Throw in more PA-RISC assembler.

crypto/Makefile

@@ -70,14 +70,13 @@ uplink.o:	$(TOP)/ms/uplink.c applink.o
 uplink-x86.s:	$(TOP)/ms/uplink-x86.pl
 	$(PERL) $(TOP)/ms/uplink-x86.pl $(PERLASM_SCHEME) > $@
 
-x86_64cpuid.s: x86_64cpuid.pl
-	$(PERL) x86_64cpuid.pl $(PERLASM_SCHEME) > $@
-ia64cpuid.s: ia64cpuid.S
-	$(CC) $(CFLAGS) -E ia64cpuid.S > $@
-ppccpuid.s:		ppccpuid.pl;	$(PERL) ppccpuid.pl $(PERLASM_SCHEME) $@
+x86_64cpuid.s:	x86_64cpuid.pl;	$(PERL) x86_64cpuid.pl $(PERLASM_SCHEME) > $@
+ia64cpuid.s:	ia64cpuid.S;	$(CC) $(CFLAGS) -E ia64cpuid.S > $@
+ppccpuid.s:	ppccpuid.pl;	$(PERL) ppccpuid.pl $(PERLASM_SCHEME) $@
+pariscid.s:	pariscid.pl;	$(PERL) pariscid.pl $(PERLASM_SCHEME) $@
 
 testapps:
-	[ -z "$(THIS)" ] || (	if echo $(SDIRS) | fgrep ' des '; \
+	[ -z "$(THIS)" ] || (	if expr " $(SDIRS) " : ".* des " >/dev/null; \
 				then cd des && $(MAKE) -e des; fi )
 	[ -z "$(THIS)" ] || ( cd pkcs7 && $(MAKE) -e testapps );
 	@if [ -z "$(THIS)" ]; then $(MAKE) -f $(TOP)/Makefile reflect THIS=$@; fi

crypto/aes/Makefile

@@ -64,6 +64,9 @@ aes-sparcv9.s: asm/aes-sparcv9.pl
 aes-ppc.s:	asm/aes-ppc.pl
 	$(PERL) asm/aes-ppc.pl $(PERLASM_SCHEME) $@
 
+aes-parisc.s:	asm/aes-parisc.pl
+	$(PERL) asm/aes-parisc.pl $(PERLASM_SCHEME) $@
+
 # GNU make "catch all"
 aes-%.s:	asm/aes-%.pl;	$(PERL) $< $(CFLAGS) > $@
 

crypto/bn/Makefile

@@ -100,6 +100,9 @@ pa-risc2W.o: asm/pa-risc2W.s
 pa-risc2.o: asm/pa-risc2.s
 	/usr/ccs/bin/as -o pa-risc2.o asm/pa-risc2.s
 
+parisc-mont.s:	asm/parisc-mont.pl
+	$(PERL) asm/parisc-mont.pl $(PERLASM_SCHEME) $@
+
 # ppc - AIX, Linux, MacOS X...
 bn-ppc.s:	asm/ppc.pl;	$(PERL) asm/ppc.pl $(PERLASM_SCHEME) $@
 ppc-mont.s:	asm/ppc-mont.pl;$(PERL) asm/ppc-mont.pl $(PERLASM_SCHEME) $@

crypto/pariscid.pl

+#!/usr/bin/env perl
+
+$flavour = shift;
+$output = shift;
+open STDOUT,">$output";
+
+if ($flavour =~ /64/) {
+	$LEVEL		="2.0W";
+	$SIZE_T		=8;
+	$ST		="std";
+} else {
+	$LEVEL		="1.1";
+	$SIZE_T		=4;
+	$ST		="stw";
+}
+
+$rp="%r2";
+$sp="%r30";
+$rv="%r28";
+
+$code=<<___;
+	.LEVEL	$LEVEL
+	.SPACE	\$TEXT\$
+	.SUBSPA	\$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
+
+	.EXPORT	OPENSSL_rdtsc,ENTRY
+	.ALIGN	8
+OPENSSL_rdtsc
+	.PROC
+	.CALLINFO	NO_CALLS
+	.ENTRY
+	mfctl	%cr16,$rv
+	bv	($rp)
+	.EXIT
+	nop
+	.PROCEND
+
+	.EXPORT	OPENSSL_wipe_cpu,ENTRY
+	.ALIGN	8
+OPENSSL_wipe_cpu
+	.PROC
+	.CALLINFO	NO_CALLS
+	.ENTRY
+	xor		%r0,%r0,%r1
+	fcpy,dbl	%fr0,%fr4
+	xor		%r0,%r0,%r19
+	fcpy,dbl	%fr0,%fr5
+	xor		%r0,%r0,%r20
+	fcpy,dbl	%fr0,%fr6
+	xor		%r0,%r0,%r21
+	fcpy,dbl	%fr0,%fr7
+	xor		%r0,%r0,%r22
+	fcpy,dbl	%fr0,%fr8
+	xor		%r0,%r0,%r23
+	fcpy,dbl	%fr0,%fr9
+	xor		%r0,%r0,%r24
+	fcpy,dbl	%fr0,%fr10
+	xor		%r0,%r0,%r25
+	fcpy,dbl	%fr0,%fr11
+	xor		%r0,%r0,%r26
+	fcpy,dbl	%fr0,%fr22
+	xor		%r0,%r0,%r29
+	fcpy,dbl	%fr0,%fr23
+	xor		%r0,%r0,%r31
+	fcpy,dbl	%fr0,%fr24
+	fcpy,dbl	%fr0,%fr25
+	fcpy,dbl	%fr0,%fr26
+	fcpy,dbl	%fr0,%fr27
+	fcpy,dbl	%fr0,%fr28
+	fcpy,dbl	%fr0,%fr29
+	fcpy,dbl	%fr0,%fr30
+	fcpy,dbl	%fr0,%fr31
+	bv		($rp)
+	.EXIT
+	ldo		0($sp),$rv
+	.PROCEND
+___
+{
+$inp="%r26";
+$len="%r25";
+
+$code.=<<___;
+	.EXPORT	OPENSSL_cleanse,ENTRY,ARGW0=GR,ARGW1=GR
+	.ALIGN	8
+OPENSSL_cleanse
+	.PROC
+	.CALLINFO	NO_CALLS
+	.ENTRY
+	cmpib,*>>	15,$len,Little
+	ldi		$SIZE_T-1,%r1
+
+Lalign
+	and,*<>		$inp,%r1,%r28
+	b,n		Laligned
+	stb		%r0,0($inp)
+	ldo		-1($len),$len
+	b		Lalign
+	ldo		1($inp),$inp
+
+Laligned
+	andcm		$len,%r1,%r28
+Loop
+	$ST		%r0,0($inp)
+	addib,*vnz	-$SIZE_T,%r28,Loop
+	ldo		$SIZE_T($inp),$inp
+
+	and,*<>		$len,%r1,$len
+	b,n		Ldone
+Little
+	stb		%r0,0($inp)
+	addib,*vnz	-1,$len,Little
+	ldo		1($inp),$inp
+Ldone
+	bv		($rp)
+	.EXIT
+	nop
+	.PROCEND
+___
+}
+
+$code =~ s/cmpib,\*/comib,/gm if ($SIZE_T==4);
+$code =~ s/,\*/,/gm if ($SIZE_T==4);
+print $code;
+close STDOUT;
+

crypto/rc4/Makefile

@@ -53,6 +53,9 @@ rc4-ia64.S: asm/rc4-ia64.pl
 rc4-s390x.s:	asm/rc4-s390x.pl
 	$(PERL) asm/rc4-s390x.pl > $@
 
+rc4-parisc.s:	asm/rc4-parisc.pl
+	$(PERL) asm/rc4-parisc.pl $(PERLASM_SCHEME) $@
+
 rc4-ia64.s: rc4-ia64.S
 	@case `awk '/^#define RC4_INT/{print$$NF}' $(TOP)/include/openssl/opensslconf.h` in \
 	int)	set -x; $(CC) $(CFLAGS) -DSZ=4 -E rc4-ia64.S > $@ ;; \

crypto/rc4/asm/rc4-parisc.pl

+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
+# ====================================================================
+
+# RC4 for PA-RISC.
+
+# June 2009.
+#
+# Performance is 33% better than gcc 3.2 generated code on PA-7100LC.
+# For reference, [4x] unrolled loop is >40% faster than folded one.
+# It's possible to unroll loop 8 times on PA-RISC 2.0, but improvement
+# is believed to be not sufficient to justify the effort...
+#
+# Special thanks to polarhome.com for providing HP-UX account.
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+
+$flavour = shift;
+$output = shift;
+open STDOUT,">$output";
+
+if ($flavour =~ /64/) {
+	$LEVEL		="2.0W";
+	$SIZE_T		=8;
+	$FRAME_MARKER	=80;
+	$SAVED_RP	=16;
+	$PUSH		="std";
+	$PUSHMA		="std,ma";
+	$POP		="ldd";
+	$POPMB		="ldd,mb";
+} else {
+	$LEVEL		="1.0";
+	$SIZE_T		=4;
+	$FRAME_MARKER	=48;
+	$SAVED_RP	=20;
+	$PUSH		="stw";
+	$PUSHMA		="stwm";
+	$POP		="ldw";
+	$POPMB		="ldwm";
+}
+
+$FRAME=4*$SIZE_T+$FRAME_MARKER;	# 4 saved regs + frame marker
+				#                [+ argument transfer]
+$SZ=1;				# defaults to RC4_CHAR
+if (open CONF,"<${dir}../../opensslconf.h") {
+    while(<CONF>) {
+	if (m/#\s*define\s+RC4_INT\s+(.*)/) {
+	    $SZ = ($1=~/char$/) ? 1 : 4;
+	    last;
+	}
+    }
+    close CONF;
+}
+
+if ($SZ==1) {	# RC4_CHAR
+    $LD="ldb";
+    $LDX="ldbx";
+    $MKX="addl";
+    $ST="stb";
+} else {	# RC4_INT (~5% faster than RC4_CHAR on PA-7100LC)
+    $LD="ldw";
+    $LDX="ldwx,s";
+    $MKX="sh2addl";
+    $ST="stw";
+}
+
+$key="%r26";
+$len="%r25";
+$inp="%r24";
+$out="%r23";
+
+@XX=("%r19","%r20");
+@TX=("%r21","%r22");
+$YY="%r28";
+$TY="%r29";
+
+$acc="%r1";
+$ix="%r2";
+$iy="%r3";
+$dat0="%r4";
+$dat1="%r5";
+$rem="%r6";
+$mask="%r31";
+
+sub unrolledloopbody {
+for ($i=0;$i<4;$i++) {
+$code.=<<___;
+	ldo	1($XX[0]),$XX[1]
+	`sprintf("$LDX	%$TY(%$key),%$dat1") if ($i>0)`	
+	and	$mask,$XX[1],$XX[1]
+	$LDX	$YY($key),$TY
+	$MKX	$YY,$key,$ix
+	$LDX	$XX[1]($key),$TX[1]
+	$MKX	$XX[0],$key,$iy
+	$ST	$TX[0],0($ix)
+	comclr,<> $XX[1],$YY,%r0	; conditional
+	copy	$TX[0],$TX[1]		; move
+	`sprintf("%sdep	%$dat1,%d,8,%$acc",$i==1?"z":"",8*($i-1)+7) if ($i>0)`
+	$ST	$TY,0($iy)
+	addl	$TX[0],$TY,$TY
+	addl	$TX[1],$YY,$YY
+	and	$mask,$TY,$TY
+	and	$mask,$YY,$YY
+___
+push(@TX,shift(@TX)); push(@XX,shift(@XX));	# "rotate" registers
+} }
+
+sub foldedloop {
+my ($label,$count)=@_;
+$code.=<<___;
+$label
+	$MKX	$YY,$key,$iy
+	$LDX	$YY($key),$TY
+	$MKX	$XX[0],$key,$ix
+	$ST	$TX[0],0($iy)
+	ldo	1($XX[0]),$XX[0]
+	$ST	$TY,0($ix)
+	addl	$TX[0],$TY,$TY
+	ldbx	$inp($out),$dat1
+	and	$mask,$TY,$TY
+	and	$mask,$XX[0],$XX[0]
+	$LDX	$TY($key),$acc
+	$LDX	$XX[0]($key),$TX[0]
+	ldo	1($out),$out
+	xor	$dat1,$acc,$acc
+	addl	$TX[0],$YY,$YY
+	stb	$acc,-1($out)
+	addib,<> -1,$count,$label	; $count is always small
+	and	$mask,$YY,$YY
+___
+}
+
+$code=<<___;
+	.LEVEL	$LEVEL
+	.SPACE	\$TEXT\$
+	.SUBSPA	\$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
+
+	.EXPORT	RC4,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR
+RC4
+	.PROC
+	.CALLINFO	FRAME=`$FRAME-4*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=6
+	.ENTRY
+	$PUSH	%r2,-$SAVED_RP(%sp)	; standard prologue
+	$PUSHMA	%r3,$FRAME(%sp)
+	$PUSH	%r4,`-$FRAME+1*$SIZE_T`(%sp)
+	$PUSH	%r5,`-$FRAME+2*$SIZE_T`(%sp)
+	$PUSH	%r6,`-$FRAME+3*$SIZE_T`(%sp)
+
+	cmpib,*= 0,$len,L\$abort
+	sub	$inp,$out,$inp		; distance between $inp and $out
+
+	$LD	`0*$SZ`($key),$XX[0]
+	$LD	`1*$SZ`($key),$YY
+	ldo	`2*$SZ`($key),$key
+
+	ldi	0xff,$mask
+	ldi	3,$dat0		
+
+	ldo	1($XX[0]),$XX[0]	; warm up loop
+	and	$mask,$XX[0],$XX[0]
+	$LDX	$XX[0]($key),$TX[0]
+	addl	$TX[0],$YY,$YY
+	cmpib,*>>= 6,$len,L\$oop1	; is $len large enough to bother?
+	and	$mask,$YY,$YY
+
+	and,<>	$out,$dat0,$rem		; is $out aligned?
+	b	L\$alignedout
+	subi	4,$rem,$rem
+	sub	$len,$rem,$len
+___
+&foldedloop("L\$alignout",$rem);	# process till $out is aligned
+
+$code.=<<___;
+L\$alignedout				; $len is at least 4 here
+	and,<>	$inp,$dat0,$acc		; is $inp aligned?
+	b	L\$oop4
+	sub	$inp,$acc,$rem		; align $inp
+
+	sh3addl	$acc,%r0,$acc
+	subi	32,$acc,$acc
+	mtctl	$acc,%cr11		; load %sar with vshd align factor
+	ldwx	$rem($out),$dat0
+	ldo	4($rem),$rem
+L\$oop4misalignedinp
+___
+&unrolledloopbody();
+$code.=<<___;
+	$LDX	$TY($key),$ix
+	ldwx	$rem($out),$dat1
+	ldo	-4($len),$len
+	or	$ix,$acc,$acc		; last piece, no need to dep
+	vshd	$dat0,$dat1,$iy		; align data
+	copy	$dat1,$dat0
+	xor	$iy,$acc,$acc
+	stw	$acc,0($out)
+	cmpib,*<< 3,$len,L\$oop4misalignedinp
+	ldo	4($out),$out
+	cmpib,*= 0,$len,L\$done
+	nop
+	b	L\$oop1
+	nop
+
+	.ALIGN	8
+L\$oop4
+___
+&unrolledloopbody();
+$code.=<<___;
+	$LDX	$TY($key),$ix
+	ldwx	$inp($out),$dat0
+	ldo	-4($len),$len
+	or	$ix,$acc,$acc		; last piece, no need to dep
+	xor	$dat0,$acc,$acc
+	stw	$acc,0($out)
+	cmpib,*<< 3,$len,L\$oop4
+	ldo	4($out),$out
+	cmpib,*= 0,$len,L\$done
+	nop
+___
+&foldedloop("L\$oop1",$len);
+$code.=<<___;
+L\$done
+	$POP	`-$FRAME-$SAVED_RP`(%sp),%r2
+	ldo	-1($XX[0]),$XX[0]	; chill out loop
+	sub	$YY,$TX[0],$YY
+	and	$mask,$XX[0],$XX[0]
+	and	$mask,$YY,$YY
+	$ST	$XX[0],`-2*$SZ`($key)
+	$ST	$YY,`-1*$SZ`($key)
+	$POP	`-$FRAME+1*$SIZE_T`(%sp),%r4
+	$POP	`-$FRAME+2*$SIZE_T`(%sp),%r5
+	$POP	`-$FRAME+3*$SIZE_T`(%sp),%r6
+L\$abort
+	bv	(%r2)
+	.EXIT
+	$POPMB	-$FRAME(%sp),%r3
+	.PROCEND
+___
+
+$code.=<<___;
+
+	.EXPORT	RC4_set_key,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
+	.ALIGN	8
+RC4_set_key
+	.PROC
+	.CALLINFO	NO_CALLS
+	.ENTRY
+	$ST	%r0,`0*$SZ`($key)
+	$ST	%r0,`1*$SZ`($key)
+	ldo	`2*$SZ`($key),$key
+	copy	%r0,@XX[0]
+L\$1st
+	$ST	@XX[0],0($key)
+	ldo	1(@XX[0]),@XX[0]
+	bb,>=	@XX[0],`31-8`,L\$1st	; @XX[0]<256
+	ldo	$SZ($key),$key
+
+	ldo	`-256*$SZ`($key),$key	; rewind $key
+	addl	$len,$inp,$inp		; $inp to point at the end
+	sub	%r0,$len,%r23		; inverse index
+	copy	%r0,@XX[0]
+	copy	%r0,@XX[1]
+	ldi	0xff,$mask
+
+L\$2nd
+	$LDX	@XX[0]($key),@TX[0]
+	ldbx	%r23($inp),@TX[1]
+	addi,nuv 1,%r23,%r23		; increment and conditional
+	sub	%r0,$len,%r23		; inverse index
+	addl	@TX[0],@XX[1],@XX[1]
+	addl	@TX[1],@XX[1],@XX[1]
+	and	$mask,@XX[1],@XX[1]
+	$MKX	@XX[0],$key,$TY
+	$LDX	@XX[1]($key),@TX[1]
+	$MKX	@XX[1],$key,$YY
+	ldo	1(@XX[0]),@XX[0]
+	$ST	@TX[0],0($YY)
+	bb,>=	@XX[0],`31-8`,L\$2nd	; @XX[0]<256
+	$ST	@TX[1],0($TY)
+
+	bv,n	(%r2)
+	.EXIT
+	nop
+	.PROCEND
+
+	.EXPORT	RC4_options,ENTRY
+	.ALIGN	8
+RC4_options
+	.PROC
+	.CALLINFO	NO_CALLS
+	.ENTRY
+	blr	%r0,%r28
+	ldi	3,%r1
+L\$pic
+	andcm	%r28,%r1,%r28
+	bv	(%r2)
+	.EXIT
+	ldo	L\$opts-L\$pic(%r28),%r28
+	.PROCEND
+	.ALIGN	8
+L\$opts
+	.STRINGZ "rc4(4x,`$SZ==1?"char":"int"`)"
+	.STRINGZ "RC4 for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
+___
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+$code =~ s/cmpib,\*/comib,/gm if ($SIZE_T==4);
+
+print $code;
+close STDOUT;