|
 |
22d461 |
Upstream-Status: Backport [
|
|
|
22d461 |
https://github.com/openssl/openssl/commit/f596bbe4da779b56eea34d96168b557d78e1149,
|
|
|
22d461 |
https://github.com/openssl/openssl/commit/7e1f3ffcc5bc15fb9a12b9e3bb202f544c6ed5aa,
|
|
|
22d461 |
hunks in crypto/ppccap.c from https://github.com/openssl/openssl/commit/f5485b97b6c9977c0d39c7669b9f97a879312447
|
|
|
22d461 |
]
|
|
|
22d461 |
diff --git a/crypto/chacha/asm/chachap10-ppc.pl b/crypto/chacha/asm/chachap10-ppc.pl
|
|
|
22d461 |
new file mode 100755
|
|
|
22d461 |
index 0000000..36e9a8d
|
|
|
22d461 |
--- /dev/null
|
|
|
22d461 |
+++ b/crypto/chacha/asm/chachap10-ppc.pl
|
|
|
22d461 |
@@ -0,0 +1,1288 @@
|
|
|
22d461 |
+#! /usr/bin/env perl
|
|
|
22d461 |
+# Copyright 2016-2020 The OpenSSL Project Authors. All Rights Reserved.
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# Licensed under the Apache License 2.0 (the "License"). You may not use
|
|
|
22d461 |
+# this file except in compliance with the License. You can obtain a copy
|
|
|
22d461 |
+# in the file LICENSE in the source distribution or at
|
|
|
22d461 |
+# https://www.openssl.org/source/license.html
|
|
|
22d461 |
+
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# ====================================================================
|
|
|
22d461 |
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
|
|
|
22d461 |
+# project. The module is, however, dual licensed under OpenSSL and
|
|
|
22d461 |
+# CRYPTOGAMS licenses depending on where you obtain it. For further
|
|
|
22d461 |
+# details see http://www.openssl.org/~appro/cryptogams/.
|
|
|
22d461 |
+# ====================================================================
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# October 2015
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# ChaCha20 for PowerPC/AltiVec.
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# June 2018
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# Add VSX 2.07 code path. Original 3xAltiVec+1xIALU is well-suited for
|
|
|
22d461 |
+# processors that can't issue more than one vector instruction per
|
|
|
22d461 |
+# cycle. But POWER8 (and POWER9) can issue a pair, and vector-only 4x
|
|
|
22d461 |
+# interleave would perform better. Incidentally PowerISA 2.07 (first
|
|
|
22d461 |
+# implemented by POWER8) defined new usable instructions, hence 4xVSX
|
|
|
22d461 |
+# code path...
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# Performance in cycles per byte out of large buffer.
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# IALU/gcc-4.x 3xAltiVec+1xIALU 4xVSX
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# Freescale e300 13.6/+115% - -
|
|
|
22d461 |
+# PPC74x0/G4e 6.81/+310% 3.81 -
|
|
|
22d461 |
+# PPC970/G5 9.29/+160% ? -
|
|
|
22d461 |
+# POWER7 8.62/+61% 3.35 -
|
|
|
22d461 |
+# POWER8 8.70/+51% 2.91 2.09
|
|
|
22d461 |
+# POWER9 8.80/+29% 4.44(*) 2.45(**)
|
|
|
22d461 |
+#
|
|
|
22d461 |
+# (*) this is trade-off result, it's possible to improve it, but
|
|
|
22d461 |
+# then it would negatively affect all others;
|
|
|
22d461 |
+# (**) POWER9 seems to be "allergic" to mixing vector and integer
|
|
|
22d461 |
+# instructions, which is why switch to vector-only code pays
|
|
|
22d461 |
+# off that much;
|
|
|
22d461 |
+
|
|
|
22d461 |
+# $output is the last argument if it looks like a file (it has an extension)
|
|
|
22d461 |
+# $flavour is the first argument if it doesn't look like a file
|
|
|
22d461 |
+$output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
|
|
|
22d461 |
+$flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
|
|
|
22d461 |
+
|
|
|
22d461 |
+if ($flavour =~ /64/) {
|
|
|
22d461 |
+ $SIZE_T =8;
|
|
|
22d461 |
+ $LRSAVE =2*$SIZE_T;
|
|
|
22d461 |
+ $STU ="stdu";
|
|
|
22d461 |
+ $POP ="ld";
|
|
|
22d461 |
+ $PUSH ="std";
|
|
|
22d461 |
+ $UCMP ="cmpld";
|
|
|
22d461 |
+} elsif ($flavour =~ /32/) {
|
|
|
22d461 |
+ $SIZE_T =4;
|
|
|
22d461 |
+ $LRSAVE =$SIZE_T;
|
|
|
22d461 |
+ $STU ="stwu";
|
|
|
22d461 |
+ $POP ="lwz";
|
|
|
22d461 |
+ $PUSH ="stw";
|
|
|
22d461 |
+ $UCMP ="cmplw";
|
|
|
22d461 |
+} else { die "nonsense $flavour"; }
|
|
|
22d461 |
+
|
|
|
22d461 |
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? 1 : 0;
|
|
|
22d461 |
+
|
|
|
22d461 |
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
|
|
|
22d461 |
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
|
|
|
22d461 |
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
|
|
|
22d461 |
+die "can't locate ppc-xlate.pl";
|
|
|
22d461 |
+
|
|
|
22d461 |
+open STDOUT,"| $^X $xlate $flavour \"$output\""
|
|
|
22d461 |
+ or die "can't call $xlate: $!";
|
|
|
22d461 |
+
|
|
|
22d461 |
+$LOCALS=6*$SIZE_T;
|
|
|
22d461 |
+$FRAME=$LOCALS+64+18*$SIZE_T; # 64 is for local variables
|
|
|
22d461 |
+
|
|
|
22d461 |
+sub AUTOLOAD() # thunk [simplified] x86-style perlasm
|
|
|
22d461 |
+{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://; $opcode =~ s/_/\./;
|
|
|
22d461 |
+ $code .= "\t$opcode\t".join(',',@_)."\n";
|
|
|
22d461 |
+}
|
|
|
22d461 |
+
|
|
|
22d461 |
+my $sp = "r1";
|
|
|
22d461 |
+
|
|
|
22d461 |
+my ($out,$inp,$len,$key,$ctr) = map("r$_",(3..7));
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+{{{
|
|
|
22d461 |
+my ($xa0,$xa1,$xa2,$xa3, $xb0,$xb1,$xb2,$xb3,
|
|
|
22d461 |
+ $xc0,$xc1,$xc2,$xc3, $xd0,$xd1,$xd2,$xd3) = map("v$_",(0..15));
|
|
|
22d461 |
+my @K = map("v$_",(16..19));
|
|
|
22d461 |
+my $CTR = "v26";
|
|
|
22d461 |
+my ($xt0,$xt1,$xt2,$xt3) = map("v$_",(27..30));
|
|
|
22d461 |
+my ($sixteen,$twelve,$eight,$seven) = ($xt0,$xt1,$xt2,$xt3);
|
|
|
22d461 |
+my $beperm = "v31";
|
|
|
22d461 |
+
|
|
|
22d461 |
+my ($x00,$x10,$x20,$x30) = (0, map("r$_",(8..10)));
|
|
|
22d461 |
+
|
|
|
22d461 |
+my $FRAME=$LOCALS+64+7*16; # 7*16 is for v26-v31 offload
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+sub VSX_lane_ROUND_4x {
|
|
|
22d461 |
+my ($a0,$b0,$c0,$d0)=@_;
|
|
|
22d461 |
+my ($a1,$b1,$c1,$d1)=map(($_&~3)+(($_+1)&3),($a0,$b0,$c0,$d0));
|
|
|
22d461 |
+my ($a2,$b2,$c2,$d2)=map(($_&~3)+(($_+1)&3),($a1,$b1,$c1,$d1));
|
|
|
22d461 |
+my ($a3,$b3,$c3,$d3)=map(($_&~3)+(($_+1)&3),($a2,$b2,$c2,$d2));
|
|
|
22d461 |
+my @x=map("\"v$_\"",(0..15));
|
|
|
22d461 |
+
|
|
|
22d461 |
+ (
|
|
|
22d461 |
+ "&vadduwm (@x[$a0],@x[$a0],@x[$b0])", # Q1
|
|
|
22d461 |
+ "&vadduwm (@x[$a1],@x[$a1],@x[$b1])", # Q2
|
|
|
22d461 |
+ "&vadduwm (@x[$a2],@x[$a2],@x[$b2])", # Q3
|
|
|
22d461 |
+ "&vadduwm (@x[$a3],@x[$a3],@x[$b3])", # Q4
|
|
|
22d461 |
+ "&vxor (@x[$d0],@x[$d0],@x[$a0])",
|
|
|
22d461 |
+ "&vxor (@x[$d1],@x[$d1],@x[$a1])",
|
|
|
22d461 |
+ "&vxor (@x[$d2],@x[$d2],@x[$a2])",
|
|
|
22d461 |
+ "&vxor (@x[$d3],@x[$d3],@x[$a3])",
|
|
|
22d461 |
+ "&vrlw (@x[$d0],@x[$d0],'$sixteen')",
|
|
|
22d461 |
+ "&vrlw (@x[$d1],@x[$d1],'$sixteen')",
|
|
|
22d461 |
+ "&vrlw (@x[$d2],@x[$d2],'$sixteen')",
|
|
|
22d461 |
+ "&vrlw (@x[$d3],@x[$d3],'$sixteen')",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$c0],@x[$c0],@x[$d0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c1],@x[$c1],@x[$d1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c2],@x[$c2],@x[$d2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c3],@x[$c3],@x[$d3])",
|
|
|
22d461 |
+ "&vxor (@x[$b0],@x[$b0],@x[$c0])",
|
|
|
22d461 |
+ "&vxor (@x[$b1],@x[$b1],@x[$c1])",
|
|
|
22d461 |
+ "&vxor (@x[$b2],@x[$b2],@x[$c2])",
|
|
|
22d461 |
+ "&vxor (@x[$b3],@x[$b3],@x[$c3])",
|
|
|
22d461 |
+ "&vrlw (@x[$b0],@x[$b0],'$twelve')",
|
|
|
22d461 |
+ "&vrlw (@x[$b1],@x[$b1],'$twelve')",
|
|
|
22d461 |
+ "&vrlw (@x[$b2],@x[$b2],'$twelve')",
|
|
|
22d461 |
+ "&vrlw (@x[$b3],@x[$b3],'$twelve')",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$a0],@x[$a0],@x[$b0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a1],@x[$a1],@x[$b1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a2],@x[$a2],@x[$b2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a3],@x[$a3],@x[$b3])",
|
|
|
22d461 |
+ "&vxor (@x[$d0],@x[$d0],@x[$a0])",
|
|
|
22d461 |
+ "&vxor (@x[$d1],@x[$d1],@x[$a1])",
|
|
|
22d461 |
+ "&vxor (@x[$d2],@x[$d2],@x[$a2])",
|
|
|
22d461 |
+ "&vxor (@x[$d3],@x[$d3],@x[$a3])",
|
|
|
22d461 |
+ "&vrlw (@x[$d0],@x[$d0],'$eight')",
|
|
|
22d461 |
+ "&vrlw (@x[$d1],@x[$d1],'$eight')",
|
|
|
22d461 |
+ "&vrlw (@x[$d2],@x[$d2],'$eight')",
|
|
|
22d461 |
+ "&vrlw (@x[$d3],@x[$d3],'$eight')",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$c0],@x[$c0],@x[$d0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c1],@x[$c1],@x[$d1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c2],@x[$c2],@x[$d2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c3],@x[$c3],@x[$d3])",
|
|
|
22d461 |
+ "&vxor (@x[$b0],@x[$b0],@x[$c0])",
|
|
|
22d461 |
+ "&vxor (@x[$b1],@x[$b1],@x[$c1])",
|
|
|
22d461 |
+ "&vxor (@x[$b2],@x[$b2],@x[$c2])",
|
|
|
22d461 |
+ "&vxor (@x[$b3],@x[$b3],@x[$c3])",
|
|
|
22d461 |
+ "&vrlw (@x[$b0],@x[$b0],'$seven')",
|
|
|
22d461 |
+ "&vrlw (@x[$b1],@x[$b1],'$seven')",
|
|
|
22d461 |
+ "&vrlw (@x[$b2],@x[$b2],'$seven')",
|
|
|
22d461 |
+ "&vrlw (@x[$b3],@x[$b3],'$seven')"
|
|
|
22d461 |
+ );
|
|
|
22d461 |
+}
|
|
|
22d461 |
+
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+
|
|
|
22d461 |
+.globl .ChaCha20_ctr32_vsx_p10
|
|
|
22d461 |
+.align 5
|
|
|
22d461 |
+.ChaCha20_ctr32_vsx_p10:
|
|
|
22d461 |
+ ${UCMP}i $len,255
|
|
|
22d461 |
+ bgt ChaCha20_ctr32_vsx_8x
|
|
|
22d461 |
+ $STU $sp,-$FRAME($sp)
|
|
|
22d461 |
+ mflr r0
|
|
|
22d461 |
+ li r10,`15+$LOCALS+64`
|
|
|
22d461 |
+ li r11,`31+$LOCALS+64`
|
|
|
22d461 |
+ mfspr r12,256
|
|
|
22d461 |
+ stvx v26,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ stvx v27,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ stvx v28,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ stvx v29,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ stvx v30,r10,$sp
|
|
|
22d461 |
+ stvx v31,r11,$sp
|
|
|
22d461 |
+ stw r12,`$FRAME-4`($sp) # save vrsave
|
|
|
22d461 |
+ li r12,-4096+63
|
|
|
22d461 |
+ $PUSH r0, `$FRAME+$LRSAVE`($sp)
|
|
|
22d461 |
+ mtspr 256,r12 # preserve 29 AltiVec registers
|
|
|
22d461 |
+
|
|
|
22d461 |
+ bl Lconsts # returns pointer Lsigma in r12
|
|
|
22d461 |
+ lvx_4w @K[0],0,r12 # load sigma
|
|
|
22d461 |
+ addi r12,r12,0x70
|
|
|
22d461 |
+ li $x10,16
|
|
|
22d461 |
+ li $x20,32
|
|
|
22d461 |
+ li $x30,48
|
|
|
22d461 |
+ li r11,64
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w @K[1],0,$key # load key
|
|
|
22d461 |
+ lvx_4w @K[2],$x10,$key
|
|
|
22d461 |
+ lvx_4w @K[3],0,$ctr # load counter
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xt0
|
|
|
22d461 |
+ lvx_4w $xt1,r11,r12
|
|
|
22d461 |
+ vspltw $CTR,@K[3],0
|
|
|
22d461 |
+ vsldoi @K[3],@K[3],$xt0,4
|
|
|
22d461 |
+ vsldoi @K[3],$xt0,@K[3],12 # clear @K[3].word[0]
|
|
|
22d461 |
+ vadduwm $CTR,$CTR,$xt1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?lvsl $beperm,0,$x10 # 0x00..0f
|
|
|
22d461 |
+ be?vspltisb $xt0,3 # 0x03..03
|
|
|
22d461 |
+ be?vxor $beperm,$beperm,$xt0 # swap bytes within words
|
|
|
22d461 |
+
|
|
|
22d461 |
+ li r0,10 # inner loop counter
|
|
|
22d461 |
+ mtctr r0
|
|
|
22d461 |
+ b Loop_outer_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+.align 5
|
|
|
22d461 |
+Loop_outer_vsx:
|
|
|
22d461 |
+ lvx $xa0,$x00,r12 # load [smashed] sigma
|
|
|
22d461 |
+ lvx $xa1,$x10,r12
|
|
|
22d461 |
+ lvx $xa2,$x20,r12
|
|
|
22d461 |
+ lvx $xa3,$x30,r12
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltw $xb0,@K[1],0 # smash the key
|
|
|
22d461 |
+ vspltw $xb1,@K[1],1
|
|
|
22d461 |
+ vspltw $xb2,@K[1],2
|
|
|
22d461 |
+ vspltw $xb3,@K[1],3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltw $xc0,@K[2],0
|
|
|
22d461 |
+ vspltw $xc1,@K[2],1
|
|
|
22d461 |
+ vspltw $xc2,@K[2],2
|
|
|
22d461 |
+ vspltw $xc3,@K[2],3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmr $xd0,$CTR # smash the counter
|
|
|
22d461 |
+ vspltw $xd1,@K[3],1
|
|
|
22d461 |
+ vspltw $xd2,@K[3],2
|
|
|
22d461 |
+ vspltw $xd3,@K[3],3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltisw $sixteen,-16 # synthesize constants
|
|
|
22d461 |
+ vspltisw $twelve,12
|
|
|
22d461 |
+ vspltisw $eight,8
|
|
|
22d461 |
+ vspltisw $seven,7
|
|
|
22d461 |
+
|
|
|
22d461 |
+Loop_vsx_4x:
|
|
|
22d461 |
+___
|
|
|
22d461 |
+ foreach (&VSX_lane_ROUND_4x(0, 4, 8,12)) { eval; }
|
|
|
22d461 |
+ foreach (&VSX_lane_ROUND_4x(0, 5,10,15)) { eval; }
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+
|
|
|
22d461 |
+ bdnz Loop_vsx_4x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xd0,$xd0,$CTR
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xt0,$xa0,$xa1 # transpose data
|
|
|
22d461 |
+ vmrgew $xt1,$xa2,$xa3
|
|
|
22d461 |
+ vmrgow $xa0,$xa0,$xa1
|
|
|
22d461 |
+ vmrgow $xa2,$xa2,$xa3
|
|
|
22d461 |
+ vmrgew $xt2,$xb0,$xb1
|
|
|
22d461 |
+ vmrgew $xt3,$xb2,$xb3
|
|
|
22d461 |
+ vpermdi $xa1,$xa0,$xa2,0b00
|
|
|
22d461 |
+ vpermdi $xa3,$xa0,$xa2,0b11
|
|
|
22d461 |
+ vpermdi $xa0,$xt0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xa2,$xt0,$xt1,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgow $xb0,$xb0,$xb1
|
|
|
22d461 |
+ vmrgow $xb2,$xb2,$xb3
|
|
|
22d461 |
+ vmrgew $xt0,$xc0,$xc1
|
|
|
22d461 |
+ vmrgew $xt1,$xc2,$xc3
|
|
|
22d461 |
+ vpermdi $xb1,$xb0,$xb2,0b00
|
|
|
22d461 |
+ vpermdi $xb3,$xb0,$xb2,0b11
|
|
|
22d461 |
+ vpermdi $xb0,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xb2,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgow $xc0,$xc0,$xc1
|
|
|
22d461 |
+ vmrgow $xc2,$xc2,$xc3
|
|
|
22d461 |
+ vmrgew $xt2,$xd0,$xd1
|
|
|
22d461 |
+ vmrgew $xt3,$xd2,$xd3
|
|
|
22d461 |
+ vpermdi $xc1,$xc0,$xc2,0b00
|
|
|
22d461 |
+ vpermdi $xc3,$xc0,$xc2,0b11
|
|
|
22d461 |
+ vpermdi $xc0,$xt0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xc2,$xt0,$xt1,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgow $xd0,$xd0,$xd1
|
|
|
22d461 |
+ vmrgow $xd2,$xd2,$xd3
|
|
|
22d461 |
+ vspltisw $xt0,4
|
|
|
22d461 |
+ vadduwm $CTR,$CTR,$xt0 # next counter value
|
|
|
22d461 |
+ vpermdi $xd1,$xd0,$xd2,0b00
|
|
|
22d461 |
+ vpermdi $xd3,$xd0,$xd2,0b11
|
|
|
22d461 |
+ vpermdi $xd0,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xd2,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa0,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb0,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc0,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd0,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa1,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb1,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc1,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd1,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa2,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb2,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc2,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd2,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa3,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb3,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc3,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd3,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ mtctr r0
|
|
|
22d461 |
+ bne Loop_outer_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+Ldone_vsx:
|
|
|
22d461 |
+ lwz r12,`$FRAME-4`($sp) # pull vrsave
|
|
|
22d461 |
+ li r10,`15+$LOCALS+64`
|
|
|
22d461 |
+ li r11,`31+$LOCALS+64`
|
|
|
22d461 |
+ $POP r0, `$FRAME+$LRSAVE`($sp)
|
|
|
22d461 |
+ mtspr 256,r12 # restore vrsave
|
|
|
22d461 |
+ lvx v26,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ lvx v27,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ lvx v28,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ lvx v29,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ lvx v30,r10,$sp
|
|
|
22d461 |
+ lvx v31,r11,$sp
|
|
|
22d461 |
+ mtlr r0
|
|
|
22d461 |
+ addi $sp,$sp,$FRAME
|
|
|
22d461 |
+ blr
|
|
|
22d461 |
+
|
|
|
22d461 |
+.align 4
|
|
|
22d461 |
+Ltail_vsx:
|
|
|
22d461 |
+ addi r11,$sp,$LOCALS
|
|
|
22d461 |
+ mtctr $len
|
|
|
22d461 |
+ stvx_4w $xa0,$x00,r11 # offload block to stack
|
|
|
22d461 |
+ stvx_4w $xb0,$x10,r11
|
|
|
22d461 |
+ stvx_4w $xc0,$x20,r11
|
|
|
22d461 |
+ stvx_4w $xd0,$x30,r11
|
|
|
22d461 |
+ subi r12,r11,1 # prepare for *++ptr
|
|
|
22d461 |
+ subi $inp,$inp,1
|
|
|
22d461 |
+ subi $out,$out,1
|
|
|
22d461 |
+
|
|
|
22d461 |
+Loop_tail_vsx:
|
|
|
22d461 |
+ lbzu r6,1(r12)
|
|
|
22d461 |
+ lbzu r7,1($inp)
|
|
|
22d461 |
+ xor r6,r6,r7
|
|
|
22d461 |
+ stbu r6,1($out)
|
|
|
22d461 |
+ bdnz Loop_tail_vsx
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $K[0],$x00,r11 # wipe copy of the block
|
|
|
22d461 |
+ stvx_4w $K[0],$x10,r11
|
|
|
22d461 |
+ stvx_4w $K[0],$x20,r11
|
|
|
22d461 |
+ stvx_4w $K[0],$x30,r11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ b Ldone_vsx
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+ .byte 0,12,0x04,1,0x80,0,5,0
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+.size .ChaCha20_ctr32_vsx_p10,.-.ChaCha20_ctr32_vsx_p10
|
|
|
22d461 |
+___
|
|
|
22d461 |
+}}}
|
|
|
22d461 |
+
|
|
|
22d461 |
+##This is 8 block in parallel implementation. The heart of chacha round uses vector instruction that has access to
|
|
|
22d461 |
+# vsr[32+X]. To perform the 8 parallel block we tend to use all 32 register to hold the 8 block info.
|
|
|
22d461 |
+# WE need to store few register value on side, so we can use VSR{32+X} for few vector instructions used in round op and hold intermediate value.
|
|
|
22d461 |
+# WE use the VSR[0]-VSR[31] for holding intermediate value and perform 8 block in parallel.
|
|
|
22d461 |
+#
|
|
|
22d461 |
+{{{
|
|
|
22d461 |
+#### ($out,$inp,$len,$key,$ctr) = map("r$_",(3..7));
|
|
|
22d461 |
+my ($xa0,$xa1,$xa2,$xa3, $xb0,$xb1,$xb2,$xb3,
|
|
|
22d461 |
+ $xc0,$xc1,$xc2,$xc3, $xd0,$xd1,$xd2,$xd3,
|
|
|
22d461 |
+ $xa4,$xa5,$xa6,$xa7, $xb4,$xb5,$xb6,$xb7,
|
|
|
22d461 |
+ $xc4,$xc5,$xc6,$xc7, $xd4,$xd5,$xd6,$xd7) = map("v$_",(0..31));
|
|
|
22d461 |
+my ($xcn4,$xcn5,$xcn6,$xcn7, $xdn4,$xdn5,$xdn6,$xdn7) = map("v$_",(8..15));
|
|
|
22d461 |
+my ($xan0,$xbn0,$xcn0,$xdn0) = map("v$_",(0..3));
|
|
|
22d461 |
+my @K = map("v$_",27,(24..26));
|
|
|
22d461 |
+my ($xt0,$xt1,$xt2,$xt3,$xt4) = map("v$_",23,(28..31));
|
|
|
22d461 |
+my $xr0 = "v4";
|
|
|
22d461 |
+my $CTR0 = "v22";
|
|
|
22d461 |
+my $CTR1 = "v5";
|
|
|
22d461 |
+my $beperm = "v31";
|
|
|
22d461 |
+my ($x00,$x10,$x20,$x30) = (0, map("r$_",(8..10)));
|
|
|
22d461 |
+my ($xv0,$xv1,$xv2,$xv3,$xv4,$xv5,$xv6,$xv7) = map("v$_",(0..7));
|
|
|
22d461 |
+my ($xv8,$xv9,$xv10,$xv11,$xv12,$xv13,$xv14,$xv15,$xv16,$xv17) = map("v$_",(8..17));
|
|
|
22d461 |
+my ($xv18,$xv19,$xv20,$xv21) = map("v$_",(18..21));
|
|
|
22d461 |
+my ($xv22,$xv23,$xv24,$xv25,$xv26) = map("v$_",(22..26));
|
|
|
22d461 |
+
|
|
|
22d461 |
+my $FRAME=$LOCALS+64+9*16; # 8*16 is for v24-v31 offload
|
|
|
22d461 |
+
|
|
|
22d461 |
+sub VSX_lane_ROUND_8x {
|
|
|
22d461 |
+my ($a0,$b0,$c0,$d0,$a4,$b4,$c4,$d4)=@_;
|
|
|
22d461 |
+my ($a1,$b1,$c1,$d1)=map(($_&~3)+(($_+1)&3),($a0,$b0,$c0,$d0));
|
|
|
22d461 |
+my ($a2,$b2,$c2,$d2)=map(($_&~3)+(($_+1)&3),($a1,$b1,$c1,$d1));
|
|
|
22d461 |
+my ($a3,$b3,$c3,$d3)=map(($_&~3)+(($_+1)&3),($a2,$b2,$c2,$d2));
|
|
|
22d461 |
+my ($a5,$b5,$c5,$d5)=map(($_&~3)+(($_+1)&3),($a4,$b4,$c4,$d4));
|
|
|
22d461 |
+my ($a6,$b6,$c6,$d6)=map(($_&~3)+(($_+1)&3),($a5,$b5,$c5,$d5));
|
|
|
22d461 |
+my ($a7,$b7,$c7,$d7)=map(($_&~3)+(($_+1)&3),($a6,$b6,$c6,$d6));
|
|
|
22d461 |
+my ($xv8,$xv9,$xv10,$xv11,$xv12,$xv13,$xv14,$xv15,$xv16,$xv17) = map("\"v$_\"",(8..17));
|
|
|
22d461 |
+my @x=map("\"v$_\"",(0..31));
|
|
|
22d461 |
+
|
|
|
22d461 |
+ (
|
|
|
22d461 |
+ "&vxxlor ($xv15 ,@x[$c7],@x[$c7])", #copy v30 to v13
|
|
|
22d461 |
+ "&vxxlorc (@x[$c7], $xv9,$xv9)",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$a0],@x[$a0],@x[$b0])", # Q1
|
|
|
22d461 |
+ "&vadduwm (@x[$a1],@x[$a1],@x[$b1])", # Q2
|
|
|
22d461 |
+ "&vadduwm (@x[$a2],@x[$a2],@x[$b2])", # Q3
|
|
|
22d461 |
+ "&vadduwm (@x[$a3],@x[$a3],@x[$b3])", # Q4
|
|
|
22d461 |
+ "&vadduwm (@x[$a4],@x[$a4],@x[$b4])", # Q1
|
|
|
22d461 |
+ "&vadduwm (@x[$a5],@x[$a5],@x[$b5])", # Q2
|
|
|
22d461 |
+ "&vadduwm (@x[$a6],@x[$a6],@x[$b6])", # Q3
|
|
|
22d461 |
+ "&vadduwm (@x[$a7],@x[$a7],@x[$b7])", # Q4
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxor (@x[$d0],@x[$d0],@x[$a0])",
|
|
|
22d461 |
+ "&vxor (@x[$d1],@x[$d1],@x[$a1])",
|
|
|
22d461 |
+ "&vxor (@x[$d2],@x[$d2],@x[$a2])",
|
|
|
22d461 |
+ "&vxor (@x[$d3],@x[$d3],@x[$a3])",
|
|
|
22d461 |
+ "&vxor (@x[$d4],@x[$d4],@x[$a4])",
|
|
|
22d461 |
+ "&vxor (@x[$d5],@x[$d5],@x[$a5])",
|
|
|
22d461 |
+ "&vxor (@x[$d6],@x[$d6],@x[$a6])",
|
|
|
22d461 |
+ "&vxor (@x[$d7],@x[$d7],@x[$a7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vrlw (@x[$d0],@x[$d0],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d1],@x[$d1],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d2],@x[$d2],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d3],@x[$d3],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d4],@x[$d4],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d5],@x[$d5],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d6],@x[$d6],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d7],@x[$d7],@x[$c7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxxlor ($xv13 ,@x[$a7],@x[$a7])",
|
|
|
22d461 |
+ "&vxxlorc (@x[$c7], $xv15,$xv15)",
|
|
|
22d461 |
+ "&vxxlorc (@x[$a7], $xv10,$xv10)",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$c0],@x[$c0],@x[$d0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c1],@x[$c1],@x[$d1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c2],@x[$c2],@x[$d2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c3],@x[$c3],@x[$d3])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c4],@x[$c4],@x[$d4])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c5],@x[$c5],@x[$d5])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c6],@x[$c6],@x[$d6])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c7],@x[$c7],@x[$d7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxor (@x[$b0],@x[$b0],@x[$c0])",
|
|
|
22d461 |
+ "&vxor (@x[$b1],@x[$b1],@x[$c1])",
|
|
|
22d461 |
+ "&vxor (@x[$b2],@x[$b2],@x[$c2])",
|
|
|
22d461 |
+ "&vxor (@x[$b3],@x[$b3],@x[$c3])",
|
|
|
22d461 |
+ "&vxor (@x[$b4],@x[$b4],@x[$c4])",
|
|
|
22d461 |
+ "&vxor (@x[$b5],@x[$b5],@x[$c5])",
|
|
|
22d461 |
+ "&vxor (@x[$b6],@x[$b6],@x[$c6])",
|
|
|
22d461 |
+ "&vxor (@x[$b7],@x[$b7],@x[$c7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vrlw (@x[$b0],@x[$b0],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b1],@x[$b1],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b2],@x[$b2],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b3],@x[$b3],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b4],@x[$b4],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b5],@x[$b5],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b6],@x[$b6],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b7],@x[$b7],@x[$a7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxxlorc (@x[$a7], $xv13,$xv13)",
|
|
|
22d461 |
+ "&vxxlor ($xv15 ,@x[$c7],@x[$c7])",
|
|
|
22d461 |
+ "&vxxlorc (@x[$c7], $xv11,$xv11)",
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$a0],@x[$a0],@x[$b0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a1],@x[$a1],@x[$b1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a2],@x[$a2],@x[$b2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a3],@x[$a3],@x[$b3])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a4],@x[$a4],@x[$b4])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a5],@x[$a5],@x[$b5])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a6],@x[$a6],@x[$b6])",
|
|
|
22d461 |
+ "&vadduwm (@x[$a7],@x[$a7],@x[$b7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxor (@x[$d0],@x[$d0],@x[$a0])",
|
|
|
22d461 |
+ "&vxor (@x[$d1],@x[$d1],@x[$a1])",
|
|
|
22d461 |
+ "&vxor (@x[$d2],@x[$d2],@x[$a2])",
|
|
|
22d461 |
+ "&vxor (@x[$d3],@x[$d3],@x[$a3])",
|
|
|
22d461 |
+ "&vxor (@x[$d4],@x[$d4],@x[$a4])",
|
|
|
22d461 |
+ "&vxor (@x[$d5],@x[$d5],@x[$a5])",
|
|
|
22d461 |
+ "&vxor (@x[$d6],@x[$d6],@x[$a6])",
|
|
|
22d461 |
+ "&vxor (@x[$d7],@x[$d7],@x[$a7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vrlw (@x[$d0],@x[$d0],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d1],@x[$d1],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d2],@x[$d2],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d3],@x[$d3],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d4],@x[$d4],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d5],@x[$d5],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d6],@x[$d6],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$d7],@x[$d7],@x[$c7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxxlorc (@x[$c7], $xv15,$xv15)",
|
|
|
22d461 |
+ "&vxxlor ($xv13 ,@x[$a7],@x[$a7])",
|
|
|
22d461 |
+ "&vxxlorc (@x[$a7], $xv12,$xv12)",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vadduwm (@x[$c0],@x[$c0],@x[$d0])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c1],@x[$c1],@x[$d1])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c2],@x[$c2],@x[$d2])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c3],@x[$c3],@x[$d3])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c4],@x[$c4],@x[$d4])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c5],@x[$c5],@x[$d5])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c6],@x[$c6],@x[$d6])",
|
|
|
22d461 |
+ "&vadduwm (@x[$c7],@x[$c7],@x[$d7])",
|
|
|
22d461 |
+ "&vxor (@x[$b0],@x[$b0],@x[$c0])",
|
|
|
22d461 |
+ "&vxor (@x[$b1],@x[$b1],@x[$c1])",
|
|
|
22d461 |
+ "&vxor (@x[$b2],@x[$b2],@x[$c2])",
|
|
|
22d461 |
+ "&vxor (@x[$b3],@x[$b3],@x[$c3])",
|
|
|
22d461 |
+ "&vxor (@x[$b4],@x[$b4],@x[$c4])",
|
|
|
22d461 |
+ "&vxor (@x[$b5],@x[$b5],@x[$c5])",
|
|
|
22d461 |
+ "&vxor (@x[$b6],@x[$b6],@x[$c6])",
|
|
|
22d461 |
+ "&vxor (@x[$b7],@x[$b7],@x[$c7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b0],@x[$b0],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b1],@x[$b1],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b2],@x[$b2],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b3],@x[$b3],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b4],@x[$b4],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b5],@x[$b5],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b6],@x[$b6],@x[$a7])",
|
|
|
22d461 |
+ "&vrlw (@x[$b7],@x[$b7],@x[$a7])",
|
|
|
22d461 |
+
|
|
|
22d461 |
+ "&vxxlorc (@x[$a7], $xv13,$xv13)",
|
|
|
22d461 |
+ );
|
|
|
22d461 |
+}
|
|
|
22d461 |
+
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+
|
|
|
22d461 |
+.globl .ChaCha20_ctr32_vsx_8x
|
|
|
22d461 |
+.align 5
|
|
|
22d461 |
+.ChaCha20_ctr32_vsx_8x:
|
|
|
22d461 |
+ $STU $sp,-$FRAME($sp)
|
|
|
22d461 |
+ mflr r0
|
|
|
22d461 |
+ li r10,`15+$LOCALS+64`
|
|
|
22d461 |
+ li r11,`31+$LOCALS+64`
|
|
|
22d461 |
+ mfspr r12,256
|
|
|
22d461 |
+ stvx v24,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ stvx v25,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ stvx v26,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ stvx v27,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ stvx v28,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ stvx v29,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ stvx v30,r10,$sp
|
|
|
22d461 |
+ stvx v31,r11,$sp
|
|
|
22d461 |
+ stw r12,`$FRAME-4`($sp) # save vrsave
|
|
|
22d461 |
+ li r12,-4096+63
|
|
|
22d461 |
+ $PUSH r0, `$FRAME+$LRSAVE`($sp)
|
|
|
22d461 |
+ mtspr 256,r12 # preserve 29 AltiVec registers
|
|
|
22d461 |
+
|
|
|
22d461 |
+ bl Lconsts # returns pointer Lsigma in r12
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w @K[0],0,r12 # load sigma
|
|
|
22d461 |
+ addi r12,r12,0x70
|
|
|
22d461 |
+ li $x10,16
|
|
|
22d461 |
+ li $x20,32
|
|
|
22d461 |
+ li $x30,48
|
|
|
22d461 |
+ li r11,64
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltisw $xa4,-16 # synthesize constants
|
|
|
22d461 |
+ vspltisw $xb4,12 # synthesize constants
|
|
|
22d461 |
+ vspltisw $xc4,8 # synthesize constants
|
|
|
22d461 |
+ vspltisw $xd4,7 # synthesize constants
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx $xa0,$x00,r12 # load [smashed] sigma
|
|
|
22d461 |
+ lvx $xa1,$x10,r12
|
|
|
22d461 |
+ lvx $xa2,$x20,r12
|
|
|
22d461 |
+ lvx $xa3,$x30,r12
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlor $xv9 ,$xa4,$xa4 #save shift val in vr9-12
|
|
|
22d461 |
+ vxxlor $xv10 ,$xb4,$xb4
|
|
|
22d461 |
+ vxxlor $xv11 ,$xc4,$xc4
|
|
|
22d461 |
+ vxxlor $xv12 ,$xd4,$xd4
|
|
|
22d461 |
+ vxxlor $xv22 ,$xa0,$xa0 #save sigma in vr22-25
|
|
|
22d461 |
+ vxxlor $xv23 ,$xa1,$xa1
|
|
|
22d461 |
+ vxxlor $xv24 ,$xa2,$xa2
|
|
|
22d461 |
+ vxxlor $xv25 ,$xa3,$xa3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w @K[1],0,$key # load key
|
|
|
22d461 |
+ lvx_4w @K[2],$x10,$key
|
|
|
22d461 |
+ lvx_4w @K[3],0,$ctr # load counter
|
|
|
22d461 |
+ vspltisw $xt3,4
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xt2
|
|
|
22d461 |
+ lvx_4w $xt1,r11,r12
|
|
|
22d461 |
+ vspltw $xa2,@K[3],0 #save the original count after spltw
|
|
|
22d461 |
+ vsldoi @K[3],@K[3],$xt2,4
|
|
|
22d461 |
+ vsldoi @K[3],$xt2,@K[3],12 # clear @K[3].word[0]
|
|
|
22d461 |
+ vadduwm $xt1,$xa2,$xt1
|
|
|
22d461 |
+ vadduwm $xt3,$xt1,$xt3 # next counter value
|
|
|
22d461 |
+ vspltw $xa0,@K[2],2 # save the K[2] spltw 2 and save v8.
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?lvsl $beperm,0,$x10 # 0x00..0f
|
|
|
22d461 |
+ be?vspltisb $xt0,3 # 0x03..03
|
|
|
22d461 |
+ be?vxor $beperm,$beperm,$xt0 # swap bytes within words
|
|
|
22d461 |
+ be?vxxlor $xv26 ,$beperm,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlor $xv0 ,@K[0],@K[0] # K0,k1,k2 to vr0,1,2
|
|
|
22d461 |
+ vxxlor $xv1 ,@K[1],@K[1]
|
|
|
22d461 |
+ vxxlor $xv2 ,@K[2],@K[2]
|
|
|
22d461 |
+ vxxlor $xv3 ,@K[3],@K[3]
|
|
|
22d461 |
+ vxxlor $xv4 ,$xt1,$xt1 #CTR ->4, CTR+4-> 5
|
|
|
22d461 |
+ vxxlor $xv5 ,$xt3,$xt3
|
|
|
22d461 |
+ vxxlor $xv8 ,$xa0,$xa0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ li r0,10 # inner loop counter
|
|
|
22d461 |
+ mtctr r0
|
|
|
22d461 |
+ b Loop_outer_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+.align 5
|
|
|
22d461 |
+Loop_outer_vsx_8x:
|
|
|
22d461 |
+ vxxlorc $xa0,$xv22,$xv22 # load [smashed] sigma
|
|
|
22d461 |
+ vxxlorc $xa1,$xv23,$xv23
|
|
|
22d461 |
+ vxxlorc $xa2,$xv24,$xv24
|
|
|
22d461 |
+ vxxlorc $xa3,$xv25,$xv25
|
|
|
22d461 |
+ vxxlorc $xa4,$xv22,$xv22
|
|
|
22d461 |
+ vxxlorc $xa5,$xv23,$xv23
|
|
|
22d461 |
+ vxxlorc $xa6,$xv24,$xv24
|
|
|
22d461 |
+ vxxlorc $xa7,$xv25,$xv25
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltw $xb0,@K[1],0 # smash the key
|
|
|
22d461 |
+ vspltw $xb1,@K[1],1
|
|
|
22d461 |
+ vspltw $xb2,@K[1],2
|
|
|
22d461 |
+ vspltw $xb3,@K[1],3
|
|
|
22d461 |
+ vspltw $xb4,@K[1],0 # smash the key
|
|
|
22d461 |
+ vspltw $xb5,@K[1],1
|
|
|
22d461 |
+ vspltw $xb6,@K[1],2
|
|
|
22d461 |
+ vspltw $xb7,@K[1],3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltw $xc0,@K[2],0
|
|
|
22d461 |
+ vspltw $xc1,@K[2],1
|
|
|
22d461 |
+ vspltw $xc2,@K[2],2
|
|
|
22d461 |
+ vspltw $xc3,@K[2],3
|
|
|
22d461 |
+ vspltw $xc4,@K[2],0
|
|
|
22d461 |
+ vspltw $xc7,@K[2],3
|
|
|
22d461 |
+ vspltw $xc5,@K[2],1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc $xd0,$xv4,$xv4 # smash the counter
|
|
|
22d461 |
+ vspltw $xd1,@K[3],1
|
|
|
22d461 |
+ vspltw $xd2,@K[3],2
|
|
|
22d461 |
+ vspltw $xd3,@K[3],3
|
|
|
22d461 |
+ vxxlorc $xd4,$xv5,$xv5 # smash the counter
|
|
|
22d461 |
+ vspltw $xd5,@K[3],1
|
|
|
22d461 |
+ vspltw $xd6,@K[3],2
|
|
|
22d461 |
+ vspltw $xd7,@K[3],3
|
|
|
22d461 |
+ vxxlorc $xc6,$xv8,$xv8 #copy of vlspt k[2],2 is in v8.v26 ->k[3] so need to wait until k3 is done
|
|
|
22d461 |
+
|
|
|
22d461 |
+Loop_vsx_8x:
|
|
|
22d461 |
+___
|
|
|
22d461 |
+ foreach (&VSX_lane_ROUND_8x(0,4, 8,12,16,20,24,28)) { eval; }
|
|
|
22d461 |
+ foreach (&VSX_lane_ROUND_8x(0,5,10,15,16,21,26,31)) { eval; }
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+
|
|
|
22d461 |
+ bdnz Loop_vsx_8x
|
|
|
22d461 |
+ vxxlor $xv13 ,$xd4,$xd4 # save the register vr24-31
|
|
|
22d461 |
+ vxxlor $xv14 ,$xd5,$xd5 #
|
|
|
22d461 |
+ vxxlor $xv15 ,$xd6,$xd6 #
|
|
|
22d461 |
+ vxxlor $xv16 ,$xd7,$xd7 #
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlor $xv18 ,$xc4,$xc4 #
|
|
|
22d461 |
+ vxxlor $xv19 ,$xc5,$xc5 #
|
|
|
22d461 |
+ vxxlor $xv20 ,$xc6,$xc6 #
|
|
|
22d461 |
+ vxxlor $xv21 ,$xc7,$xc7 #
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlor $xv6 ,$xb6,$xb6 # save vr23, so we get 8 regs
|
|
|
22d461 |
+ vxxlor $xv7 ,$xb7,$xb7 # save vr23, so we get 8 regs
|
|
|
22d461 |
+ be?vxxlorc $beperm,$xv26,$xv26 # copy back the the beperm.
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc @K[0],$xv0,$xv0 #27
|
|
|
22d461 |
+ vxxlorc @K[1],$xv1,$xv1 #24
|
|
|
22d461 |
+ vxxlorc @K[2],$xv2,$xv2 #25
|
|
|
22d461 |
+ vxxlorc @K[3],$xv3,$xv3 #26
|
|
|
22d461 |
+ vxxlorc $CTR0,$xv4,$xv4
|
|
|
22d461 |
+###changing to vertical
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xt0,$xa0,$xa1 # transpose data
|
|
|
22d461 |
+ vmrgew $xt1,$xa2,$xa3
|
|
|
22d461 |
+ vmrgow $xa0,$xa0,$xa1
|
|
|
22d461 |
+ vmrgow $xa2,$xa2,$xa3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xt2,$xb0,$xb1
|
|
|
22d461 |
+ vmrgew $xt3,$xb2,$xb3
|
|
|
22d461 |
+ vmrgow $xb0,$xb0,$xb1
|
|
|
22d461 |
+ vmrgow $xb2,$xb2,$xb3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xd0,$xd0,$CTR0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vpermdi $xa1,$xa0,$xa2,0b00
|
|
|
22d461 |
+ vpermdi $xa3,$xa0,$xa2,0b11
|
|
|
22d461 |
+ vpermdi $xa0,$xt0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xa2,$xt0,$xt1,0b11
|
|
|
22d461 |
+ vpermdi $xb1,$xb0,$xb2,0b00
|
|
|
22d461 |
+ vpermdi $xb3,$xb0,$xb2,0b11
|
|
|
22d461 |
+ vpermdi $xb0,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xb2,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xt0,$xc0,$xc1
|
|
|
22d461 |
+ vmrgew $xt1,$xc2,$xc3
|
|
|
22d461 |
+ vmrgow $xc0,$xc0,$xc1
|
|
|
22d461 |
+ vmrgow $xc2,$xc2,$xc3
|
|
|
22d461 |
+ vmrgew $xt2,$xd0,$xd1
|
|
|
22d461 |
+ vmrgew $xt3,$xd2,$xd3
|
|
|
22d461 |
+ vmrgow $xd0,$xd0,$xd1
|
|
|
22d461 |
+ vmrgow $xd2,$xd2,$xd3
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vpermdi $xc1,$xc0,$xc2,0b00
|
|
|
22d461 |
+ vpermdi $xc3,$xc0,$xc2,0b11
|
|
|
22d461 |
+ vpermdi $xc0,$xt0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xc2,$xt0,$xt1,0b11
|
|
|
22d461 |
+ vpermdi $xd1,$xd0,$xd2,0b00
|
|
|
22d461 |
+ vpermdi $xd3,$xd0,$xd2,0b11
|
|
|
22d461 |
+ vpermdi $xd0,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xd2,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltisw $xt0,8
|
|
|
22d461 |
+ vadduwm $CTR0,$CTR0,$xt0 # next counter value
|
|
|
22d461 |
+ vxxlor $xv4 ,$CTR0,$CTR0 #CTR+4-> 5
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa0,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb0,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc0,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd0,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa1,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb1,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc1,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd1,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa2,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb2,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc2,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd2,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xa0,$xa3,@K[0]
|
|
|
22d461 |
+ vadduwm $xb0,$xb3,@K[1]
|
|
|
22d461 |
+ vadduwm $xc0,$xc3,@K[2]
|
|
|
22d461 |
+ vadduwm $xd0,$xd3,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xa0,$xa0,$xa0,$beperm
|
|
|
22d461 |
+ be?vperm $xb0,$xb0,$xb0,$beperm
|
|
|
22d461 |
+ be?vperm $xc0,$xc0,$xc0,$beperm
|
|
|
22d461 |
+ be?vperm $xd0,$xd0,$xd0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xt0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xt0,$xt0,$xa0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xb0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xc0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xd0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xt0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+#blk4-7: 24:31 remain the same as we can use the same logic above . Reg a4-b7 remain same.Load c4,d7--> position 8-15.we can reuse vr24-31.
|
|
|
22d461 |
+#VR0-3 : are used to load temp value, vr4 --> as xr0 instead of xt0.
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc $CTR1 ,$xv5,$xv5
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc $xcn4 ,$xv18,$xv18
|
|
|
22d461 |
+ vxxlorc $xcn5 ,$xv19,$xv19
|
|
|
22d461 |
+ vxxlorc $xcn6 ,$xv20,$xv20
|
|
|
22d461 |
+ vxxlorc $xcn7 ,$xv21,$xv21
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc $xdn4 ,$xv13,$xv13
|
|
|
22d461 |
+ vxxlorc $xdn5 ,$xv14,$xv14
|
|
|
22d461 |
+ vxxlorc $xdn6 ,$xv15,$xv15
|
|
|
22d461 |
+ vxxlorc $xdn7 ,$xv16,$xv16
|
|
|
22d461 |
+ vadduwm $xdn4,$xdn4,$CTR1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxxlorc $xb6 ,$xv6,$xv6
|
|
|
22d461 |
+ vxxlorc $xb7 ,$xv7,$xv7
|
|
|
22d461 |
+#use xa1->xr0, as xt0...in the block 4-7
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xr0,$xa4,$xa5 # transpose data
|
|
|
22d461 |
+ vmrgew $xt1,$xa6,$xa7
|
|
|
22d461 |
+ vmrgow $xa4,$xa4,$xa5
|
|
|
22d461 |
+ vmrgow $xa6,$xa6,$xa7
|
|
|
22d461 |
+ vmrgew $xt2,$xb4,$xb5
|
|
|
22d461 |
+ vmrgew $xt3,$xb6,$xb7
|
|
|
22d461 |
+ vmrgow $xb4,$xb4,$xb5
|
|
|
22d461 |
+ vmrgow $xb6,$xb6,$xb7
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vpermdi $xa5,$xa4,$xa6,0b00
|
|
|
22d461 |
+ vpermdi $xa7,$xa4,$xa6,0b11
|
|
|
22d461 |
+ vpermdi $xa4,$xr0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xa6,$xr0,$xt1,0b11
|
|
|
22d461 |
+ vpermdi $xb5,$xb4,$xb6,0b00
|
|
|
22d461 |
+ vpermdi $xb7,$xb4,$xb6,0b11
|
|
|
22d461 |
+ vpermdi $xb4,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xb6,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vmrgew $xr0,$xcn4,$xcn5
|
|
|
22d461 |
+ vmrgew $xt1,$xcn6,$xcn7
|
|
|
22d461 |
+ vmrgow $xcn4,$xcn4,$xcn5
|
|
|
22d461 |
+ vmrgow $xcn6,$xcn6,$xcn7
|
|
|
22d461 |
+ vmrgew $xt2,$xdn4,$xdn5
|
|
|
22d461 |
+ vmrgew $xt3,$xdn6,$xdn7
|
|
|
22d461 |
+ vmrgow $xdn4,$xdn4,$xdn5
|
|
|
22d461 |
+ vmrgow $xdn6,$xdn6,$xdn7
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vpermdi $xcn5,$xcn4,$xcn6,0b00
|
|
|
22d461 |
+ vpermdi $xcn7,$xcn4,$xcn6,0b11
|
|
|
22d461 |
+ vpermdi $xcn4,$xr0,$xt1,0b00
|
|
|
22d461 |
+ vpermdi $xcn6,$xr0,$xt1,0b11
|
|
|
22d461 |
+ vpermdi $xdn5,$xdn4,$xdn6,0b00
|
|
|
22d461 |
+ vpermdi $xdn7,$xdn4,$xdn6,0b11
|
|
|
22d461 |
+ vpermdi $xdn4,$xt2,$xt3,0b00
|
|
|
22d461 |
+ vpermdi $xdn6,$xt2,$xt3,0b11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vspltisw $xr0,8
|
|
|
22d461 |
+ vadduwm $CTR1,$CTR1,$xr0 # next counter value
|
|
|
22d461 |
+ vxxlor $xv5 ,$CTR1,$CTR1 #CTR+4-> 5
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xan0,$xa4,@K[0]
|
|
|
22d461 |
+ vadduwm $xbn0,$xb4,@K[1]
|
|
|
22d461 |
+ vadduwm $xcn0,$xcn4,@K[2]
|
|
|
22d461 |
+ vadduwm $xdn0,$xdn4,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xan0,$xa4,$xa4,$beperm
|
|
|
22d461 |
+ be?vperm $xbn0,$xb4,$xb4,$beperm
|
|
|
22d461 |
+ be?vperm $xcn0,$xcn4,$xcn4,$beperm
|
|
|
22d461 |
+ be?vperm $xdn0,$xdn4,$xdn4,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x_1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xr0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xr0,$xr0,$xan0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xbn0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xcn0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xdn0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xr0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xan0,$xa5,@K[0]
|
|
|
22d461 |
+ vadduwm $xbn0,$xb5,@K[1]
|
|
|
22d461 |
+ vadduwm $xcn0,$xcn5,@K[2]
|
|
|
22d461 |
+ vadduwm $xdn0,$xdn5,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xan0,$xan0,$xan0,$beperm
|
|
|
22d461 |
+ be?vperm $xbn0,$xbn0,$xbn0,$beperm
|
|
|
22d461 |
+ be?vperm $xcn0,$xcn0,$xcn0,$beperm
|
|
|
22d461 |
+ be?vperm $xdn0,$xdn0,$xdn0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x_1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xr0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xr0,$xr0,$xan0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xbn0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xcn0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xdn0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xr0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xan0,$xa6,@K[0]
|
|
|
22d461 |
+ vadduwm $xbn0,$xb6,@K[1]
|
|
|
22d461 |
+ vadduwm $xcn0,$xcn6,@K[2]
|
|
|
22d461 |
+ vadduwm $xdn0,$xdn6,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xan0,$xan0,$xan0,$beperm
|
|
|
22d461 |
+ be?vperm $xbn0,$xbn0,$xbn0,$beperm
|
|
|
22d461 |
+ be?vperm $xcn0,$xcn0,$xcn0,$beperm
|
|
|
22d461 |
+ be?vperm $xdn0,$xdn0,$xdn0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x_1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xr0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xr0,$xr0,$xan0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xbn0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xcn0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xdn0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xr0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vadduwm $xan0,$xa7,@K[0]
|
|
|
22d461 |
+ vadduwm $xbn0,$xb7,@K[1]
|
|
|
22d461 |
+ vadduwm $xcn0,$xcn7,@K[2]
|
|
|
22d461 |
+ vadduwm $xdn0,$xdn7,@K[3]
|
|
|
22d461 |
+
|
|
|
22d461 |
+ be?vperm $xan0,$xan0,$xan0,$beperm
|
|
|
22d461 |
+ be?vperm $xbn0,$xbn0,$xbn0,$beperm
|
|
|
22d461 |
+ be?vperm $xcn0,$xcn0,$xcn0,$beperm
|
|
|
22d461 |
+ be?vperm $xdn0,$xdn0,$xdn0,$beperm
|
|
|
22d461 |
+
|
|
|
22d461 |
+ ${UCMP}i $len,0x40
|
|
|
22d461 |
+ blt Ltail_vsx_8x_1
|
|
|
22d461 |
+
|
|
|
22d461 |
+ lvx_4w $xr0,$x00,$inp
|
|
|
22d461 |
+ lvx_4w $xt1,$x10,$inp
|
|
|
22d461 |
+ lvx_4w $xt2,$x20,$inp
|
|
|
22d461 |
+ lvx_4w $xt3,$x30,$inp
|
|
|
22d461 |
+
|
|
|
22d461 |
+ vxor $xr0,$xr0,$xan0
|
|
|
22d461 |
+ vxor $xt1,$xt1,$xbn0
|
|
|
22d461 |
+ vxor $xt2,$xt2,$xcn0
|
|
|
22d461 |
+ vxor $xt3,$xt3,$xdn0
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $xr0,$x00,$out
|
|
|
22d461 |
+ stvx_4w $xt1,$x10,$out
|
|
|
22d461 |
+ addi $inp,$inp,0x40
|
|
|
22d461 |
+ stvx_4w $xt2,$x20,$out
|
|
|
22d461 |
+ subi $len,$len,0x40
|
|
|
22d461 |
+ stvx_4w $xt3,$x30,$out
|
|
|
22d461 |
+ addi $out,$out,0x40
|
|
|
22d461 |
+ beq Ldone_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ mtctr r0
|
|
|
22d461 |
+ bne Loop_outer_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+Ldone_vsx_8x:
|
|
|
22d461 |
+ lwz r12,`$FRAME-4`($sp) # pull vrsave
|
|
|
22d461 |
+ li r10,`15+$LOCALS+64`
|
|
|
22d461 |
+ li r11,`31+$LOCALS+64`
|
|
|
22d461 |
+ $POP r0, `$FRAME+$LRSAVE`($sp)
|
|
|
22d461 |
+ mtspr 256,r12 # restore vrsave
|
|
|
22d461 |
+ lvx v24,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ lvx v25,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ lvx v26,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ lvx v27,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ lvx v28,r10,$sp
|
|
|
22d461 |
+ addi r10,r10,32
|
|
|
22d461 |
+ lvx v29,r11,$sp
|
|
|
22d461 |
+ addi r11,r11,32
|
|
|
22d461 |
+ lvx v30,r10,$sp
|
|
|
22d461 |
+ lvx v31,r11,$sp
|
|
|
22d461 |
+ mtlr r0
|
|
|
22d461 |
+ addi $sp,$sp,$FRAME
|
|
|
22d461 |
+ blr
|
|
|
22d461 |
+
|
|
|
22d461 |
+.align 4
|
|
|
22d461 |
+Ltail_vsx_8x:
|
|
|
22d461 |
+ addi r11,$sp,$LOCALS
|
|
|
22d461 |
+ mtctr $len
|
|
|
22d461 |
+ stvx_4w $xa0,$x00,r11 # offload block to stack
|
|
|
22d461 |
+ stvx_4w $xb0,$x10,r11
|
|
|
22d461 |
+ stvx_4w $xc0,$x20,r11
|
|
|
22d461 |
+ stvx_4w $xd0,$x30,r11
|
|
|
22d461 |
+ subi r12,r11,1 # prepare for *++ptr
|
|
|
22d461 |
+ subi $inp,$inp,1
|
|
|
22d461 |
+ subi $out,$out,1
|
|
|
22d461 |
+ bl Loop_tail_vsx_8x
|
|
|
22d461 |
+Ltail_vsx_8x_1:
|
|
|
22d461 |
+ addi r11,$sp,$LOCALS
|
|
|
22d461 |
+ mtctr $len
|
|
|
22d461 |
+ stvx_4w $xan0,$x00,r11 # offload block to stack
|
|
|
22d461 |
+ stvx_4w $xbn0,$x10,r11
|
|
|
22d461 |
+ stvx_4w $xcn0,$x20,r11
|
|
|
22d461 |
+ stvx_4w $xdn0,$x30,r11
|
|
|
22d461 |
+ subi r12,r11,1 # prepare for *++ptr
|
|
|
22d461 |
+ subi $inp,$inp,1
|
|
|
22d461 |
+ subi $out,$out,1
|
|
|
22d461 |
+ bl Loop_tail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+Loop_tail_vsx_8x:
|
|
|
22d461 |
+ lbzu r6,1(r12)
|
|
|
22d461 |
+ lbzu r7,1($inp)
|
|
|
22d461 |
+ xor r6,r6,r7
|
|
|
22d461 |
+ stbu r6,1($out)
|
|
|
22d461 |
+ bdnz Loop_tail_vsx_8x
|
|
|
22d461 |
+
|
|
|
22d461 |
+ stvx_4w $K[0],$x00,r11 # wipe copy of the block
|
|
|
22d461 |
+ stvx_4w $K[0],$x10,r11
|
|
|
22d461 |
+ stvx_4w $K[0],$x20,r11
|
|
|
22d461 |
+ stvx_4w $K[0],$x30,r11
|
|
|
22d461 |
+
|
|
|
22d461 |
+ b Ldone_vsx_8x
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+ .byte 0,12,0x04,1,0x80,0,5,0
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+.size .ChaCha20_ctr32_vsx_8x,.-.ChaCha20_ctr32_vsx_8x
|
|
|
22d461 |
+___
|
|
|
22d461 |
+}}}
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+.align 5
|
|
|
22d461 |
+Lconsts:
|
|
|
22d461 |
+ mflr r0
|
|
|
22d461 |
+ bcl 20,31,\$+4
|
|
|
22d461 |
+ mflr r12 #vvvvv "distance between . and Lsigma
|
|
|
22d461 |
+ addi r12,r12,`64-8`
|
|
|
22d461 |
+ mtlr r0
|
|
|
22d461 |
+ blr
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+ .byte 0,12,0x14,0,0,0,0,0
|
|
|
22d461 |
+ .space `64-9*4`
|
|
|
22d461 |
+Lsigma:
|
|
|
22d461 |
+ .long 0x61707865,0x3320646e,0x79622d32,0x6b206574
|
|
|
22d461 |
+ .long 1,0,0,0
|
|
|
22d461 |
+ .long 2,0,0,0
|
|
|
22d461 |
+ .long 3,0,0,0
|
|
|
22d461 |
+ .long 4,0,0,0
|
|
|
22d461 |
+___
|
|
|
22d461 |
+$code.=<<___ if ($LITTLE_ENDIAN);
|
|
|
22d461 |
+ .long 0x0e0f0c0d,0x0a0b0809,0x06070405,0x02030001
|
|
|
22d461 |
+ .long 0x0d0e0f0c,0x090a0b08,0x05060704,0x01020300
|
|
|
22d461 |
+___
|
|
|
22d461 |
+$code.=<<___ if (!$LITTLE_ENDIAN); # flipped words
|
|
|
22d461 |
+ .long 0x02030001,0x06070405,0x0a0b0809,0x0e0f0c0d
|
|
|
22d461 |
+ .long 0x01020300,0x05060704,0x090a0b08,0x0d0e0f0c
|
|
|
22d461 |
+___
|
|
|
22d461 |
+$code.=<<___;
|
|
|
22d461 |
+ .long 0x61707865,0x61707865,0x61707865,0x61707865
|
|
|
22d461 |
+ .long 0x3320646e,0x3320646e,0x3320646e,0x3320646e
|
|
|
22d461 |
+ .long 0x79622d32,0x79622d32,0x79622d32,0x79622d32
|
|
|
22d461 |
+ .long 0x6b206574,0x6b206574,0x6b206574,0x6b206574
|
|
|
22d461 |
+ .long 0,1,2,3
|
|
|
22d461 |
+ .long 0x03020100,0x07060504,0x0b0a0908,0x0f0e0d0c
|
|
|
22d461 |
+.asciz "ChaCha20 for PowerPC/AltiVec, CRYPTOGAMS by <appro\@openssl.org>"
|
|
|
22d461 |
+.align 2
|
|
|
22d461 |
+___
|
|
|
22d461 |
+
|
|
|
22d461 |
+foreach (split("\n",$code)) {
|
|
|
22d461 |
+ s/\`([^\`]*)\`/eval $1/ge;
|
|
|
22d461 |
+
|
|
|
22d461 |
+ # instructions prefixed with '?' are endian-specific and need
|
|
|
22d461 |
+ # to be adjusted accordingly...
|
|
|
22d461 |
+ if ($flavour !~ /le$/) { # big-endian
|
|
|
22d461 |
+ s/be\?// or
|
|
|
22d461 |
+ s/le\?/#le#/ or
|
|
|
22d461 |
+ s/\?lvsr/lvsl/ or
|
|
|
22d461 |
+ s/\?lvsl/lvsr/ or
|
|
|
22d461 |
+ s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/ or
|
|
|
22d461 |
+ s/vrldoi(\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9]+)/vsldoi$1$2$2 16-$3/;
|
|
|
22d461 |
+ } else { # little-endian
|
|
|
22d461 |
+ s/le\?// or
|
|
|
22d461 |
+ s/be\?/#be#/ or
|
|
|
22d461 |
+ s/\?([a-z]+)/$1/ or
|
|
|
22d461 |
+ s/vrldoi(\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9]+)/vsldoi$1$2$2 $3/;
|
|
|
22d461 |
+ }
|
|
|
22d461 |
+
|
|
|
22d461 |
+ print $_,"\n";
|
|
|
22d461 |
+}
|
|
|
22d461 |
+
|
|
|
22d461 |
+close STDOUT or die "error closing STDOUT: $!";
|
|
|
22d461 |
diff --git a/crypto/chacha/build.info b/crypto/chacha/build.info
|
|
|
22d461 |
index c12cb9c..2a819b2 100644
|
|
|
22d461 |
--- a/crypto/chacha/build.info
|
|
|
22d461 |
+++ b/crypto/chacha/build.info
|
|
|
22d461 |
@@ -12,7 +12,7 @@ IF[{- !$disabled{asm} -}]
|
|
|
22d461 |
$CHACHAASM_armv4=chacha-armv4.S
|
|
|
22d461 |
$CHACHAASM_aarch64=chacha-armv8.S
|
|
|
22d461 |
|
|
|
22d461 |
- $CHACHAASM_ppc32=chacha_ppc.c chacha-ppc.s
|
|
|
22d461 |
+ $CHACHAASM_ppc32=chacha_ppc.c chacha-ppc.s chachap10-ppc.s
|
|
|
22d461 |
$CHACHAASM_ppc64=$CHACHAASM_ppc32
|
|
|
22d461 |
|
|
|
22d461 |
$CHACHAASM_c64xplus=chacha-c64xplus.s
|
|
|
22d461 |
@@ -29,6 +29,7 @@ SOURCE[../../libcrypto]=$CHACHAASM
|
|
|
1ac26c |
GENERATE[chacha-x86.S]=asm/chacha-x86.pl
|
|
|
22d461 |
GENERATE[chacha-x86_64.s]=asm/chacha-x86_64.pl
|
|
|
22d461 |
GENERATE[chacha-ppc.s]=asm/chacha-ppc.pl
|
|
|
22d461 |
+GENERATE[chachap10-ppc.s]=asm/chachap10-ppc.pl
|
|
|
22d461 |
GENERATE[chacha-armv4.S]=asm/chacha-armv4.pl
|
|
|
22d461 |
INCLUDE[chacha-armv4.o]=..
|
|
|
22d461 |
GENERATE[chacha-armv8.S]=asm/chacha-armv8.pl
|
|
|
22d461 |
diff --git a/crypto/chacha/chacha_ppc.c b/crypto/chacha/chacha_ppc.c
|
|
|
22d461 |
index 5319040..f99cca8 100644
|
|
|
22d461 |
--- a/crypto/chacha/chacha_ppc.c
|
|
|
22d461 |
+++ b/crypto/chacha/chacha_ppc.c
|
|
|
22d461 |
@@ -23,13 +23,18 @@ void ChaCha20_ctr32_vmx(unsigned char *out, const unsigned char *inp,
|
|
|
22d461 |
void ChaCha20_ctr32_vsx(unsigned char *out, const unsigned char *inp,
|
|
|
22d461 |
size_t len, const unsigned int key[8],
|
|
|
22d461 |
const unsigned int counter[4]);
|
|
|
22d461 |
+void ChaCha20_ctr32_vsx_p10(unsigned char *out, const unsigned char *inp,
|
|
|
22d461 |
+ size_t len, const unsigned int key[8],
|
|
|
22d461 |
+ const unsigned int counter[4]);
|
|
|
22d461 |
void ChaCha20_ctr32(unsigned char *out, const unsigned char *inp,
|
|
|
22d461 |
size_t len, const unsigned int key[8],
|
|
|
22d461 |
const unsigned int counter[4])
|
|
|
22d461 |
{
|
|
|
22d461 |
- OPENSSL_ppccap_P & PPC_CRYPTO207
|
|
|
22d461 |
- ? ChaCha20_ctr32_vsx(out, inp, len, key, counter)
|
|
|
22d461 |
- : OPENSSL_ppccap_P & PPC_ALTIVEC
|
|
|
22d461 |
- ? ChaCha20_ctr32_vmx(out, inp, len, key, counter)
|
|
|
22d461 |
- : ChaCha20_ctr32_int(out, inp, len, key, counter);
|
|
|
22d461 |
+ OPENSSL_ppccap_P & PPC_BRD31
|
|
|
22d461 |
+ ? ChaCha20_ctr32_vsx_p10(out, inp, len, key, counter)
|
|
|
22d461 |
+ :OPENSSL_ppccap_P & PPC_CRYPTO207
|
|
|
22d461 |
+ ? ChaCha20_ctr32_vsx(out, inp, len, key, counter)
|
|
|
22d461 |
+ : OPENSSL_ppccap_P & PPC_ALTIVEC
|
|
|
22d461 |
+ ? ChaCha20_ctr32_vmx(out, inp, len, key, counter)
|
|
|
22d461 |
+ : ChaCha20_ctr32_int(out, inp, len, key, counter);
|
|
|
22d461 |
}
|
|
|
22d461 |
diff --git a/crypto/perlasm/ppc-xlate.pl b/crypto/perlasm/ppc-xlate.pl
|
|
|
22d461 |
index 2ee4440..4590340 100755
|
|
|
22d461 |
--- a/crypto/perlasm/ppc-xlate.pl
|
|
|
22d461 |
+++ b/crypto/perlasm/ppc-xlate.pl
|
|
|
22d461 |
@@ -293,6 +293,14 @@ my $vpermdi = sub { # xxpermdi
|
|
|
22d461 |
$dm = oct($dm) if ($dm =~ /^0/);
|
|
|
22d461 |
" .long ".sprintf "0x%X",(60<<26)|($vrt<<21)|($vra<<16)|($vrb<<11)|($dm<<8)|(10<<3)|7;
|
|
|
22d461 |
};
|
|
|
22d461 |
+my $vxxlor = sub { # xxlor
|
|
|
22d461 |
+ my ($f, $vrt, $vra, $vrb) = @_;
|
|
|
22d461 |
+ " .long ".sprintf "0x%X",(60<<26)|($vrt<<21)|($vra<<16)|($vrb<<11)|(146<<3)|6;
|
|
|
22d461 |
+};
|
|
|
22d461 |
+my $vxxlorc = sub { # xxlor
|
|
|
22d461 |
+ my ($f, $vrt, $vra, $vrb) = @_;
|
|
|
22d461 |
+ " .long ".sprintf "0x%X",(60<<26)|($vrt<<21)|($vra<<16)|($vrb<<11)|(146<<3)|1;
|
|
|
22d461 |
+};
|
|
|
22d461 |
|
|
|
22d461 |
# PowerISA 2.07 stuff
|
|
|
22d461 |
sub vcrypto_op {
|
|
|
22d461 |
@@ -377,6 +385,15 @@ my $addex = sub {
|
|
|
22d461 |
};
|
|
|
22d461 |
my $vmsumudm = sub { vfour_vsr(@_, 35); };
|
|
|
22d461 |
|
|
|
22d461 |
+# PowerISA 3.1 stuff
|
|
|
22d461 |
+my $brd = sub {
|
|
|
22d461 |
+ my ($f, $ra, $rs) = @_;
|
|
|
22d461 |
+ " .long ".sprintf "0x%X",(31<<26)|($rs<<21)|($ra<<16)|(187<<1);
|
|
|
22d461 |
+};
|
|
|
22d461 |
+my $vsrq = sub { vcrypto_op(@_, 517); };
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
while($line=<>) {
|
|
|
22d461 |
|
|
|
22d461 |
$line =~ s|[#!;].*$||; # get rid of asm-style comments...
|
|
|
22d461 |
diff --git a/crypto/ppccap.c b/crypto/ppccap.c
|
|
|
22d461 |
index 8bcfed2..664627c 100644
|
|
|
22d461 |
--- a/crypto/ppccap.c
|
|
|
22d461 |
+++ b/crypto/ppccap.c
|
|
|
22d461 |
@@ -45,6 +45,7 @@ void OPENSSL_ppc64_probe(void);
|
|
|
22d461 |
void OPENSSL_altivec_probe(void);
|
|
|
22d461 |
void OPENSSL_crypto207_probe(void);
|
|
|
22d461 |
void OPENSSL_madd300_probe(void);
|
|
|
22d461 |
+void OPENSSL_brd31_probe(void);
|
|
|
22d461 |
|
|
|
22d461 |
long OPENSSL_rdtsc_mftb(void);
|
|
|
22d461 |
long OPENSSL_rdtsc_mfspr268(void);
|
|
|
22d461 |
@@ -117,16 +118,21 @@ static unsigned long getauxval(unsigned long key)
|
|
|
22d461 |
#endif
|
|
|
22d461 |
|
|
|
22d461 |
/* I wish <sys/auxv.h> was universally available */
|
|
|
22d461 |
-#define HWCAP 16 /* AT_HWCAP */
|
|
|
22d461 |
+#ifndef AT_HWCAP
|
|
|
22d461 |
+# define AT_HWCAP 16 /* AT_HWCAP */
|
|
|
22d461 |
+#endif
|
|
|
22d461 |
#define HWCAP_PPC64 (1U << 30)
|
|
|
22d461 |
#define HWCAP_ALTIVEC (1U << 28)
|
|
|
22d461 |
#define HWCAP_FPU (1U << 27)
|
|
|
22d461 |
#define HWCAP_POWER6_EXT (1U << 9)
|
|
|
22d461 |
#define HWCAP_VSX (1U << 7)
|
|
|
22d461 |
|
|
|
22d461 |
-#define HWCAP2 26 /* AT_HWCAP2 */
|
|
|
22d461 |
+#ifndef AT_HWCAP2
|
|
|
22d461 |
+# define AT_HWCAP2 26 /* AT_HWCAP2 */
|
|
|
22d461 |
+#endif
|
|
|
22d461 |
#define HWCAP_VEC_CRYPTO (1U << 25)
|
|
|
22d461 |
#define HWCAP_ARCH_3_00 (1U << 23)
|
|
|
22d461 |
+#define HWCAP_ARCH_3_1 (1U << 18)
|
|
|
22d461 |
|
|
|
22d461 |
# if defined(__GNUC__) && __GNUC__>=2
|
|
|
22d461 |
__attribute__ ((constructor))
|
|
|
22d461 |
@@ -187,6 +193,9 @@ void OPENSSL_cpuid_setup(void)
|
|
|
22d461 |
if (__power_set(0xffffffffU<<17)) /* POWER9 and later */
|
|
|
22d461 |
OPENSSL_ppccap_P |= PPC_MADD300;
|
|
|
22d461 |
|
|
|
22d461 |
+ if (__power_set(0xffffffffU<<18)) /* POWER10 and later */
|
|
|
22d461 |
+ OPENSSL_ppccap_P |= PPC_BRD31;
|
|
|
22d461 |
+
|
|
|
22d461 |
return;
|
|
|
22d461 |
# endif
|
|
|
22d461 |
#endif
|
|
|
22d461 |
@@ -215,8 +224,8 @@ void OPENSSL_cpuid_setup(void)
|
|
|
22d461 |
|
|
|
22d461 |
#ifdef OSSL_IMPLEMENT_GETAUXVAL
|
|
|
22d461 |
{
|
|
|
22d461 |
- unsigned long hwcap = getauxval(HWCAP);
|
|
|
22d461 |
- unsigned long hwcap2 = getauxval(HWCAP2);
|
|
|
22d461 |
+ unsigned long hwcap = getauxval(AT_HWCAP);
|
|
|
22d461 |
+ unsigned long hwcap2 = getauxval(AT_HWCAP2);
|
|
|
22d461 |
|
|
|
22d461 |
if (hwcap & HWCAP_FPU) {
|
|
|
22d461 |
OPENSSL_ppccap_P |= PPC_FPU;
|
|
|
22d461 |
@@ -242,6 +251,10 @@ void OPENSSL_cpuid_setup(void)
|
|
|
22d461 |
if (hwcap2 & HWCAP_ARCH_3_00) {
|
|
|
22d461 |
OPENSSL_ppccap_P |= PPC_MADD300;
|
|
|
22d461 |
}
|
|
|
22d461 |
+
|
|
|
22d461 |
+ if (hwcap2 & HWCAP_ARCH_3_1) {
|
|
|
22d461 |
+ OPENSSL_ppccap_P |= PPC_BRD31;
|
|
|
22d461 |
+ }
|
|
|
22d461 |
}
|
|
|
22d461 |
#endif
|
|
|
22d461 |
|
|
|
22d461 |
@@ -263,7 +276,7 @@ void OPENSSL_cpuid_setup(void)
|
|
|
22d461 |
sigaction(SIGILL, &ill_act, &ill_oact);
|
|
|
22d461 |
|
|
|
22d461 |
#ifndef OSSL_IMPLEMENT_GETAUXVAL
|
|
|
22d461 |
- if (sigsetjmp(ill_jmp,1) == 0) {
|
|
|
22d461 |
+ if (sigsetjmp(ill_jmp, 1) == 0) {
|
|
|
22d461 |
OPENSSL_fpu_probe();
|
|
|
22d461 |
OPENSSL_ppccap_P |= PPC_FPU;
|
|
|
22d461 |
|
|
|
22d461 |
diff --git a/crypto/ppccpuid.pl b/crypto/ppccpuid.pl
|
|
|
22d461 |
index c6555df..706164a 100755
|
|
|
22d461 |
--- a/crypto/ppccpuid.pl
|
|
|
22d461 |
+++ b/crypto/ppccpuid.pl
|
|
|
22d461 |
@@ -81,6 +81,17 @@ $code=<<___;
|
|
|
22d461 |
.long 0
|
|
|
22d461 |
.byte 0,12,0x14,0,0,0,0,0
|
|
|
22d461 |
|
|
|
22d461 |
+.globl .OPENSSL_brd31_probe
|
|
|
22d461 |
+.align 4
|
|
|
22d461 |
+.OPENSSL_brd31_probe:
|
|
|
22d461 |
+ xor r0,r0,r0
|
|
|
22d461 |
+ brd r3,r0
|
|
|
22d461 |
+ blr
|
|
|
22d461 |
+ .long 0
|
|
|
22d461 |
+ .byte 0,12,0x14,0,0,0,0,0
|
|
|
22d461 |
+.size .OPENSSL_brd31_probe,.-.OPENSSL_brd31_probe
|
|
|
22d461 |
+
|
|
|
22d461 |
+
|
|
|
22d461 |
.globl .OPENSSL_wipe_cpu
|
|
|
22d461 |
.align 4
|
|
|
22d461 |
.OPENSSL_wipe_cpu:
|
|
|
22d461 |
diff --git a/include/crypto/ppc_arch.h b/include/crypto/ppc_arch.h
|
|
|
22d461 |
index 3b3ce4b..fcc846c 100644
|
|
|
22d461 |
--- a/include/crypto/ppc_arch.h
|
|
|
22d461 |
+++ b/include/crypto/ppc_arch.h
|
|
|
22d461 |
@@ -24,5 +24,6 @@ extern unsigned int OPENSSL_ppccap_P;
|
|
|
22d461 |
# define PPC_MADD300 (1<<4)
|
|
|
22d461 |
# define PPC_MFTB (1<<5)
|
|
|
22d461 |
# define PPC_MFSPR268 (1<<6)
|
|
|
22d461 |
+# define PPC_BRD31 (1<<7)
|
|
|
22d461 |
|
|
|
22d461 |
#endif
|