Blob Blame Raw
diff -up openssl-1.1.1b/crypto/chacha/asm/chacha-s390x.pl.s390x-update openssl-1.1.1b/crypto/chacha/asm/chacha-s390x.pl
--- openssl-1.1.1b/crypto/chacha/asm/chacha-s390x.pl.s390x-update	2019-02-26 15:15:30.000000000 +0100
+++ openssl-1.1.1b/crypto/chacha/asm/chacha-s390x.pl	2019-05-06 10:54:00.035367605 +0200
@@ -1,5 +1,5 @@
 #! /usr/bin/env perl
-# Copyright 2016 The OpenSSL Project Authors. All Rights Reserved.
+# Copyright 2016-2019 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
@@ -20,41 +20,53 @@
 #
 # 3 times faster than compiler-generated code.
 
-$flavour = shift;
+#
+# August 2018
+#
+# Add vx code path: 4x"vertical".
+#
+# Copyright IBM Corp. 2018
+# Author: Patrick Steuer <patrick.steuer@de.ibm.com>
+
+#
+# February 2019
+#
+# Add 6x"horizontal" VX implementation. It's ~25% faster than IBM's
+# 4x"vertical" submission [on z13] and >3 faster than scalar code.
+# But to harness overheads revert to transliteration of VSX code path
+# from chacha-ppc module, which is also 4x"vertical", to handle inputs
+# not longer than 256 bytes.
+
+use strict;
+use FindBin qw($Bin);
+use lib "$Bin/../..";
+use perlasm::s390x qw(:DEFAULT :VX AUTOLOAD LABEL INCLUDE);
 
+my $flavour = shift;
+
+my ($z,$SIZE_T);
 if ($flavour =~ /3[12]/) {
+	$z=0;	# S/390 ABI
 	$SIZE_T=4;
-	$g="";
 } else {
+	$z=1;	# zSeries ABI
 	$SIZE_T=8;
-	$g="g";
 }
 
+my $output;
 while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
-open STDOUT,">$output";
-
-sub AUTOLOAD()		# thunk [simplified] x86-style perlasm
-{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://;
-    $code .= "\t$opcode\t".join(',',@_)."\n";
-}
 
 my $sp="%r15";
-
 my $stdframe=16*$SIZE_T+4*8;
-my $frame=$stdframe+4*20;
-
-my ($out,$inp,$len,$key,$counter)=map("%r$_",(2..6));
 
+sub ROUND {
 my @x=map("%r$_",(0..7,"x","x","x","x",(10..13)));
 my @t=map("%r$_",(8,9));
-
-sub ROUND {
 my ($a0,$b0,$c0,$d0)=@_;
 my ($a1,$b1,$c1,$d1)=map(($_&~3)+(($_+1)&3),($a0,$b0,$c0,$d0));
 my ($a2,$b2,$c2,$d2)=map(($_&~3)+(($_+1)&3),($a1,$b1,$c1,$d1));
 my ($a3,$b3,$c3,$d3)=map(($_&~3)+(($_+1)&3),($a2,$b2,$c2,$d2));
-my ($xc,$xc_)=map("\"$_\"",@t);
-my @x=map("\"$_\"",@x);
+my ($xc,$xc_)=map("$_",@t);
 
 	# Consider order in which variables are addressed by their
 	# index:
@@ -78,249 +90,967 @@ my @x=map("\"$_\"",@x);
 	# 'c' stores and loads in the middle, but none in the beginning
 	# or end.
 
-	(
-	"&alr	(@x[$a0],@x[$b0])",	# Q1
-	 "&alr	(@x[$a1],@x[$b1])",	# Q2
-	"&xr	(@x[$d0],@x[$a0])",
-	 "&xr	(@x[$d1],@x[$a1])",
-	"&rll	(@x[$d0],@x[$d0],16)",
-	 "&rll	(@x[$d1],@x[$d1],16)",
-
-	"&alr	($xc,@x[$d0])",
-	 "&alr	($xc_,@x[$d1])",
-	"&xr	(@x[$b0],$xc)",
-	 "&xr	(@x[$b1],$xc_)",
-	"&rll	(@x[$b0],@x[$b0],12)",
-	 "&rll	(@x[$b1],@x[$b1],12)",
-
-	"&alr	(@x[$a0],@x[$b0])",
-	 "&alr	(@x[$a1],@x[$b1])",
-	"&xr	(@x[$d0],@x[$a0])",
-	 "&xr	(@x[$d1],@x[$a1])",
-	"&rll	(@x[$d0],@x[$d0],8)",
-	 "&rll	(@x[$d1],@x[$d1],8)",
-
-	"&alr	($xc,@x[$d0])",
-	 "&alr	($xc_,@x[$d1])",
-	"&xr	(@x[$b0],$xc)",
-	 "&xr	(@x[$b1],$xc_)",
-	"&rll	(@x[$b0],@x[$b0],7)",
-	 "&rll	(@x[$b1],@x[$b1],7)",
-
-	"&stm	($xc,$xc_,'$stdframe+4*8+4*$c0($sp)')",	# reload pair of 'c's
-	"&lm	($xc,$xc_,'$stdframe+4*8+4*$c2($sp)')",
-
-	"&alr	(@x[$a2],@x[$b2])",	# Q3
-	 "&alr	(@x[$a3],@x[$b3])",	# Q4
-	"&xr	(@x[$d2],@x[$a2])",
-	 "&xr	(@x[$d3],@x[$a3])",
-	"&rll	(@x[$d2],@x[$d2],16)",
-	 "&rll	(@x[$d3],@x[$d3],16)",
-
-	"&alr	($xc,@x[$d2])",
-	 "&alr	($xc_,@x[$d3])",
-	"&xr	(@x[$b2],$xc)",
-	 "&xr	(@x[$b3],$xc_)",
-	"&rll	(@x[$b2],@x[$b2],12)",
-	 "&rll	(@x[$b3],@x[$b3],12)",
-
-	"&alr	(@x[$a2],@x[$b2])",
-	 "&alr	(@x[$a3],@x[$b3])",
-	"&xr	(@x[$d2],@x[$a2])",
-	 "&xr	(@x[$d3],@x[$a3])",
-	"&rll	(@x[$d2],@x[$d2],8)",
-	 "&rll	(@x[$d3],@x[$d3],8)",
-
-	"&alr	($xc,@x[$d2])",
-	 "&alr	($xc_,@x[$d3])",
-	"&xr	(@x[$b2],$xc)",
-	 "&xr	(@x[$b3],$xc_)",
-	"&rll	(@x[$b2],@x[$b2],7)",
-	 "&rll	(@x[$b3],@x[$b3],7)"
-	);
-}
-
-$code.=<<___;
-.text
-
-.globl	ChaCha20_ctr32
-.type	ChaCha20_ctr32,\@function
-.align	32
-ChaCha20_ctr32:
-	lt${g}r	$len,$len			# $len==0?
-	bzr	%r14
-	a${g}hi	$len,-64
-	l${g}hi	%r1,-$frame
-	stm${g}	%r6,%r15,`6*$SIZE_T`($sp)
-	sl${g}r	$out,$inp			# difference
-	la	$len,0($inp,$len)		# end of input minus 64
-	larl	%r7,.Lsigma
-	lgr	%r0,$sp
-	la	$sp,0(%r1,$sp)
-	st${g}	%r0,0($sp)
-
-	lmg	%r8,%r11,0($key)		# load key
-	lmg	%r12,%r13,0($counter)		# load counter
-	lmg	%r6,%r7,0(%r7)			# load sigma constant
-
-	la	%r14,0($inp)
-	st${g}	$out,$frame+3*$SIZE_T($sp)
-	st${g}	$len,$frame+4*$SIZE_T($sp)
-	stmg	%r6,%r13,$stdframe($sp)		# copy key schedule to stack
-	srlg	@x[12],%r12,32			# 32-bit counter value
-	j	.Loop_outer
-
-.align	16
-.Loop_outer:
-	lm	@x[0],@x[7],$stdframe+4*0($sp)		# load x[0]-x[7]
-	lm	@t[0],@t[1],$stdframe+4*10($sp)		# load x[10]-x[11]
-	lm	@x[13],@x[15],$stdframe+4*13($sp)	# load x[13]-x[15]
-	stm	@t[0],@t[1],$stdframe+4*8+4*10($sp)	# offload x[10]-x[11]
-	lm	@t[0],@t[1],$stdframe+4*8($sp)		# load x[8]-x[9]
-	st	@x[12],$stdframe+4*12($sp)		# save counter
-	st${g}	%r14,$frame+2*$SIZE_T($sp)		# save input pointer
-	lhi	%r14,10
-	j	.Loop
-
-.align	4
-.Loop:
-___
-	foreach (&ROUND(0, 4, 8,12)) { eval; }
-	foreach (&ROUND(0, 5,10,15)) { eval; }
-$code.=<<___;
-	brct	%r14,.Loop
-
-	l${g}	%r14,$frame+2*$SIZE_T($sp)		# pull input pointer
-	stm	@t[0],@t[1],$stdframe+4*8+4*8($sp)	# offload x[8]-x[9]
-	lm${g}	@t[0],@t[1],$frame+3*$SIZE_T($sp)
-
-	al	@x[0],$stdframe+4*0($sp)	# accumulate key schedule
-	al	@x[1],$stdframe+4*1($sp)
-	al	@x[2],$stdframe+4*2($sp)
-	al	@x[3],$stdframe+4*3($sp)
-	al	@x[4],$stdframe+4*4($sp)
-	al	@x[5],$stdframe+4*5($sp)
-	al	@x[6],$stdframe+4*6($sp)
-	al	@x[7],$stdframe+4*7($sp)
-	lrvr	@x[0],@x[0]
-	lrvr	@x[1],@x[1]
-	lrvr	@x[2],@x[2]
-	lrvr	@x[3],@x[3]
-	lrvr	@x[4],@x[4]
-	lrvr	@x[5],@x[5]
-	lrvr	@x[6],@x[6]
-	lrvr	@x[7],@x[7]
-	al	@x[12],$stdframe+4*12($sp)
-	al	@x[13],$stdframe+4*13($sp)
-	al	@x[14],$stdframe+4*14($sp)
-	al	@x[15],$stdframe+4*15($sp)
-	lrvr	@x[12],@x[12]
-	lrvr	@x[13],@x[13]
-	lrvr	@x[14],@x[14]
-	lrvr	@x[15],@x[15]
-
-	la	@t[0],0(@t[0],%r14)		# reconstruct output pointer
-	cl${g}r	%r14,@t[1]
-	jh	.Ltail
-
-	x	@x[0],4*0(%r14)			# xor with input
-	x	@x[1],4*1(%r14)
-	st	@x[0],4*0(@t[0])		# store output
-	x	@x[2],4*2(%r14)
-	st	@x[1],4*1(@t[0])
-	x	@x[3],4*3(%r14)
-	st	@x[2],4*2(@t[0])
-	x	@x[4],4*4(%r14)
-	st	@x[3],4*3(@t[0])
-	 lm	@x[0],@x[3],$stdframe+4*8+4*8($sp)	# load x[8]-x[11]
-	x	@x[5],4*5(%r14)
-	st	@x[4],4*4(@t[0])
-	x	@x[6],4*6(%r14)
-	 al	@x[0],$stdframe+4*8($sp)
-	st	@x[5],4*5(@t[0])
-	x	@x[7],4*7(%r14)
-	 al	@x[1],$stdframe+4*9($sp)
-	st	@x[6],4*6(@t[0])
-	x	@x[12],4*12(%r14)
-	 al	@x[2],$stdframe+4*10($sp)
-	st	@x[7],4*7(@t[0])
-	x	@x[13],4*13(%r14)
-	 al	@x[3],$stdframe+4*11($sp)
-	st	@x[12],4*12(@t[0])
-	x	@x[14],4*14(%r14)
-	st	@x[13],4*13(@t[0])
-	x	@x[15],4*15(%r14)
-	st	@x[14],4*14(@t[0])
-	 lrvr	@x[0],@x[0]
-	st	@x[15],4*15(@t[0])
-	 lrvr	@x[1],@x[1]
-	 lrvr	@x[2],@x[2]
-	 lrvr	@x[3],@x[3]
-	lhi	@x[12],1
-	 x	@x[0],4*8(%r14)
-	al	@x[12],$stdframe+4*12($sp)	# increment counter
-	 x	@x[1],4*9(%r14)
-	 st	@x[0],4*8(@t[0])
-	 x	@x[2],4*10(%r14)
-	 st	@x[1],4*9(@t[0])
-	 x	@x[3],4*11(%r14)
-	 st	@x[2],4*10(@t[0])
-	 st	@x[3],4*11(@t[0])
-
-	cl${g}r	%r14,@t[1]			# done yet?
-	la	%r14,64(%r14)
-	jl	.Loop_outer
-
-.Ldone:
-	xgr	%r0,%r0
-	xgr	%r1,%r1
-	xgr	%r2,%r2
-	xgr	%r3,%r3
-	stmg	%r0,%r3,$stdframe+4*4($sp)	# wipe key copy
-	stmg	%r0,%r3,$stdframe+4*12($sp)
-
-	lm${g}	%r6,%r15,`$frame+6*$SIZE_T`($sp)
-	br	%r14
-
-.align	16
-.Ltail:
-	la	@t[1],64($t[1])
-	stm	@x[0],@x[7],$stdframe+4*0($sp)
-	sl${g}r	@t[1],%r14
-	lm	@x[0],@x[3],$stdframe+4*8+4*8($sp)
-	l${g}hi	@x[6],0
-	stm	@x[12],@x[15],$stdframe+4*12($sp)
-	al	@x[0],$stdframe+4*8($sp)
-	al	@x[1],$stdframe+4*9($sp)
-	al	@x[2],$stdframe+4*10($sp)
-	al	@x[3],$stdframe+4*11($sp)
-	lrvr	@x[0],@x[0]
-	lrvr	@x[1],@x[1]
-	lrvr	@x[2],@x[2]
-	lrvr	@x[3],@x[3]
-	stm	@x[0],@x[3],$stdframe+4*8($sp)
-
-.Loop_tail:
-	llgc	@x[4],0(@x[6],%r14)
-	llgc	@x[5],$stdframe(@x[6],$sp)
-	xr	@x[5],@x[4]
-	stc	@x[5],0(@x[6],@t[0])
-	la	@x[6],1(@x[6])
-	brct	@t[1],.Loop_tail
-
-	j	.Ldone
-.size	ChaCha20_ctr32,.-ChaCha20_ctr32
-
-.align	32
-.Lsigma:
-.long	0x61707865,0x3320646e,0x79622d32,0x6b206574	# endian-neutral
-.asciz	"ChaCha20 for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-.align	4
-___
+	alr	(@x[$a0],@x[$b0]);	# Q1
+	 alr	(@x[$a1],@x[$b1]);	# Q2
+	xr	(@x[$d0],@x[$a0]);
+	 xr	(@x[$d1],@x[$a1]);
+	rll	(@x[$d0],@x[$d0],16);
+	 rll	(@x[$d1],@x[$d1],16);
+
+	alr	($xc,@x[$d0]);
+	 alr	($xc_,@x[$d1]);
+	xr	(@x[$b0],$xc);
+	 xr	(@x[$b1],$xc_);
+	rll	(@x[$b0],@x[$b0],12);
+	 rll	(@x[$b1],@x[$b1],12);
+
+	alr	(@x[$a0],@x[$b0]);
+	 alr	(@x[$a1],@x[$b1]);
+	xr	(@x[$d0],@x[$a0]);
+	 xr	(@x[$d1],@x[$a1]);
+	rll	(@x[$d0],@x[$d0],8);
+	 rll	(@x[$d1],@x[$d1],8);
+
+	alr	($xc,@x[$d0]);
+	 alr	($xc_,@x[$d1]);
+	xr	(@x[$b0],$xc);
+	 xr	(@x[$b1],$xc_);
+	rll	(@x[$b0],@x[$b0],7);
+	 rll	(@x[$b1],@x[$b1],7);
+
+	stm	($xc,$xc_,"$stdframe+4*8+4*$c0($sp)");	# reload pair of 'c's
+	lm	($xc,$xc_,"$stdframe+4*8+4*$c2($sp)");
+
+	alr	(@x[$a2],@x[$b2]);	# Q3
+	 alr	(@x[$a3],@x[$b3]);	# Q4
+	xr	(@x[$d2],@x[$a2]);
+	 xr	(@x[$d3],@x[$a3]);
+	rll	(@x[$d2],@x[$d2],16);
+	 rll	(@x[$d3],@x[$d3],16);
+
+	alr	($xc,@x[$d2]);
+	 alr	($xc_,@x[$d3]);
+	xr	(@x[$b2],$xc);
+	 xr	(@x[$b3],$xc_);
+	rll	(@x[$b2],@x[$b2],12);
+	 rll	(@x[$b3],@x[$b3],12);
+
+	alr	(@x[$a2],@x[$b2]);
+	 alr	(@x[$a3],@x[$b3]);
+	xr	(@x[$d2],@x[$a2]);
+	 xr	(@x[$d3],@x[$a3]);
+	rll	(@x[$d2],@x[$d2],8);
+	 rll	(@x[$d3],@x[$d3],8);
+
+	alr	($xc,@x[$d2]);
+	 alr	($xc_,@x[$d3]);
+	xr	(@x[$b2],$xc);
+	 xr	(@x[$b3],$xc_);
+	rll	(@x[$b2],@x[$b2],7);
+	 rll	(@x[$b3],@x[$b3],7);
+}
+
+sub VX_lane_ROUND {
+my ($a0,$b0,$c0,$d0)=@_;
+my ($a1,$b1,$c1,$d1)=map(($_&~3)+(($_+1)&3),($a0,$b0,$c0,$d0));
+my ($a2,$b2,$c2,$d2)=map(($_&~3)+(($_+1)&3),($a1,$b1,$c1,$d1));
+my ($a3,$b3,$c3,$d3)=map(($_&~3)+(($_+1)&3),($a2,$b2,$c2,$d2));
+my @x=map("%v$_",(0..15));
 
-foreach (split("\n",$code)) {
-	s/\`([^\`]*)\`/eval $1/ge;
+	vaf	(@x[$a0],@x[$a0],@x[$b0]);	# Q1
+	vx	(@x[$d0],@x[$d0],@x[$a0]);
+	verllf	(@x[$d0],@x[$d0],16);
+	vaf	(@x[$a1],@x[$a1],@x[$b1]);	# Q2
+	vx	(@x[$d1],@x[$d1],@x[$a1]);
+	verllf	(@x[$d1],@x[$d1],16);
+	vaf	(@x[$a2],@x[$a2],@x[$b2]);	# Q3
+	vx	(@x[$d2],@x[$d2],@x[$a2]);
+	verllf	(@x[$d2],@x[$d2],16);
+	vaf	(@x[$a3],@x[$a3],@x[$b3]);	# Q4
+	vx	(@x[$d3],@x[$d3],@x[$a3]);
+	verllf	(@x[$d3],@x[$d3],16);
+
+	vaf	(@x[$c0],@x[$c0],@x[$d0]);
+	vx	(@x[$b0],@x[$b0],@x[$c0]);
+	verllf	(@x[$b0],@x[$b0],12);
+	vaf	(@x[$c1],@x[$c1],@x[$d1]);
+	vx	(@x[$b1],@x[$b1],@x[$c1]);
+	verllf	(@x[$b1],@x[$b1],12);
+	vaf	(@x[$c2],@x[$c2],@x[$d2]);
+	vx	(@x[$b2],@x[$b2],@x[$c2]);
+	verllf	(@x[$b2],@x[$b2],12);
+	vaf	(@x[$c3],@x[$c3],@x[$d3]);
+	vx	(@x[$b3],@x[$b3],@x[$c3]);
+	verllf	(@x[$b3],@x[$b3],12);
+
+	vaf	(@x[$a0],@x[$a0],@x[$b0]);
+	vx	(@x[$d0],@x[$d0],@x[$a0]);
+	verllf	(@x[$d0],@x[$d0],8);
+	vaf	(@x[$a1],@x[$a1],@x[$b1]);
+	vx	(@x[$d1],@x[$d1],@x[$a1]);
+	verllf	(@x[$d1],@x[$d1],8);
+	vaf	(@x[$a2],@x[$a2],@x[$b2]);
+	vx	(@x[$d2],@x[$d2],@x[$a2]);
+	verllf	(@x[$d2],@x[$d2],8);
+	vaf	(@x[$a3],@x[$a3],@x[$b3]);
+	vx	(@x[$d3],@x[$d3],@x[$a3]);
+	verllf	(@x[$d3],@x[$d3],8);
+
+	vaf	(@x[$c0],@x[$c0],@x[$d0]);
+	vx	(@x[$b0],@x[$b0],@x[$c0]);
+	verllf	(@x[$b0],@x[$b0],7);
+	vaf	(@x[$c1],@x[$c1],@x[$d1]);
+	vx	(@x[$b1],@x[$b1],@x[$c1]);
+	verllf	(@x[$b1],@x[$b1],7);
+	vaf	(@x[$c2],@x[$c2],@x[$d2]);
+	vx	(@x[$b2],@x[$b2],@x[$c2]);
+	verllf	(@x[$b2],@x[$b2],7);
+	vaf	(@x[$c3],@x[$c3],@x[$d3]);
+	vx	(@x[$b3],@x[$b3],@x[$c3]);
+	verllf	(@x[$b3],@x[$b3],7);
+}
 
-	print $_,"\n";
+sub VX_ROUND {
+my @a=@_[0..5];
+my @b=@_[6..11];
+my @c=@_[12..17];
+my @d=@_[18..23];
+my $odd=@_[24];
+
+	vaf		(@a[$_],@a[$_],@b[$_]) for (0..5);
+	vx		(@d[$_],@d[$_],@a[$_]) for (0..5);
+	verllf		(@d[$_],@d[$_],16) for (0..5);
+
+	vaf		(@c[$_],@c[$_],@d[$_]) for (0..5);
+	vx		(@b[$_],@b[$_],@c[$_]) for (0..5);
+	verllf		(@b[$_],@b[$_],12) for (0..5);
+
+	vaf		(@a[$_],@a[$_],@b[$_]) for (0..5);
+	vx		(@d[$_],@d[$_],@a[$_]) for (0..5);
+	verllf		(@d[$_],@d[$_],8) for (0..5);
+
+	vaf		(@c[$_],@c[$_],@d[$_]) for (0..5);
+	vx		(@b[$_],@b[$_],@c[$_]) for (0..5);
+	verllf		(@b[$_],@b[$_],7) for (0..5);
+
+	vsldb		(@c[$_],@c[$_],@c[$_],8) for (0..5);
+	vsldb		(@b[$_],@b[$_],@b[$_],$odd?12:4) for (0..5);
+	vsldb		(@d[$_],@d[$_],@d[$_],$odd?4:12) for (0..5);
 }
-close STDOUT;
+
+PERLASM_BEGIN($output);
+
+INCLUDE	("s390x_arch.h");
+TEXT	();
+
+################
+# void ChaCha20_ctr32(unsigned char *out, const unsigned char *inp, size_t len,
+#                     const unsigned int key[8], const unsigned int counter[4])
+my ($out,$inp,$len,$key,$counter)=map("%r$_",(2..6));
+{
+my $frame=$stdframe+4*20;
+my @x=map("%r$_",(0..7,"x","x","x","x",(10..13)));
+my @t=map("%r$_",(8,9));
+
+GLOBL	("ChaCha20_ctr32");
+TYPE	("ChaCha20_ctr32","\@function");
+ALIGN	(32);
+LABEL	("ChaCha20_ctr32");
+	larl	("%r1","OPENSSL_s390xcap_P");
+
+	lghi	("%r0",64);
+&{$z?	\&ltgr:\&ltr}	($len,$len);		# len==0?
+	bzr	("%r14");
+	lg	("%r1","S390X_STFLE+16(%r1)");
+&{$z?	\&clgr:\&clr}	($len,"%r0");
+	jle	(".Lshort");
+
+	tmhh	("%r1",0x4000);			# check for vx bit
+	jnz	(".LChaCha20_ctr32_vx");
+
+LABEL	(".Lshort");
+&{$z?	\&aghi:\&ahi}	($len,-64);
+&{$z?	\&lghi:\&lhi}	("%r1",-$frame);
+&{$z?	\&stmg:\&stm}	("%r6","%r15","6*$SIZE_T($sp)");
+&{$z?	\&slgr:\&slr}	($out,$inp);	# difference
+	la	($len,"0($inp,$len)");	# end of input minus 64
+	larl	("%r7",".Lsigma");
+	lgr	("%r0",$sp);
+	la	($sp,"0(%r1,$sp)");
+&{$z?	\&stg:\&st}	("%r0","0($sp)");
+
+	lmg	("%r8","%r11","0($key)");	# load key
+	lmg	("%r12","%r13","0($counter)");	# load counter
+	lmg	("%r6","%r7","0(%r7)");	# load sigma constant
+
+	la	("%r14","0($inp)");
+&{$z?	\&stg:\&st}	($out,"$frame+3*$SIZE_T($sp)");
+&{$z?	\&stg:\&st}	($len,"$frame+4*$SIZE_T($sp)");
+	stmg	("%r6","%r13","$stdframe($sp)");# copy key schedule to stack
+	srlg	(@x[12],"%r12",32);	# 32-bit counter value
+	j	(".Loop_outer");
+
+ALIGN	(16);
+LABEL	(".Loop_outer");
+	lm	(@x[0],@x[7],"$stdframe+4*0($sp)");	# load x[0]-x[7]
+	lm	(@t[0],@t[1],"$stdframe+4*10($sp)");	# load x[10]-x[11]
+	lm	(@x[13],@x[15],"$stdframe+4*13($sp)");	# load x[13]-x[15]
+	stm	(@t[0],@t[1],"$stdframe+4*8+4*10($sp)");# offload x[10]-x[11]
+	lm	(@t[0],@t[1],"$stdframe+4*8($sp)");	# load x[8]-x[9]
+	st	(@x[12],"$stdframe+4*12($sp)");	# save counter
+&{$z?	\&stg:\&st}	("%r14","$frame+2*$SIZE_T($sp)");# save input pointer
+	lhi	("%r14",10);
+	j	(".Loop");
+
+ALIGN	(4);
+LABEL	(".Loop");
+	ROUND	(0, 4, 8,12);
+	ROUND	(0, 5,10,15);
+	brct	("%r14",".Loop");
+
+&{$z?	\&lg:\&l}	("%r14","$frame+2*$SIZE_T($sp)");# pull input pointer
+	stm	(@t[0],@t[1],"$stdframe+4*8+4*8($sp)");	# offload x[8]-x[9]
+&{$z?	\&lmg:\&lm}	(@t[0],@t[1],"$frame+3*$SIZE_T($sp)");
+
+	al	(@x[0],"$stdframe+4*0($sp)");	# accumulate key schedule
+	al	(@x[1],"$stdframe+4*1($sp)");
+	al	(@x[2],"$stdframe+4*2($sp)");
+	al	(@x[3],"$stdframe+4*3($sp)");
+	al	(@x[4],"$stdframe+4*4($sp)");
+	al	(@x[5],"$stdframe+4*5($sp)");
+	al	(@x[6],"$stdframe+4*6($sp)");
+	al	(@x[7],"$stdframe+4*7($sp)");
+	lrvr	(@x[0],@x[0]);
+	lrvr	(@x[1],@x[1]);
+	lrvr	(@x[2],@x[2]);
+	lrvr	(@x[3],@x[3]);
+	lrvr	(@x[4],@x[4]);
+	lrvr	(@x[5],@x[5]);
+	lrvr	(@x[6],@x[6]);
+	lrvr	(@x[7],@x[7]);
+	al	(@x[12],"$stdframe+4*12($sp)");
+	al	(@x[13],"$stdframe+4*13($sp)");
+	al	(@x[14],"$stdframe+4*14($sp)");
+	al	(@x[15],"$stdframe+4*15($sp)");
+	lrvr	(@x[12],@x[12]);
+	lrvr	(@x[13],@x[13]);
+	lrvr	(@x[14],@x[14]);
+	lrvr	(@x[15],@x[15]);
+
+	la	(@t[0],"0(@t[0],%r14)");	# reconstruct output pointer
+&{$z?	\&clgr:\&clr}	("%r14",@t[1]);
+	jh	(".Ltail");
+
+	x	(@x[0],"4*0(%r14)");	# xor with input
+	x	(@x[1],"4*1(%r14)");
+	st	(@x[0],"4*0(@t[0])");	# store output
+	x	(@x[2],"4*2(%r14)");
+	st	(@x[1],"4*1(@t[0])");
+	x	(@x[3],"4*3(%r14)");
+	st	(@x[2],"4*2(@t[0])");
+	x	(@x[4],"4*4(%r14)");
+	st	(@x[3],"4*3(@t[0])");
+	 lm	(@x[0],@x[3],"$stdframe+4*8+4*8($sp)");	# load x[8]-x[11]
+	x	(@x[5],"4*5(%r14)");
+	st	(@x[4],"4*4(@t[0])");
+	x	(@x[6],"4*6(%r14)");
+	 al	(@x[0],"$stdframe+4*8($sp)");
+	st	(@x[5],"4*5(@t[0])");
+	x	(@x[7],"4*7(%r14)");
+	 al	(@x[1],"$stdframe+4*9($sp)");
+	st	(@x[6],"4*6(@t[0])");
+	x	(@x[12],"4*12(%r14)");
+	 al	(@x[2],"$stdframe+4*10($sp)");
+	st	(@x[7],"4*7(@t[0])");
+	x	(@x[13],"4*13(%r14)");
+	 al	(@x[3],"$stdframe+4*11($sp)");
+	st	(@x[12],"4*12(@t[0])");
+	x	(@x[14],"4*14(%r14)");
+	st	(@x[13],"4*13(@t[0])");
+	x	(@x[15],"4*15(%r14)");
+	st	(@x[14],"4*14(@t[0])");
+	 lrvr	(@x[0],@x[0]);
+	st	(@x[15],"4*15(@t[0])");
+	 lrvr	(@x[1],@x[1]);
+	 lrvr	(@x[2],@x[2]);
+	 lrvr	(@x[3],@x[3]);
+	lhi	(@x[12],1);
+	 x	(@x[0],"4*8(%r14)");
+	al	(@x[12],"$stdframe+4*12($sp)");	# increment counter
+	 x	(@x[1],"4*9(%r14)");
+	 st	(@x[0],"4*8(@t[0])");
+	 x	(@x[2],"4*10(%r14)");
+	 st	(@x[1],"4*9(@t[0])");
+	 x	(@x[3],"4*11(%r14)");
+	 st	(@x[2],"4*10(@t[0])");
+	 st	(@x[3],"4*11(@t[0])");
+
+&{$z?	\&clgr:\&clr}	("%r14",@t[1]);	# done yet?
+	la	("%r14","64(%r14)");
+	jl	(".Loop_outer");
+
+LABEL	(".Ldone");
+	xgr	("%r0","%r0");
+	xgr	("%r1","%r1");
+	xgr	("%r2","%r2");
+	xgr	("%r3","%r3");
+	stmg	("%r0","%r3","$stdframe+4*4($sp)");	# wipe key copy
+	stmg	("%r0","%r3","$stdframe+4*12($sp)");
+
+&{$z?	\&lmg:\&lm}	("%r6","%r15","$frame+6*$SIZE_T($sp)");
+	br	("%r14");
+
+ALIGN	(16);
+LABEL	(".Ltail");
+	la	(@t[1],"64($t[1])");
+	stm	(@x[0],@x[7],"$stdframe+4*0($sp)");
+&{$z?	\&slgr:\&slr}	(@t[1],"%r14");
+	lm	(@x[0],@x[3],"$stdframe+4*8+4*8($sp)");
+&{$z?	\&lghi:\&lhi}	(@x[6],0);
+	stm	(@x[12],@x[15],"$stdframe+4*12($sp)");
+	al	(@x[0],"$stdframe+4*8($sp)");
+	al	(@x[1],"$stdframe+4*9($sp)");
+	al	(@x[2],"$stdframe+4*10($sp)");
+	al	(@x[3],"$stdframe+4*11($sp)");
+	lrvr	(@x[0],@x[0]);
+	lrvr	(@x[1],@x[1]);
+	lrvr	(@x[2],@x[2]);
+	lrvr	(@x[3],@x[3]);
+	stm	(@x[0],@x[3],"$stdframe+4*8($sp)");
+
+LABEL	(".Loop_tail");
+	llgc	(@x[4],"0(@x[6],%r14)");
+	llgc	(@x[5],"$stdframe(@x[6],$sp)");
+	xr	(@x[5],@x[4]);
+	stc	(@x[5],"0(@x[6],@t[0])");
+	la	(@x[6],"1(@x[6])");
+	brct	(@t[1],".Loop_tail");
+
+	j	(".Ldone");
+SIZE	("ChaCha20_ctr32",".-ChaCha20_ctr32");
+}
+
+########################################################################
+# 4x"vertical" layout minimizes amount of instructions, but pipeline
+# runs underutilized [because of vector instructions' high latency].
+# On the other hand minimum amount of data it takes to fully utilize
+# the pipeline is higher, so that effectively, short inputs would be
+# processed slower. Hence this code path targeting <=256 bytes lengths.
+#
+{
+my ($xa0,$xa1,$xa2,$xa3, $xb0,$xb1,$xb2,$xb3,
+    $xc0,$xc1,$xc2,$xc3, $xd0,$xd1,$xd2,$xd3)=map("%v$_",(0..15));
+my @K=map("%v$_",(16..19));
+my $CTR="%v26";
+my ($xt0,$xt1,$xt2,$xt3)=map("%v$_",(27..30));
+my $beperm="%v31";
+my ($x00,$x10,$x20,$x30)=(0,map("r$_",(8..10)));
+my $FRAME=$stdframe+4*16;
+
+ALIGN	(32);
+LABEL	("ChaCha20_ctr32_4x");
+LABEL	(".LChaCha20_ctr32_4x");
+&{$z?	\&stmg:\&stm}	("%r6","%r7","6*$SIZE_T($sp)");
+if (!$z) {
+	std	("%f4","16*$SIZE_T+2*8($sp)");
+	std	("%f6","16*$SIZE_T+3*8($sp)");
+}
+&{$z?	\&lghi:\&lhi}	("%r1",-$FRAME);
+	lgr	("%r0",$sp);
+	la	($sp,"0(%r1,$sp)");
+&{$z?	\&stg:\&st}	("%r0","0($sp)");	# back-chain
+if ($z) {
+	std	("%f8","$stdframe+8*0($sp)");
+	std	("%f9","$stdframe+8*1($sp)");
+	std	("%f10","$stdframe+8*2($sp)");
+	std	("%f11","$stdframe+8*3($sp)");
+	std	("%f12","$stdframe+8*4($sp)");
+	std	("%f13","$stdframe+8*5($sp)");
+	std	("%f14","$stdframe+8*6($sp)");
+	std	("%f15","$stdframe+8*7($sp)");
+}
+	larl	("%r7",".Lsigma");
+	lhi	("%r0",10);
+	lhi	("%r1",0);
+
+	vl	(@K[0],"0(%r7)");		# load sigma
+	vl	(@K[1],"0($key)");		# load key
+	vl	(@K[2],"16($key)");
+	vl	(@K[3],"0($counter)");		# load counter
+
+	vl	($beperm,"0x40(%r7)");
+	vl	($xt1,"0x50(%r7)");
+	vrepf	($CTR,@K[3],0);
+	vlvgf	(@K[3],"%r1",0);		# clear @K[3].word[0]
+	vaf	($CTR,$CTR,$xt1);
+
+#LABEL	(".Loop_outer_4x");
+	vlm	($xa0,$xa3,"0x60(%r7)");	# load [smashed] sigma
+
+	vrepf	($xb0,@K[1],0);			# smash the key
+	vrepf	($xb1,@K[1],1);
+	vrepf	($xb2,@K[1],2);
+	vrepf	($xb3,@K[1],3);
+
+	vrepf	($xc0,@K[2],0);
+	vrepf	($xc1,@K[2],1);
+	vrepf	($xc2,@K[2],2);
+	vrepf	($xc3,@K[2],3);
+
+	vlr	($xd0,$CTR);
+	vrepf	($xd1,@K[3],1);
+	vrepf	($xd2,@K[3],2);
+	vrepf	($xd3,@K[3],3);
+
+LABEL	(".Loop_4x");
+	VX_lane_ROUND(0, 4, 8,12);
+	VX_lane_ROUND(0, 5,10,15);
+	brct	("%r0",".Loop_4x");
+
+	vaf	($xd0,$xd0,$CTR);
+
+	vmrhf	($xt0,$xa0,$xa1);		# transpose data
+	vmrhf	($xt1,$xa2,$xa3);
+	vmrlf	($xt2,$xa0,$xa1);
+	vmrlf	($xt3,$xa2,$xa3);
+	vpdi	($xa0,$xt0,$xt1,0b0000);
+	vpdi	($xa1,$xt0,$xt1,0b0101);
+	vpdi	($xa2,$xt2,$xt3,0b0000);
+	vpdi	($xa3,$xt2,$xt3,0b0101);
+
+	vmrhf	($xt0,$xb0,$xb1);
+	vmrhf	($xt1,$xb2,$xb3);
+	vmrlf	($xt2,$xb0,$xb1);
+	vmrlf	($xt3,$xb2,$xb3);
+	vpdi	($xb0,$xt0,$xt1,0b0000);
+	vpdi	($xb1,$xt0,$xt1,0b0101);
+	vpdi	($xb2,$xt2,$xt3,0b0000);
+	vpdi	($xb3,$xt2,$xt3,0b0101);
+
+	vmrhf	($xt0,$xc0,$xc1);
+	vmrhf	($xt1,$xc2,$xc3);
+	vmrlf	($xt2,$xc0,$xc1);
+	vmrlf	($xt3,$xc2,$xc3);
+	vpdi	($xc0,$xt0,$xt1,0b0000);
+	vpdi	($xc1,$xt0,$xt1,0b0101);
+	vpdi	($xc2,$xt2,$xt3,0b0000);
+	vpdi	($xc3,$xt2,$xt3,0b0101);
+
+	vmrhf	($xt0,$xd0,$xd1);
+	vmrhf	($xt1,$xd2,$xd3);
+	vmrlf	($xt2,$xd0,$xd1);
+	vmrlf	($xt3,$xd2,$xd3);
+	vpdi	($xd0,$xt0,$xt1,0b0000);
+	vpdi	($xd1,$xt0,$xt1,0b0101);
+	vpdi	($xd2,$xt2,$xt3,0b0000);
+	vpdi	($xd3,$xt2,$xt3,0b0101);
+
+	#vrepif	($xt0,4);
+	#vaf	($CTR,$CTR,$xt0);		# next counter value
+
+	vaf	($xa0,$xa0,@K[0]);
+	vaf	($xb0,$xb0,@K[1]);
+	vaf	($xc0,$xc0,@K[2]);
+	vaf	($xd0,$xd0,@K[3]);
+
+	vperm	($xa0,$xa0,$xa0,$beperm);
+	vperm	($xb0,$xb0,$xb0,$beperm);
+	vperm	($xc0,$xc0,$xc0,$beperm);
+	vperm	($xd0,$xd0,$xd0,$beperm);
+
+	#&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	#jl	(".Ltail_4x");
+
+	vlm	($xt0,$xt3,"0($inp)");
+
+	vx	($xt0,$xt0,$xa0);
+	vx	($xt1,$xt1,$xb0);
+	vx	($xt2,$xt2,$xc0);
+	vx	($xt3,$xt3,$xd0);
+
+	vstm	($xt0,$xt3,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	#je	(".Ldone_4x");
+
+	vaf	($xa0,$xa1,@K[0]);
+	vaf	($xb0,$xb1,@K[1]);
+	vaf	($xc0,$xc1,@K[2]);
+	vaf	($xd0,$xd1,@K[3]);
+
+	vperm	($xa0,$xa0,$xa0,$beperm);
+	vperm	($xb0,$xb0,$xb0,$beperm);
+	vperm	($xc0,$xc0,$xc0,$beperm);
+	vperm	($xd0,$xd0,$xd0,$beperm);
+
+&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	jl	(".Ltail_4x");
+
+	vlm	($xt0,$xt3,"0($inp)");
+
+	vx	($xt0,$xt0,$xa0);
+	vx	($xt1,$xt1,$xb0);
+	vx	($xt2,$xt2,$xc0);
+	vx	($xt3,$xt3,$xd0);
+
+	vstm	($xt0,$xt3,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_4x");
+
+	vaf	($xa0,$xa2,@K[0]);
+	vaf	($xb0,$xb2,@K[1]);
+	vaf	($xc0,$xc2,@K[2]);
+	vaf	($xd0,$xd2,@K[3]);
+
+	vperm	($xa0,$xa0,$xa0,$beperm);
+	vperm	($xb0,$xb0,$xb0,$beperm);
+	vperm	($xc0,$xc0,$xc0,$beperm);
+	vperm	($xd0,$xd0,$xd0,$beperm);
+
+&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	jl	(".Ltail_4x");
+
+	vlm	($xt0,$xt3,"0($inp)");
+
+	vx	($xt0,$xt0,$xa0);
+	vx	($xt1,$xt1,$xb0);
+	vx	($xt2,$xt2,$xc0);
+	vx	($xt3,$xt3,$xd0);
+
+	vstm	($xt0,$xt3,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_4x");
+
+	vaf	($xa0,$xa3,@K[0]);
+	vaf	($xb0,$xb3,@K[1]);
+	vaf	($xc0,$xc3,@K[2]);
+	vaf	($xd0,$xd3,@K[3]);
+
+	vperm	($xa0,$xa0,$xa0,$beperm);
+	vperm	($xb0,$xb0,$xb0,$beperm);
+	vperm	($xc0,$xc0,$xc0,$beperm);
+	vperm	($xd0,$xd0,$xd0,$beperm);
+
+&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	jl	(".Ltail_4x");
+
+	vlm	($xt0,$xt3,"0($inp)");
+
+	vx	($xt0,$xt0,$xa0);
+	vx	($xt1,$xt1,$xb0);
+	vx	($xt2,$xt2,$xc0);
+	vx	($xt3,$xt3,$xd0);
+
+	vstm	($xt0,$xt3,"0($out)");
+
+	#la	$inp,0x40($inp));
+	#la	$out,0x40($out));
+	#lhi	%r0,10);
+	#&{$z?	\&aghi:\&ahi}	$len,-0x40);
+	#jne	.Loop_outer_4x);
+
+LABEL	(".Ldone_4x");
+if (!$z) {
+	ld	("%f4","$FRAME+16*$SIZE_T+2*8($sp)");
+	ld	("%f6","$FRAME+16*$SIZE_T+3*8($sp)");
+} else {
+	ld	("%f8","$stdframe+8*0($sp)");
+	ld	("%f9","$stdframe+8*1($sp)");
+	ld	("%f10","$stdframe+8*2($sp)");
+	ld	("%f11","$stdframe+8*3($sp)");
+	ld	("%f12","$stdframe+8*4($sp)");
+	ld	("%f13","$stdframe+8*5($sp)");
+	ld	("%f14","$stdframe+8*6($sp)");
+	ld	("%f15","$stdframe+8*7($sp)");
+}
+&{$z?	\&lmg:\&lm}	("%r6","%r7","$FRAME+6*$SIZE_T($sp)");
+	la	($sp,"$FRAME($sp)");
+	br	("%r14");
+
+ALIGN	(16);
+LABEL	(".Ltail_4x");
+if (!$z) {
+	vlr	($xt0,$xb0);
+	ld	("%f4","$FRAME+16*$SIZE_T+2*8($sp)");
+	ld	("%f6","$FRAME+16*$SIZE_T+3*8($sp)");
+
+	vst	($xa0,"$stdframe+0x00($sp)");
+	vst	($xt0,"$stdframe+0x10($sp)");
+	vst	($xc0,"$stdframe+0x20($sp)");
+	vst	($xd0,"$stdframe+0x30($sp)");
+} else {
+	vlr	($xt0,$xc0);
+	ld	("%f8","$stdframe+8*0($sp)");
+	ld	("%f9","$stdframe+8*1($sp)");
+	ld	("%f10","$stdframe+8*2($sp)");
+	ld	("%f11","$stdframe+8*3($sp)");
+	vlr	($xt1,$xd0);
+	ld	("%f12","$stdframe+8*4($sp)");
+	ld	("%f13","$stdframe+8*5($sp)");
+	ld	("%f14","$stdframe+8*6($sp)");
+	ld	("%f15","$stdframe+8*7($sp)");
+
+	vst	($xa0,"$stdframe+0x00($sp)");
+	vst	($xb0,"$stdframe+0x10($sp)");
+	vst	($xt0,"$stdframe+0x20($sp)");
+	vst	($xt1,"$stdframe+0x30($sp)");
+}
+	lghi	("%r1",0);
+
+LABEL	(".Loop_tail_4x");
+	llgc	("%r5","0(%r1,$inp)");
+	llgc	("%r6","$stdframe(%r1,$sp)");
+	xr	("%r6","%r5");
+	stc	("%r6","0(%r1,$out)");
+	la	("%r1","1(%r1)");
+	brct	($len,".Loop_tail_4x");
+
+&{$z?	\&lmg:\&lm}	("%r6","%r7","$FRAME+6*$SIZE_T($sp)");
+	la	($sp,"$FRAME($sp)");
+	br	("%r14");
+SIZE	("ChaCha20_ctr32_4x",".-ChaCha20_ctr32_4x");
+}
+
+########################################################################
+# 6x"horizontal" layout is optimal fit for the platform in its current
+# shape, more specifically for given vector instructions' latency. Well,
+# computational part of 8x"vertical" would be faster, but it consumes
+# all registers and dealing with that will diminish the return...
+#
+{
+my ($a0,$b0,$c0,$d0, $a1,$b1,$c1,$d1,
+    $a2,$b2,$c2,$d2, $a3,$b3,$c3,$d3,
+    $a4,$b4,$c4,$d4, $a5,$b5,$c5,$d5)=map("%v$_",(0..23));
+my @K=map("%v$_",(27,24..26));
+my ($t0,$t1,$t2,$t3)=map("%v$_",27..30);
+my $beperm="%v31";
+my $FRAME=$stdframe + 4*16;
+
+GLOBL	("ChaCha20_ctr32_vx");
+ALIGN	(32);
+LABEL	("ChaCha20_ctr32_vx");
+LABEL	(".LChaCha20_ctr32_vx");
+&{$z?	\&clgfi:\&clfi}	($len,256);
+	jle	(".LChaCha20_ctr32_4x");
+&{$z?	\&stmg:\&stm}	("%r6","%r7","6*$SIZE_T($sp)");
+if (!$z) {
+	std	("%f4","16*$SIZE_T+2*8($sp)");
+	std	("%f6","16*$SIZE_T+3*8($sp)");
+}
+&{$z?	\&lghi:\&lhi}	("%r1",-$FRAME);
+	lgr	("%r0",$sp);
+	la	($sp,"0(%r1,$sp)");
+&{$z?	\&stg:\&st}	("%r0","0($sp)");	# back-chain
+if ($z) {
+	std	("%f8","$FRAME-8*8($sp)");
+	std	("%f9","$FRAME-8*7($sp)");
+	std	("%f10","$FRAME-8*6($sp)");
+	std	("%f11","$FRAME-8*5($sp)");
+	std	("%f12","$FRAME-8*4($sp)");
+	std	("%f13","$FRAME-8*3($sp)");
+	std	("%f14","$FRAME-8*2($sp)");
+	std	("%f15","$FRAME-8*1($sp)");
+}
+	larl	("%r7",".Lsigma");
+	lhi	("%r0",10);
+
+	vlm	(@K[1],@K[2],"0($key)");	# load key
+	vl	(@K[3],"0($counter)");		# load counter
+
+	vlm	(@K[0],"$beperm","0(%r7)");	# load sigma, increments, ...
+
+LABEL	(".Loop_outer_vx");
+	vlr	($a0,@K[0]);
+	vlr	($b0,@K[1]);
+	vlr	($a1,@K[0]);
+	vlr	($b1,@K[1]);
+	vlr	($a2,@K[0]);
+	vlr	($b2,@K[1]);
+	vlr	($a3,@K[0]);
+	vlr	($b3,@K[1]);
+	vlr	($a4,@K[0]);
+	vlr	($b4,@K[1]);
+	vlr	($a5,@K[0]);
+	vlr	($b5,@K[1]);
+
+	vlr	($d0,@K[3]);
+	vaf	($d1,@K[3],$t1);		# K[3]+1
+	vaf	($d2,@K[3],$t2);		# K[3]+2
+	vaf	($d3,@K[3],$t3);		# K[3]+3
+	vaf	($d4,$d2,$t2);			# K[3]+4
+	vaf	($d5,$d2,$t3);			# K[3]+5
+
+	vlr	($c0,@K[2]);
+	vlr	($c1,@K[2]);
+	vlr	($c2,@K[2]);
+	vlr	($c3,@K[2]);
+	vlr	($c4,@K[2]);
+	vlr	($c5,@K[2]);
+
+	vlr	($t1,$d1);
+	vlr	($t2,$d2);
+	vlr	($t3,$d3);
+
+ALIGN	(4);
+LABEL	(".Loop_vx");
+
+	VX_ROUND($a0,$a1,$a2,$a3,$a4,$a5,
+		 $b0,$b1,$b2,$b3,$b4,$b5,
+		 $c0,$c1,$c2,$c3,$c4,$c5,
+		 $d0,$d1,$d2,$d3,$d4,$d5,
+		 0);
+
+	VX_ROUND($a0,$a1,$a2,$a3,$a4,$a5,
+		 $b0,$b1,$b2,$b3,$b4,$b5,
+		 $c0,$c1,$c2,$c3,$c4,$c5,
+		 $d0,$d1,$d2,$d3,$d4,$d5,
+		 1);
+
+	brct	("%r0",".Loop_vx");
+
+	vaf	($a0,$a0,@K[0]);
+	vaf	($b0,$b0,@K[1]);
+	vaf	($c0,$c0,@K[2]);
+	vaf	($d0,$d0,@K[3]);
+	vaf	($a1,$a1,@K[0]);
+	vaf	($d1,$d1,$t1);			# +K[3]+1
+
+	vperm	($a0,$a0,$a0,$beperm);
+	vperm	($b0,$b0,$b0,$beperm);
+	vperm	($c0,$c0,$c0,$beperm);
+	vperm	($d0,$d0,$d0,$beperm);
+
+&{$z?	\&clgfi:\&clfi}	($len,0x40);
+	jl	(".Ltail_vx");
+
+	vaf	($d2,$d2,$t2);			# +K[3]+2
+	vaf	($d3,$d3,$t3);			# +K[3]+3
+	vlm	($t0,$t3,"0($inp)");
+
+	vx	($a0,$a0,$t0);
+	vx	($b0,$b0,$t1);
+	vx	($c0,$c0,$t2);
+	vx	($d0,$d0,$t3);
+
+	vlm	(@K[0],$t3,"0(%r7)");		# re-load sigma and increments
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_vx");
+
+	vaf	($b1,$b1,@K[1]);
+	vaf	($c1,$c1,@K[2]);
+
+	vperm	($a0,$a1,$a1,$beperm);
+	vperm	($b0,$b1,$b1,$beperm);
+	vperm	($c0,$c1,$c1,$beperm);
+	vperm	($d0,$d1,$d1,$beperm);
+
+&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	jl	(".Ltail_vx");
+
+	vlm	($a1,$d1,"0($inp)");
+
+	vx	($a0,$a0,$a1);
+	vx	($b0,$b0,$b1);
+	vx	($c0,$c0,$c1);
+	vx	($d0,$d0,$d1);
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_vx");
+
+	vaf	($a2,$a2,@K[0]);
+	vaf	($b2,$b2,@K[1]);
+	vaf	($c2,$c2,@K[2]);
+
+	vperm	($a0,$a2,$a2,$beperm);
+	vperm	($b0,$b2,$b2,$beperm);
+	vperm	($c0,$c2,$c2,$beperm);
+	vperm	($d0,$d2,$d2,$beperm);
+
+&{$z?	\&clgfi:\&clfi}	($len,0x40);
+	jl	(".Ltail_vx");
+
+	vlm	($a1,$d1,"0($inp)");
+
+	vx	($a0,$a0,$a1);
+	vx	($b0,$b0,$b1);
+	vx	($c0,$c0,$c1);
+	vx	($d0,$d0,$d1);
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_vx");
+
+	vaf	($a3,$a3,@K[0]);
+	vaf	($b3,$b3,@K[1]);
+	vaf	($c3,$c3,@K[2]);
+	vaf	($d2,@K[3],$t3);		# K[3]+3
+
+	vperm	($a0,$a3,$a3,$beperm);
+	vperm	($b0,$b3,$b3,$beperm);
+	vperm	($c0,$c3,$c3,$beperm);
+	vperm	($d0,$d3,$d3,$beperm);
+
+&{$z?	\&clgfi:\&clfi}	($len,0x40);
+	jl	(".Ltail_vx");
+
+	vaf	($d3,$d2,$t1);			# K[3]+4
+	vlm	($a1,$d1,"0($inp)");
+
+	vx	($a0,$a0,$a1);
+	vx	($b0,$b0,$b1);
+	vx	($c0,$c0,$c1);
+	vx	($d0,$d0,$d1);
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_vx");
+
+	vaf	($a4,$a4,@K[0]);
+	vaf	($b4,$b4,@K[1]);
+	vaf	($c4,$c4,@K[2]);
+	vaf	($d4,$d4,$d3);			# +K[3]+4
+	vaf	($d3,$d3,$t1);			# K[3]+5
+	vaf	(@K[3],$d2,$t3);		# K[3]+=6
+
+	vperm	($a0,$a4,$a4,$beperm);
+	vperm	($b0,$b4,$b4,$beperm);
+	vperm	($c0,$c4,$c4,$beperm);
+	vperm	($d0,$d4,$d4,$beperm);
+
+&{$z?	\&clgfi:\&clfi}	($len,0x40);
+	jl	(".Ltail_vx");
+
+	vlm	($a1,$d1,"0($inp)");
+
+	vx	($a0,$a0,$a1);
+	vx	($b0,$b0,$b1);
+	vx	($c0,$c0,$c1);
+	vx	($d0,$d0,$d1);
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	je	(".Ldone_vx");
+
+	vaf	($a5,$a5,@K[0]);
+	vaf	($b5,$b5,@K[1]);
+	vaf	($c5,$c5,@K[2]);
+	vaf	($d5,$d5,$d3);			# +K[3]+5
+
+	vperm	($a0,$a5,$a5,$beperm);
+	vperm	($b0,$b5,$b5,$beperm);
+	vperm	($c0,$c5,$c5,$beperm);
+	vperm	($d0,$d5,$d5,$beperm);
+
+&{$z?	\&clgfi:\&clfi} ($len,0x40);
+	jl	(".Ltail_vx");
+
+	vlm	($a1,$d1,"0($inp)");
+
+	vx	($a0,$a0,$a1);
+	vx	($b0,$b0,$b1);
+	vx	($c0,$c0,$c1);
+	vx	($d0,$d0,$d1);
+
+	vstm	($a0,$d0,"0($out)");
+
+	la	($inp,"0x40($inp)");
+	la	($out,"0x40($out)");
+	lhi	("%r0",10);
+&{$z?	\&aghi:\&ahi}	($len,-0x40);
+	jne	(".Loop_outer_vx");
+
+LABEL	(".Ldone_vx");
+if (!$z) {
+	ld	("%f4","$FRAME+16*$SIZE_T+2*8($sp)");
+	ld	("%f6","$FRAME+16*$SIZE_T+3*8($sp)");
+} else {
+	ld	("%f8","$FRAME-8*8($sp)");
+	ld	("%f9","$FRAME-8*7($sp)");
+	ld	("%f10","$FRAME-8*6($sp)");
+	ld	("%f11","$FRAME-8*5($sp)");
+	ld	("%f12","$FRAME-8*4($sp)");
+	ld	("%f13","$FRAME-8*3($sp)");
+	ld	("%f14","$FRAME-8*2($sp)");
+	ld	("%f15","$FRAME-8*1($sp)");
+}
+&{$z?	\&lmg:\&lm}	("%r6","%r7","$FRAME+6*$SIZE_T($sp)");
+	la	($sp,"$FRAME($sp)");
+	br	("%r14");
+
+ALIGN	(16);
+LABEL	(".Ltail_vx");
+if (!$z) {
+	ld	("%f4","$FRAME+16*$SIZE_T+2*8($sp)");
+	ld	("%f6","$FRAME+16*$SIZE_T+3*8($sp)");
+} else {
+	ld	("%f8","$FRAME-8*8($sp)");
+	ld	("%f9","$FRAME-8*7($sp)");
+	ld	("%f10","$FRAME-8*6($sp)");
+	ld	("%f11","$FRAME-8*5($sp)");
+	ld	("%f12","$FRAME-8*4($sp)");
+	ld	("%f13","$FRAME-8*3($sp)");
+	ld	("%f14","$FRAME-8*2($sp)");
+	ld	("%f15","$FRAME-8*1($sp)");
+}
+	vstm	($a0,$d0,"$stdframe($sp)");
+	lghi	("%r1",0);
+
+LABEL	(".Loop_tail_vx");
+	llgc	("%r5","0(%r1,$inp)");
+	llgc	("%r6","$stdframe(%r1,$sp)");
+	xr	("%r6","%r5");
+	stc	("%r6","0(%r1,$out)");
+	la	("%r1","1(%r1)");
+	brct	($len,".Loop_tail_vx");
+
+&{$z?	\&lmg:\&lm}	("%r6","%r7","$FRAME+6*$SIZE_T($sp)");
+	la	($sp,"$FRAME($sp)");
+	br	("%r14");
+SIZE	("ChaCha20_ctr32_vx",".-ChaCha20_ctr32_vx");
+}
+################
+
+ALIGN	(32);
+LABEL	(".Lsigma");
+LONG	(0x61707865,0x3320646e,0x79622d32,0x6b206574);	# endian-neutral sigma
+LONG	(1,0,0,0);
+LONG	(2,0,0,0);
+LONG	(3,0,0,0);
+LONG	(0x03020100,0x07060504,0x0b0a0908,0x0f0e0d0c);	# byte swap
+
+LONG	(0,1,2,3);
+LONG	(0x61707865,0x61707865,0x61707865,0x61707865);	# smashed sigma
+LONG	(0x3320646e,0x3320646e,0x3320646e,0x3320646e);
+LONG	(0x79622d32,0x79622d32,0x79622d32,0x79622d32);
+LONG	(0x6b206574,0x6b206574,0x6b206574,0x6b206574);
+
+ASCIZ	("\"ChaCha20 for s390x, CRYPTOGAMS by <appro\@openssl.org>\"");
+ALIGN	(4);
+
+PERLASM_END();
diff -up openssl-1.1.1b/crypto/perlasm/s390x.pm.s390x-update openssl-1.1.1b/crypto/perlasm/s390x.pm
--- openssl-1.1.1b/crypto/perlasm/s390x.pm.s390x-update	2019-05-06 10:54:00.037367571 +0200
+++ openssl-1.1.1b/crypto/perlasm/s390x.pm	2019-05-06 10:54:00.038367554 +0200
@@ -0,0 +1,3060 @@
+#!/usr/bin/env perl
+# Copyright 2018 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
+
+# Copyright IBM Corp. 2018
+# Author: Patrick Steuer <patrick.steuer@de.ibm.com>
+
+package perlasm::s390x;
+
+use strict;
+use warnings;
+use Carp qw(confess);
+use Exporter qw(import);
+
+our @EXPORT=qw(PERLASM_BEGIN PERLASM_END);
+our @EXPORT_OK=qw(AUTOLOAD LABEL INCLUDE stfle);
+our %EXPORT_TAGS=(
+	MSA => [qw(kmac km kmc kimd klmd)],
+	MSA4 => [qw(kmf kmo pcc kmctr)],
+	MSA5 => [qw(ppno prno)],
+	MSA8 => [qw(kma)],
+	VX => [qw(vgef vgeg vgbm vzero vone vgm vgmb vgmh vgmf vgmg
+	    vl vlr vlrep vlrepb vlreph vlrepf vlrepg vleb vleh vlef vleg vleib
+	    vleih vleif vleig vlgv vlgvb vlgvh vlgvf vlgvg vllez vllezb vllezh
+	    vllezf vllezg vlm vlbb vlvg vlvgb vlvgh vlvgf vlvgg vlvgp
+	    vll vmrh vmrhb vmrhh vmrhf vmrhg vmrl vmrlb vmrlh vmrlf vmrlg vpk
+	    vpkh vpkf vpkg vpks vpksh vpksf vpksg vpkshs vpksfs vpksgs vpkls
+	    vpklsh vpklsf vpklsg vpklshs vpklsfs vpklsgs vperm vpdi vrep vrepb
+	    vreph vrepf vrepg vrepi vrepib vrepih vrepif vrepig vscef vsceg
+	    vsel vseg vsegb vsegh vsegf vst vsteb vsteh vstef vsteg vstm vstl
+	    vuph vuphb vuphh vuphf vuplh vuplhb vuplhh vuplhf vupl vuplb vuplhw
+	    vuplf vupll vupllb vupllh vupllf va vab vah vaf vag vaq vacc vaccb
+	    vacch vaccf vaccg vaccq vac vacq vaccc vacccq vn vnc vavg vavgb
+	    vavgh vavgf vavgg vavgl vavglb vavglh vavglf vavglg vcksm vec_ vecb
+	    vech vecf vecg vecl veclb veclh veclf veclg vceq vceqb vceqh vceqf
+	    vceqg vceqbs vceqhs vceqfs vceqgs vch vchb vchh vchf vchg vchbs
+	    vchhs vchfs vchgs vchl vchlb vchlh vchlf vchlg vchlbs vchlhs vchlfs
+	    vchlgs vclz vclzb vclzh vclzf vclzg vctz vctzb vctzh vctzf vctzg
+	    vx vgfm vgfmb vgfmh vgfmf vgfmg vgfma vgfmab vgfmah vgfmaf vgfmag
+	    vlc vlcb vlch vlcf vlcg vlp vlpb vlph vlpf vlpg vmx vmxb vmxh vmxf
+	    vmxg vmxl vmxlb vmxlh vmxlf vmxlg vmn vmnb vmnh vmnf vmng vmnl
+	    vmnlb vmnlh vmnlf vmnlg vmal vmalb vmalhw vmalf vmah vmahb vmahh
+	    vmahf vmalh vmalhb vmalhh vmalhf vmae vmaeb vmaeh vmaef vmale
+	    vmaleb vmaleh vmalef vmao vmaob vmaoh vmaof vmalo vmalob vmaloh
+	    vmalof vmh vmhb vmhh vmhf vmlh vmlhb vmlhh vmlhf vml vmlb vmlhw
+	    vmlf vme vmeb vmeh vmef vmle vmleb vmleh vmlef vmo vmob vmoh vmof
+	    vmlo vmlob vmloh vmlof vno vnot vo vpopct verllv verllvb verllvh
+	    verllvf verllvg verll verllb verllh verllf verllg verim verimb
+	    verimh verimf verimg veslv veslvb veslvh veslvf veslvg vesl veslb
+	    veslh veslf veslg vesrav vesravb vesravh vesravf vesravg vesra
+	    vesrab vesrah vesraf vesrag vesrlv vesrlvb vesrlvh vesrlvf vesrlvg
+	    vesrl vesrlb vesrlh vesrlf vesrlg vsl vslb vsldb vsra vsrab vsrl
+	    vsrlb vs vsb vsh vsf vsg vsq vscbi vscbib vscbih vscbif vscbig
+	    vscbiq vsbi vsbiq vsbcbi vsbcbiq vsumg vsumgh vsumgf vsumq vsumqf
+	    vsumqg vsum vsumb vsumh vtm vfae vfaeb vfaeh vfaef vfaebs vfaehs
+	    vfaefs vfaezb vfaezh vfaezf vfaezbs vfaezhs vfaezfs vfee vfeeb
+	    vfeeh vfeef vfeebs vfeehs vfeefs vfeezb vfeezh vfeezf vfeezbs
+	    vfeezhs vfeezfs vfene vfeneb vfeneh vfenef vfenebs vfenehs vfenefs
+	    vfenezb vfenezh vfenezf vfenezbs vfenezhs vfenezfs vistr vistrb
+	    vistrh vistrf vistrbs vistrhs vistrfs vstrc vstrcb vstrch vstrcf
+	    vstrcbs vstrchs vstrcfs vstrczb vstrczh vstrczf vstrczbs vstrczhs
+	    vstrczfs vfa vfadb wfadb wfc wfcdb wfk wfkdb vfce vfcedb wfcedb
+	    vfcedbs wfcedbs vfch vfchdb wfchdb vfchdbs wfchdbs vfche vfchedb
+	    wfchedb vfchedbs wfchedbs vcdg vcdgb wcdgb vcdlg vcdlgb wcdlgb vcgd
+	    vcgdb wcgdb vclgd vclgdb wclgdb vfd vfddb wfddb vfi vfidb wfidb
+	    vlde vldeb wldeb vled vledb wledb vfm vfmdb wfmdb vfma vfmadb
+	    wfmadb vfms vfmsdb wfmsdb vfpso vfpsodb wfpsodb vflcdb wflcdb
+	    vflndb wflndb vflpdb wflpdb vfsq vfsqdb wfsqdb vfs vfsdb wfsdb
+	    vftci vftcidb wftcidb)],
+	VXE => [qw(vbperm vllezlf vmsl vmslg vnx vnn voc vpopctb vpopcth
+	    vpopctf vpopctg vfasb wfasb wfaxb wfcsb wfcxb wfksb wfkxb vfcesb
+	    vfcesbs wfcesb wfcesbs wfcexb wfcexbs vfchsb vfchsbs wfchsb wfchsbs
+	    wfchxb wfchxbs vfchesb vfchesbs wfchesb wfchesbs wfchexb wfchexbs
+	    vfdsb wfdsb wfdxb vfisb wfisb wfixb vfll vflls wflls wflld vflr
+	    vflrd wflrd wflrx vfmax vfmaxsb vfmaxdb wfmaxsb wfmaxdb wfmaxxb
+	    vfmin vfminsb vfmindb wfminsb wfmindb wfminxb vfmsb wfmsb wfmxb
+	    vfnma vfnms vfmasb wfmasb wfmaxb vfmssb wfmssb wfmsxb vfnmasb
+	    vfnmadb wfnmasb wfnmadb wfnmaxb vfnmssb vfnmsdb wfnmssb wfnmsdb
+	    wfnmsxb vfpsosb wfpsosb vflcsb wflcsb vflnsb wflnsb vflpsb wflpsb
+	    vfpsoxb wfpsoxb vflcxb wflcxb vflnxb wflnxb vflpxb wflpxb vfsqsb
+	    wfsqsb wfsqxb vfssb wfssb wfsxb vftcisb wftcisb wftcixb)],
+	VXD => [qw(vlrlr vlrl vstrlr vstrl vap vcp vcvb vcvbg vcvd vcvdg vdp
+	    vlip vmp vmsp vpkz vpsop vrp vsdp vsrp vsp vtp vupkz)],
+);
+Exporter::export_ok_tags(qw(MSA MSA4 MSA5 MSA8 VX VXE VXD));
+
+our $AUTOLOAD;
+
+my $GR='(?:%r)?([0-9]|1[0-5])';
+my $VR='(?:%v)?([0-9]|1[0-9]|2[0-9]|3[0-1])';
+
+my ($file,$out);
+
+sub PERLASM_BEGIN
+{
+	($file,$out)=(shift,"");
+}
+sub PERLASM_END
+{
+	if (defined($file)) {
+		open(my $fd,'>',$file)||die("can't open $file: $!");
+		print({$fd}$out);
+		close($fd);
+	} else {
+		print($out);
+	}
+}
+
+sub AUTOLOAD {
+	confess(err("PARSE")) if (grep(!defined($_),@_));
+	my $token;
+	for ($AUTOLOAD) {
+		$token=".$1" if (/^.*::([A-Z_]+)$/);	# uppercase: directive
+		$token="\t$1" if (/^.*::([a-z]+)$/);	# lowercase: mnemonic
+		confess(err("PARSE")) if (!defined($token));
+	}
+	$token.="\t" if ($#_>=0);
+	$out.=$token.join(',',@_)."\n";
+}
+
+sub LABEL {						# label directive
+	confess(err("ARGNUM")) if ($#_!=0);
+	my ($label)=@_;
+	$out.="$label:\n";
+}
+
+sub INCLUDE {
+	confess(err("ARGNUM")) if ($#_!=0);
+	my ($file)=@_;
+	$out.="#include \"$file\"\n";
+}
+
+#
+# Mnemonics
+#
+
+sub stfle {
+	confess(err("ARGNUM")) if ($#_!=0);
+	S(0xb2b0,@_);
+}
+
+# MSA
+
+sub kmac {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb91e,@_);
+}
+
+sub km {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb92e,@_);
+}
+
+sub kmc {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb92f,@_);
+}
+
+sub kimd {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb93e,@_);
+}
+
+sub klmd {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb93f,@_);
+}
+
+# MSA4
+
+sub kmf {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb92a,@_);
+}
+
+sub kmo {
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb92b,@_);
+}
+
+sub pcc {
+	confess(err("ARGNUM")) if ($#_!=-1);
+	RRE(0xb92c,@_);
+}
+
+sub kmctr {
+	confess(err("ARGNUM")) if ($#_!=2);
+	RRFb(0xb92d,@_);
+}
+
+# MSA5
+
+sub prno {
+	ppno(@_);
+}
+
+sub ppno {						# deprecated, use prno
+	confess(err("ARGNUM")) if ($#_!=1);
+	RRE(0xb93c,@_);
+}
+
+# MSA8
+
+sub kma {
+	confess(err("ARGNUM")) if ($#_!=2);
+	RRFb(0xb929,@_);
+}
+
+# VX - Support Instructions
+
+sub vgef {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRV(0xe713,@_);
+}
+sub vgeg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRV(0xe712,@_);
+}
+
+sub vgbm {
+	confess(err("ARGNUM")) if ($#_!=1);
+	VRIa(0xe744,@_);
+}
+sub vzero {
+	vgbm(@_,0);
+}
+sub vone {
+	vgbm(@_,0xffff);
+}
+
+sub vgm {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRIb(0xe746,@_);
+}
+sub vgmb {
+	vgm(@_,0);
+}
+sub vgmh {
+	vgm(@_,1);
+}
+sub vgmf {
+	vgm(@_,2);
+}
+sub vgmg {
+	vgm(@_,3);
+}
+
+sub vl {
+	confess(err("ARGNUM")) if ($#_<1||$#_>2);
+	VRX(0xe706,@_);
+}
+
+sub vlr {
+	confess(err("ARGNUM")) if ($#_!=1);
+	VRRa(0xe756,@_);
+}
+
+sub vlrep {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe705,@_);
+}
+sub vlrepb {
+	vlrep(@_,0);
+}
+sub vlreph {
+	vlrep(@_,1);
+}
+sub vlrepf {
+	vlrep(@_,2);
+}
+sub vlrepg {
+	vlrep(@_,3);
+}
+
+sub vleb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe700,@_);
+}
+sub vleh {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe701,@_);
+}
+sub vlef {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe703,@_);
+}
+sub vleg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe702,@_);
+}
+
+sub vleib {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIa(0xe740,@_);
+}
+sub vleih {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIa(0xe741,@_);
+}
+sub vleif {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIa(0xe743,@_);
+}
+sub vleig {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIa(0xe742,@_);
+}
+
+sub vlgv {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSc(0xe721,@_);
+}
+sub vlgvb {
+	vlgv(@_,0);
+}
+sub vlgvh {
+	vlgv(@_,1);
+}
+sub vlgvf {
+	vlgv(@_,2);
+}
+sub vlgvg {
+	vlgv(@_,3);
+}
+
+sub vllez {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe704,@_);
+}
+sub vllezb {
+	vllez(@_,0);
+}
+sub vllezh {
+	vllez(@_,1);
+}
+sub vllezf {
+	vllez(@_,2);
+}
+sub vllezg {
+	vllez(@_,3);
+}
+
+sub vlm {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	VRSa(0xe736,@_);
+}
+
+sub vlbb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe707,@_);
+}
+
+sub vlvg {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSb(0xe722,@_);
+}
+sub vlvgb {
+	vlvg(@_,0);
+}
+sub vlvgh {
+	vlvg(@_,1);
+}
+sub vlvgf {
+	vlvg(@_,2);
+}
+sub vlvgg {
+	vlvg(@_,3);
+}
+
+sub vlvgp {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRf(0xe762,@_);
+}
+
+sub vll {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRSb(0xe737,@_);
+}
+
+sub vmrh {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe761,@_);
+}
+sub vmrhb {
+	vmrh(@_,0);
+}
+sub vmrhh {
+	vmrh(@_,1);
+}
+sub vmrhf {
+	vmrh(@_,2);
+}
+sub vmrhg {
+	vmrh(@_,3);
+}
+
+sub vmrl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe760,@_);
+}
+sub vmrlb {
+	vmrl(@_,0);
+}
+sub vmrlh {
+	vmrl(@_,1);
+}
+sub vmrlf {
+	vmrl(@_,2);
+}
+sub vmrlg {
+	vmrl(@_,3);
+}
+
+sub vpk {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe794,@_);
+}
+sub vpkh {
+	vpk(@_,1);
+}
+sub vpkf {
+	vpk(@_,2);
+}
+sub vpkg {
+	vpk(@_,3);
+}
+
+sub vpks {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRb(0xe797,@_);
+}
+sub vpksh {
+	vpks(@_,1,0);
+}
+sub vpksf {
+	vpks(@_,2,0);
+}
+sub vpksg {
+	vpks(@_,3,0);
+}
+sub vpkshs {
+	vpks(@_,1,1);
+}
+sub vpksfs {
+	vpks(@_,2,1);
+}
+sub vpksgs {
+	vpks(@_,3,1);
+}
+
+sub vpkls {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRb(0xe795,@_);
+}
+sub vpklsh {
+	vpkls(@_,1,0);
+}
+sub vpklsf {
+	vpkls(@_,2,0);
+}
+sub vpklsg {
+	vpkls(@_,3,0);
+}
+sub vpklshs {
+	vpkls(@_,1,1);
+}
+sub vpklsfs {
+	vpkls(@_,2,1);
+}
+sub vpklsgs {
+	vpkls(@_,3,1);
+}
+
+sub vperm {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRe(0xe78c,@_);
+}
+
+sub vpdi {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe784,@_);
+}
+
+sub vrep {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRIc(0xe74d,@_);
+}
+sub vrepb {
+	vrep(@_,0);
+}
+sub vreph {
+	vrep(@_,1);
+}
+sub vrepf {
+	vrep(@_,2);
+}
+sub vrepg {
+	vrep(@_,3);
+}
+
+sub vrepi {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIa(0xe745,@_);
+}
+sub vrepib {
+	vrepi(@_,0);
+}
+sub vrepih {
+	vrepi(@_,1);
+}
+sub vrepif {
+	vrepi(@_,2);
+}
+sub vrepig {
+	vrepi(@_,3);
+}
+
+sub vscef {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRV(0xe71b,@_);
+}
+sub vsceg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRV(0xe71a,@_);
+}
+
+sub vsel {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRe(0xe78d,@_);
+}
+
+sub vseg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe75f,@_);
+}
+sub vsegb {
+	vseg(@_,0);
+}
+sub vsegh {
+	vseg(@_,1);
+}
+sub vsegf {
+	vseg(@_,2);
+}
+
+sub vst {
+	confess(err("ARGNUM")) if ($#_<1||$#_>2);
+	VRX(0xe70e,@_);
+}
+
+sub vsteb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe708,@_);
+}
+sub vsteh {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe709,@_);
+}
+sub vstef {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe70b,@_);
+}
+sub vsteg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRX(0xe70a,@_);
+}
+
+sub vstm {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	VRSa(0xe73e,@_);
+}
+
+sub vstl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRSb(0xe73f,@_);
+}
+
+sub vuph {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7d7,@_);
+}
+sub vuphb {
+	vuph(@_,0);
+}
+sub vuphh {
+	vuph(@_,1);
+}
+sub vuphf {
+	vuph(@_,2);
+}
+
+sub vuplh {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7d5,@_);
+}
+sub vuplhb {
+	vuplh(@_,0);
+}
+sub vuplhh {
+	vuplh(@_,1);
+}
+sub vuplhf {
+	vuplh(@_,2);
+}
+
+sub vupl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7d6,@_);
+}
+sub vuplb {
+	vupl(@_,0);
+}
+sub vuplhw {
+	vupl(@_,1);
+}
+sub vuplf {
+	vupl(@_,2);
+}
+
+sub vupll {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7d4,@_);
+}
+sub vupllb {
+	vupll(@_,0);
+}
+sub vupllh {
+	vupll(@_,1);
+}
+sub vupllf {
+	vupll(@_,2);
+}
+
+# VX - Integer Instructions
+
+sub va {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f3,@_);
+}
+sub vab {
+	va(@_,0);
+}
+sub vah {
+	va(@_,1);
+}
+sub vaf {
+	va(@_,2);
+}
+sub vag {
+	va(@_,3);
+}
+sub vaq {
+	va(@_,4);
+}
+
+sub vacc {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f1,@_);
+}
+sub vaccb {
+	vacc(@_,0);
+}
+sub vacch {
+	vacc(@_,1);
+}
+sub vaccf {
+	vacc(@_,2);
+}
+sub vaccg {
+	vacc(@_,3);
+}
+sub vaccq {
+	vacc(@_,4);
+}
+
+sub vac {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7bb,@_);
+}
+sub vacq {
+	vac(@_,4);
+}
+
+sub vaccc {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7b9,@_);
+}
+sub vacccq {
+	vaccc(@_,4);
+}
+
+sub vn {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe768,@_);
+}
+
+sub vnc {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe769,@_);
+}
+
+sub vavg {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f2,@_);
+}
+sub vavgb {
+	vavg(@_,0);
+}
+sub vavgh {
+	vavg(@_,1);
+}
+sub vavgf {
+	vavg(@_,2);
+}
+sub vavgg {
+	vavg(@_,3);
+}
+
+sub vavgl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f0,@_);
+}
+sub vavglb {
+	vavgl(@_,0);
+}
+sub vavglh {
+	vavgl(@_,1);
+}
+sub vavglf {
+	vavgl(@_,2);
+}
+sub vavglg {
+	vavgl(@_,3);
+}
+
+sub vcksm {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe766,@_);
+}
+
+sub vec_ {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7db,@_);
+}
+sub vecb {
+	vec_(@_,0);
+}
+sub vech {
+	vec_(@_,1);
+}
+sub vecf {
+	vec_(@_,2);
+}
+sub vecg {
+	vec_(@_,3);
+}
+
+sub vecl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7d9,@_);
+}
+sub veclb {
+	vecl(@_,0);
+}
+sub veclh {
+	vecl(@_,1);
+}
+sub veclf {
+	vecl(@_,2);
+}
+sub veclg {
+	vecl(@_,3);
+}
+
+sub vceq {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRb(0xe7f8,@_);
+}
+sub vceqb {
+	vceq(@_,0,0);
+}
+sub vceqh {
+	vceq(@_,1,0);
+}
+sub vceqf {
+	vceq(@_,2,0);
+}
+sub vceqg {
+	vceq(@_,3,0);
+}
+sub vceqbs {
+	vceq(@_,0,1);
+}
+sub vceqhs {
+	vceq(@_,1,1);
+}
+sub vceqfs {
+	vceq(@_,2,1);
+}
+sub vceqgs {
+	vceq(@_,3,1);
+}
+
+sub vch {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRb(0xe7fb,@_);
+}
+sub vchb {
+	vch(@_,0,0);
+}
+sub vchh {
+	vch(@_,1,0);
+}
+sub vchf {
+	vch(@_,2,0);
+}
+sub vchg {
+	vch(@_,3,0);
+}
+sub vchbs {
+	vch(@_,0,1);
+}
+sub vchhs {
+	vch(@_,1,1);
+}
+sub vchfs {
+	vch(@_,2,1);
+}
+sub vchgs {
+	vch(@_,3,1);
+}
+
+sub vchl {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRb(0xe7f9,@_);
+}
+sub vchlb {
+	vchl(@_,0,0);
+}
+sub vchlh {
+	vchl(@_,1,0);
+}
+sub vchlf {
+	vchl(@_,2,0);
+}
+sub vchlg {
+	vchl(@_,3,0);
+}
+sub vchlbs {
+	vchl(@_,0,1);
+}
+sub vchlhs {
+	vchl(@_,1,1);
+}
+sub vchlfs {
+	vchl(@_,2,1);
+}
+sub vchlgs {
+	vchl(@_,3,1);
+}
+
+sub vclz {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe753,@_);
+}
+sub vclzb {
+	vclz(@_,0);
+}
+sub vclzh {
+	vclz(@_,1);
+}
+sub vclzf {
+	vclz(@_,2);
+}
+sub vclzg {
+	vclz(@_,3);
+}
+
+sub vctz {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe752,@_);
+}
+sub vctzb {
+	vctz(@_,0);
+}
+sub vctzh {
+	vctz(@_,1);
+}
+sub vctzf {
+	vctz(@_,2);
+}
+sub vctzg {
+	vctz(@_,3);
+}
+
+sub vx {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76d,@_);
+}
+
+sub vgfm {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7b4,@_);
+}
+sub vgfmb {
+	vgfm(@_,0);
+}
+sub vgfmh {
+	vgfm(@_,1);
+}
+sub vgfmf {
+	vgfm(@_,2);
+}
+sub vgfmg {
+	vgfm(@_,3);
+}
+
+sub vgfma {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7bc,@_);
+}
+sub vgfmab {
+	vgfma(@_,0);
+}
+sub vgfmah {
+	vgfma(@_,1);
+}
+sub vgfmaf {
+	vgfma(@_,2);
+}
+sub vgfmag {
+	vgfma(@_,3);
+}
+
+sub vlc {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7de,@_);
+}
+sub vlcb {
+	vlc(@_,0);
+}
+sub vlch {
+	vlc(@_,1);
+}
+sub vlcf {
+	vlc(@_,2);
+}
+sub vlcg {
+	vlc(@_,3);
+}
+
+sub vlp {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe7df,@_);
+}
+sub vlpb {
+	vlp(@_,0);
+}
+sub vlph {
+	vlp(@_,1);
+}
+sub vlpf {
+	vlp(@_,2);
+}
+sub vlpg {
+	vlp(@_,3);
+}
+
+sub vmx {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7ff,@_);
+}
+sub vmxb {
+	vmx(@_,0);
+}
+sub vmxh {
+	vmx(@_,1);
+}
+sub vmxf {
+	vmx(@_,2);
+}
+sub vmxg {
+	vmx(@_,3);
+}
+
+sub vmxl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7fd,@_);
+}
+sub vmxlb {
+	vmxl(@_,0);
+}
+sub vmxlh {
+	vmxl(@_,1);
+}
+sub vmxlf {
+	vmxl(@_,2);
+}
+sub vmxlg {
+	vmxl(@_,3);
+}
+
+sub vmn {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7fe,@_);
+}
+sub vmnb {
+	vmn(@_,0);
+}
+sub vmnh {
+	vmn(@_,1);
+}
+sub vmnf {
+	vmn(@_,2);
+}
+sub vmng {
+	vmn(@_,3);
+}
+
+sub vmnl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7fc,@_);
+}
+sub vmnlb {
+	vmnl(@_,0);
+}
+sub vmnlh {
+	vmnl(@_,1);
+}
+sub vmnlf {
+	vmnl(@_,2);
+}
+sub vmnlg {
+	vmnl(@_,3);
+}
+
+sub vmal {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7aa,@_);
+}
+sub vmalb {
+	vmal(@_,0);
+}
+sub vmalhw {
+	vmal(@_,1);
+}
+sub vmalf {
+	vmal(@_,2);
+}
+
+sub vmah {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7ab,@_);
+}
+sub vmahb {
+	vmah(@_,0);
+}
+sub vmahh {
+	vmah(@_,1);
+}
+sub vmahf {
+	vmah(@_,2);
+}
+
+sub vmalh {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7a9,@_);
+}
+sub vmalhb {
+	vmalh(@_,0);
+}
+sub vmalhh {
+	vmalh(@_,1);
+}
+sub vmalhf {
+	vmalh(@_,2);
+}
+
+sub vmae {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7ae,@_);
+}
+sub vmaeb {
+	vmae(@_,0);
+}
+sub vmaeh {
+	vmae(@_,1);
+}
+sub vmaef {
+	vmae(@_,2);
+}
+
+sub vmale {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7ac,@_);
+}
+sub vmaleb {
+	vmale(@_,0);
+}
+sub vmaleh {
+	vmale(@_,1);
+}
+sub vmalef {
+	vmale(@_,2);
+}
+
+sub vmao {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7af,@_);
+}
+sub vmaob {
+	vmao(@_,0);
+}
+sub vmaoh {
+	vmao(@_,1);
+}
+sub vmaof {
+	vmao(@_,2);
+}
+
+sub vmalo {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7ad,@_);
+}
+sub vmalob {
+	vmalo(@_,0);
+}
+sub vmaloh {
+	vmalo(@_,1);
+}
+sub vmalof {
+	vmalo(@_,2);
+}
+
+sub vmh {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a3,@_);
+}
+sub vmhb {
+	vmh(@_,0);
+}
+sub vmhh {
+	vmh(@_,1);
+}
+sub vmhf {
+	vmh(@_,2);
+}
+
+sub vmlh {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a1,@_);
+}
+sub vmlhb {
+	vmlh(@_,0);
+}
+sub vmlhh {
+	vmlh(@_,1);
+}
+sub vmlhf {
+	vmlh(@_,2);
+}
+
+sub vml {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a2,@_);
+}
+sub vmlb {
+	vml(@_,0);
+}
+sub vmlhw {
+	vml(@_,1);
+}
+sub vmlf {
+	vml(@_,2);
+}
+
+sub vme {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a6,@_);
+}
+sub vmeb {
+	vme(@_,0);
+}
+sub vmeh {
+	vme(@_,1);
+}
+sub vmef {
+	vme(@_,2);
+}
+
+sub vmle {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a4,@_);
+}
+sub vmleb {
+	vmle(@_,0);
+}
+sub vmleh {
+	vmle(@_,1);
+}
+sub vmlef {
+	vmle(@_,2);
+}
+
+sub vmo {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a7,@_);
+}
+sub vmob {
+	vmo(@_,0);
+}
+sub vmoh {
+	vmo(@_,1);
+}
+sub vmof {
+	vmo(@_,2);
+}
+
+sub vmlo {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7a5,@_);
+}
+sub vmlob {
+	vmlo(@_,0);
+}
+sub vmloh {
+	vmlo(@_,1);
+}
+sub vmlof {
+	vmlo(@_,2);
+}
+
+sub vno {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76b,@_);
+}
+sub vnot {
+	vno(@_,$_[1]);
+}
+
+sub vo {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76a,@_);
+}
+
+sub vpopct {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRa(0xe750,@_);
+}
+
+sub verllv {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe773,@_);
+}
+sub verllvb {
+	verllv(@_,0);
+}
+sub verllvh {
+	verllv(@_,1);
+}
+sub verllvf {
+	verllv(@_,2);
+}
+sub verllvg {
+	verllv(@_,3);
+}
+
+sub verll {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSa(0xe733,@_);
+}
+sub verllb {
+	verll(@_,0);
+}
+sub verllh {
+	verll(@_,1);
+}
+sub verllf {
+	verll(@_,2);
+}
+sub verllg {
+	verll(@_,3);
+}
+
+sub verim {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRId(0xe772,@_);
+}
+sub verimb {
+	verim(@_,0);
+}
+sub verimh {
+	verim(@_,1);
+}
+sub verimf {
+	verim(@_,2);
+}
+sub verimg {
+	verim(@_,3);
+}
+
+sub veslv {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe770,@_);
+}
+sub veslvb {
+	veslv(@_,0);
+}
+sub veslvh {
+	veslv(@_,1);
+}
+sub veslvf {
+	veslv(@_,2);
+}
+sub veslvg {
+	veslv(@_,3);
+}
+
+sub vesl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSa(0xe730,@_);
+}
+sub veslb {
+	vesl(@_,0);
+}
+sub veslh {
+	vesl(@_,1);
+}
+sub veslf {
+	vesl(@_,2);
+}
+sub veslg {
+	vesl(@_,3);
+}
+
+sub vesrav {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe77a,@_);
+}
+sub vesravb {
+	vesrav(@_,0);
+}
+sub vesravh {
+	vesrav(@_,1);
+}
+sub vesravf {
+	vesrav(@_,2);
+}
+sub vesravg {
+	vesrav(@_,3);
+}
+
+sub vesra {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSa(0xe73a,@_);
+}
+sub vesrab {
+	vesra(@_,0);
+}
+sub vesrah {
+	vesra(@_,1);
+}
+sub vesraf {
+	vesra(@_,2);
+}
+sub vesrag {
+	vesra(@_,3);
+}
+
+sub vesrlv {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe778,@_);
+}
+sub vesrlvb {
+	vesrlv(@_,0);
+}
+sub vesrlvh {
+	vesrlv(@_,1);
+}
+sub vesrlvf {
+	vesrlv(@_,2);
+}
+sub vesrlvg {
+	vesrlv(@_,3);
+}
+
+sub vesrl {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRSa(0xe738,@_);
+}
+sub vesrlb {
+	vesrl(@_,0);
+}
+sub vesrlh {
+	vesrl(@_,1);
+}
+sub vesrlf {
+	vesrl(@_,2);
+}
+sub vesrlg {
+	vesrl(@_,3);
+}
+
+sub vsl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe774,@_);
+}
+
+sub vslb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe775,@_);
+}
+
+sub vsldb {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRId(0xe777,@_);
+}
+
+sub vsra {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe77e,@_);
+}
+
+sub vsrab {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe77f,@_);
+}
+
+sub vsrl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe77c,@_);
+}
+
+sub vsrlb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe77d,@_);
+}
+
+sub vs {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f7,@_);
+}
+sub vsb {
+	vs(@_,0);
+}
+sub vsh {
+	vs(@_,1);
+}
+sub vsf {
+	vs(@_,2);
+}
+sub vsg {
+	vs(@_,3);
+}
+sub vsq {
+	vs(@_,4);
+}
+
+sub vscbi {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe7f5,@_);
+}
+sub vscbib {
+	vscbi(@_,0);
+}
+sub vscbih {
+	vscbi(@_,1);
+}
+sub vscbif {
+	vscbi(@_,2);
+}
+sub vscbig {
+	vscbi(@_,3);
+}
+sub vscbiq {
+	vscbi(@_,4);
+}
+
+sub vsbi {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7bf,@_);
+}
+sub vsbiq {
+	vsbi(@_,4);
+}
+
+sub vsbcbi {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRd(0xe7bd,@_);
+}
+sub vsbcbiq {
+	vsbcbi(@_,4);
+}
+
+sub vsumg {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe765,@_);
+}
+sub vsumgh {
+	vsumg(@_,1);
+}
+sub vsumgf {
+	vsumg(@_,2);
+}
+
+sub vsumq {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe767,@_);
+}
+sub vsumqf {
+	vsumq(@_,2);
+}
+sub vsumqg {
+	vsumq(@_,3);
+}
+
+sub vsum {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRc(0xe764,@_);
+}
+sub vsumb {
+	vsum(@_,0);
+}
+sub vsumh {
+	vsum(@_,1);
+}
+
+sub vtm {
+	confess(err("ARGNUM")) if ($#_!=1);
+	VRRa(0xe7d8,@_);
+}
+
+# VX - String Instructions
+
+sub vfae {
+	confess(err("ARGNUM")) if ($#_<3||$#_>4);
+	VRRb(0xe782,@_);
+}
+sub vfaeb {
+	vfae(@_[0..2],0,$_[3]);
+}
+sub vfaeh {
+	vfae(@_[0..2],1,$_[3]);
+}
+sub vfaef {
+	vfae(@_[0..2],2,$_[3]);
+}
+sub vfaebs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],0,0x1|$_[3]);
+}
+sub vfaehs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],1,0x1|$_[3]);
+}
+sub vfaefs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],2,0x1|$_[3]);
+}
+sub vfaezb {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],0,0x2|$_[3]);
+}
+sub vfaezh {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],1,0x2|$_[3]);
+}
+sub vfaezf {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],2,0x2|$_[3]);
+}
+sub vfaezbs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],0,0x3|$_[3]);
+}
+sub vfaezhs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],1,0x3|$_[3]);
+}
+sub vfaezfs {
+	$_[3]=0 if (!defined($_[3]));
+	vfae(@_[0..2],2,0x3|$_[3]);
+}
+
+sub vfee {
+	confess(err("ARGNUM")) if ($#_<3||$#_>4);
+	VRRb(0xe780,@_);
+}
+sub vfeeb {
+	vfee(@_[0..2],0,$_[3]);
+}
+sub vfeeh {
+	vfee(@_[0..2],1,$_[3]);
+}
+sub vfeef {
+	vfee(@_[0..2],2,$_[3]);
+}
+sub vfeebs {
+	vfee(@_,0,1);
+}
+sub vfeehs {
+	vfee(@_,1,1);
+}
+sub vfeefs {
+	vfee(@_,2,1);
+}
+sub vfeezb {
+	vfee(@_,0,2);
+}
+sub vfeezh {
+	vfee(@_,1,2);
+}
+sub vfeezf {
+	vfee(@_,2,2);
+}
+sub vfeezbs {
+	vfee(@_,0,3);
+}
+sub vfeezhs {
+	vfee(@_,1,3);
+}
+sub vfeezfs {
+	vfee(@_,2,3);
+}
+
+sub vfene {
+	confess(err("ARGNUM")) if ($#_<3||$#_>4);
+	VRRb(0xe781,@_);
+}
+sub vfeneb {
+	vfene(@_[0..2],0,$_[3]);
+}
+sub vfeneh {
+	vfene(@_[0..2],1,$_[3]);
+}
+sub vfenef {
+	vfene(@_[0..2],2,$_[3]);
+}
+sub vfenebs {
+	vfene(@_,0,1);
+}
+sub vfenehs {
+	vfene(@_,1,1);
+}
+sub vfenefs {
+	vfene(@_,2,1);
+}
+sub vfenezb {
+	vfene(@_,0,2);
+}
+sub vfenezh {
+	vfene(@_,1,2);
+}
+sub vfenezf {
+	vfene(@_,2,2);
+}
+sub vfenezbs {
+	vfene(@_,0,3);
+}
+sub vfenezhs {
+	vfene(@_,1,3);
+}
+sub vfenezfs {
+	vfene(@_,2,3);
+}
+
+sub vistr {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	VRRa(0xe75c,@_[0..2],0,$_[3]);
+}
+sub vistrb {
+	vistr(@_[0..1],0,$_[2]);
+}
+sub vistrh {
+	vistr(@_[0..1],1,$_[2]);
+}
+sub vistrf {
+	vistr(@_[0..1],2,$_[2]);
+}
+sub vistrbs {
+	vistr(@_,0,1);
+}
+sub vistrhs {
+	vistr(@_,1,1);
+}
+sub vistrfs {
+	vistr(@_,2,1);
+}
+
+sub vstrc {
+	confess(err("ARGNUM")) if ($#_<4||$#_>5);
+	VRRd(0xe78a,@_);
+}
+sub vstrcb {
+	vstrc(@_[0..3],0,$_[4]);
+}
+sub vstrch {
+	vstrc(@_[0..3],1,$_[4]);
+}
+sub vstrcf {
+	vstrc(@_[0..3],2,$_[4]);
+}
+sub vstrcbs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],0,0x1|$_[4]);
+}
+sub vstrchs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],1,0x1|$_[4]);
+}
+sub vstrcfs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],2,0x1|$_[4]);
+}
+sub vstrczb {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],0,0x2|$_[4]);
+}
+sub vstrczh {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],1,0x2|$_[4]);
+}
+sub vstrczf {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],2,0x2|$_[4]);
+}
+sub vstrczbs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],0,0x3|$_[4]);
+}
+sub vstrczhs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],1,0x3|$_[4]);
+}
+sub vstrczfs {
+	$_[4]=0 if (!defined($_[4]));
+	vstrc(@_[0..3],2,0x3|$_[4]);
+}
+
+# VX - Floating-point Instructions
+
+sub vfa {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRc(0xe7e3,@_);
+}
+sub vfadb {
+	vfa(@_,3,0);
+}
+sub wfadb {
+	vfa(@_,3,8);
+}
+
+sub wfc {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRa(0xe7cb,@_);
+}
+sub wfcdb {
+	wfc(@_,3,0);
+}
+
+sub wfk {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRa(0xe7ca,@_);
+}
+sub wfksb {
+	wfk(@_,2,0);
+}
+sub wfkdb {
+	wfk(@_,3,0);
+}
+sub wfkxb {
+	wfk(@_,4,0);
+}
+
+sub vfce {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRc(0xe7e8,@_);
+}
+sub vfcedb {
+	vfce(@_,3,0,0);
+}
+sub vfcedbs {
+	vfce(@_,3,0,1);
+}
+sub wfcedb {
+	vfce(@_,3,8,0);
+}
+sub wfcedbs {
+	vfce(@_,3,8,1);
+}
+
+sub vfch {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRc(0xe7eb,@_);
+}
+sub vfchdb {
+	vfch(@_,3,0,0);
+}
+sub vfchdbs {
+	vfch(@_,3,0,1);
+}
+sub wfchdb {
+	vfch(@_,3,8,0);
+}
+sub wfchdbs {
+	vfch(@_,3,8,1);
+}
+
+sub vfche {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRc(0xe7ea,@_);
+}
+sub vfchedb {
+	vfche(@_,3,0,0);
+}
+sub vfchedbs {
+	vfche(@_,3,0,1);
+}
+sub wfchedb {
+	vfche(@_,3,8,0);
+}
+sub wfchedbs {
+	vfche(@_,3,8,1);
+}
+
+sub vcdg {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c3,@_);
+}
+sub vcdgb {
+	vcdg(@_[0..1],3,@_[2..3]);
+}
+sub wcdgb {
+	vcdg(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vcdlg {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c1,@_);
+}
+sub vcdlgb {
+	vcdlg(@_[0..1],3,@_[2..3]);
+}
+sub wcdlgb {
+	vcdlg(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vcgd {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c2,@_);
+}
+sub vcgdb {
+	vcgd(@_[0..1],3,@_[2..3]);
+}
+sub wcgdb {
+	vcgd(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vclgd {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c0,@_);
+}
+sub vclgdb {
+	vclgd(@_[0..1],3,@_[2..3]);
+}
+sub wclgdb {
+	vclgd(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vfd {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRc(0xe7e5,@_);
+}
+sub vfddb {
+	vfd(@_,3,0);
+}
+sub wfddb {
+	vfd(@_,3,8);
+}
+
+sub vfi {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c7,@_);
+}
+sub vfidb {
+	vfi(@_[0..1],3,@_[2..3]);
+}
+sub wfidb {
+	vfi(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vlde {	# deprecated, use vfll
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRa(0xe7c4,@_);
+}
+sub vldeb {	# deprecated, use vflls
+	vlde(@_,2,0);
+}
+sub wldeb {	# deprecated, use wflls
+	vlde(@_,2,8);
+}
+
+sub vled {	# deprecated, use vflr
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7c5,@_);
+}
+sub vledb {	# deprecated, use vflrd
+	vled(@_[0..1],3,@_[2..3]);
+}
+sub wledb {	# deprecated, use wflrd
+	vled(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+
+sub vfm {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRc(0xe7e7,@_);
+}
+sub vfmdb {
+	vfm(@_,3,0);
+}
+sub wfmdb {
+	vfm(@_,3,8);
+}
+
+sub vfma {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRe(0xe78f,@_);
+}
+sub vfmadb {
+	vfma(@_,0,3);
+}
+sub wfmadb {
+	vfma(@_,8,3);
+}
+
+sub vfms {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRe(0xe78e,@_);
+}
+sub vfmsdb {
+	vfms(@_,0,3);
+}
+sub wfmsdb {
+	vfms(@_,8,3);
+}
+
+sub vfpso {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRa(0xe7cc,@_);
+}
+sub vfpsodb {
+	vfpso(@_[0..1],3,0,$_[2]);
+}
+sub wfpsodb {
+	vfpso(@_[0..1],3,8,$_[2]);
+}
+sub vflcdb {
+	vfpso(@_,3,0,0);
+}
+sub wflcdb {
+	vfpso(@_,3,8,0);
+}
+sub vflndb {
+	vfpso(@_,3,0,1);
+}
+sub wflndb {
+	vfpso(@_,3,8,1);
+}
+sub vflpdb {
+	vfpso(@_,3,0,2);
+}
+sub wflpdb {
+	vfpso(@_,3,8,2);
+}
+
+sub vfsq {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRRa(0xe7ce,@_);
+}
+sub vfsqdb {
+	vfsq(@_,3,0);
+}
+sub wfsqdb {
+	vfsq(@_,3,8);
+}
+
+sub vfs {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRRc(0xe7e2,@_);
+}
+sub vfsdb {
+	vfs(@_,3,0);
+}
+sub wfsdb {
+	vfs(@_,3,8);
+}
+
+sub vftci {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIe(0xe74a,@_);
+}
+sub vftcidb {
+	vftci(@_,3,0);
+}
+sub wftcidb {
+	vftci(@_,3,8);
+}
+
+# VXE - Support Instructions
+
+sub vbperm {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe785,@_);
+}
+
+sub vllezlf {
+	vllez(@_,6);
+}
+
+# VXE - Integer Instructions
+
+sub vmsl {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRd(0xe7b8,@_);
+}
+sub vmslg {
+	vmsl(@_[0..3],3,$_[4]);
+}
+
+sub vnx {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76c,@_);
+}
+
+sub vnn {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76e,@_);
+}
+
+sub voc {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRc(0xe76f,@_);
+}
+
+sub vpopctb {
+	vpopct(@_,0);
+}
+sub vpopcth {
+	vpopct(@_,1);
+}
+sub vpopctf {
+	vpopct(@_,2);
+}
+sub vpopctg {
+	vpopct(@_,3);
+}
+
+# VXE - Floating-Point Instructions
+
+sub vfasb {
+	vfa(@_,2,0);
+}
+sub wfasb {
+	vfa(@_,2,8);
+}
+sub wfaxb {
+	vfa(@_,4,8);
+}
+
+sub wfcsb {
+	wfc(@_,2,0);
+}
+sub wfcxb {
+	wfc(@_,4,0);
+}
+
+sub vfcesb {
+	vfce(@_,2,0,0);
+}
+sub vfcesbs {
+	vfce(@_,2,0,1);
+}
+sub wfcesb {
+	vfce(@_,2,8,0);
+}
+sub wfcesbs {
+	vfce(@_,2,8,1);
+}
+sub wfcexb {
+	vfce(@_,4,8,0);
+}
+sub wfcexbs {
+	vfce(@_,4,8,1);
+}
+
+sub vfchsb {
+	vfch(@_,2,0,0);
+}
+sub vfchsbs {
+	vfch(@_,2,0,1);
+}
+sub wfchsb {
+	vfch(@_,2,8,0);
+}
+sub wfchsbs {
+	vfch(@_,2,8,1);
+}
+sub wfchxb {
+	vfch(@_,4,8,0);
+}
+sub wfchxbs {
+	vfch(@_,4,8,1);
+}
+
+sub vfchesb {
+	vfche(@_,2,0,0);
+}
+sub vfchesbs {
+	vfche(@_,2,0,1);
+}
+sub wfchesb {
+	vfche(@_,2,8,0);
+}
+sub wfchesbs {
+	vfche(@_,2,8,1);
+}
+sub wfchexb {
+	vfche(@_,4,8,0);
+}
+sub wfchexbs {
+	vfche(@_,4,8,1);
+}
+
+sub vfdsb {
+	vfd(@_,2,0);
+}
+sub wfdsb {
+	vfd(@_,2,8);
+}
+sub wfdxb {
+	vfd(@_,4,8);
+}
+
+sub vfisb {
+	vfi(@_[0..1],2,@_[2..3]);
+}
+sub wfisb {
+	vfi(@_[0..1],2,0x8|$_[2],$_[3]);
+}
+sub wfixb {
+	vfi(@_[0..1],4,0x8|$_[2],$_[3]);
+}
+
+sub vfll {
+	vlde(@_);
+}
+sub vflls {
+	vfll(@_,2,0);
+}
+sub wflls {
+	vfll(@_,2,8);
+}
+sub wflld {
+	vfll(@_,3,8);
+}
+
+sub vflr {
+	vled(@_);
+}
+sub vflrd {
+	vflr(@_[0..1],3,@_[2..3]);
+}
+sub wflrd {
+	vflr(@_[0..1],3,0x8|$_[2],$_[3]);
+}
+sub wflrx {
+	vflr(@_[0..1],4,0x8|$_[2],$_[3]);
+}
+
+sub vfmax {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRc(0xe7ef,@_);
+}
+sub vfmaxsb {
+	vfmax(@_[0..2],2,0,$_[3]);
+}
+sub vfmaxdb {
+	vfmax(@_[0..2],3,0,$_[3]);
+}
+sub wfmaxsb {
+	vfmax(@_[0..2],2,8,$_[3]);
+}
+sub wfmaxdb {
+	vfmax(@_[0..2],3,8,$_[3]);
+}
+sub wfmaxxb {
+	vfmax(@_[0..2],4,8,$_[3]);
+}
+
+sub vfmin {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRc(0xe7ee,@_);
+}
+sub vfminsb {
+	vfmin(@_[0..2],2,0,$_[5]);
+}
+sub vfmindb {
+	vfmin(@_[0..2],3,0,$_[5]);
+}
+sub wfminsb {
+	vfmin(@_[0..2],2,8,$_[5]);
+}
+sub wfmindb {
+	vfmin(@_[0..2],3,8,$_[5]);
+}
+sub wfminxb {
+	vfmin(@_[0..2],4,8,$_[5]);
+}
+
+sub vfmsb {
+	vfm(@_,2,0);
+}
+sub wfmsb {
+	vfm(@_,2,8);
+}
+sub wfmxb {
+	vfm(@_,4,8);
+}
+
+sub vfmasb {
+	vfma(@_,0,2);
+}
+sub wfmasb {
+	vfma(@_,8,2);
+}
+sub wfmaxb {
+	vfma(@_,8,4);
+}
+
+sub vfmssb {
+	vfms(@_,0,2);
+}
+sub wfmssb {
+	vfms(@_,8,2);
+}
+sub wfmsxb {
+	vfms(@_,8,4);
+}
+
+sub vfnma {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRe(0xe79f,@_);
+}
+sub vfnmasb {
+	vfnma(@_,0,2);
+}
+sub vfnmadb {
+	vfnma(@_,0,3);
+}
+sub wfnmasb {
+	vfnma(@_,8,2);
+}
+sub wfnmadb {
+	vfnma(@_,8,3);
+}
+sub wfnmaxb {
+	vfnma(@_,8,4);
+}
+
+sub vfnms {
+	confess(err("ARGNUM")) if ($#_!=5);
+	VRRe(0xe79e,@_);
+}
+sub vfnmssb {
+	vfnms(@_,0,2);
+}
+sub vfnmsdb {
+	vfnms(@_,0,3);
+}
+sub wfnmssb {
+	vfnms(@_,8,2);
+}
+sub wfnmsdb {
+	vfnms(@_,8,3);
+}
+sub wfnmsxb {
+	vfnms(@_,8,4);
+}
+
+sub vfpsosb {
+	vfpso(@_[0..1],2,0,$_[2]);
+}
+sub wfpsosb {
+	vfpso(@_[0..1],2,8,$_[2]);
+}
+sub vflcsb {
+	vfpso(@_,2,0,0);
+}
+sub wflcsb {
+	vfpso(@_,2,8,0);
+}
+sub vflnsb {
+	vfpso(@_,2,0,1);
+}
+sub wflnsb {
+	vfpso(@_,2,8,1);
+}
+sub vflpsb {
+	vfpso(@_,2,0,2);
+}
+sub wflpsb {
+	vfpso(@_,2,8,2);
+}
+sub vfpsoxb {
+	vfpso(@_[0..1],4,0,$_[2]);
+}
+sub wfpsoxb {
+	vfpso(@_[0..1],4,8,$_[2]);
+}
+sub vflcxb {
+	vfpso(@_,4,0,0);
+}
+sub wflcxb {
+	vfpso(@_,4,8,0);
+}
+sub vflnxb {
+	vfpso(@_,4,0,1);
+}
+sub wflnxb {
+	vfpso(@_,4,8,1);
+}
+sub vflpxb {
+	vfpso(@_,4,0,2);
+}
+sub wflpxb {
+	vfpso(@_,4,8,2);
+}
+
+sub vfsqsb {
+	vfsq(@_,2,0);
+}
+sub wfsqsb {
+	vfsq(@_,2,8);
+}
+sub wfsqxb {
+	vfsq(@_,4,8);
+}
+
+sub vfssb {
+	vfs(@_,2,0);
+}
+sub wfssb {
+	vfs(@_,2,8);
+}
+sub wfsxb {
+	vfs(@_,4,8);
+}
+
+sub vftcisb {
+	vftci(@_,2,0);
+}
+sub wftcisb {
+	vftci(@_,2,8);
+}
+sub wftcixb {
+	vftci(@_,4,8);
+}
+
+# VXD - Support Instructions
+
+sub vlrlr {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRSd(0xe637,@_);
+}
+
+sub vlrl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VSI(0xe635,@_);
+}
+
+sub vstrlr {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRSd(0xe63f,@_);
+}
+
+sub vstrl {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VSI(0xe63d,@_);
+}
+
+sub vap {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe671,@_);
+}
+
+sub vcp {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRh(0xe677,@_);
+}
+
+sub vcvb {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRi(0xe650,@_);
+}
+
+sub vcvbg {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRRi(0xe652,@_);
+}
+
+sub vcvd {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRIi(0xe658,@_);
+}
+
+sub vcvdg {
+	confess(err("ARGNUM")) if ($#_!=3);
+	VRIi(0xe65a,@_);
+}
+
+sub vdp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe67a,@_);
+}
+
+sub vlip {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VRIh(0xe649,@_);
+}
+
+sub vmp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe678,@_);
+}
+
+sub vmsp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe679,@_);
+}
+
+sub vpkz {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VSI(0xe634,@_);
+}
+
+sub vpsop {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIg(0xe65b,@_);
+}
+
+sub vrp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe67b,@_);
+}
+
+sub vsdp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe67e,@_);
+}
+
+sub vsrp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIg(0xe659,@_);
+}
+
+sub vsp {
+	confess(err("ARGNUM")) if ($#_!=4);
+	VRIf(0xe673,@_);
+}
+
+sub vtp {
+	confess(err("ARGNUM")) if ($#_!=0);
+	VRRg(0xe65f,@_);
+}
+
+sub vupkz {
+	confess(err("ARGNUM")) if ($#_!=2);
+	VSI(0xe63c,@_);
+}
+
+#
+# Instruction Formats
+#
+
+sub RRE {
+	confess(err("ARGNUM")) if ($#_<0||2<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$r1,$r2)=(shift,get_R(shift),get_R(shift));
+
+	$out.="\t.long\t".sprintf("%#010x",($opcode<<16|$r1<<4|$r2));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub RRFb {
+	confess(err("ARGNUM")) if ($#_<3||4<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$r1,$r3,$r2,$m4)=(shift,get_R(shift),get_R(shift)
+	    ,get_R(shift),get_M(shift));
+
+	$out.="\t.long\t"
+	    .sprintf("%#010x",($opcode<<16|$r3<<12|$m4<<8|$r1<<4|$r2));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub S {
+	confess(err("ARGNUM")) if ($#_<0||1<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$d2,$b2)=(shift,get_DB(shift));
+
+	$out.="\t.long\t".sprintf("%#010x",($opcode<<16|$b2<<12|$d2));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIa {
+	confess(err("ARGNUM")) if ($#_<2||3<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$i2,$m3)=(shift,get_V(shift),get_I(shift,16),
+	    get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)).",";
+	$out.=sprintf("%#06x",$i2).",";
+	$out.=sprintf("%#06x",($m3<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIb {
+	confess(err("ARGNUM")) if ($#_!=4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$i2,$i3,$m4)=(shift,get_V(shift),get_I(shift,8),
+	    ,get_I(shift,8),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)).",";
+	$out.=sprintf("%#06x",($i2<<8|$i3)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIc {
+	confess(err("ARGNUM")) if ($#_!=4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v3,$i2,$m4)=(shift,get_V(shift),get_V(shift),
+	    ,get_I(shift,16),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|($v3&0xf)).",";
+	$out.=sprintf("%#06x",$i2).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRId {
+	confess(err("ARGNUM")) if ($#_<4||$#_>5);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$i4,$m5)=(shift,get_V(shift),get_V(shift),
+	    ,get_V(shift),get_I(shift,8),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|($v2&0xf)).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$i4)).",";
+	$out.=sprintf("%#06x",($m5<<12|RXB($v1,$v2,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIe {
+	confess(err("ARGNUM")) if ($#_!=5);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$i3,$m4,$m5)=(shift,get_V(shift),get_V(shift),
+	    ,get_I(shift,12),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|($v2&0xf)).",";
+	$out.=sprintf("%#06x",($i3<<4|$m5)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIf {
+	confess(err("ARGNUM")) if ($#_!=5);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$i4,$m5)=(shift,get_V(shift),get_V(shift),
+	    ,get_V(shift),get_I(shift,8),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|($v2&0xf)).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$m5<<4)|$i4>>4).",";
+	$out.=sprintf("%#06x",(($i4&0xf)<<12|RXB($v1,$v2,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIg {
+	confess(err("ARGNUM")) if ($#_!=5);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$i3,$i4,$m5)=(shift,get_V(shift),get_V(shift),
+	    ,get_I(shift,8),get_I(shift,8),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|($v2&0xf)).",";
+	$out.=sprintf("%#06x",($i4<<8|$m5<<4|$i3>>4)).",";
+	$out.=sprintf("%#06x",(($i3&0xf)<<12|RXB($v1,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIh {
+	confess(err("ARGNUM")) if ($#_!=3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$i2,$i3)=(shift,get_V(shift),get_I(shift,16),
+	    get_I(shift,4));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)).",";
+	$out.=sprintf("%#06x",$i2).",";
+	$out.=sprintf("%#06x",($i3<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRIi {
+	confess(err("ARGNUM")) if ($#_!=4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$r2,$i3,$m4)=(shift,get_V(shift),get_R(shift),
+	    ,get_I(shift,8),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4)|$r2).",";
+	$out.=sprintf("%#06x",($m4<<4|$i3>>4)).",";
+	$out.=sprintf("%#06x",(($i3&0xf)<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRa {
+	confess(err("ARGNUM")) if ($#_<2||5<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$m3,$m4,$m5)=(shift,get_V(shift),get_V(shift),
+	    get_M(shift),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",($m5<<4|$m4)).",";
+	$out.=sprintf("%#06x",($m3<<12|RXB($v1,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRb {
+	confess(err("ARGNUM")) if ($#_<3||5<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$m4,$m5)=(shift,get_V(shift),get_V(shift),
+	    get_V(shift),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$m5<<4)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1,$v2,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRc {
+	confess(err("ARGNUM")) if ($#_<3||6<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$m4,$m5,$m6)=(shift,get_V(shift),get_V(shift),
+	    get_V(shift),get_M(shift),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$m6<<4|$m5)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1,$v2,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRd {
+	confess(err("ARGNUM")) if ($#_<4||6<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$v4,$m5,$m6)=(shift,get_V(shift),get_V(shift),
+	    get_V(shift),get_V(shift),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$m5<<8|$m6<<4)).",";
+	$out.=sprintf("%#06x",(($v4&0xf)<<12|RXB($v1,$v2,$v3,$v4)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRe {
+	confess(err("ARGNUM")) if ($#_<4||6<$#_);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$v3,$v4,$m5,$m6)=(shift,get_V(shift),get_V(shift),
+	    get_V(shift),get_V(shift),get_M(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",(($v3&0xf)<<12|$m6<<8|$m5)).",";
+	$out.=sprintf("%#06x",(($v4&0xf)<<12|RXB($v1,$v2,$v3,$v4)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRf {
+	confess(err("ARGNUM")) if ($#_!=3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$r2,$r3)=(shift,get_V(shift),get_R(shift),
+	    get_R(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|$r2)).",";
+	$out.=sprintf("%#06x",($r3<<12)).",";
+	$out.=sprintf("%#06x",(RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRg {
+	confess(err("ARGNUM")) if ($#_!=1);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1)=(shift,get_V(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf))).",";
+	$out.=sprintf("%#06x",0x0000).",";
+	$out.=sprintf("%#06x",(RXB(0,$v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRh {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v2,$m3)=(shift,get_V(shift),get_V(shift),
+	    get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf))).",";
+	$out.=sprintf("%#06x",(($v2&0xf)<<12|$m3<<4)).",";
+	$out.=sprintf("%#06x",(RXB(0,$v1,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRRi {
+	confess(err("ARGNUM")) if ($#_!=3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$r1,$v2,$m3)=(shift,get_R(shift),get_V(shift),
+	    get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|$r1<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",($m3<<4))."\,";
+	$out.=sprintf("%#06x",(RXB(0,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRSa {
+	confess(err("ARGNUM")) if ($#_<3||$#_>4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$v3,$d2,$b2,$m4)=(shift,get_V(shift),get_V(shift),
+	    get_DB(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v3&0xf))).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRSb {
+	confess(err("ARGNUM")) if ($#_<3||$#_>4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$r3,$d2,$b2,$m4)=(shift,get_V(shift),get_R(shift),
+	    get_DB(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|$r3)).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRSc {
+	confess(err("ARGNUM")) if ($#_!=4);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$r1,$v3,$d2,$b2,$m4)=(shift,get_R(shift),get_V(shift),
+	    get_DB(shift),get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|$r1<<4|($v3&0xf))).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",($m4<<12|RXB(0,$v3)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRSd {
+	confess(err("ARGNUM")) if ($#_!=3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$r3,$d2,$b2)=(shift,get_V(shift),get_R(shift),
+	    get_DB(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|$r3)).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",(($v1&0xf)<<12|RXB(0,0,0,$v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRV {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$d2,$v2,$b2,$m3)=(shift,get_V(shift),get_DVB(shift),
+	    get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($v2&0xf))).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",($m3<<12|RXB($v1,$v2)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VRX {
+	confess(err("ARGNUM")) if ($#_<2||$#_>3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$d2,$x2,$b2,$m3)=(shift,get_V(shift),get_DXB(shift),
+	    get_M(shift));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|($v1&0xf)<<4|($x2))).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",($m3<<12|RXB($v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+sub VSI {
+	confess(err("ARGNUM")) if ($#_!=3);
+	my $ops=join(',',@_[1..$#_]);
+	my $memn=(caller(1))[3];
+	$memn=~s/^.*:://;
+	my ($opcode,$v1,$d2,$b2,$i3)=(shift,get_V(shift),get_DB(shift),
+	    get_I(shift,8));
+
+	$out.="\t.word\t";
+	$out.=sprintf("%#06x",($opcode&0xff00|$i3)).",";
+	$out.=sprintf("%#06x",($b2<<12|$d2)).",";
+	$out.=sprintf("%#06x",(($v1&0xf)<<12|RXB(0,0,0,$v1)<<8|$opcode&0xff));
+	$out.="\t# $memn\t$ops\n"
+}
+
+#
+# Internal
+#
+
+sub get_R {
+	confess(err("ARGNUM")) if ($#_!=0);
+	my $r;
+
+	for (shift) {
+		if (!defined) {
+			$r=0;
+		} elsif (/^$GR$/) {
+			$r=$1;
+		} else {
+			confess(err("PARSE"));
+		}
+	}
+	confess(err("ARGRANGE")) if ($r&~0xf);
+
+	return $r;
+}
+
+sub get_V {
+	confess(err("ARGNUM")) if ($#_!=0);
+	my $v;
+
+	for (shift) {
+		if (!defined) {
+			$v=0;
+		} elsif (/^$VR$/) {
+			$v=$1;
+		} else {
+			confess(err("PARSE"));
+		}
+	}
+	confess(err("ARGRANGE")) if ($v&~0x1f);
+
+	return $v;
+}
+
+sub get_I {
+	confess(err("ARGNUM")) if ($#_!=1);
+	my ($i,$bits)=(shift,shift);
+
+	$i=defined($i)?(eval($i)):(0);
+	confess(err("PARSE")) if (!defined($i));
+	confess(err("ARGRANGE")) if (abs($i)&~(2**$bits-1));
+
+	return $i&(2**$bits-1);
+}
+
+sub get_M {
+	confess(err("ARGNUM")) if ($#_!=0);
+	my $m=shift;
+
+	$m=defined($m)?(eval($m)):(0);
+	confess(err("PARSE")) if (!defined($m));
+	confess(err("ARGRANGE")) if ($m&~0xf);
+
+	return $m;
+}
+
+sub get_DB
+{
+	confess(err("ARGNUM")) if ($#_!=0);
+	my ($d,$b);
+
+	for (shift) {
+		if (!defined) {
+			($d,$b)=(0,0);
+		} elsif (/^(.+)\($GR\)$/) {
+			($d,$b)=(eval($1),$2);
+			confess(err("PARSE")) if (!defined($d));
+		} elsif (/^(.+)$/) {
+			($d,$b)=(eval($1),0);
+			confess(err("PARSE")) if (!defined($d));
+		} else {
+			confess(err("PARSE"));
+		}
+	}
+	confess(err("ARGRANGE")) if ($d&~0xfff||$b&~0xf);
+
+	return ($d,$b);
+}
+
+sub get_DVB
+{
+	confess(err("ARGNUM")) if ($#_!=0);
+	my ($d,$v,$b);
+
+	for (shift) {
+		if (!defined) {
+			($d,$v,$b)=(0,0,0);
+		} elsif (/^(.+)\($VR,$GR\)$/) {
+			($d,$v,$b)=(eval($1),$2,$3);
+			confess(err("PARSE")) if (!defined($d));
+		} elsif (/^(.+)\($GR\)$/) {
+			($d,$v,$b)=(eval($1),0,$2);
+			confess(err("PARSE")) if (!defined($d));
+		} elsif (/^(.+)$/) {
+			($d,$v,$b)=(eval($1),0,0);
+			confess(err("PARSE")) if (!defined($d));
+		} else {
+			confess(err("PARSE"));
+		}
+	}
+	confess(err("ARGRANGE")) if ($d&~0xfff||$v&~0x1f||$b&~0xf);
+
+	return ($d,$v,$b);
+}
+
+sub get_DXB
+{
+	confess(err("ARGNUM")) if ($#_!=0);
+	my ($d,$x,$b);
+
+	for (shift) {
+		if (!defined) {
+			($d,$x,$b)=(0,0,0);
+		} elsif (/^(.+)\($GR,$GR\)$/) {
+			($d,$x,$b)=(eval($1),$2,$3);
+			confess(err("PARSE")) if (!defined($d));
+		} elsif (/^(.+)\($GR\)$/) {
+			($d,$x,$b)=(eval($1),0,$2);
+			confess(err("PARSE")) if (!defined($d));
+		} elsif (/^(.+)$/) {
+			($d,$x,$b)=(eval($1),0,0);
+			confess(err("PARSE")) if (!defined($d));
+		} else {
+			confess(err("PARSE"));
+		}
+	}
+	confess(err("ARGRANGE")) if ($d&~0xfff||$x&~0xf||$b&~0xf);
+
+	return ($d,$x,$b);
+}
+
+sub RXB
+{
+	confess(err("ARGNUM")) if ($#_<0||3<$#_);
+	my $rxb=0;
+
+	$rxb|=0x08 if (defined($_[0])&&($_[0]&0x10));
+	$rxb|=0x04 if (defined($_[1])&&($_[1]&0x10));
+	$rxb|=0x02 if (defined($_[2])&&($_[2]&0x10));
+	$rxb|=0x01 if (defined($_[3])&&($_[3]&0x10));
+
+	return $rxb;
+}
+
+sub err {
+	my %ERR		=
+	(
+		ARGNUM	=>	'Wrong number of arguments',
+		ARGRANGE=>	'Argument out of range',
+		PARSE	=>	'Parse error',
+	);
+	confess($ERR{ARGNUM}) if ($#_!=0);
+
+	return $ERR{$_[0]};
+}
+
+1;
diff -up openssl-1.1.1b/crypto/poly1305/asm/poly1305-s390x.pl.s390x-update openssl-1.1.1b/crypto/poly1305/asm/poly1305-s390x.pl
--- openssl-1.1.1b/crypto/poly1305/asm/poly1305-s390x.pl.s390x-update	2019-02-26 15:15:30.000000000 +0100
+++ openssl-1.1.1b/crypto/poly1305/asm/poly1305-s390x.pl	2019-05-06 10:54:00.036367588 +0200
@@ -24,204 +24,961 @@
 #
 # On side note, z13 enables vector base 2^26 implementation...
 
-$flavour = shift;
+#
+# January 2019
+#
+# Add vx code path (base 2^26).
+#
+# Copyright IBM Corp. 2019
+# Author: Patrick Steuer <patrick.steuer@de.ibm.com>
+
+#
+# January 2019
+#
+# Add vector base 2^26 implementation. It's problematic to accurately
+# measure performance, because reference system is hardly idle. But
+# it's sub-cycle, i.e. less than 1 cycle per processed byte, and it's
+# >=20% faster than IBM's submission on long inputs, and much faster on
+# short ones, because calculation of key powers is postponed till we
+# know that input is long enough to justify the additional overhead.
+
+use strict;
+use FindBin qw($Bin);
+use lib "$Bin/../..";
+use perlasm::s390x qw(:DEFAULT :VX AUTOLOAD LABEL INCLUDE);
+
+my $flavour = shift;
 
+my ($z,$SIZE_T);
 if ($flavour =~ /3[12]/) {
+	$z=0;	# S/390 ABI
 	$SIZE_T=4;
-	$g="";
 } else {
+	$z=1;	# zSeries ABI
 	$SIZE_T=8;
-	$g="g";
 }
 
+my $output;
 while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
-open STDOUT,">$output";
 
-$sp="%r15";
+my $stdframe=16*$SIZE_T+4*8;
+my $sp="%r15";
 
 my ($ctx,$inp,$len,$padbit) = map("%r$_",(2..5));
 
-$code.=<<___;
-.text
+PERLASM_BEGIN($output);
 
-.globl	poly1305_init
-.type	poly1305_init,\@function
-.align	16
-poly1305_init:
-	lghi	%r0,0
-	lghi	%r1,-1
-	stg	%r0,0($ctx)		# zero hash value
-	stg	%r0,8($ctx)
-	stg	%r0,16($ctx)
-
-	cl${g}r	$inp,%r0
-	je	.Lno_key
-
-	lrvg	%r4,0($inp)		# load little-endian key
-	lrvg	%r5,8($inp)
-
-	nihl	%r1,0xffc0		# 0xffffffc0ffffffff
-	srlg	%r0,%r1,4		# 0x0ffffffc0fffffff
-	srlg	%r1,%r1,4
-	nill	%r1,0xfffc		# 0x0ffffffc0ffffffc
-
-	ngr	%r4,%r0
-	ngr	%r5,%r1
-
-	stg	%r4,32($ctx)
-	stg	%r5,40($ctx)
-
-.Lno_key:
-	lghi	%r2,0
-	br	%r14
-.size	poly1305_init,.-poly1305_init
-___
+INCLUDE	("s390x_arch.h");
+TEXT	();
+
+################
+# static void poly1305_init(void *ctx, const unsigned char key[16])
+{
+GLOBL	("poly1305_init");
+TYPE	("poly1305_init","\@function");
+ALIGN	(16);
+LABEL	("poly1305_init");
+	lghi	("%r0",0);
+	lghi	("%r1",-1);
+	stg	("%r0","0($ctx)");		# zero hash value
+	stg	("%r0","8($ctx)");
+	stg	("%r0","16($ctx)");
+	st	("%r0","24($ctx)");		# clear is_base2_26
+	lgr	("%r5",$ctx);			# reassign $ctx
+	lghi	("%r2",0);
+
+&{$z?	\&clgr:\&clr}	($inp,"%r0");
+	je	(".Lno_key");
+
+	lrvg	("%r2","0($inp)");		# load little-endian key
+	lrvg	("%r3","8($inp)");
+
+	nihl	("%r1",0xffc0);			# 0xffffffc0ffffffff
+	srlg	("%r0","%r1",4);		# 0x0ffffffc0fffffff
+	srlg	("%r1","%r1",4);
+	nill	("%r1",0xfffc);			# 0x0ffffffc0ffffffc
+
+	ngr	("%r2","%r0");
+	ngr	("%r3","%r1");
+
+	stmg	("%r2","%r3","32(%r5)");
+
+	larl	("%r1","OPENSSL_s390xcap_P");
+	lg	("%r0","16(%r1)");
+	srlg	("%r0","%r0",62);
+	nill	("%r0",1);			# extract vx bit
+	lcgr	("%r0","%r0");
+	larl	("%r1",".Lpoly1305_blocks");
+	larl	("%r2",".Lpoly1305_blocks_vx");
+	larl	("%r3",".Lpoly1305_emit");
+&{$z?	\&xgr:\&xr}	("%r2","%r1");		# select between scalar and vector
+&{$z?	\&ngr:\&nr}	("%r2","%r0");
+&{$z?	\&xgr:\&xr}	("%r2","%r1");
+&{$z?	\&stmg:\&stm}	("%r2","%r3","0(%r4)");
+	lghi	("%r2",1);
+LABEL	(".Lno_key");
+	br	("%r14");
+SIZE	("poly1305_init",".-poly1305_init");
+}
+
+################
+# static void poly1305_blocks(void *ctx, const unsigned char *inp,
+#                             size_t len, u32 padbit)
 {
 my ($d0hi,$d0lo,$d1hi,$d1lo,$t0,$h0,$t1,$h1,$h2) = map("%r$_",(6..14));
 my ($r0,$r1,$s1) = map("%r$_",(0..2));
 
-$code.=<<___;
-.globl	poly1305_blocks
-.type	poly1305_blocks,\@function
-.align	16
-poly1305_blocks:
-	srl${g}	$len,4			# fixed-up in 64-bit build
-	lghi	%r0,0
-	cl${g}r	$len,%r0
-	je	.Lno_data
-
-	stm${g}	%r6,%r14,`6*$SIZE_T`($sp)
-
-	llgfr   $padbit,$padbit		# clear upper half, much needed with
-					# non-64-bit ABI
-	lg	$r0,32($ctx)		# load key
-	lg	$r1,40($ctx)
-
-	lg	$h0,0($ctx)		# load hash value
-	lg	$h1,8($ctx)
-	lg	$h2,16($ctx)
-
-	st$g	$ctx,`2*$SIZE_T`($sp)	# off-load $ctx
-	srlg	$s1,$r1,2
-	algr	$s1,$r1			# s1 = r1 + r1>>2
-	j	.Loop
-
-.align	16
-.Loop:
-	lrvg	$d0lo,0($inp)		# load little-endian input
-	lrvg	$d1lo,8($inp)
-	la	$inp,16($inp)
-
-	algr	$d0lo,$h0		# accumulate input
-	alcgr	$d1lo,$h1
-
-	lgr	$h0,$d0lo
-	mlgr	$d0hi,$r0		# h0*r0	  -> $d0hi:$d0lo
-	lgr	$h1,$d1lo
-	mlgr	$d1hi,$s1		# h1*5*r1 -> $d1hi:$d1lo
-
-	mlgr	$t0,$r1			# h0*r1   -> $t0:$h0
-	mlgr	$t1,$r0			# h1*r0   -> $t1:$h1
-	alcgr	$h2,$padbit
-
-	algr	$d0lo,$d1lo
-	lgr	$d1lo,$h2
-	alcgr	$d0hi,$d1hi
-	lghi	$d1hi,0
-
-	algr	$h1,$h0
-	alcgr	$t1,$t0
-
-	msgr	$d1lo,$s1		# h2*s1
-	msgr	$h2,$r0			# h2*r0
-
-	algr	$h1,$d1lo
-	alcgr	$t1,$d1hi		# $d1hi is zero
-
-	algr	$h1,$d0hi
-	alcgr	$h2,$t1
-
-	lghi	$h0,-4			# final reduction step
-	ngr	$h0,$h2
-	srlg	$t0,$h2,2
-	algr	$h0,$t0
-	lghi	$t1,3
-	ngr	$h2,$t1
-
-	algr	$h0,$d0lo
-	alcgr	$h1,$d1hi		# $d1hi is still zero
-	alcgr	$h2,$d1hi		# $d1hi is still zero
-
-	brct$g	$len,.Loop
-
-	l$g	$ctx,`2*$SIZE_T`($sp)	# restore $ctx
-
-	stg	$h0,0($ctx)		# store hash value
-	stg	$h1,8($ctx)
-	stg	$h2,16($ctx)
-
-	lm${g}	%r6,%r14,`6*$SIZE_T`($sp)
-.Lno_data:
-	br	%r14
-.size	poly1305_blocks,.-poly1305_blocks
-___
+GLOBL	("poly1305_blocks");
+TYPE	("poly1305_blocks","\@function");
+ALIGN	(16);
+LABEL	("poly1305_blocks");
+LABEL	(".Lpoly1305_blocks");
+&{$z?	\&ltgr:\&ltr}	("%r0",$len);
+	jz	(".Lno_data");
+
+&{$z?	\&stmg:\&stm}	("%r6","%r14","6*$SIZE_T($sp)");
+
+	lg	($h0,"0($ctx)");		# load hash value
+	lg	($h1,"8($ctx)");
+	lg	($h2,"16($ctx)");
+
+LABEL	(".Lpoly1305_blocks_entry");
+if ($z) {
+	srlg	($len,$len,4);
+} else {
+	srl	($len,4);
+}
+	llgfr   ($padbit,$padbit);		# clear upper half, much needed with
+						# non-64-bit ABI
+	lg	($r0,"32($ctx)");		# load key
+	lg	($r1,"40($ctx)");
+
+&{$z?	\&stg:\&st}	($ctx,"2*$SIZE_T($sp)");	# off-load $ctx
+	srlg	($s1,$r1,2);
+	algr	($s1,$r1);			# s1 = r1 + r1>>2
+	j	(".Loop");
+
+ALIGN	(16);
+LABEL	(".Loop");
+	lrvg	($d0lo,"0($inp)");		# load little-endian input
+	lrvg	($d1lo,"8($inp)");
+	la	($inp,"16($inp)");
+
+	algr	($d0lo,$h0);			# accumulate input
+	alcgr	($d1lo,$h1);
+	alcgr	($h2,$padbit);
+
+	lgr	($h0,$d0lo);
+	mlgr	($d0hi,$r0);			# h0*r0	  -> $d0hi:$d0lo
+	lgr	($h1,$d1lo);
+	mlgr	($d1hi,$s1);			# h1*5*r1 -> $d1hi:$d1lo
+
+	mlgr	($t0,$r1);			# h0*r1   -> $t0:$h0
+	mlgr	($t1,$r0);			# h1*r0   -> $t1:$h1
+
+	algr	($d0lo,$d1lo);
+	lgr	($d1lo,$h2);
+	alcgr	($d0hi,$d1hi);
+	lghi	($d1hi,0);
+
+	algr	($h1,$h0);
+	alcgr	($t1,$t0);
+
+	msgr	($d1lo,$s1);			# h2*s1
+	msgr	($h2,$r0);			# h2*r0
+
+	algr	($h1,$d1lo);
+	alcgr	($t1,$d1hi);			# $d1hi is zero
+
+	algr	($h1,$d0hi);
+	alcgr	($h2,$t1);
+
+	lghi	($h0,-4);			# final reduction step
+	ngr	($h0,$h2);
+	srlg	($t0,$h2,2);
+	algr	($h0,$t0);
+	lghi	($t1,3);
+	ngr	($h2,$t1);
+
+	algr	($h0,$d0lo);
+	alcgr	($h1,$d1hi);			# $d1hi is still zero
+	alcgr	($h2,$d1hi);			# $d1hi is still zero
+
+&{$z?	\&brctg:\&brct}	($len,".Loop");
+
+&{$z?	\&lg:\&l}	($ctx,"2*$SIZE_T($sp)");# restore $ctx
+
+	stg	($h0,"0($ctx)");		# store hash value
+	stg	($h1,"8($ctx)");
+	stg	($h2,"16($ctx)");
+
+&{$z?	\&lmg:\&lm}	("%r6","%r14","6*$SIZE_T($sp)");
+LABEL	(".Lno_data");
+	br	("%r14");
+SIZE	("poly1305_blocks",".-poly1305_blocks");
 }
+
+################
+# static void poly1305_blocks_vx(void *ctx, const unsigned char *inp,
+#                                size_t len, u32 padbit)
+{
+my ($H0, $H1, $H2, $H3, $H4) = map("%v$_",(0..4));
+my ($I0, $I1, $I2, $I3, $I4) = map("%v$_",(5..9));
+my ($R0, $R1, $S1, $R2, $S2) = map("%v$_",(10..14));
+my      ($R3, $S3, $R4, $S4) = map("%v$_",(15..18));
+my ($ACC0, $ACC1, $ACC2, $ACC3, $ACC4) = map("%v$_",(19..23));
+my      ($T1, $T2, $T3, $T4) = map("%v$_",(24..27));
+my ($mask26,$bswaplo,$bswaphi,$bswapmi) = map("%v$_",(28..31));
+
+my ($d2,$d0,$h0,$d1,$h1,$h2)=map("%r$_",(9..14));
+
+TYPE	("poly1305_blocks_vx","\@function");
+ALIGN	(16);
+LABEL	("poly1305_blocks_vx");
+LABEL	(".Lpoly1305_blocks_vx");
+&{$z?	\&clgfi:\&clfi} ($len,128);
+	jhe	("__poly1305_blocks_vx");
+
+&{$z?	\&stmg:\&stm}	("%r6","%r14","6*$SIZE_T($sp)");
+
+	lg	($d0,"0($ctx)");
+	lg	($d1,"8($ctx)");
+	lg	($d2,"16($ctx)");
+
+	llgfr	("%r0",$d0);			# base 2^26 -> base 2^64
+	srlg	($h0,$d0,32);
+	llgfr	("%r1",$d1);
+	srlg	($h1,$d1,32);
+	srlg	($h2,$d2,32);
+
+	sllg	("%r0","%r0",26);
+	algr	($h0,"%r0");
+	sllg	("%r0",$h1,52);
+	srlg	($h1,$h1,12);
+	sllg	("%r1","%r1",14);
+	algr	($h0,"%r0");
+	alcgr	($h1,"%r1");
+	sllg	("%r0",$h2,40);
+	srlg	($h2,$h2,24);
+	lghi	("%r1",0);
+	algr	($h1,"%r0");
+	alcgr	($h2,"%r1");
+
+	llgf	("%r0","24($ctx)");		# is_base2_26
+	lcgr	("%r0","%r0");
+
+	xgr	($h0,$d0);			# choose between radixes
+	xgr	($h1,$d1);
+	xgr	($h2,$d2);
+	ngr	($h0,"%r0");
+	ngr	($h1,"%r0");
+	ngr	($h2,"%r0");
+	xgr	($h0,$d0);
+	xgr	($h1,$d1);
+	xgr	($h2,$d2);
+
+	lhi	("%r0",0);
+	st	("%r0","24($ctx)");		# clear is_base2_26
+
+	j	(".Lpoly1305_blocks_entry");
+SIZE	("poly1305_blocks_vx",".-poly1305_blocks_vx");
+
+TYPE	("__poly1305_mul","\@function");
+ALIGN	(16);
+LABEL	("__poly1305_mul");
+	vmlof		($ACC0,$H0,$R0);
+	vmlof		($ACC1,$H0,$R1);
+	vmlof		($ACC2,$H0,$R2);
+	vmlof		($ACC3,$H0,$R3);
+	vmlof		($ACC4,$H0,$R4);
+
+	vmalof		($ACC0,$H1,$S4,$ACC0);
+	vmalof		($ACC1,$H1,$R0,$ACC1);
+	vmalof		($ACC2,$H1,$R1,$ACC2);
+	vmalof		($ACC3,$H1,$R2,$ACC3);
+	vmalof		($ACC4,$H1,$R3,$ACC4);
+
+	vmalof		($ACC0,$H2,$S3,$ACC0);
+	vmalof		($ACC1,$H2,$S4,$ACC1);
+	vmalof		($ACC2,$H2,$R0,$ACC2);
+	vmalof		($ACC3,$H2,$R1,$ACC3);
+	vmalof		($ACC4,$H2,$R2,$ACC4);
+
+	vmalof		($ACC0,$H3,$S2,$ACC0);
+	vmalof		($ACC1,$H3,$S3,$ACC1);
+	vmalof		($ACC2,$H3,$S4,$ACC2);
+	vmalof		($ACC3,$H3,$R0,$ACC3);
+	vmalof		($ACC4,$H3,$R1,$ACC4);
+
+	vmalof		($ACC0,$H4,$S1,$ACC0);
+	vmalof		($ACC1,$H4,$S2,$ACC1);
+	vmalof		($ACC2,$H4,$S3,$ACC2);
+	vmalof		($ACC3,$H4,$S4,$ACC3);
+	vmalof		($ACC4,$H4,$R0,$ACC4);
+
+	################################################################
+	# lazy reduction
+
+	vesrlg		($H4,$ACC3,26);
+	vesrlg		($H1,$ACC0,26);
+	vn		($H3,$ACC3,$mask26);
+	vn		($H0,$ACC0,$mask26);
+	vag		($H4,$H4,$ACC4);	# h3 -> h4
+	vag		($H1,$H1,$ACC1);	# h0 -> h1
+
+	vesrlg		($ACC4,$H4,26);
+	vesrlg		($ACC1,$H1,26);
+	vn		($H4,$H4,$mask26);
+	vn		($H1,$H1,$mask26);
+	vag		($H0,$H0,$ACC4);
+	vag		($H2,$ACC2,$ACC1);	# h1 -> h2
+
+	veslg		($ACC4,$ACC4,2);	# <<2
+	vesrlg		($ACC2,$H2,26);
+	vn		($H2,$H2,$mask26);
+	vag		($H0,$H0,$ACC4);	# h4 -> h0
+	vag		($H3,$H3,$ACC2);	# h2 -> h3
+
+	vesrlg		($ACC0,$H0,26);
+	vesrlg		($ACC3,$H3,26);
+	vn		($H0,$H0,$mask26);
+	vn		($H3,$H3,$mask26);
+	vag		($H1,$H1,$ACC0);	# h0 -> h1
+	vag		($H4,$H4,$ACC3);	# h3 -> h4
+	br		("%r14");
+SIZE	("__poly1305_mul",".-__poly1305_mul");
+
+TYPE	("__poly1305_blocks_vx","\@function");
+ALIGN	(16);
+LABEL	("__poly1305_blocks_vx");
+&{$z?	\&lgr:\&lr}	("%r0",$sp);
+&{$z?	\&stmg:\&stm}	("%r10","%r15","10*$SIZE_T($sp)");
+if (!$z) {
+	std	("%f4","16*$SIZE_T+2*8($sp)");
+	std	("%f6","16*$SIZE_T+3*8($sp)");
+	ahi	($sp,-$stdframe);
+	st	("%r0","0($sp)");		# back-chain
+
+	llgfr	($len,$len);			# so that srlg works on $len
+} else {
+	aghi	($sp,"-($stdframe+8*8)");
+	stg	("%r0","0($sp)");		# back-chain
+
+	std	("%f8","$stdframe+0*8($sp)");
+	std	("%f9","$stdframe+1*8($sp)");
+	std	("%f10","$stdframe+2*8($sp)");
+	std	("%f11","$stdframe+3*8($sp)");
+	std	("%f12","$stdframe+4*8($sp)");
+	std	("%f13","$stdframe+5*8($sp)");
+	std	("%f14","$stdframe+6*8($sp)");
+	std	("%f15","$stdframe+7*8($sp)");
+}
+	larl	("%r1",".Lconst");
+	vgmg	($mask26,38,63);
+	vlm	($bswaplo,$bswapmi,"16(%r1)");
+
+	&lt	("%r0","24($ctx)");		# is_base2_26?
+	jnz	(".Lskip_init");
+
+	lg	($h0,"32($ctx)");		# load key base 2^64
+	lg	($h1,"40($ctx)");
+
+	risbg	($d0,$h0,38,0x80+63,38);	# base 2^64 -> 2^26
+	srlg	($d1,$h0,52);
+	risbg	($h0,$h0,38,0x80+63,0);
+	vlvgg	($R0,$h0,0);
+	risbg	($d1,$h1,38,51,12);
+	vlvgg	($R1,$d0,0);
+	risbg	($d0,$h1,38,63,50);
+	vlvgg	($R2,$d1,0);
+	srlg	($d1,$h1,40);
+	vlvgg	($R3,$d0,0);
+	vlvgg	($R4,$d1,0);
+
+	veslg	($S1,$R1,2);
+	veslg	($S2,$R2,2);
+	veslg	($S3,$R3,2);
+	veslg	($S4,$R4,2);
+	vlr	($H0,$R0);
+	vlr	($H1,$R1);
+	vlr	($H2,$R2);
+	vlr	($H3,$R3);
+	vlr	($H4,$R4);
+	vag	($S1,$S1,$R1);			# * 5
+	vag	($S2,$S2,$R2);
+	vag	($S3,$S3,$R3);
+	vag	($S4,$S4,$R4);
+
+	brasl	("%r14","__poly1305_mul");	# r^1:- * r^1:-
+
+	vpdi	($R0,$H0,$R0,0);		# r^2:r^1
+	vpdi	($R1,$H1,$R1,0);
+	vpdi	($R2,$H2,$R2,0);
+	vpdi	($R3,$H3,$R3,0);
+	vpdi	($R4,$H4,$R4,0);
+	vpdi	($H0,$H0,$H0,0);		# r^2:r^2
+	vpdi	($H1,$H1,$H1,0);
+	vpdi	($H2,$H2,$H2,0);
+	vpdi	($H3,$H3,$H3,0);
+	vpdi	($H4,$H4,$H4,0);
+	veslg	($S1,$R1,2);
+	veslg	($S2,$R2,2);
+	veslg	($S3,$R3,2);
+	veslg	($S4,$R4,2);
+	vag	($S1,$S1,$R1);			# * 5
+	vag	($S2,$S2,$R2);
+	vag	($S3,$S3,$R3);
+	vag	($S4,$S4,$R4);
+
+	brasl	("%r14,__poly1305_mul");	# r^2:r^2 * r^2:r^1
+
+	vl	($I0,"0(%r1)");			# borrow $I0
+	vperm	($R0,$R0,$H0,$I0);		# r^2:r^4:r^1:r^3
+	vperm	($R1,$R1,$H1,$I0);
+	vperm	($R2,$R2,$H2,$I0);
+	vperm	($R3,$R3,$H3,$I0);
+	vperm	($R4,$R4,$H4,$I0);
+	veslf	($S1,$R1,2);
+	veslf	($S2,$R2,2);
+	veslf	($S3,$R3,2);
+	veslf	($S4,$R4,2);
+	vaf	($S1,$S1,$R1);			# * 5
+	vaf	($S2,$S2,$R2);
+	vaf	($S3,$S3,$R3);
+	vaf	($S4,$S4,$R4);
+
+	lg	($h0,"0($ctx)");		# load hash base 2^64
+	lg	($h1,"8($ctx)");
+	lg	($h2,"16($ctx)");
+
+	vzero	($H0);
+	vzero	($H1);
+	vzero	($H2);
+	vzero	($H3);
+	vzero	($H4);
+
+	risbg	($d0,$h0,38,0x80+63,38);	# base 2^64 -> 2^26
+	srlg	($d1,$h0,52);
+	risbg	($h0,$h0,38,0x80+63,0);
+	vlvgg	($H0,$h0,0);
+	risbg	($d1,$h1,38,51,12);
+	vlvgg	($H1,$d0,0);
+	risbg	($d0,$h1,38,63,50);
+	vlvgg	($H2,$d1,0);
+	srlg	($d1,$h1,40);
+	vlvgg	($H3,$d0,0);
+	risbg	($d1,$h2,37,39,24);
+	vlvgg	($H4,$d1,0);
+
+	lhi	("%r0",1);
+	st	("%r0","24($ctx)");		# set is_base2_26
+
+	vstm	($R0,$S4,"48($ctx)");		# save key schedule base 2^26
+
+	vpdi	($R0,$R0,$R0,0);		# broadcast r^2:r^4
+	vpdi	($R1,$R1,$R1,0);
+	vpdi	($S1,$S1,$S1,0);
+	vpdi	($R2,$R2,$R2,0);
+	vpdi	($S2,$S2,$S2,0);
+	vpdi	($R3,$R3,$R3,0);
+	vpdi	($S3,$S3,$S3,0);
+	vpdi	($R4,$R4,$R4,0);
+	vpdi	($S4,$S4,$S4,0);
+
+	j	(".Loaded_hash");
+
+ALIGN	(16);
+LABEL	(".Lskip_init");
+	vllezf	($H0,"0($ctx)");		# load hash base 2^26
+	vllezf	($H1,"4($ctx)");
+	vllezf	($H2,"8($ctx)");
+	vllezf	($H3,"12($ctx)");
+	vllezf	($H4,"16($ctx)");
+
+	vlrepg	($R0,"0x30($ctx)");		# broadcast r^2:r^4
+	vlrepg	($R1,"0x40($ctx)");
+	vlrepg	($S1,"0x50($ctx)");
+	vlrepg	($R2,"0x60($ctx)");
+	vlrepg	($S2,"0x70($ctx)");
+	vlrepg	($R3,"0x80($ctx)");
+	vlrepg	($S3,"0x90($ctx)");
+	vlrepg	($R4,"0xa0($ctx)");
+	vlrepg	($S4,"0xb0($ctx)");
+
+LABEL	(".Loaded_hash");
+	vzero	($I1);
+	vzero	($I3);
+
+	vlm	($T1,$T4,"0x00($inp)");		# load first input block
+	la	($inp,"0x40($inp)");
+	vgmg	($mask26,6,31);
+	vgmf	($I4,5,5);			# padbit<<2
+
+	vperm	($I0,$T3,$T4,$bswaplo);
+	vperm	($I2,$T3,$T4,$bswapmi);
+	vperm	($T3,$T3,$T4,$bswaphi);
+
+	verimg	($I1,$I0,$mask26,6);		# >>26
+	veslg	($I0,$I0,32);
+	veslg	($I2,$I2,28);			# >>4
+	verimg	($I3,$T3,$mask26,18);		# >>14
+	verimg	($I4,$T3,$mask26,58);		# >>38
+	vn	($I0,$I0,$mask26);
+	vn	($I2,$I2,$mask26);
+	vesrlf	($I4,$I4,2);			# >>2
+
+	vgmg	($mask26,38,63);
+	vperm	($T3,$T1,$T2,$bswaplo);
+	vperm	($T4,$T1,$T2,$bswaphi);
+	vperm	($T2,$T1,$T2,$bswapmi);
+
+	verimg	($I0,$T3,$mask26,0);
+	verimg	($I1,$T3,$mask26,38);		# >>26
+	verimg	($I2,$T2,$mask26,60);		# >>4
+	verimg	($I3,$T4,$mask26,50);		# >>14
+	vesrlg	($T4,$T4,40);
+	vo	($I4,$I4,$T4);
+
+	srlg	("%r0",$len,6);
+&{$z?	\&aghi:\&ahi}	("%r0",-1);
+
+ALIGN	(16);
+LABEL	(".Loop_vx");
+	vmlef		($ACC0,$I0,$R0);
+	vmlef		($ACC1,$I0,$R1);
+	vmlef		($ACC2,$I0,$R2);
+	vmlef		($ACC3,$I0,$R3);
+	vmlef		($ACC4,$I0,$R4);
+
+	vmalef		($ACC0,$I1,$S4,$ACC0);
+	vmalef		($ACC1,$I1,$R0,$ACC1);
+	vmalef		($ACC2,$I1,$R1,$ACC2);
+	vmalef		($ACC3,$I1,$R2,$ACC3);
+	vmalef		($ACC4,$I1,$R3,$ACC4);
+
+	 vaf		($H2,$H2,$I2);
+	 vaf		($H0,$H0,$I0);
+	 vaf		($H3,$H3,$I3);
+	 vaf		($H1,$H1,$I1);
+	 vaf		($H4,$H4,$I4);
+
+	vmalef		($ACC0,$I2,$S3,$ACC0);
+	vmalef		($ACC1,$I2,$S4,$ACC1);
+	vmalef		($ACC2,$I2,$R0,$ACC2);
+	vmalef		($ACC3,$I2,$R1,$ACC3);
+	vmalef		($ACC4,$I2,$R2,$ACC4);
+
+	 vlm		($T1,$T4,"0x00($inp)");	# load next input block
+	 la		($inp,"0x40($inp)");
+	 vgmg		($mask26,6,31);
+
+	vmalef		($ACC0,$I3,$S2,$ACC0);
+	vmalef		($ACC1,$I3,$S3,$ACC1);
+	vmalef		($ACC2,$I3,$S4,$ACC2);
+	vmalef		($ACC3,$I3,$R0,$ACC3);
+	vmalef		($ACC4,$I3,$R1,$ACC4);
+
+	 vperm		($I0,$T3,$T4,$bswaplo);
+	 vperm		($I2,$T3,$T4,$bswapmi);
+	 vperm		($T3,$T3,$T4,$bswaphi);
+
+	vmalef		($ACC0,$I4,$S1,$ACC0);
+	vmalef		($ACC1,$I4,$S2,$ACC1);
+	vmalef		($ACC2,$I4,$S3,$ACC2);
+	vmalef		($ACC3,$I4,$S4,$ACC3);
+	vmalef		($ACC4,$I4,$R0,$ACC4);
+
+	 verimg		($I1,$I0,$mask26,6);	# >>26
+	 veslg		($I0,$I0,32);
+	 veslg		($I2,$I2,28);		# >>4
+	 verimg		($I3,$T3,$mask26,18);	# >>14
+
+	vmalof		($ACC0,$H0,$R0,$ACC0);
+	vmalof		($ACC1,$H0,$R1,$ACC1);
+	vmalof		($ACC2,$H0,$R2,$ACC2);
+	vmalof		($ACC3,$H0,$R3,$ACC3);
+	vmalof		($ACC4,$H0,$R4,$ACC4);
+
+	 vgmf		($I4,5,5);		# padbit<<2
+	 verimg		($I4,$T3,$mask26,58);	# >>38
+	 vn		($I0,$I0,$mask26);
+	 vn		($I2,$I2,$mask26);
+	 vesrlf		($I4,$I4,2);		# >>2
+
+	vmalof		($ACC0,$H1,$S4,$ACC0);
+	vmalof		($ACC1,$H1,$R0,$ACC1);
+	vmalof		($ACC2,$H1,$R1,$ACC2);
+	vmalof		($ACC3,$H1,$R2,$ACC3);
+	vmalof		($ACC4,$H1,$R3,$ACC4);
+
+	 vgmg		($mask26,38,63);
+	 vperm		($T3,$T1,$T2,$bswaplo);
+	 vperm		($T4,$T1,$T2,$bswaphi);
+	 vperm		($T2,$T1,$T2,$bswapmi);
+
+	vmalof		($ACC0,$H2,$S3,$ACC0);
+	vmalof		($ACC1,$H2,$S4,$ACC1);
+	vmalof		($ACC2,$H2,$R0,$ACC2);
+	vmalof		($ACC3,$H2,$R1,$ACC3);
+	vmalof		($ACC4,$H2,$R2,$ACC4);
+
+	 verimg		($I0,$T3,$mask26,0);
+	 verimg		($I1,$T3,$mask26,38);	# >>26
+	 verimg		($I2,$T2,$mask26,60);	# >>4
+
+	vmalof		($ACC0,$H3,$S2,$ACC0);
+	vmalof		($ACC1,$H3,$S3,$ACC1);
+	vmalof		($ACC2,$H3,$S4,$ACC2);
+	vmalof		($ACC3,$H3,$R0,$ACC3);
+	vmalof		($ACC4,$H3,$R1,$ACC4);
+
+	 verimg		($I3,$T4,$mask26,50);	# >>14
+	 vesrlg		($T4,$T4,40);
+	 vo		($I4,$I4,$T4);
+
+	vmalof		($ACC0,$H4,$S1,$ACC0);
+	vmalof		($ACC1,$H4,$S2,$ACC1);
+	vmalof		($ACC2,$H4,$S3,$ACC2);
+	vmalof		($ACC3,$H4,$S4,$ACC3);
+	vmalof		($ACC4,$H4,$R0,$ACC4);
+
+	################################################################
+	# lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
+	# and P. Schwabe
+
+	vesrlg		($H4,$ACC3,26);
+	vesrlg		($H1,$ACC0,26);
+	vn		($H3,$ACC3,$mask26);
+	vn		($H0,$ACC0,$mask26);
+	vag		($H4,$H4,$ACC4);	# h3 -> h4
+	vag		($H1,$H1,$ACC1);	# h0 -> h1
+
+	vesrlg		($ACC4,$H4,26);
+	vesrlg		($ACC1,$H1,26);
+	vn		($H4,$H4,$mask26);
+	vn		($H1,$H1,$mask26);
+	vag		($H0,$H0,$ACC4);
+	vag		($H2,$ACC2,$ACC1);	# h1 -> h2
+
+	veslg		($ACC4,$ACC4,2);	# <<2
+	vesrlg		($ACC2,$H2,26);
+	vn		($H2,$H2,$mask26);
+	vag		($H0,$H0,$ACC4);	# h4 -> h0
+	vag		($H3,$H3,$ACC2);	# h2 -> h3
+
+	vesrlg		($ACC0,$H0,26);
+	vesrlg		($ACC3,$H3,26);
+	vn		($H0,$H0,$mask26);
+	vn		($H3,$H3,$mask26);
+	vag		($H1,$H1,$ACC0);	# h0 -> h1
+	vag		($H4,$H4,$ACC3);	# h3 -> h4
+
+&{$z?	\&brctg:\&brct}	("%r0",".Loop_vx");
+
+	vlm	($R0,$S4,"48($ctx)");		# load all powers
+
+	lghi	("%r0",0x30);
+&{$z?	\&lcgr:\&lcr}	($len,$len);
+&{$z?	\&ngr:\&nr}	($len,"%r0");
+&{$z?	\&slgr:\&slr}	($inp,$len);
+
+LABEL	(".Last");
+	vmlef	($ACC0,$I0,$R0);
+	vmlef	($ACC1,$I0,$R1);
+	vmlef	($ACC2,$I0,$R2);
+	vmlef	($ACC3,$I0,$R3);
+	vmlef	($ACC4,$I0,$R4);
+
+	vmalef	($ACC0,$I1,$S4,$ACC0);
+	vmalef	($ACC1,$I1,$R0,$ACC1);
+	vmalef	($ACC2,$I1,$R1,$ACC2);
+	vmalef	($ACC3,$I1,$R2,$ACC3);
+	vmalef	($ACC4,$I1,$R3,$ACC4);
+
+	 vaf	($H0,$H0,$I0);
+	 vaf	($H1,$H1,$I1);
+	 vaf	($H2,$H2,$I2);
+	 vaf	($H3,$H3,$I3);
+	 vaf	($H4,$H4,$I4);
+
+	vmalef	($ACC0,$I2,$S3,$ACC0);
+	vmalef	($ACC1,$I2,$S4,$ACC1);
+	vmalef	($ACC2,$I2,$R0,$ACC2);
+	vmalef	($ACC3,$I2,$R1,$ACC3);
+	vmalef	($ACC4,$I2,$R2,$ACC4);
+
+	vmalef	($ACC0,$I3,$S2,$ACC0);
+	vmalef	($ACC1,$I3,$S3,$ACC1);
+	vmalef	($ACC2,$I3,$S4,$ACC2);
+	vmalef	($ACC3,$I3,$R0,$ACC3);
+	vmalef	($ACC4,$I3,$R1,$ACC4);
+
+	vmalef	($ACC0,$I4,$S1,$ACC0);
+	vmalef	($ACC1,$I4,$S2,$ACC1);
+	vmalef	($ACC2,$I4,$S3,$ACC2);
+	vmalef	($ACC3,$I4,$S4,$ACC3);
+	vmalef	($ACC4,$I4,$R0,$ACC4);
+
+	vmalof	($ACC0,$H0,$R0,$ACC0);
+	vmalof	($ACC1,$H0,$R1,$ACC1);
+	vmalof	($ACC2,$H0,$R2,$ACC2);
+	vmalof	($ACC3,$H0,$R3,$ACC3);
+	vmalof	($ACC4,$H0,$R4,$ACC4);
+
+	vmalof	($ACC0,$H1,$S4,$ACC0);
+	vmalof	($ACC1,$H1,$R0,$ACC1);
+	vmalof	($ACC2,$H1,$R1,$ACC2);
+	vmalof	($ACC3,$H1,$R2,$ACC3);
+	vmalof	($ACC4,$H1,$R3,$ACC4);
+
+	vmalof	($ACC0,$H2,$S3,$ACC0);
+	vmalof	($ACC1,$H2,$S4,$ACC1);
+	vmalof	($ACC2,$H2,$R0,$ACC2);
+	vmalof	($ACC3,$H2,$R1,$ACC3);
+	vmalof	($ACC4,$H2,$R2,$ACC4);
+
+	vmalof	($ACC0,$H3,$S2,$ACC0);
+	vmalof	($ACC1,$H3,$S3,$ACC1);
+	vmalof	($ACC2,$H3,$S4,$ACC2);
+	vmalof	($ACC3,$H3,$R0,$ACC3);
+	vmalof	($ACC4,$H3,$R1,$ACC4);
+
+	vmalof	($ACC0,$H4,$S1,$ACC0);
+	vmalof	($ACC1,$H4,$S2,$ACC1);
+	vmalof	($ACC2,$H4,$S3,$ACC2);
+	vmalof	($ACC3,$H4,$S4,$ACC3);
+	vmalof	($ACC4,$H4,$R0,$ACC4);
+
+	################################################################
+	# horizontal addition
+
+	vzero	($H0);
+	vsumqg	($ACC0,$ACC0,$H0);
+	vsumqg	($ACC1,$ACC1,$H0);
+	vsumqg	($ACC2,$ACC2,$H0);
+	vsumqg	($ACC3,$ACC3,$H0);
+	vsumqg	($ACC4,$ACC4,$H0);
+
+	################################################################
+	# lazy reduction
+
+	vesrlg	($H4,$ACC3,26);
+	vesrlg	($H1,$ACC0,26);
+	vn	($H3,$ACC3,$mask26);
+	vn	($H0,$ACC0,$mask26);
+	vag	($H4,$H4,$ACC4);		# h3 -> h4
+	vag	($H1,$H1,$ACC1);		# h0 -> h1
+
+	vesrlg	($ACC4,$H4,26);
+	vesrlg	($ACC1,$H1,26);
+	vn	($H4,$H4,$mask26);
+	vn	($H1,$H1,$mask26);
+	vag	($H0,$H0,$ACC4);
+	vag	($H2,$ACC2,$ACC1);		# h1 -> h2
+
+	veslg	($ACC4,$ACC4,2);		# <<2
+	vesrlg	($ACC2,$H2,26);
+	vn	($H2,$H2,$mask26);
+	vag	($H0,$H0,$ACC4);		# h4 -> h0
+	vag	($H3,$H3,$ACC2);		# h2 -> h3
+
+	vesrlg	($ACC0,$H0,26);
+	vesrlg	($ACC3,$H3,26);
+	vn	($H0,$H0,$mask26);
+	vn	($H3,$H3,$mask26);
+	vag	($H1,$H1,$ACC0);		# h0 -> h1
+	vag	($H4,$H4,$ACC3);		# h3 -> h4
+
+&{$z?	\&clgfi:\&clfi} ($len,0);
+	je	(".Ldone");
+
+	vlm	($T1,$T4,"0x00($inp)");		# load last partial block
+	vgmg	($mask26,6,31);
+	vgmf	($I4,5,5);			# padbit<<2
+
+	vperm	($I0,$T3,$T4,$bswaplo);
+	vperm	($I2,$T3,$T4,$bswapmi);
+	vperm	($T3,$T3,$T4,$bswaphi);
+
+	vl	($ACC0,"0x30($len,%r1)");	# borrow $ACC0,1
+	vl	($ACC1,"0x60($len,%r1)");
+
+	verimg	($I1,$I0,$mask26,6);		# >>26
+	veslg	($I0,$I0,32);
+	veslg	($I2,$I2,28);			# >>4
+	verimg	($I3,$T3,$mask26,18);		# >>14
+	verimg	($I4,$T3,$mask26,58);		# >>38
+	vn	($I0,$I0,$mask26);
+	vn	($I2,$I2,$mask26);
+	vesrlf	($I4,$I4,2);			# >>2
+
+	vgmg	($mask26,38,63);
+	vperm	($T3,$T1,$T2,$bswaplo);
+	vperm	($T4,$T1,$T2,$bswaphi);
+	vperm	($T2,$T1,$T2,$bswapmi);
+
+	verimg	($I0,$T3,$mask26,0);
+	verimg	($I1,$T3,$mask26,38);		# >>26
+	verimg	($I2,$T2,$mask26,60);		# >>4
+	verimg	($I3,$T4,$mask26,50);		# >>14
+	vesrlg	($T4,$T4,40);
+	vo	($I4,$I4,$T4);
+
+	vperm	($H0,$H0,$H0,$ACC0);		# move hash to right lane
+	vn	($I0,$I0,$ACC1);		# mask redundant lane[s]
+	vperm	($H1,$H1,$H1,$ACC0);
+	vn	($I1,$I1,$ACC1);
+	vperm	($H2,$H2,$H2,$ACC0);
+	vn	($I2,$I2,$ACC1);
+	vperm	($H3,$H3,$H3,$ACC0);
+	vn	($I3,$I3,$ACC1);
+	vperm	($H4,$H4,$H4,$ACC0);
+	vn	($I4,$I4,$ACC1);
+
+	vaf	($I0,$I0,$H0);			# accumulate hash
+	vzero	($H0);				# wipe hash value
+	vaf	($I1,$I1,$H1);
+	vzero	($H1);
+	vaf	($I2,$I2,$H2);
+	vzero	($H2);
+	vaf	($I3,$I3,$H3);
+	vzero	($H3);
+	vaf	($I4,$I4,$H4);
+	vzero	($H4);
+
+&{$z?	\&lghi:\&lhi}	($len,0);
+	j	(".Last");
+	# I don't bother to tell apart cases when only one multiplication
+	# pass is sufficient, because I argue that mispredicted branch
+	# penalties are comparable to overhead of sometimes redundant
+	# multiplication pass...
+
+LABEL	(".Ldone");
+	vstef	($H0,"0($ctx)",3);		# store hash base 2^26
+	vstef	($H1,"4($ctx)",3);
+	vstef	($H2,"8($ctx)",3);
+	vstef	($H3,"12($ctx)",3);
+	vstef	($H4,"16($ctx)",3);
+
+if ($z) {
+	ld	("%f8","$stdframe+0*8($sp)");
+	ld	("%f9","$stdframe+1*8($sp)");
+	ld	("%f10","$stdframe+2*8($sp)");
+	ld	("%f11","$stdframe+3*8($sp)");
+	ld	("%f12","$stdframe+4*8($sp)");
+	ld	("%f13","$stdframe+5*8($sp)");
+	ld	("%f14","$stdframe+6*8($sp)");
+	ld	("%f15","$stdframe+7*8($sp)");
+&{$z?	\&lmg:\&lm}	("%r10","%r15","$stdframe+8*8+10*$SIZE_T($sp)");
+} else {
+	ld	("%f4","$stdframe+16*$SIZE_T+2*8($sp)");
+	ld	("%f6","$stdframe+16*$SIZE_T+3*8($sp)");
+&{$z?	\&lmg:\&lm}	("%r10","%r15","$stdframe+10*$SIZE_T($sp)");
+}
+	br	("%r14");
+SIZE	("__poly1305_blocks_vx",".-__poly1305_blocks_vx");
+}
+
+################
+# static void poly1305_emit(void *ctx, unsigned char mac[16],
+#                           const u32 nonce[4])
 {
 my ($mac,$nonce)=($inp,$len);
-my ($h0,$h1,$h2,$d0,$d1)=map("%r$_",(5..9));
+my ($h0,$h1,$h2,$d0,$d1,$d2)=map("%r$_",(5..10));
 
-$code.=<<___;
-.globl	poly1305_emit
-.type	poly1305_emit,\@function
-.align	16
-poly1305_emit:
-	stm${g}	%r6,%r9,`6*$SIZE_T`($sp)
-
-	lg	$h0,0($ctx)
-	lg	$h1,8($ctx)
-	lg	$h2,16($ctx)
-
-	lghi	%r0,5
-	lghi	%r1,0
-	lgr	$d0,$h0
-	lgr	$d1,$h1
-
-	algr	$h0,%r0			# compare to modulus
-	alcgr	$h1,%r1
-	alcgr	$h2,%r1
-
-	srlg	$h2,$h2,2		# did it borrow/carry?
-	slgr	%r1,$h2			# 0-$h2>>2
-	lg	$h2,0($nonce)		# load nonce
-	lghi	%r0,-1
-	lg	$ctx,8($nonce)
-	xgr	%r0,%r1			# ~%r1
-
-	ngr	$h0,%r1
-	ngr	$d0,%r0
-	ngr	$h1,%r1
-	ngr	$d1,%r0
-	ogr	$h0,$d0
-	rllg	$d0,$h2,32		# flip nonce words
-	ogr	$h1,$d1
-	rllg	$d1,$ctx,32
-
-	algr	$h0,$d0			# accumulate nonce
-	alcgr	$h1,$d1
-
-	strvg	$h0,0($mac)		# write little-endian result
-	strvg	$h1,8($mac)
-
-	lm${g}	%r6,%r9,`6*$SIZE_T`($sp)
-	br	%r14
-.size	poly1305_emit,.-poly1305_emit
-
-.string	"Poly1305 for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-___
+GLOBL	("poly1305_emit");
+TYPE	("poly1305_emit","\@function");
+ALIGN	(16);
+LABEL	("poly1305_emit");
+LABEL	(".Lpoly1305_emit");
+&{$z?	\&stmg:\&stm}	("%r6","%r10","6*$SIZE_T($sp)");
+
+	lg	($d0,"0($ctx)");
+	lg	($d1,"8($ctx)");
+	lg	($d2,"16($ctx)");
+
+	llgfr	("%r0",$d0);			# base 2^26 -> base 2^64
+	srlg	($h0,$d0,32);
+	llgfr	("%r1",$d1);
+	srlg	($h1,$d1,32);
+	srlg	($h2,$d2,32);
+
+	sllg	("%r0","%r0",26);
+	algr	($h0,"%r0");
+	sllg	("%r0",$h1,52);
+	srlg	($h1,$h1,12);
+	sllg	("%r1","%r1",14);
+	algr	($h0,"%r0");
+	alcgr	($h1,"%r1");
+	sllg	("%r0",$h2,40);
+	srlg	($h2,$h2,24);
+	lghi	("%r1",0);
+	algr	($h1,"%r0");
+	alcgr	($h2,"%r1");
+
+	llgf	("%r0","24($ctx)");		# is_base2_26
+	lcgr	("%r0","%r0");
+
+	xgr	($h0,$d0);			# choose between radixes
+	xgr	($h1,$d1);
+	xgr	($h2,$d2);
+	ngr	($h0,"%r0");
+	ngr	($h1,"%r0");
+	ngr	($h2,"%r0");
+	xgr	($h0,$d0);
+	xgr	($h1,$d1);
+	xgr	($h2,$d2);
+
+	lghi	("%r0",5);
+	lgr	($d0,$h0);
+	lgr	($d1,$h1);
+
+	algr	($h0,"%r0");			# compare to modulus
+	alcgr	($h1,"%r1");
+	alcgr	($h2,"%r1");
+
+	srlg	($h2,$h2,2);			# did it borrow/carry?
+	slgr	("%r1",$h2);				# 0-$h2>>2
+	lg	($d2,"0($nonce)");		# load nonce
+	lg	($ctx,"8($nonce)");
+
+	xgr	($h0,$d0);
+	xgr	($h1,$d1);
+	ngr	($h0,"%r1");
+	ngr	($h1,"%r1");
+	xgr	($h0,$d0);
+	rllg	($d0,$d2,32);			# flip nonce words
+	xgr	($h1,$d1);
+	rllg	($d1,$ctx,32);
+
+	algr	($h0,$d0);			# accumulate nonce
+	alcgr	($h1,$d1);
+
+	strvg	($h0,"0($mac)");		# write little-endian result
+	strvg	($h1,"8($mac)");
+
+&{$z?	\&lmg:\&lm}	("%r6","%r10","6*$SIZE_T($sp)");
+	br	("%r14");
+SIZE	("poly1305_emit",".-poly1305_emit");
 }
 
-$code =~ s/\`([^\`]*)\`/eval $1/gem;
-$code =~ s/\b(srlg\s+)(%r[0-9]+\s*,)\s*([0-9]+)/$1$2$2$3/gm;
+################
+
+ALIGN	(16);
+LABEL	(".Lconst");
+LONG	(0x04050607,0x14151617,0x0c0d0e0f,0x1c1d1e1f);	# merge odd
+LONG	(0x07060504,0x03020100,0x17161514,0x13121110);	# byte swap masks
+LONG	(0x0f0e0d0c,0x0b0a0908,0x1f1e1d1c,0x1b1a1918);
+LONG	(0x00000000,0x09080706,0x00000000,0x19181716);
+
+LONG	(0x00000000,0x00000000,0x00000000,0x0c0d0e0f);	# magic tail masks
+LONG	(0x0c0d0e0f,0x00000000,0x00000000,0x00000000);
+LONG	(0x00000000,0x00000000,0x0c0d0e0f,0x00000000);
+
+LONG	(0xffffffff,0x00000000,0xffffffff,0xffffffff);
+LONG	(0xffffffff,0x00000000,0xffffffff,0x00000000);
+LONG	(0x00000000,0x00000000,0xffffffff,0x00000000);
+
+STRING	("\"Poly1305 for s390x, CRYPTOGAMS by <appro\@openssl.org>\"");
 
-print $code;
-close STDOUT;
+PERLASM_END();
diff -up openssl-1.1.1b/crypto/poly1305/build.info.s390x-update openssl-1.1.1b/crypto/poly1305/build.info
--- openssl-1.1.1b/crypto/poly1305/build.info.s390x-update	2019-05-06 10:54:00.036367588 +0200
+++ openssl-1.1.1b/crypto/poly1305/build.info	2019-05-06 10:56:14.964105164 +0200
@@ -18,6 +18,7 @@ INCLUDE[poly1305-armv8.o]=..
 GENERATE[poly1305-mips.S]=asm/poly1305-mips.pl $(PERLASM_SCHEME)
 INCLUDE[poly1305-mips.o]=..
 GENERATE[poly1305-s390x.S]=asm/poly1305-s390x.pl $(PERLASM_SCHEME)
+INCLUDE[poly1305-s390x.o]=..
 
 BEGINRAW[Makefile(unix)]
 {- $builddir -}/poly1305-%.S:	{- $sourcedir -}/asm/poly1305-%.pl