#!/usr/bin/env perl
-# Copyright 2017 The OpenSSL Project Authors. All Rights Reserved.
+# Copyright 2017-2018 The OpenSSL Project Authors. All Rights Reserved.
#
-# Licensed under the OpenSSL license (the "License"). You may not use
+# Licensed under the Apache License 2.0 (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
# is the keyword, i.e. "almost 1/2" is kind of complexity cap [for the
# time being]...
#
+# Reduce per-round instruction count in Thumb-2 case by 16%. This is
+# achieved by folding ldr/str pairs to their double-word counterparts.
+# Theoretically this should have improved performance on single-issue
+# cores, such as Cortex-A5/A7, by 19%. Reality is a bit different, as
+# usual...
+#
########################################################################
# Numbers are cycles per processed byte. Non-NEON results account even
# for input bit interleaving.
#
-# r=1088(*), NEON
+# r=1088(*) Thumb-2(**) NEON
#
# ARM11xx 82/+150%
-# Cortex-A5 88/+160%, 36
-# Cortex-A7 78/+160%, 34
-# Cortex-A8 51/+230%, 30
-# Cortex-A9 53/+210%, 26
-# Cortex-A15 42/+160%, 18
-# Snapdragon S4 43/+210%, 24
+# Cortex-A5 88/+160%, 86, 36
+# Cortex-A7 78/+160%, 68, 34
+# Cortex-A8 51/+230%, 57, 30
+# Cortex-A9 53/+210%, 51, 26
+# Cortex-A15 42/+160%, 38, 18
+# Snapdragon S4 43/+210%, 38, 24
#
# (*) Corresponds to SHA3-256. Percentage after slash is improvement
# over compiler-generated KECCAK_2X reference code.
+# (**) Thumb-2 results for Cortex-A5/A7 are likely to apply even to
+# Cortex-Mx, x>=3. Otherwise, non-NEON results for NEON-capable
+# processors are presented mostly for reference purposes.
+
+$flavour = shift;
+if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; }
+else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} }
+
+if ($flavour && $flavour ne "void") {
+ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+ ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
+ ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
+ die "can't locate arm-xlate.pl";
+
+ open STDOUT,"| \"$^X\" $xlate $flavour $output";
+} else {
+ open STDOUT,">$output";
+}
my @C = map("r$_",(0..9));
my @E = map("r$_",(10..12,14));
my @T = map([ 8*$_, 8*($_+1), 8*($_+2), 8*($_+3), 8*($_+4) ], (30,35,40,45,50));
$code.=<<___;
-.text
+#include "arm_arch.h"
#if defined(__thumb2__)
.syntax unified
.code 32
#endif
+.text
+
.type iotas32, %object
.align 5
iotas32:
str lr,[sp,#440]
eor @E[1],@E[1],@E[1]
str @E[1],[sp,#444]
- b .Lround
+ b .Lround2x
.align 4
-.Lround:
+.Lround2x:
___
sub Round {
my (@A,@R); (@A[0..4],@R) = @_;
$code.=<<___;
ldmia @E[2],{@C[0]-@C[3]} @ A[0][0..1]
ldmia @E[0],{@E[0]-@E[2],@E[3]} @ A[1][0..1]
+#ifdef __thumb2__
+ eor @C[0],@C[0],@E[0]
+ eor @C[1],@C[1],@E[1]
+ eor @C[2],@C[2],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[1][2]]
+ eor @C[3],@C[3],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[1][3]]
+ eor @C[4],@C[4],@E[0]
+ eor @C[5],@C[5],@E[1]
+ eor @C[6],@C[6],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[1][4]]
+ eor @C[7],@C[7],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[2][0]]
+ eor @C[8],@C[8],@E[0]
+ eor @C[9],@C[9],@E[1]
+ eor @C[0],@C[0],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[2][1]]
+ eor @C[1],@C[1],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[2][2]]
+ eor @C[2],@C[2],@E[0]
+ eor @C[3],@C[3],@E[1]
+ eor @C[4],@C[4],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[2][3]]
+ eor @C[5],@C[5],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[2][4]]
+ eor @C[6],@C[6],@E[0]
+ eor @C[7],@C[7],@E[1]
+ eor @C[8],@C[8],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[3][0]]
+ eor @C[9],@C[9],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[3][1]]
+ eor @C[0],@C[0],@E[0]
+ eor @C[1],@C[1],@E[1]
+ eor @C[2],@C[2],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[3][2]]
+ eor @C[3],@C[3],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[3][3]]
+ eor @C[4],@C[4],@E[0]
+ eor @C[5],@C[5],@E[1]
+ eor @C[6],@C[6],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[3][4]]
+ eor @C[7],@C[7],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[4][0]]
+ eor @C[8],@C[8],@E[0]
+ eor @C[9],@C[9],@E[1]
+ eor @C[0],@C[0],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[4][1]]
+ eor @C[1],@C[1],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[0][2]]
+ eor @C[2],@C[2],@E[0]
+ eor @C[3],@C[3],@E[1]
+ eor @C[4],@C[4],@E[2]
+ ldrd @E[0],@E[1],[sp,#$A[0][3]]
+ eor @C[5],@C[5],@E[3]
+ ldrd @E[2],@E[3],[sp,#$A[0][4]]
+#else
eor @C[0],@C[0],@E[0]
add @E[0],sp,#$A[1][2]
eor @C[1],@C[1],@E[1]
eor @C[4],@C[4],@E[2]
eor @C[5],@C[5],@E[3]
ldmia @E[0],{@E[0]-@E[2],@E[3]} @ A[0][3..4]
+#endif
eor @C[6],@C[6],@E[0]
eor @C[7],@C[7],@E[1]
eor @C[8],@C[8],@E[2]
eor @C[9],@C[9],@E[3]
eor @E[0],@C[0],@C[5],ror#32-1 @ E[0] = ROL64(C[2], 1) ^ C[0];
+ str.l @E[0],[sp,#$D[1]] @ D[1] = E[0]
eor @E[1],@C[1],@C[4]
- str @E[0],[sp,#$D[1]] @ D[1] = E[0]
+ str.h @E[1],[sp,#$D[1]+4]
eor @E[2],@C[6],@C[1],ror#32-1 @ E[1] = ROL64(C[0], 1) ^ C[3];
- str @E[1],[sp,#$D[1]+4]
eor @E[3],@C[7],@C[0]
- str @E[2],[sp,#$D[4]] @ D[4] = E[1]
+ str.l @E[2],[sp,#$D[4]] @ D[4] = E[1]
eor @C[0],@C[8],@C[3],ror#32-1 @ C[0] = ROL64(C[1], 1) ^ C[4];
- str @E[3],[sp,#$D[4]+4]
+ str.h @E[3],[sp,#$D[4]+4]
eor @C[1],@C[9],@C[2]
- str @C[0],[sp,#$D[0]] @ D[0] = C[0]
+ str.l @C[0],[sp,#$D[0]] @ D[0] = C[0]
eor @C[2],@C[2],@C[7],ror#32-1 @ C[1] = ROL64(C[3], 1) ^ C[1];
- str @C[1],[sp,#$D[0]+4]
+ ldr.l @C[7],[sp,#$A[3][3]]
eor @C[3],@C[3],@C[6]
- ldr @C[7],[sp,#$A[3][3]]
- str @C[2],[sp,#$D[2]] @ D[2] = C[1]
+ str.h @C[1],[sp,#$D[0]+4]
+ ldr.h @C[6],[sp,#$A[3][3]+4]
+ str.l @C[2],[sp,#$D[2]] @ D[2] = C[1]
eor @C[4],@C[4],@C[9],ror#32-1 @ C[2] = ROL64(C[4], 1) ^ C[2];
- ldr @C[6],[sp,#$A[3][3]+4]
- str @C[3],[sp,#$D[2]+4]
+ str.h @C[3],[sp,#$D[2]+4]
eor @C[5],@C[5],@C[8]
- ldr @C[8],[sp,#$A[4][4]]
- ldr @C[9],[sp,#$A[4][4]+4]
- str @C[4],[sp,#$D[3]] @ D[3] = C[2]
+ ldr.l @C[8],[sp,#$A[4][4]]
+ ldr.h @C[9],[sp,#$A[4][4]+4]
+ str.l @C[4],[sp,#$D[3]] @ D[3] = C[2]
eor @C[7],@C[7],@C[4]
- str @C[5],[sp,#$D[3]+4]
+ str.h @C[5],[sp,#$D[3]+4]
eor @C[6],@C[6],@C[5]
- ldr @C[4],[sp,#$A[0][0]]
+ ldr.l @C[4],[sp,#$A[0][0]]
@ ror @C[7],@C[7],#32-10 @ C[3] = ROL64(A[3][3] ^ C[2], rhotates[3][3]); /* D[3] */
@ ror @C[6],@C[6],#32-11
+ ldr.h @C[5],[sp,#$A[0][0]+4]
eor @C[8],@C[8],@E[2]
- ldr @C[5],[sp,#$A[0][0]+4]
eor @C[9],@C[9],@E[3]
- ldr @E[2],[sp,#$A[2][2]]
+ ldr.l @E[2],[sp,#$A[2][2]]
eor @C[0],@C[0],@C[4]
- ldr @E[3],[sp,#$A[2][2]+4]
+ ldr.h @E[3],[sp,#$A[2][2]+4]
@ ror @C[8],@C[8],#32-7 @ C[4] = ROL64(A[4][4] ^ E[1], rhotates[4][4]); /* D[4] */
@ ror @C[9],@C[9],#32-7
eor @C[1],@C[1],@C[5] @ C[0] = A[0][0] ^ C[0]; /* rotate by 0 */ /* D[0] */
eor @E[2],@E[2],@C[2]
- ldr @C[2],[sp,#$A[1][1]]
+ ldr.l @C[2],[sp,#$A[1][1]]
eor @E[3],@E[3],@C[3]
- ldr @C[3],[sp,#$A[1][1]+4]
+ ldr.h @C[3],[sp,#$A[1][1]+4]
ror @C[5],@E[2],#32-21 @ C[2] = ROL64(A[2][2] ^ C[1], rhotates[2][2]); /* D[2] */
ldr @E[2],[sp,#444] @ load counter
eor @C[2],@C[2],@E[0]
ldmia @E[3],{@E[0],@E[1]} @ iotas[i]
___
$code.=<<___ if ($A[0][0] == $T[0][0]);
- ldr @E[0],[@E[3],#8] @ iotas[i].lo
+ ldr.l @E[0],[@E[3],#8] @ iotas[i].lo
add @E[2],@E[2],#16
- ldr @E[1],[@E[3],#12] @ iotas[i].hi
+ ldr.h @E[1],[@E[3],#12] @ iotas[i].hi
cmp @E[2],#192
str @E[2],[sp,#444] @ store counter
___
eor @E[3],@E[3],@C[1]
eor @E[0],@E[0],@E[2]
eor @E[1],@E[1],@E[3]
- str @E[0],[sp,#$R[0][0]] @ R[0][0] = C[0] ^ (~C[1] & C[2]) ^ iotas[i];
+ str.l @E[0],[sp,#$R[0][0]] @ R[0][0] = C[0] ^ (~C[1] & C[2]) ^ iotas[i];
bic @E[2],@C[6],@C[4],ror#11
- str @E[1],[sp,#$R[0][0]+4]
+ str.h @E[1],[sp,#$R[0][0]+4]
bic @E[3],@C[7],@C[5],ror#10
bic @E[0],@C[8],@C[6],ror#32-(11-7)
bic @E[1],@C[9],@C[7],ror#32-(10-7)
eor @E[2],@C[2],@E[2],ror#32-11
+ str.l @E[2],[sp,#$R[0][1]] @ R[0][1] = C[1] ^ (~C[2] & C[3]);
eor @E[3],@C[3],@E[3],ror#32-10
- str @E[2],[sp,#$R[0][1]] @ R[0][1] = C[1] ^ (~C[2] & C[3]);
+ str.h @E[3],[sp,#$R[0][1]+4]
eor @E[0],@C[4],@E[0],ror#32-7
- str @E[3],[sp,#$R[0][1]+4]
eor @E[1],@C[5],@E[1],ror#32-7
- str @E[0],[sp,#$R[0][2]] @ R[0][2] = C[2] ^ (~C[3] & C[4]);
+ str.l @E[0],[sp,#$R[0][2]] @ R[0][2] = C[2] ^ (~C[3] & C[4]);
bic @E[2],@C[0],@C[8],ror#32-7
- str @E[1],[sp,#$R[0][2]+4]
+ str.h @E[1],[sp,#$R[0][2]+4]
bic @E[3],@C[1],@C[9],ror#32-7
eor @E[2],@E[2],@C[6],ror#32-11
+ str.l @E[2],[sp,#$R[0][3]] @ R[0][3] = C[3] ^ (~C[4] & C[0]);
eor @E[3],@E[3],@C[7],ror#32-10
- str @E[2],[sp,#$R[0][3]] @ R[0][3] = C[3] ^ (~C[4] & C[0]);
+ str.h @E[3],[sp,#$R[0][3]+4]
bic @E[0],@C[2],@C[0]
- str @E[3],[sp,#$R[0][3]+4]
add @E[3],sp,#$D[3]
+ ldr.l @C[0],[sp,#$A[0][3]] @ A[0][3]
bic @E[1],@C[3],@C[1]
- ldr @C[0],[sp,#$A[0][3]] @ A[0][3]
+ ldr.h @C[1],[sp,#$A[0][3]+4]
eor @E[0],@E[0],@C[8],ror#32-7
- ldr @C[1],[sp,#$A[0][3]+4]
eor @E[1],@E[1],@C[9],ror#32-7
- str @E[0],[sp,#$R[0][4]] @ R[0][4] = C[4] ^ (~C[0] & C[1]);
+ str.l @E[0],[sp,#$R[0][4]] @ R[0][4] = C[4] ^ (~C[0] & C[1]);
add @C[9],sp,#$D[0]
- str @E[1],[sp,#$R[0][4]+4]
+ str.h @E[1],[sp,#$R[0][4]+4]
ldmia @E[3],{@E[0]-@E[2],@E[3]} @ D[3..4]
ldmia @C[9],{@C[6]-@C[9]} @ D[0..1]
- ldr @C[2],[sp,#$A[1][4]] @ A[1][4]
+ ldr.l @C[2],[sp,#$A[1][4]] @ A[1][4]
eor @C[0],@C[0],@E[0]
- ldr @C[3],[sp,#$A[1][4]+4]
+ ldr.h @C[3],[sp,#$A[1][4]+4]
eor @C[1],@C[1],@E[1]
@ ror @C[0],@C[0],#32-14 @ C[0] = ROL64(A[0][3] ^ D[3], rhotates[0][3]);
- ldr @E[0],[sp,#$A[3][1]] @ A[3][1]
+ ldr.l @E[0],[sp,#$A[3][1]] @ A[3][1]
@ ror @C[1],@C[1],#32-14
- ldr @E[1],[sp,#$A[3][1]+4]
+ ldr.h @E[1],[sp,#$A[3][1]+4]
eor @C[2],@C[2],@E[2]
- ldr @C[4],[sp,#$A[2][0]] @ A[2][0]
+ ldr.l @C[4],[sp,#$A[2][0]] @ A[2][0]
eor @C[3],@C[3],@E[3]
- ldr @C[5],[sp,#$A[2][0]+4]
+ ldr.h @C[5],[sp,#$A[2][0]+4]
@ ror @C[2],@C[2],#32-10 @ C[1] = ROL64(A[1][4] ^ D[4], rhotates[1][4]);
@ ror @C[3],@C[3],#32-10
eor @C[6],@C[6],@C[4]
- ldr @E[2],[sp,#$D[2]] @ D[2]
+ ldr.l @E[2],[sp,#$D[2]] @ D[2]
eor @C[7],@C[7],@C[5]
- ldr @E[3],[sp,#$D[2]+4]
+ ldr.h @E[3],[sp,#$D[2]+4]
ror @C[5],@C[6],#32-1 @ C[2] = ROL64(A[2][0] ^ D[0], rhotates[2][0]);
ror @C[4],@C[7],#32-2
eor @E[0],@E[0],@C[8]
- ldr @C[8],[sp,#$A[4][2]] @ A[4][2]
+ ldr.l @C[8],[sp,#$A[4][2]] @ A[4][2]
eor @E[1],@E[1],@C[9]
- ldr @C[9],[sp,#$A[4][2]+4]
+ ldr.h @C[9],[sp,#$A[4][2]+4]
ror @C[7],@E[0],#32-22 @ C[3] = ROL64(A[3][1] ^ D[1], rhotates[3][1]);
ror @C[6],@E[1],#32-23
ror @C[8],@E[3],#32-31
eor @E[0],@E[0],@C[0],ror#32-14
eor @E[1],@E[1],@C[1],ror#32-14
- str @E[0],[sp,#$R[1][0]] @ R[1][0] = C[0] ^ (~C[1] & C[2])
+ str.l @E[0],[sp,#$R[1][0]] @ R[1][0] = C[0] ^ (~C[1] & C[2])
bic @E[2],@C[6],@C[4]
- str @E[1],[sp,#$R[1][0]+4]
+ str.h @E[1],[sp,#$R[1][0]+4]
bic @E[3],@C[7],@C[5]
eor @E[2],@E[2],@C[2],ror#32-10
+ str.l @E[2],[sp,#$R[1][1]] @ R[1][1] = C[1] ^ (~C[2] & C[3]);
eor @E[3],@E[3],@C[3],ror#32-10
- str @E[2],[sp,#$R[1][1]] @ R[1][1] = C[1] ^ (~C[2] & C[3]);
+ str.h @E[3],[sp,#$R[1][1]+4]
bic @E[0],@C[8],@C[6]
- str @E[3],[sp,#$R[1][1]+4]
bic @E[1],@C[9],@C[7]
bic @E[2],@C[0],@C[8],ror#14
bic @E[3],@C[1],@C[9],ror#14
eor @E[0],@E[0],@C[4]
eor @E[1],@E[1],@C[5]
- str @E[0],[sp,#$R[1][2]] @ R[1][2] = C[2] ^ (~C[3] & C[4]);
- bic @E[0],@C[2],@C[0],ror#32-(14-10)
+ str.l @E[0],[sp,#$R[1][2]] @ R[1][2] = C[2] ^ (~C[3] & C[4]);
+ bic @C[2],@C[2],@C[0],ror#32-(14-10)
+ str.h @E[1],[sp,#$R[1][2]+4]
eor @E[2],@C[6],@E[2],ror#32-14
- str @E[1],[sp,#$R[1][2]+4]
bic @E[1],@C[3],@C[1],ror#32-(14-10)
+ str.l @E[2],[sp,#$R[1][3]] @ R[1][3] = C[3] ^ (~C[4] & C[0]);
eor @E[3],@C[7],@E[3],ror#32-14
- str @E[2],[sp,#$R[1][3]] @ R[1][3] = C[3] ^ (~C[4] & C[0]);
+ str.h @E[3],[sp,#$R[1][3]+4]
add @E[2],sp,#$D[1]
- str @E[3],[sp,#$R[1][3]+4]
- ldr @C[1],[sp,#$A[0][1]] @ A[0][1]
- eor @E[0],@C[8],@E[0],ror#32-10
- ldr @C[0],[sp,#$A[0][1]+4]
+ ldr.l @C[1],[sp,#$A[0][1]] @ A[0][1]
+ eor @E[0],@C[8],@C[2],ror#32-10
+ ldr.h @C[0],[sp,#$A[0][1]+4]
eor @E[1],@C[9],@E[1],ror#32-10
- str @E[0],[sp,#$R[1][4]] @ R[1][4] = C[4] ^ (~C[0] & C[1]);
- add @C[9],sp,#$D[3]
- str @E[1],[sp,#$R[1][4]+4]
+ str.l @E[0],[sp,#$R[1][4]] @ R[1][4] = C[4] ^ (~C[0] & C[1]);
+ str.h @E[1],[sp,#$R[1][4]+4]
+ add @C[9],sp,#$D[3]
ldmia @E[2],{@E[0]-@E[2],@E[3]} @ D[1..2]
- ldr @C[2],[sp,#$A[1][2]] @ A[1][2]
- ldr @C[3],[sp,#$A[1][2]+4]
+ ldr.l @C[2],[sp,#$A[1][2]] @ A[1][2]
+ ldr.h @C[3],[sp,#$A[1][2]+4]
ldmia @C[9],{@C[6]-@C[9]} @ D[3..4]
eor @C[1],@C[1],@E[0]
- ldr @C[4],[sp,#$A[2][3]] @ A[2][3]
+ ldr.l @C[4],[sp,#$A[2][3]] @ A[2][3]
eor @C[0],@C[0],@E[1]
- ldr @C[5],[sp,#$A[2][3]+4]
+ ldr.h @C[5],[sp,#$A[2][3]+4]
ror @C[0],@C[0],#32-1 @ C[0] = ROL64(A[0][1] ^ D[1], rhotates[0][1]);
eor @C[2],@C[2],@E[2]
- ldr @E[0],[sp,#$A[3][4]] @ A[3][4]
+ ldr.l @E[0],[sp,#$A[3][4]] @ A[3][4]
eor @C[3],@C[3],@E[3]
- ldr @E[1],[sp,#$A[3][4]+4]
+ ldr.h @E[1],[sp,#$A[3][4]+4]
@ ror @C[2],@C[2],#32-3 @ C[1] = ROL64(A[1][2] ^ D[2], rhotates[1][2]);
- ldr @E[2],[sp,#$D[0]] @ D[0]
+ ldr.l @E[2],[sp,#$D[0]] @ D[0]
@ ror @C[3],@C[3],#32-3
- ldr @E[3],[sp,#$D[0]+4]
+ ldr.h @E[3],[sp,#$D[0]+4]
eor @C[4],@C[4],@C[6]
eor @C[5],@C[5],@C[7]
@ ror @C[4],@C[7],#32-13 @ [track reverse order below]
eor @E[0],@E[0],@C[8]
- ldr @C[8],[sp,#$A[4][0]] @ A[4][0]
+ ldr.l @C[8],[sp,#$A[4][0]] @ A[4][0]
eor @E[1],@E[1],@C[9]
- ldr @C[9],[sp,#$A[4][0]+4]
+ ldr.h @C[9],[sp,#$A[4][0]+4]
ror @C[6],@E[0],#32-4 @ C[3] = ROL64(A[3][4] ^ D[4], rhotates[3][4]);
ror @C[7],@E[1],#32-4
bic @E[3],@C[7],@C[4],ror#32-12
eor @E[0],@C[0],@E[0],ror#32-13
eor @E[1],@C[1],@E[1],ror#32-12
- str @E[0],[sp,#$R[2][0]] @ R[2][0] = C[0] ^ (~C[1] & C[2])
+ str.l @E[0],[sp,#$R[2][0]] @ R[2][0] = C[0] ^ (~C[1] & C[2])
eor @E[2],@E[2],@C[2],ror#32-3
- str @E[1],[sp,#$R[2][0]+4]
+ str.h @E[1],[sp,#$R[2][0]+4]
eor @E[3],@E[3],@C[3],ror#32-3
- str @E[2],[sp,#$R[2][1]] @ R[2][1] = C[1] ^ (~C[2] & C[3]);
+ str.l @E[2],[sp,#$R[2][1]] @ R[2][1] = C[1] ^ (~C[2] & C[3]);
bic @E[0],@C[8],@C[6]
- str @E[3],[sp,#$R[2][1]+4]
bic @E[1],@C[9],@C[7]
+ str.h @E[3],[sp,#$R[2][1]+4]
eor @E[0],@E[0],@C[5],ror#32-13
eor @E[1],@E[1],@C[4],ror#32-12
- str @E[0],[sp,#$R[2][2]] @ R[2][2] = C[2] ^ (~C[3] & C[4]);
+ str.l @E[0],[sp,#$R[2][2]] @ R[2][2] = C[2] ^ (~C[3] & C[4]);
bic @E[2],@C[0],@C[8]
- str @E[1],[sp,#$R[2][2]+4]
+ str.h @E[1],[sp,#$R[2][2]+4]
bic @E[3],@C[1],@C[9]
eor @E[2],@E[2],@C[6]
eor @E[3],@E[3],@C[7]
- str @E[2],[sp,#$R[2][3]] @ R[2][3] = C[3] ^ (~C[4] & C[0]);
+ str.l @E[2],[sp,#$R[2][3]] @ R[2][3] = C[3] ^ (~C[4] & C[0]);
bic @E[0],@C[2],@C[0],ror#3
- str @E[3],[sp,#$R[2][3]+4]
+ str.h @E[3],[sp,#$R[2][3]+4]
bic @E[1],@C[3],@C[1],ror#3
- ldr @C[1],[sp,#$A[0][4]] @ A[0][4] [in reverse order]
+ ldr.l @C[1],[sp,#$A[0][4]] @ A[0][4] [in reverse order]
eor @E[0],@C[8],@E[0],ror#32-3
- ldr @C[0],[sp,#$A[0][4]+4]
+ ldr.h @C[0],[sp,#$A[0][4]+4]
eor @E[1],@C[9],@E[1],ror#32-3
- str @E[0],[sp,#$R[2][4]] @ R[2][4] = C[4] ^ (~C[0] & C[1]);
+ str.l @E[0],[sp,#$R[2][4]] @ R[2][4] = C[4] ^ (~C[0] & C[1]);
add @C[9],sp,#$D[1]
- str @E[1],[sp,#$R[2][4]+4]
+ str.h @E[1],[sp,#$R[2][4]+4]
- ldr @E[0],[sp,#$D[4]] @ D[4]
- ldr @E[1],[sp,#$D[4]+4]
- ldr @E[2],[sp,#$D[0]] @ D[0]
- ldr @E[3],[sp,#$D[0]+4]
+ ldr.l @E[0],[sp,#$D[4]] @ D[4]
+ ldr.h @E[1],[sp,#$D[4]+4]
+ ldr.l @E[2],[sp,#$D[0]] @ D[0]
+ ldr.h @E[3],[sp,#$D[0]+4]
ldmia @C[9],{@C[6]-@C[9]} @ D[1..2]
eor @C[1],@C[1],@E[0]
- ldr @C[2],[sp,#$A[1][0]] @ A[1][0]
+ ldr.l @C[2],[sp,#$A[1][0]] @ A[1][0]
eor @C[0],@C[0],@E[1]
- ldr @C[3],[sp,#$A[1][0]+4]
+ ldr.h @C[3],[sp,#$A[1][0]+4]
@ ror @C[1],@E[0],#32-13 @ C[0] = ROL64(A[0][4] ^ D[4], rhotates[0][4]);
- ldr @C[4],[sp,#$A[2][1]] @ A[2][1]
+ ldr.l @C[4],[sp,#$A[2][1]] @ A[2][1]
@ ror @C[0],@E[1],#32-14 @ [was loaded in reverse order]
- ldr @C[5],[sp,#$A[2][1]+4]
+ ldr.h @C[5],[sp,#$A[2][1]+4]
eor @C[2],@C[2],@E[2]
- ldr @E[0],[sp,#$A[3][2]] @ A[3][2]
+ ldr.l @E[0],[sp,#$A[3][2]] @ A[3][2]
eor @C[3],@C[3],@E[3]
- ldr @E[1],[sp,#$A[3][2]+4]
+ ldr.h @E[1],[sp,#$A[3][2]+4]
@ ror @C[2],@C[2],#32-18 @ C[1] = ROL64(A[1][0] ^ D[0], rhotates[1][0]);
- ldr @E[2],[sp,#$D[3]] @ D[3]
+ ldr.l @E[2],[sp,#$D[3]] @ D[3]
@ ror @C[3],@C[3],#32-18
- ldr @E[3],[sp,#$D[3]+4]
+ ldr.h @E[3],[sp,#$D[3]+4]
eor @C[6],@C[6],@C[4]
eor @C[7],@C[7],@C[5]
ror @C[5],@C[7],#32-5
eor @E[0],@E[0],@C[8]
- ldr @C[8],[sp,#$A[4][3]] @ A[4][3]
+ ldr.l @C[8],[sp,#$A[4][3]] @ A[4][3]
eor @E[1],@E[1],@C[9]
- ldr @C[9],[sp,#$A[4][3]+4]
+ ldr.h @C[9],[sp,#$A[4][3]+4]
ror @C[7],@E[0],#32-7 @ C[3] = ROL64(A[3][2] ^ D[2], rhotates[3][2]);
ror @C[6],@E[1],#32-8
bic @E[1],@C[5],@C[3],ror#32-18
eor @E[0],@E[0],@C[0],ror#32-14
eor @E[1],@E[1],@C[1],ror#32-13
- str @E[0],[sp,#$R[3][0]] @ R[3][0] = C[0] ^ (~C[1] & C[2])
+ str.l @E[0],[sp,#$R[3][0]] @ R[3][0] = C[0] ^ (~C[1] & C[2])
bic @E[2],@C[6],@C[4]
- str @E[1],[sp,#$R[3][0]+4]
+ str.h @E[1],[sp,#$R[3][0]+4]
bic @E[3],@C[7],@C[5]
eor @E[2],@E[2],@C[2],ror#32-18
+ str.l @E[2],[sp,#$R[3][1]] @ R[3][1] = C[1] ^ (~C[2] & C[3]);
eor @E[3],@E[3],@C[3],ror#32-18
- str @E[2],[sp,#$R[3][1]] @ R[3][1] = C[1] ^ (~C[2] & C[3]);
+ str.h @E[3],[sp,#$R[3][1]+4]
bic @E[0],@C[8],@C[6]
- str @E[3],[sp,#$R[3][1]+4]
bic @E[1],@C[9],@C[7]
bic @E[2],@C[0],@C[8],ror#14
bic @E[3],@C[1],@C[9],ror#13
eor @E[0],@E[0],@C[4]
eor @E[1],@E[1],@C[5]
- str @E[0],[sp,#$R[3][2]] @ R[3][2] = C[2] ^ (~C[3] & C[4]);
- bic @E[0],@C[2],@C[0],ror#18-14
+ str.l @E[0],[sp,#$R[3][2]] @ R[3][2] = C[2] ^ (~C[3] & C[4]);
+ bic @C[2],@C[2],@C[0],ror#18-14
+ str.h @E[1],[sp,#$R[3][2]+4]
eor @E[2],@C[6],@E[2],ror#32-14
- str @E[1],[sp,#$R[3][2]+4]
bic @E[1],@C[3],@C[1],ror#18-13
eor @E[3],@C[7],@E[3],ror#32-13
- str @E[2],[sp,#$R[3][3]] @ R[3][3] = C[3] ^ (~C[4] & C[0]);
- str @E[3],[sp,#$R[3][3]+4]
+ str.l @E[2],[sp,#$R[3][3]] @ R[3][3] = C[3] ^ (~C[4] & C[0]);
+ str.h @E[3],[sp,#$R[3][3]+4]
add @E[3],sp,#$D[2]
- ldr @C[0],[sp,#$A[0][2]] @ A[0][2]
- eor @E[0],@C[8],@E[0],ror#32-18
- ldr @C[1],[sp,#$A[0][2]+4]
+ ldr.l @C[0],[sp,#$A[0][2]] @ A[0][2]
+ eor @E[0],@C[8],@C[2],ror#32-18
+ ldr.h @C[1],[sp,#$A[0][2]+4]
eor @E[1],@C[9],@E[1],ror#32-18
- str @E[0],[sp,#$R[3][4]] @ R[3][4] = C[4] ^ (~C[0] & C[1]);
- str @E[1],[sp,#$R[3][4]+4]
+ str.l @E[0],[sp,#$R[3][4]] @ R[3][4] = C[4] ^ (~C[0] & C[1]);
+ str.h @E[1],[sp,#$R[3][4]+4]
ldmia @E[3],{@E[0]-@E[2],@E[3]} @ D[2..3]
- ldr @C[2],[sp,#$A[1][3]] @ A[1][3]
- ldr @C[3],[sp,#$A[1][3]+4]
- ldr @C[6],[sp,#$D[4]] @ D[4]
- ldr @C[7],[sp,#$D[4]+4]
+ ldr.l @C[2],[sp,#$A[1][3]] @ A[1][3]
+ ldr.h @C[3],[sp,#$A[1][3]+4]
+ ldr.l @C[6],[sp,#$D[4]] @ D[4]
+ ldr.h @C[7],[sp,#$D[4]+4]
eor @C[0],@C[0],@E[0]
- ldr @C[4],[sp,#$A[2][4]] @ A[2][4]
+ ldr.l @C[4],[sp,#$A[2][4]] @ A[2][4]
eor @C[1],@C[1],@E[1]
- ldr @C[5],[sp,#$A[2][4]+4]
+ ldr.h @C[5],[sp,#$A[2][4]+4]
@ ror @C[0],@C[0],#32-31 @ C[0] = ROL64(A[0][2] ^ D[2], rhotates[0][2]);
- ldr @C[8],[sp,#$D[0]] @ D[0]
+ ldr.l @C[8],[sp,#$D[0]] @ D[0]
@ ror @C[1],@C[1],#32-31
- ldr @C[9],[sp,#$D[0]+4]
+ ldr.h @C[9],[sp,#$D[0]+4]
eor @E[2],@E[2],@C[2]
- ldr @E[0],[sp,#$A[3][0]] @ A[3][0]
+ ldr.l @E[0],[sp,#$A[3][0]] @ A[3][0]
eor @E[3],@E[3],@C[3]
- ldr @E[1],[sp,#$A[3][0]+4]
+ ldr.h @E[1],[sp,#$A[3][0]+4]
ror @C[3],@E[2],#32-27 @ C[1] = ROL64(A[1][3] ^ D[3], rhotates[1][3]);
- ldr @E[2],[sp,#$D[1]] @ D[1]
+ ldr.l @E[2],[sp,#$D[1]] @ D[1]
ror @C[2],@E[3],#32-28
- ldr @E[3],[sp,#$D[1]+4]
+ ldr.h @E[3],[sp,#$D[1]+4]
eor @C[6],@C[6],@C[4]
eor @C[7],@C[7],@C[5]
ror @C[4],@C[7],#32-20
eor @E[0],@E[0],@C[8]
- ldr @C[8],[sp,#$A[4][1]] @ A[4][1]
+ ldr.l @C[8],[sp,#$A[4][1]] @ A[4][1]
eor @E[1],@E[1],@C[9]
- ldr @C[9],[sp,#$A[4][1]+4]
+ ldr.h @C[9],[sp,#$A[4][1]+4]
ror @C[7],@E[0],#32-20 @ C[3] = ROL64(A[3][0] ^ D[0], rhotates[3][0]);
ror @C[6],@E[1],#32-21
bic @E[0],@C[4],@C[2]
bic @E[1],@C[5],@C[3]
eor @E[0],@E[0],@C[0],ror#32-31
+ str.l @E[0],[sp,#$R[4][0]] @ R[4][0] = C[0] ^ (~C[1] & C[2])
eor @E[1],@E[1],@C[1],ror#32-31
- str @E[0],[sp,#$R[4][0]] @ R[4][0] = C[0] ^ (~C[1] & C[2])
+ str.h @E[1],[sp,#$R[4][0]+4]
bic @E[2],@C[6],@C[4]
- str @E[1],[sp,#$R[4][0]+4]
bic @E[3],@C[7],@C[5]
eor @E[2],@E[2],@C[2]
eor @E[3],@E[3],@C[3]
- str @E[2],[sp,#$R[4][1]] @ R[4][1] = C[1] ^ (~C[2] & C[3]);
+ str.l @E[2],[sp,#$R[4][1]] @ R[4][1] = C[1] ^ (~C[2] & C[3]);
bic @E[0],@C[8],@C[6],ror#1
- str @E[3],[sp,#$R[4][1]+4]
+ str.h @E[3],[sp,#$R[4][1]+4]
bic @E[1],@C[9],@C[7],ror#1
bic @E[2],@C[0],@C[8],ror#31-1
bic @E[3],@C[1],@C[9],ror#31-1
eor @C[4],@C[4],@E[0],ror#32-1
+ str.l @C[4],[sp,#$R[4][2]] @ R[4][2] = C[2] ^= (~C[3] & C[4]);
eor @C[5],@C[5],@E[1],ror#32-1
- str @C[4],[sp,#$R[4][2]] @ R[4][2] = C[2] ^= (~C[3] & C[4]);
+ str.h @C[5],[sp,#$R[4][2]+4]
eor @C[6],@C[6],@E[2],ror#32-31
- str @C[5],[sp,#$R[4][2]+4]
eor @C[7],@C[7],@E[3],ror#32-31
- str @C[6],[sp,#$R[4][3]] @ R[4][3] = C[3] ^= (~C[4] & C[0]);
+ str.l @C[6],[sp,#$R[4][3]] @ R[4][3] = C[3] ^= (~C[4] & C[0]);
bic @E[0],@C[2],@C[0],ror#32-31
- str @C[7],[sp,#$R[4][3]+4]
+ str.h @C[7],[sp,#$R[4][3]+4]
bic @E[1],@C[3],@C[1],ror#32-31
add @E[2],sp,#$R[0][0]
eor @C[8],@E[0],@C[8],ror#32-1
add @E[0],sp,#$R[1][0]
eor @C[9],@E[1],@C[9],ror#32-1
- str @C[8],[sp,#$R[4][4]] @ R[4][4] = C[4] ^= (~C[0] & C[1]);
- str @C[9],[sp,#$R[4][4]+4]
+ str.l @C[8],[sp,#$R[4][4]] @ R[4][4] = C[4] ^= (~C[0] & C[1]);
+ str.h @C[9],[sp,#$R[4][4]+4]
___
}
Round(@A,@T);
Round(@T,@A);
$code.=<<___;
- blo .Lround
+ blo .Lround2x
+#if __ARM_ARCH__>=5
ldr pc,[sp,#440]
+#else
+ ldr lr,[sp,#440]
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
.size KeccakF1600_int,.-KeccakF1600_int
.type KeccakF1600, %function
stmia @E[1], {@C[0]-@C[9]}
add sp,sp,#440+20
+#if __ARM_ARCH__>=5
ldmia sp!,{r4-r11,pc}
+#else
+ ldmia sp!,{r4-r11,lr}
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
.size KeccakF1600,.-KeccakF1600
___
{ my ($A_flat,$inp,$len,$bsz) = map("r$_",(10..12,14));
.Labsorb_abort:
add sp,sp,#456+32
mov r0,$len @ return value
+#if __ARM_ARCH__>=5
ldmia sp!,{r4-r12,pc}
+#else
+ ldmia sp!,{r4-r12,lr}
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
.size SHA3_absorb,.-SHA3_absorb
___
}
.align 4
.Lsqueeze_done:
add sp,sp,#24
+#if __ARM_ARCH__>=5
ldmia sp!,{r4-r10,pc}
+#else
+ ldmia sp!,{r4-r10,lr}
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
.size SHA3_squeeze,.-SHA3_squeeze
___
}
$code.=<<___;
+#if __ARM_MAX_ARCH__>=7
.fpu neon
.type iotas64, %object
.align 4
.Loop_neon:
@ Theta
- vst1.64 {q4}, [r0:64] @ offload A[0..1][4]
+ vst1.64 {q4}, [r0,:64] @ offload A[0..1][4]
veor q13, q0, q5 @ A[0..1][0]^A[2..3][0]
- vst1.64 {d18}, [r1:64] @ offload A[2][4]
+ vst1.64 {d18}, [r1,:64] @ offload A[2][4]
veor q14, q1, q6 @ A[0..1][1]^A[2..3][1]
veor q15, q2, q7 @ A[0..1][2]^A[2..3][2]
veor d26, d26, d27 @ C[0]=A[0][0]^A[1][0]^A[2][0]^A[3][0]
veor d16, d16, d28 @ A[2][3] ^= C[2]
veor d17, d17, d28 @ A[3][3] ^= C[2]
veor d23, d23, d28 @ A[4][3] ^= C[2]
- vld1.64 {q4}, [r0:64] @ restore A[0..1][4]
+ vld1.64 {q4}, [r0,:64] @ restore A[0..1][4]
vmov d28, d29
- vld1.64 {d18}, [r1:64] @ restore A[2][4]
+ vld1.64 {d18}, [r1,:64] @ restore A[2][4]
veor q2, q2, q13 @ A[0..1][2] ^= D[2]
veor q7, q7, q13 @ A[2..3][2] ^= D[2]
veor d22, d22, d27 @ A[4][2] ^= D[2]
veor q13, q13, q0 @ A[0..1][0] ^ (~A[0..1][1] & A[0..1][2])
veor q14, q14, q1 @ A[0..1][1] ^ (~A[0..1][2] & A[0..1][3])
veor q2, q2, q15 @ A[0..1][2] ^= (~A[0..1][3] & A[0..1][4])
- vst1.64 {q13}, [r0:64] @ offload A[0..1][0]
+ vst1.64 {q13}, [r0,:64] @ offload A[0..1][0]
vbic q13, q0, q4
vbic q15, q1, q0
vmov q1, q14 @ A[0..1][1]
vmov q14, q10 @ A[4][0..1]
veor q9, q9, q13 @ A[2..3][4] ^= (~A[2..3][0] & A[2..3][1])
- vld1.64 d25, [r2:64]! @ Iota[i++]
+ vld1.64 d25, [r2,:64]! @ Iota[i++]
vbic d26, d22, d21
vbic d27, d23, d22
- vld1.64 {q0}, [r0:64] @ restore A[0..1][0]
+ vld1.64 {q0}, [r0,:64] @ restore A[0..1][0]
veor d20, d20, d26 @ A[4][0] ^= (~A[4][1] & A[4][2])
vbic d26, d24, d23
veor d21, d21, d27 @ A[4][1] ^= (~A[4][2] & A[4][3])
subs r3, r3, #1
bne .Loop_neon
- bx lr
+ ret
.size KeccakF1600_neon,.-KeccakF1600_neon
.global SHA3_absorb_neon
mov r5, r2 @ len
mov r6, r3 @ bsz
- vld1.32 {d0}, [r0:64]! @ A[0][0]
- vld1.32 {d2}, [r0:64]! @ A[0][1]
- vld1.32 {d4}, [r0:64]! @ A[0][2]
- vld1.32 {d6}, [r0:64]! @ A[0][3]
- vld1.32 {d8}, [r0:64]! @ A[0][4]
-
- vld1.32 {d1}, [r0:64]! @ A[1][0]
- vld1.32 {d3}, [r0:64]! @ A[1][1]
- vld1.32 {d5}, [r0:64]! @ A[1][2]
- vld1.32 {d7}, [r0:64]! @ A[1][3]
- vld1.32 {d9}, [r0:64]! @ A[1][4]
-
- vld1.32 {d10}, [r0:64]! @ A[2][0]
- vld1.32 {d12}, [r0:64]! @ A[2][1]
- vld1.32 {d14}, [r0:64]! @ A[2][2]
- vld1.32 {d16}, [r0:64]! @ A[2][3]
- vld1.32 {d18}, [r0:64]! @ A[2][4]
-
- vld1.32 {d11}, [r0:64]! @ A[3][0]
- vld1.32 {d13}, [r0:64]! @ A[3][1]
- vld1.32 {d15}, [r0:64]! @ A[3][2]
- vld1.32 {d17}, [r0:64]! @ A[3][3]
- vld1.32 {d19}, [r0:64]! @ A[3][4]
-
- vld1.32 {d20-d23}, [r0:64]! @ A[4][0..3]
- vld1.32 {d24}, [r0:64] @ A[4][4]
+ vld1.32 {d0}, [r0,:64]! @ A[0][0]
+ vld1.32 {d2}, [r0,:64]! @ A[0][1]
+ vld1.32 {d4}, [r0,:64]! @ A[0][2]
+ vld1.32 {d6}, [r0,:64]! @ A[0][3]
+ vld1.32 {d8}, [r0,:64]! @ A[0][4]
+
+ vld1.32 {d1}, [r0,:64]! @ A[1][0]
+ vld1.32 {d3}, [r0,:64]! @ A[1][1]
+ vld1.32 {d5}, [r0,:64]! @ A[1][2]
+ vld1.32 {d7}, [r0,:64]! @ A[1][3]
+ vld1.32 {d9}, [r0,:64]! @ A[1][4]
+
+ vld1.32 {d10}, [r0,:64]! @ A[2][0]
+ vld1.32 {d12}, [r0,:64]! @ A[2][1]
+ vld1.32 {d14}, [r0,:64]! @ A[2][2]
+ vld1.32 {d16}, [r0,:64]! @ A[2][3]
+ vld1.32 {d18}, [r0,:64]! @ A[2][4]
+
+ vld1.32 {d11}, [r0,:64]! @ A[3][0]
+ vld1.32 {d13}, [r0,:64]! @ A[3][1]
+ vld1.32 {d15}, [r0,:64]! @ A[3][2]
+ vld1.32 {d17}, [r0,:64]! @ A[3][3]
+ vld1.32 {d19}, [r0,:64]! @ A[3][4]
+
+ vld1.32 {d20-d23}, [r0,:64]! @ A[4][0..3]
+ vld1.32 {d24}, [r0,:64] @ A[4][4]
sub r0, r0, #24*8 @ rewind
b .Loop_absorb_neon
.align 4
.Labsorbed_neon:
- vst1.32 {d0}, [r0:64]! @ A[0][0..4]
- vst1.32 {d2}, [r0:64]!
- vst1.32 {d4}, [r0:64]!
- vst1.32 {d6}, [r0:64]!
- vst1.32 {d8}, [r0:64]!
-
- vst1.32 {d1}, [r0:64]! @ A[1][0..4]
- vst1.32 {d3}, [r0:64]!
- vst1.32 {d5}, [r0:64]!
- vst1.32 {d7}, [r0:64]!
- vst1.32 {d9}, [r0:64]!
-
- vst1.32 {d10}, [r0:64]! @ A[2][0..4]
- vst1.32 {d12}, [r0:64]!
- vst1.32 {d14}, [r0:64]!
- vst1.32 {d16}, [r0:64]!
- vst1.32 {d18}, [r0:64]!
-
- vst1.32 {d11}, [r0:64]! @ A[3][0..4]
- vst1.32 {d13}, [r0:64]!
- vst1.32 {d15}, [r0:64]!
- vst1.32 {d17}, [r0:64]!
- vst1.32 {d19}, [r0:64]!
-
- vst1.32 {d20-d23}, [r0:64]! @ A[4][0..4]
- vst1.32 {d24}, [r0:64]
+ vst1.32 {d0}, [r0,:64]! @ A[0][0..4]
+ vst1.32 {d2}, [r0,:64]!
+ vst1.32 {d4}, [r0,:64]!
+ vst1.32 {d6}, [r0,:64]!
+ vst1.32 {d8}, [r0,:64]!
+
+ vst1.32 {d1}, [r0,:64]! @ A[1][0..4]
+ vst1.32 {d3}, [r0,:64]!
+ vst1.32 {d5}, [r0,:64]!
+ vst1.32 {d7}, [r0,:64]!
+ vst1.32 {d9}, [r0,:64]!
+
+ vst1.32 {d10}, [r0,:64]! @ A[2][0..4]
+ vst1.32 {d12}, [r0,:64]!
+ vst1.32 {d14}, [r0,:64]!
+ vst1.32 {d16}, [r0,:64]!
+ vst1.32 {d18}, [r0,:64]!
+
+ vst1.32 {d11}, [r0,:64]! @ A[3][0..4]
+ vst1.32 {d13}, [r0,:64]!
+ vst1.32 {d15}, [r0,:64]!
+ vst1.32 {d17}, [r0,:64]!
+ vst1.32 {d19}, [r0,:64]!
+
+ vst1.32 {d20-d23}, [r0,:64]! @ A[4][0..4]
+ vst1.32 {d24}, [r0,:64]
mov r0, r5 @ return value
vldmia sp!, {d8-d15}
vstmdb sp!, {d8-d15}
- vld1.32 {d0}, [r0:64]! @ A[0][0..4]
- vld1.32 {d2}, [r0:64]!
- vld1.32 {d4}, [r0:64]!
- vld1.32 {d6}, [r0:64]!
- vld1.32 {d8}, [r0:64]!
-
- vld1.32 {d1}, [r0:64]! @ A[1][0..4]
- vld1.32 {d3}, [r0:64]!
- vld1.32 {d5}, [r0:64]!
- vld1.32 {d7}, [r0:64]!
- vld1.32 {d9}, [r0:64]!
-
- vld1.32 {d10}, [r0:64]! @ A[2][0..4]
- vld1.32 {d12}, [r0:64]!
- vld1.32 {d14}, [r0:64]!
- vld1.32 {d16}, [r0:64]!
- vld1.32 {d18}, [r0:64]!
-
- vld1.32 {d11}, [r0:64]! @ A[3][0..4]
- vld1.32 {d13}, [r0:64]!
- vld1.32 {d15}, [r0:64]!
- vld1.32 {d17}, [r0:64]!
- vld1.32 {d19}, [r0:64]!
-
- vld1.32 {d20-d23}, [r0:64]! @ A[4][0..4]
- vld1.32 {d24}, [r0:64]
+ vld1.32 {d0}, [r0,:64]! @ A[0][0..4]
+ vld1.32 {d2}, [r0,:64]!
+ vld1.32 {d4}, [r0,:64]!
+ vld1.32 {d6}, [r0,:64]!
+ vld1.32 {d8}, [r0,:64]!
+
+ vld1.32 {d1}, [r0,:64]! @ A[1][0..4]
+ vld1.32 {d3}, [r0,:64]!
+ vld1.32 {d5}, [r0,:64]!
+ vld1.32 {d7}, [r0,:64]!
+ vld1.32 {d9}, [r0,:64]!
+
+ vld1.32 {d10}, [r0,:64]! @ A[2][0..4]
+ vld1.32 {d12}, [r0,:64]!
+ vld1.32 {d14}, [r0,:64]!
+ vld1.32 {d16}, [r0,:64]!
+ vld1.32 {d18}, [r0,:64]!
+
+ vld1.32 {d11}, [r0,:64]! @ A[3][0..4]
+ vld1.32 {d13}, [r0,:64]!
+ vld1.32 {d15}, [r0,:64]!
+ vld1.32 {d17}, [r0,:64]!
+ vld1.32 {d19}, [r0,:64]!
+
+ vld1.32 {d20-d23}, [r0,:64]! @ A[4][0..4]
+ vld1.32 {d24}, [r0,:64]
sub r0, r0, #24*8 @ rewind
bl KeccakF1600_neon
mov r12, r0 @ A_flat
- vst1.32 {d0}, [r0:64]! @ A[0][0..4]
- vst1.32 {d2}, [r0:64]!
- vst1.32 {d4}, [r0:64]!
- vst1.32 {d6}, [r0:64]!
- vst1.32 {d8}, [r0:64]!
-
- vst1.32 {d1}, [r0:64]! @ A[1][0..4]
- vst1.32 {d3}, [r0:64]!
- vst1.32 {d5}, [r0:64]!
- vst1.32 {d7}, [r0:64]!
- vst1.32 {d9}, [r0:64]!
-
- vst1.32 {d10}, [r0:64]! @ A[2][0..4]
- vst1.32 {d12}, [r0:64]!
- vst1.32 {d14}, [r0:64]!
- vst1.32 {d16}, [r0:64]!
- vst1.32 {d18}, [r0:64]!
-
- vst1.32 {d11}, [r0:64]! @ A[3][0..4]
- vst1.32 {d13}, [r0:64]!
- vst1.32 {d15}, [r0:64]!
- vst1.32 {d17}, [r0:64]!
- vst1.32 {d19}, [r0:64]!
-
- vst1.32 {d20-d23}, [r0:64]! @ A[4][0..4]
+ vst1.32 {d0}, [r0,:64]! @ A[0][0..4]
+ vst1.32 {d2}, [r0,:64]!
+ vst1.32 {d4}, [r0,:64]!
+ vst1.32 {d6}, [r0,:64]!
+ vst1.32 {d8}, [r0,:64]!
+
+ vst1.32 {d1}, [r0,:64]! @ A[1][0..4]
+ vst1.32 {d3}, [r0,:64]!
+ vst1.32 {d5}, [r0,:64]!
+ vst1.32 {d7}, [r0,:64]!
+ vst1.32 {d9}, [r0,:64]!
+
+ vst1.32 {d10}, [r0,:64]! @ A[2][0..4]
+ vst1.32 {d12}, [r0,:64]!
+ vst1.32 {d14}, [r0,:64]!
+ vst1.32 {d16}, [r0,:64]!
+ vst1.32 {d18}, [r0,:64]!
+
+ vst1.32 {d11}, [r0,:64]! @ A[3][0..4]
+ vst1.32 {d13}, [r0,:64]!
+ vst1.32 {d15}, [r0,:64]!
+ vst1.32 {d17}, [r0,:64]!
+ vst1.32 {d19}, [r0,:64]!
+
+ vst1.32 {d20-d23}, [r0,:64]! @ A[4][0..4]
mov r14, r6 @ bsz
- vst1.32 {d24}, [r0:64]
+ vst1.32 {d24}, [r0,:64]
mov r0, r12 @ rewind
vldmia sp!, {d8-d15}
.Lsqueeze_neon_done:
ldmia sp!, {r4-r6,pc}
.size SHA3_squeeze_neon,.-SHA3_squeeze_neon
+#endif
.asciz "Keccak-1600 absorb and squeeze for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
.align 2
___
-print $code;
+{
+ my %ldr, %str;
+
+ sub ldrd {
+ my ($mnemonic,$half,$reg,$ea) = @_;
+ my $op = $mnemonic eq "ldr" ? \%ldr : \%str;
+
+ if ($half eq "l") {
+ $$op{reg} = $reg;
+ $$op{ea} = $ea;
+ sprintf "#ifndef __thumb2__\n" .
+ " %s\t%s,%s\n" .
+ "#endif", $mnemonic,$reg,$ea;
+ } else {
+ sprintf "#ifndef __thumb2__\n" .
+ " %s\t%s,%s\n" .
+ "#else\n" .
+ " %sd\t%s,%s,%s\n" .
+ "#endif", $mnemonic,$reg,$ea,
+ $mnemonic,$$op{reg},$reg,$$op{ea};
+ }
+ }
+}
+
+foreach (split($/,$code)) {
+ s/\`([^\`]*)\`/eval $1/ge;
+
+ s/^\s+(ldr|str)\.([lh])\s+(r[0-9]+),\s*(\[.*)/ldrd($1,$2,$3,$4)/ge or
+ s/\b(ror|ls[rl])\s+(r[0-9]+.*)#/mov $2$1#/g or
+ s/\bret\b/bx lr/g or
+ s/\bbx\s+lr\b/.word\t0xe12fff1e/g; # make it possible to compile with -march=armv4
+
+ print $_,"\n";
+}
close STDOUT; # enforce flush
projects / openssl.git / blobdiff