X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=blobdiff_plain;f=crypto%2Fsha%2Fasm%2Fkeccak1600-armv4.pl;h=a560aa7c31f50f515bbeba49d4ea0c4d356d5c87;hp=616beae9d806c0d2e623a488ee39c2fbe049e7cc;hb=56676f877d1ec4bc49ceb45636717e0f9a7ae445;hpb=5371810714337149ca430cf866931c4c2372f1b0;ds=sidebyside

diff --git a/crypto/sha/asm/keccak1600-armv4.pl b/crypto/sha/asm/keccak1600-armv4.pl
index 616beae9d8..a560aa7c31 100644
--- a/crypto/sha/asm/keccak1600-armv4.pl
+++ b/crypto/sha/asm/keccak1600-armv4.pl
@@ -1,6 +1,63 @@
+#!/usr/bin/env perl
+# Copyright 2017 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License").  You may not use
+# this file except in compliance with the License.  You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# Keccak-1600 for ARMv4.
+#
+# June 2017.
+#
+# This is KECCAK_1X variant (see keccak1600.c) with bit interleaving.
+# How does it compare to Keccak Code Package? It's as fast, but several
+# times smaller, and is endian- and ISA-neutral. ISA neutrality means
+# that minimum ISA requirement is ARMv4, yet it can be assembled even
+# as ARMv7 Thumb-2.
+#
+########################################################################
+# Numbers are cycles per processed byte accounting even for input bit
+# interleaving.
+#
+#		r=1600(*)	r=1024
+#
+# Cortex-A7	71/+180%	103
+# Cortex-A8	48/+290%	69
+# Cortex-A15	34/+210%	49
+#
+# (*)	Not used in real life, meaningful as estimate for single sponge
+#	operation performance. Numbers after slash are improvement over
+#	compiler-generated KECCAK_1X reference code.
+
 my @C = map("r$_",(0..9));
 my @E = map("r$_",(10..12,14));
 
+########################################################################
+# Stack layout
+# ----->+-----------------------+
+#       | uint64_t A[5][5]      |
+#       | ...                   |
+# +200->+-----------------------+
+#       | uint64_t D[5]         |
+#       | ...                   |
+# +240->+-----------------------+
+#       | uint64_t T[2][5]      |
+#       | ...                   |
+# +320->+-----------------------+
+#       | saved lr              |
+# +324->+-----------------------+
+#       | loop counter          |
+# +328->+-----------------------+
+#       | ...
+
 my @A = map([ 8*$_, 8*($_+1), 8*($_+2), 8*($_+3), 8*($_+4) ], (0,5,10,15,20));
 my @D = map(8*$_, (25..29));
 my @T = map([ 8*$_, 8*($_+1), 8*($_+2), 8*($_+3), 8*($_+4) ], (30,35));
@@ -8,6 +65,13 @@ my @T = map([ 8*$_, 8*($_+1), 8*($_+2), 8*($_+3), 8*($_+4) ], (30,35));
 $code.=<<___;
 .text
 
+#if defined(__thumb2__)
+.syntax	unified
+.thumb
+#else
+.code	32
+#endif
+
 .type	iotas,%object
 .align	5
 iotas:
@@ -35,34 +99,23 @@ iotas:
 	.long	0x00000000, 0x80000088
 	.long	0x00000001, 0x00008000
 	.long	0x00000000, 0x80008082
-.size	iostas,.-iotas
+.size	iotas,.-iotas
 
-.global	KeccakF1600
-.type	KeccakF1600, %function
+.type	KeccakF1600_int, %function
 .align	5
-KeccakF1600:
-	eor	r1,r1,r1
-	stmdb	sp!,{r0,r1,r4-r12,lr}
-	sub	sp,sp,#320			@ space for A[5][5],D[5],T[2][5]
-
-	add	@E[0],r0,#$A[1][0]
-	add	@E[1],sp,#$A[1][0]
-	mov	@E[2],r0
-	ldmia	@E[0]!,{@C[0]-@C[9]}		@ copy A[5][5] to stack
-	stmia	@E[1]!,{@C[0]-@C[9]}
-	ldmia	@E[0]!,{@C[0]-@C[9]}
-	stmia	@E[1]!,{@C[0]-@C[9]}
-	ldmia	@E[0]!,{@C[0]-@C[9]}
-	stmia	@E[1]!,{@C[0]-@C[9]}
-	ldmia	@E[0],{@C[0]-@C[9]}
-	stmia	@E[1],{@C[0]-@C[9]}
-	ldmia	@E[2],{@C[0]-@C[9]}
-	stmia	sp,{@C[0]-@C[9]}
+KeccakF1600_int:
+	ldmia	sp,{@C[0]-@C[9]}		@ A[0][0..4]
 	add	@E[0],sp,#$A[1][0]
-	b	.Lround
+KeccakF1600_enter:
+	str	lr,[sp,#320]
+	eor	@E[1],@E[1],@E[1]
+	str	@E[1],[sp,#324]
+	b	.Lround_enter
 
 .align	4
 .Lround:
+	ldmia	sp,{@C[0]-@C[9]}		@ A[0][0..4]
+.Lround_enter:
 	ldmia	@E[0],{@E[0]-@E[2],@E[3]}	@ A[1][0..1]
 	eor	@C[0],@C[0],@E[0]
 	 add	@E[0],sp,#$A[1][2]
@@ -141,14 +194,14 @@ KeccakF1600:
 	eor	@C[4],@C[4],@C[9],ror#32-1	@ C[2] = ROL64(C[4], 1) ^ C[2];
 	str	@C[3],[sp,#$D[2]+4]
 	eor	@C[5],@C[5],@C[8]
+	 ldr	@C[8],[sp,#$A[3][0]]
+	 ldr	@C[9],[sp,#$A[3][0]+4]
 	str	@C[4],[sp,#$D[3]]		@ D[3] = C[2]
 	str	@C[5],[sp,#$D[3]+4]
 
-	ldr	@C[8],[sp,#$A[3][0]]
-	ldr	@C[9],[sp,#$A[3][0]+4]
 	ldr	@C[6],[sp,#$A[0][1]]
-	ldr	@C[7],[sp,#$A[0][1]+4]
 	eor	@C[8],@C[8],@C[0]
+	ldr	@C[7],[sp,#$A[0][1]+4]
 	eor	@C[9],@C[9],@C[1]
 	str	@C[8],[sp,#$T[0][0]]		@ T[0][0] = A[3][0] ^ C[0]; /* borrow T[0][0] */
 	ldr	@C[8],[sp,#$A[0][2]]
@@ -172,14 +225,14 @@ KeccakF1600:
 	eor	@C[8],@C[8],@E[2]
 	str	@C[7],[sp,#$T[0][3]+4]
 	eor	@C[9],@C[9],@E[3]
+	 ldr	@C[6],[sp,#$A[3][3]]
+	 ldr	@C[7],[sp,#$A[3][3]+4]
 	str	@C[8],[sp,#$T[0][4]]		@ T[0][4] = A[0][4] ^ E[1]; /* D[4] */
 	str	@C[9],[sp,#$T[0][4]+4]
 
-	ldr	@C[6],[sp,#$A[3][3]]
-	ldr	@C[7],[sp,#$A[3][3]+4]
 	ldr	@C[8],[sp,#$A[4][4]]
-	ldr	@C[9],[sp,#$A[4][4]+4]
 	eor	@C[4],@C[4],@C[6]
+	ldr	@C[9],[sp,#$A[4][4]+4]
 	eor	@C[5],@C[5],@C[7]
 	ror	@C[7],@C[4],#32-10		@ C[3] = ROL64(A[3][3] ^ C[2], rhotates[3][3]);   /* D[3] */
 	ldr	@C[4],[sp,#$A[0][0]]
@@ -198,25 +251,25 @@ KeccakF1600:
 	eor	@E[3],@E[3],@C[3]
 	ldr	@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,#324]			@ load counter
 	eor	@C[2],@C[2],@E[0]
 	ror	@C[4],@E[3],#32-22
-	 adr	@E[0],iotas
+	 adr	@E[3],iotas
 	eor	@C[3],@C[3],@E[1]
-	 ldr	@E[1],[sp,#320+4]		@ load counter
 	ror	@C[2],@C[2],#32-22		@ C[1] = ROL64(A[1][1] ^ E[0], rhotates[1][1]);   /* D[1] */
+	 add	@E[3],@E[3],@E[2]
 	ror	@C[3],@C[3],#32-22
 
-	add	@E[0],@E[0],@E[1]
-	ldr	@E[2],[@E[0],#0]
-	add	@E[1],@E[1],#8
-	ldr	@E[3],[@E[0],#4]
-	cmp	@E[1],#192
-	str	@E[1],[sp,#320+4]		@ store counter
+	ldr	@E[0],[@E[3],#0]		@ iotas[i].lo
+	add	@E[2],@E[2],#8
+	ldr	@E[1],[@E[3],#4]		@ iotas[i].hi
+	cmp	@E[2],#192
+	str	@E[2],[sp,#324]			@ store counter
 
-	bic	@E[0],@C[4],@C[2]
-	bic	@E[1],@C[5],@C[3]
-	eor	@E[0],@E[0],@C[0]
-	eor	@E[1],@E[1],@C[1]
+	bic	@E[2],@C[4],@C[2]
+	bic	@E[3],@C[5],@C[3]
+	eor	@E[2],@E[2],@C[0]
+	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,#$A[0][0]]		@ A[0][0] = C[0] ^ (~C[1] & C[2]) ^ iotas[i];
@@ -248,13 +301,13 @@ KeccakF1600:
 	str	@E[1],[sp,#$A[0][4]+4]
 
 	ldmia	@E[3],{@C[6]-@C[9],@E[0],@E[1],@E[2],@E[3]}	@ D[0..3]
-	ldr	@C[4],[sp,#$D[4]]
-	ldr	@C[5],[sp,#$D[4]+4]
 	ldr	@C[0],[sp,#$A[1][0]]
 	ldr	@C[1],[sp,#$A[1][0]+4]
 	ldr	@C[2],[sp,#$A[2][1]]
 	ldr	@C[3],[sp,#$A[2][1]+4]
+	ldr	@C[4],[sp,#$D[4]]
 	eor	@C[0],@C[0],@C[6]
+	ldr	@C[5],[sp,#$D[4]+4]
 	eor	@C[1],@C[1],@C[7]
 	str	@C[0],[sp,#$T[1][0]]		@ T[1][0] = A[1][0] ^ (C[3] = D[0]);
 	add	@C[0],sp,#$A[1][2]
@@ -506,27 +559,243 @@ KeccakF1600:
 	str	@E[3],[sp,#$A[4][3]+4]
 	bic	@E[1],@C[3],@C[1]
 	eor	@E[2],@E[0],@C[8]
-	 add	@E[0],sp,#$A[1][0]
 	eor	@E[3],@E[1],@C[9]
-	 ldmia	sp,{@C[0]-@C[9]}		@ A[0][0..5]
 	str	@E[2],[sp,#$A[4][4]]		@ A[4][4] = C[4] ^ (~C[0] & C[1]);
+	 add	@E[0],sp,#$A[1][0]
 	str	@E[3],[sp,#$A[4][4]+4]
 
 	blo	.Lround
 
-	ldr	@E[1],[sp,#320]			@ restore pointer to A
-	stmia	@E[1]!,{@C[0]-@C[9]}		@ copy A[5][5] from stack
+	ldr	pc,[sp,#320]
+.size	KeccakF1600_int,.-KeccakF1600_int
+
+.type	KeccakF1600, %function
+.align	5
+KeccakF1600:
+	stmdb	sp!,{r0,r4-r11,lr}
+	sub	sp,sp,#320+16			@ space for A[5][5],D[5],T[2][5],...
+
+	add	@E[0],r0,#$A[1][0]
+	add	@E[1],sp,#$A[1][0]
+	mov	@E[2],r0
+	ldmia	@E[0]!,{@C[0]-@C[9]}		@ copy A[5][5] to stack
+	stmia	@E[1]!,{@C[0]-@C[9]}
+	ldmia	@E[0]!,{@C[0]-@C[9]}
+	stmia	@E[1]!,{@C[0]-@C[9]}
+	ldmia	@E[0]!,{@C[0]-@C[9]}
+	stmia	@E[1]!,{@C[0]-@C[9]}
+	ldmia	@E[0], {@C[0]-@C[9]}
+	stmia	@E[1], {@C[0]-@C[9]}
+	ldmia	@E[2], {@C[0]-@C[9]}		@ A[0][0..4]
+	add	@E[0],sp,#$A[1][0]
+	stmia	sp,    {@C[0]-@C[9]}
+
+	bl	KeccakF1600_enter
+
+	ldr	@E[1], [sp,#320+16]		@ restore pointer to A
+	ldmia	sp,    {@C[0]-@C[9]}
+	stmia	@E[1]!,{@C[0]-@C[9]}		@ return A[5][5]
 	ldmia	@E[0]!,{@C[0]-@C[9]}
 	stmia	@E[1]!,{@C[0]-@C[9]}
 	ldmia	@E[0]!,{@C[0]-@C[9]}
 	stmia	@E[1]!,{@C[0]-@C[9]}
 	ldmia	@E[0]!,{@C[0]-@C[9]}
 	stmia	@E[1]!,{@C[0]-@C[9]}
-	ldmia	@E[0],{@C[0]-@C[9]}
-	stmia	@E[1],{@C[0]-@C[9]}
+	ldmia	@E[0], {@C[0]-@C[9]}
+	stmia	@E[1], {@C[0]-@C[9]}
 
-	add	sp,sp,#320+8
-	ldmia	sp!,{r4-r12,pc}
+	add	sp,sp,#320+20
+	ldmia	sp!,{r4-r11,pc}
 .size	KeccakF1600,.-KeccakF1600
 ___
+{ my ($hi,$lo,$i,$A_flat, $len,$bsz,$inp) = map("r$_",(5..8, 10..12));
+
+########################################################################
+# Stack layout
+# ----->+-----------------------+
+#       | uint64_t A[5][5]      |
+#       | ...                   |
+#       | ...                   |
+# +336->+-----------------------+
+#       | uint64_t *A           |
+# +340->+-----------------------+
+#       | const void *inp       |
+# +344->+-----------------------+
+#       | size_t len            |
+# +348->+-----------------------+
+#       | size_t bs             |
+# +352->+-----------------------+
+#       | ....
+
+$code.=<<___;
+.global	SHA3_absorb
+.type	SHA3_absorb,%function
+.align	5
+SHA3_absorb:
+	stmdb	sp!,{r0-r12,lr}
+	sub	sp,sp,#320+16
+
+	mov	r12,r0
+	add	r14,sp,#0
+	mov	$len,r2
+	mov	$bsz,r3
+
+	ldmia	r12!,{@C[0]-@C[9]}	@ copy A[5][5] to stack
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12, {@C[0]-@C[9]}
+	stmia	r14, {@C[0]-@C[9]}
+
+	ldr	$inp,[sp,#340]
+
+.Loop_absorb:
+	subs	r0,$len,$bsz
+	blo	.Labsorbed
+	add	$A_flat,sp,#0
+	str	r0,[sp,#344]		@ save len - bsz
+
+.Loop_block:
+	ldmia	$A_flat,{r2-r3}		@ A_flat[i]
+	ldrb	r0,[$inp,#7]!		@ inp[7]
+	mov	$i,#8
+
+.Lane_loop:
+	subs	$i,$i,#1
+	lsl	r1,r0,#24
+	blo	.Lane_done
+#ifdef	__thumb2__
+	it	ne
+	ldrbne	r0,[$inp,#-1]!
+#else
+	ldrneb	r0,[$inp,#-1]!
+#endif
+	adds	r1,r1,r1		@ sip through carry flag
+	adc	$hi,$hi,$hi
+	adds	r1,r1,r1
+	adc	$lo,$lo,$lo
+	adds	r1,r1,r1
+	adc	$hi,$hi,$hi
+	adds	r1,r1,r1
+	adc	$lo,$lo,$lo
+	adds	r1,r1,r1
+	adc	$hi,$hi,$hi
+	adds	r1,r1,r1
+	adc	$lo,$lo,$lo
+	adds	r1,r1,r1
+	adc	$hi,$hi,$hi
+	adds	r1,r1,r1
+	adc	$lo,$lo,$lo
+	b	.Lane_loop
+
+.Lane_done:
+	eor	r2,r2,$lo
+	eor	r3,r3,$hi
+	add	$inp,$inp,#8
+	stmia	$A_flat!,{r2-r3}	@ A_flat[i++] ^= BitInterleave(inp[0..7])
+	subs	$bsz,$bsz,#8
+	bhi	.Loop_block
+
+	str	$inp,[sp,#340]
+
+	bl	KeccakF1600_int
+
+	ldr	$inp,[sp,#340]
+	ldr	$len,[sp,#344]
+	ldr	$bsz,[sp,#348]
+	b	.Loop_absorb
+
+.align	4
+.Labsorbed:
+	add	r12,sp,#$A[1][0]
+	ldr	r14, [sp,#336]		@ pull pointer to A[5][5]
+	ldmia	sp,  {@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}	@ return A[5][5]
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12!,{@C[0]-@C[9]}
+	stmia	r14!,{@C[0]-@C[9]}
+	ldmia	r12, {@C[0]-@C[9]}
+	stmia	r14, {@C[0]-@C[9]}
+
+	add	sp,sp,#320+32
+	mov	r0,$len			@ return value
+	ldmia	sp!,{r4-r12,pc}
+.size	SHA3_absorb,.-SHA3_absorb
+___
+}
+{ my ($A_flat,$out,$len,$bsz, $byte,$shl) = map("r$_", (4..9));
+
+$code.=<<___;
+.global	SHA3_squeeze
+.type	SHA3_squeeze,%function
+.align	5
+SHA3_squeeze:
+	stmdb	sp!,{r4-r10,lr}
+	mov	r12,r0
+	mov	$A_flat,r0
+	mov	$out,r1
+	mov	$len,r2
+	mov	$bsz,r3
+	mov	r14,r3
+	b	.Loop_squeeze
+
+.align	4
+.Loop_squeeze:
+	ldmia	r12!,{r0,r1}		@ A_flat[i++]
+	mov	$shl,#28
+
+.Lane_squeeze:
+	lsl	r2,r0,$shl
+	lsl	r3,r1,$shl
+	eor	$byte,$byte,$byte
+	adds	r3,r3,r3		@ sip through carry flag
+	adc	$byte,$byte,$byte
+	adds	r2,r2,r2
+	adc	$byte,$byte,$byte
+	adds	r3,r3,r3
+	adc	$byte,$byte,$byte
+	adds	r2,r2,r2
+	adc	$byte,$byte,$byte
+	adds	r3,r3,r3
+	adc	$byte,$byte,$byte
+	adds	r2,r2,r2
+	adc	$byte,$byte,$byte
+	adds	r3,r3,r3
+	adc	$byte,$byte,$byte
+	adds	r2,r2,r2
+	adc	$byte,$byte,$byte
+	subs	$len,$len,#1		@ len -= 1
+	str	$byte,[$out],#1
+	beq	.Lsqueeze_done
+	subs	$shl,$shl,#4
+	bhs	.Lane_squeeze
+
+	subs	r14,r14,#8		@ bsz -= 8
+	bhi	.Loop_squeeze
+
+	mov	r0,$A_flat
+
+	bl	KeccakF1600
+
+	mov	r12,$A_flat
+	mov	r14,$bsz
+	b	.Loop_squeeze
+
+.Lsqueeze_done:
+	ldmia	sp!,{r4-r10,pc}
+.size	SHA3_squeeze,.-SHA3_squeeze
+.asciz	"Keccak-1600 absorb and squeeze for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
+.align	2
+___
+}
+
 print $code;
+
+close STDOUT; # enforce flush