Add md5-sparcv9.pl.

3 files changed, 289 insertions, 0 deletions

diff --git a/crypto/md5/Makefile b/crypto/md5/Makefile
index 7d42da4fd1..5b289213f5 100644
--- a/crypto/md5/Makefile
+++ b/crypto/md5/Makefile
@@ -52,6 +52,9 @@ md5-ia64.s: asm/md5-ia64.S
 	$(CC) $(CFLAGS) -E asm/md5-ia64.S | \
 	$(PERL) -ne 's/;\s+/;\n/g; print;' > $@
 
+md5-sparcv9.S:	asm/md5-sparcv9.pl
+	$(PERL) asm/md5-sparcv9.pl $@ $(CFLAGS)
+
 files:
 	$(PERL) $(TOP)/util/files.pl Makefile >> $(TOP)/MINFO
 
diff --git a/crypto/md5/asm/md5-sparcv9.pl b/crypto/md5/asm/md5-sparcv9.pl
new file mode 100644
index 0000000000..c9ce0b94ad
--- /dev/null
+++ b/crypto/md5/asm/md5-sparcv9.pl
@@ -0,0 +1,284 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@opensl.org> for the OpenSSL
+# 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/.
+# ====================================================================
+
+# MD5 for SPARCv9, 7.5 cycles per byte on UltraSPARC, >40% faster than
+# code generated by Sun C 5.2.
+
+$bits=32;
+for (@ARGV)	{ $bits=64 if (/\-m64/ || /\-xarch\=v9/); }
+if ($bits==64)	{ $bias=2047; $frame=192; }
+else		{ $bias=0;    $frame=112; }
+
+$output=shift;
+open STDOUT,">$output";
+
+use integer;
+
+($ctx,$inp,$len)=("%i0","%i1","%i2");	# input arguments
+
+# 64-bit values
+@X=("%o0","%o1","%o2","%o3","%o4","%o5","%o7","%g1","%g2");
+$tx="%g3";
+($AB,$CD)=("%g4","%g5");
+
+# 32-bit values
+@V=($A,$B,$C,$D)=map("%l$_",(0..3));
+($t1,$t2,$t3,$saved_asi)=map("%l$_",(4..7));
+($shr,$shl1,$shl2)=("%i3","%i4","%i5");
+
+my @K=(	0xd76aa478,0xe8c7b756,0x242070db,0xc1bdceee,
+	0xf57c0faf,0x4787c62a,0xa8304613,0xfd469501,
+	0x698098d8,0x8b44f7af,0xffff5bb1,0x895cd7be,
+	0x6b901122,0xfd987193,0xa679438e,0x49b40821,
+
+	0xf61e2562,0xc040b340,0x265e5a51,0xe9b6c7aa,
+	0xd62f105d,0x02441453,0xd8a1e681,0xe7d3fbc8,
+	0x21e1cde6,0xc33707d6,0xf4d50d87,0x455a14ed,
+	0xa9e3e905,0xfcefa3f8,0x676f02d9,0x8d2a4c8a,
+
+	0xfffa3942,0x8771f681,0x6d9d6122,0xfde5380c,
+	0xa4beea44,0x4bdecfa9,0xf6bb4b60,0xbebfbc70,
+	0x289b7ec6,0xeaa127fa,0xd4ef3085,0x04881d05,
+	0xd9d4d039,0xe6db99e5,0x1fa27cf8,0xc4ac5665,
+
+	0xf4292244,0x432aff97,0xab9423a7,0xfc93a039,
+	0x655b59c3,0x8f0ccc92,0xffeff47d,0x85845dd1,
+	0x6fa87e4f,0xfe2ce6e0,0xa3014314,0x4e0811a1,
+	0xf7537e82,0xbd3af235,0x2ad7d2bb,0xeb86d391, 0	);
+
+sub R0 {
+  my ($i,$a,$b,$c,$d) = @_;
+  my $rot = (7,12,17,22)[$i%4];
+  my $j   = ($i+1)/2;
+
+  if ($i&1) {
+    $code.=<<___;
+	 srlx	@X[$j],$shr,@X[$j]	! align X[`$i+1`]
+	and	$b,$t1,$t1		! round $i
+	 sllx	@X[$j+1],$shl1,$tx
+	add	$t2,$a,$a
+	 sllx	$tx,$shl2,$tx
+	xor	$d,$t1,$t1
+	 or	$tx,@X[$j],@X[$j]
+	 sethi	%hi(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	 or	$t2,%lo(@K[$i+1]),$t2
+	sll	$a,$rot,$t3
+	 add	@X[$j],$t2,$t2		! X[`$i+1`]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	 xor	 $b,$c,$t1
+	add	$t3,$a,$a
+___
+  } else {
+    $code.=<<___;
+	 srlx	@X[$j],32,$tx		! extract X[`2*$j+1`]
+	and	$b,$t1,$t1		! round $i
+	add	$t2,$a,$a
+	xor	$d,$t1,$t1
+	 sethi	%hi(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	 or	$t2,%lo(@K[$i+1]),$t2
+	sll	$a,$rot,$t3
+	 add	$tx,$t2,$t2		! X[`2*$j+1`]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	 xor	 $b,$c,$t1
+	add	$t3,$a,$a
+___
+  }
+}
+
+sub R0_1 {
+  my ($i,$a,$b,$c,$d) = @_;
+  my $rot = (7,12,17,22)[$i%4];
+
+$code.=<<___;
+	 srlx	@X[0],32,$tx		! extract X[1]
+	and	$b,$t1,$t1		! round $i
+	add	$t2,$a,$a
+	xor	$d,$t1,$t1
+	 sethi	%hi(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	 or	$t2,%lo(@K[$i+1]),$t2
+	sll	$a,$rot,$t3
+	 add	$tx,$t2,$t2		! X[1]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	 andn	 $b,$c,$t1
+	add	$t3,$a,$a
+___
+}
+
+sub R1 {
+  my ($i,$a,$b,$c,$d) = @_;
+  my $rot = (5,9,14,20)[$i%4];
+  my $j   = $i<31 ? (1+5*($i+1))%16 : (5+3*($i+1))%16;
+  my $xi  = @X[$j/2];
+
+$code.=<<___ if ($j&1 && ($xi=$tx));
+	 srlx	@X[$j/2],32,$xi		! extract X[$j]
+___
+$code.=<<___;
+	and	$b,$d,$t3		! round $i
+	add	$t2,$a,$a
+	or	$t3,$t1,$t1
+	 sethi	%hi(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	 or	$t2,%lo(@K[$i+1]),$t2
+	sll	$a,$rot,$t3
+	 add	$xi,$t2,$t2		! X[$j]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	 `$i<31?"andn":"xor"`	 $b,$c,$t1
+	add	$t3,$a,$a
+___
+}
+
+sub R2 {
+  my ($i,$a,$b,$c,$d) = @_;
+  my $rot = (4,11,16,23)[$i%4];
+  my $j   = $i<47 ? (5+3*($i+1))%16 : (0+7*($i+1))%16;
+  my $xi  = @X[$j/2];
+
+$code.=<<___ if ($j&1 && ($xi=$tx));
+	 srlx	@X[$j/2],32,$xi		! extract X[$j]
+___
+$code.=<<___;
+	add	$t2,$a,$a		! round $i
+	xor	$b,$t1,$t1
+	 sethi	%hi(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	 or	$t2,%lo(@K[$i+1]),$t2
+	sll	$a,$rot,$t3
+	 add	$xi,$t2,$t2		! X[$j]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	 xor	 $b,$c,$t1
+	add	$t3,$a,$a
+___
+}
+
+sub R3 {
+  my ($i,$a,$b,$c,$d) = @_;
+  my $rot = (6,10,15,21)[$i%4];
+  my $j   = (0+7*($i+1))%16;
+  my $xi  = @X[$j/2];
+
+$code.=<<___;
+	add	$t2,$a,$a		! round $i
+___
+$code.=<<___ if ($j&1 && ($xi=$tx));
+	 srlx	@X[$j/2],32,$xi		! extract X[$j]
+___
+$code.=<<___;
+	orn	$b,$d,$t1
+	 sethi	%hi(@K[$i+1]),$t2
+	xor	$c,$t1,$t1
+	 or	$t2,%lo(@K[$i+1]),$t2
+	add	$t1,$a,$a
+	sll	$a,$rot,$t3
+	 add	$xi,$t2,$t2		! X[$j]+K[`$i+1`]
+	srl	$a,32-$rot,$a
+	add	$b,$t3,$t3
+	add	$t3,$a,$a
+___
+}
+
+$code.=<<___ if ($bits==64);
+.register	%g2,#scratch
+.register	%g3,#scratch
+___
+$code.=<<___;
+.section	".text",#alloc,#execinstr
+
+.globl	md5_block_asm_data_order
+.align	32
+md5_block_asm_data_order:
+	save	%sp,-$frame,%sp
+
+	rd	%asi,$saved_asi
+	wr	%g0,0x88,%asi		! ASI_PRIMARY_LITTLE
+	and	$inp,7,$shr
+	andn	$inp,7,$inp
+
+	sll	$shr,3,$shr		! *=8
+	mov	56,$shl2
+	ld	[$ctx+0],$A
+	sub	$shl2,$shr,$shl2
+	ld	[$ctx+4],$B
+	and	$shl2,32,$shl1
+	add	$shl2,8,$shl2
+	ld	[$ctx+8],$C
+	sub	$shl2,$shl1,$shl2	! shr+shl1+shl2==64
+	ld	[$ctx+12],$D
+	nop
+
+.Loop:
+	 cmp	$shr,0			! was inp aligned?
+	ldxa	[$inp+0]%asi,@X[0]	! load little-endian input
+	ldxa	[$inp+8]%asi,@X[1]
+	ldxa	[$inp+16]%asi,@X[2]
+	ldxa	[$inp+24]%asi,@X[3]
+	ldxa	[$inp+32]%asi,@X[4]
+	 sllx	$A,32,$AB		! pack A,B
+	ldxa	[$inp+40]%asi,@X[5]
+	 sllx	$C,32,$CD		! pack C,D
+	ldxa	[$inp+48]%asi,@X[6]
+	 or	$B,$AB,$AB
+	ldxa	[$inp+56]%asi,@X[7]
+	 or	$D,$CD,$CD
+	bnz,a,pn	%icc,.+8
+	ldxa	[$inp+64]%asi,@X[8]
+
+	srlx	@X[0],$shr,@X[0]	! align X[0]
+	sllx	@X[1],$shl1,$tx
+	 sethi	%hi(@K[0]),$t2
+	sllx	$tx,$shl2,$tx
+	 or	$t2,%lo(@K[0]),$t2
+	or	$tx,@X[0],@X[0]
+	 xor	$C,$D,$t1
+	 add	@X[0],$t2,$t2		! X[0]+K[0]
+___
+	for ($i=0;$i<15;$i++)	{ &R0($i,@V);	unshift(@V,pop(@V)); }
+	for (;$i<16;$i++)	{ &R0_1($i,@V);	unshift(@V,pop(@V)); }
+	for (;$i<32;$i++)	{ &R1($i,@V);	unshift(@V,pop(@V)); }
+	for (;$i<48;$i++)	{ &R2($i,@V);	unshift(@V,pop(@V)); }
+	for (;$i<64;$i++)	{ &R3($i,@V);	unshift(@V,pop(@V)); }
+$code.=<<___;
+	srlx	$AB,32,$t1		! unpack A,B,C,D and accumulate
+	add	$inp,64,$inp		! advance inp
+	srlx	$CD,32,$t2
+	add	$t1,$A,$A
+	subcc	$len,1,$len		! done yet?
+	add	$AB,$B,$B
+	add	$t2,$C,$C
+	add	$CD,$D,$D
+	srl	$B,0,$B			! clruw	$B
+	bne	`$bits==64?"%xcc":"%icc"`,.Loop
+	srl	$D,0,$D			! clruw	$D
+
+	st	$A,[$ctx+0]		! write out ctx
+	st	$B,[$ctx+4]
+	st	$C,[$ctx+8]
+	st	$D,[$ctx+12]
+
+	wr	%g0,$saved_asi,%asi
+	ret
+	restore
+.type	md5_block_asm_data_order,#function
+.size	md5_block_asm_data_order,(.-md5_block_asm_data_order)
+
+.asciz	"MD5 block transform for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
+.align	4
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT;
diff --git a/crypto/md5/md5_locl.h b/crypto/md5/md5_locl.h
index 968d577995..37e43fe428 100644
--- a/crypto/md5/md5_locl.h
+++ b/crypto/md5/md5_locl.h
@@ -71,6 +71,8 @@
 #  define md5_block_data_order md5_block_asm_data_order
 # elif defined(__ia64) || defined(__ia64__) || defined(_M_IA64)
 #  define md5_block_data_order md5_block_asm_data_order
+# elif defined(__sparc) || defined(__sparc__)
+#  define md5_block_data_order md5_block_asm_data_order
 # endif
 #endif