+ setc $ido#b
+ mov $ido#d,260($dat)
+ jmp .Lc1stloop
+
+.align 16
+.Lw1stloop:
+ mov %eax,($dat,%rax,4)
+ add \$1,%al
+ jnc .Lw1stloop
+
+ xor $ido,$ido
+ xor $idx,$idx
+.align 16
+.Lw2ndloop:
+ mov ($dat,$ido,4),%r10d
+ add ($inp,$len,1),$idx#b
+ add %r10b,$idx#b
+ add \$1,$len
+ mov ($dat,$idx,4),%r11d
+ cmovz %rcx,$len
+ mov %r10d,($dat,$idx,4)
+ mov %r11d,($dat,$ido,4)
+ add \$1,$ido#b
+ jnc .Lw2ndloop
+ jmp .Lexit_key
+
+.align 16
+.Lc1stloop:
+ mov %al,($dat,%rax)
+ add \$1,%al
+ jnc .Lc1stloop
+
+ xor $ido,$ido
+ xor $idx,$idx
+.align 16
+.Lc2ndloop:
+ mov ($dat,$ido),%r10b
+ add ($inp,$len),$idx#b
+ add %r10b,$idx#b
+ add \$1,$len
+ mov ($dat,$idx),%r11b
+ jnz .Lcnowrap
+ mov %rcx,$len
+.Lcnowrap:
+ mov %r10b,($dat,$idx)
+ mov %r11b,($dat,$ido)
+ add \$1,$ido#b
+ jnc .Lc2ndloop
+ movl \$-1,256($dat)
+
+.align 16
+.Lexit_key:
+ xor %eax,%eax
+ mov %eax,-8($dat)
+ mov %eax,-4($dat)
+ ret
+.size RC4_set_key,.-RC4_set_key
+
+.globl RC4_options
+.type RC4_options,\@function,0
+.align 16
+RC4_options:
+ .picmeup %rax
+ lea .Lopts-.(%rax),%rax
+ mov OPENSSL_ia32cap_P(%rip),%edx
+ bt \$20,%edx
+ jnc .Ldone
+ add \$12,%rax
+ bt \$30,%edx
+ jnc .Ldone
+ add \$13,%rax
+.Ldone:
+ ret
+.align 64
+.Lopts:
+.asciz "rc4(8x,int)"
+.asciz "rc4(8x,char)"
+.asciz "rc4(1x,char)"
+.asciz "RC4 for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align 64
+.size RC4_options,.-RC4_options
+___
+
$code =~ s/#([bwd])/$1/gm;
print $code;
projects / oweals / openssl.git / commitdiff