ARM: 8119/1: crypto: sha1: add ARM NEON implementation

This patch adds ARM NEON assembly implementation of SHA-1 algorithm.

tcrypt benchmark results on Cortex-A8, sha1-arm-asm vs sha1-neon-asm:

block-size      bytes/update    old-vs-new
16              16              1.04x
64              16              1.02x
64              64              1.05x
256             16              1.03x
256             64              1.04x
256             256             1.30x
1024            16              1.03x
1024            256             1.36x
1024            1024            1.52x
2048            16              1.03x
2048            256             1.39x
2048            1024            1.55x
2048            2048            1.59x
4096            16              1.03x
4096            256             1.40x
4096            1024            1.57x
4096            4096            1.62x
8192            16              1.03x
8192            256             1.40x
8192            1024            1.58x
8192            4096            1.63x
8192            8192            1.63x

Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

1 files changed, 634 insertions, 0 deletions

diff --git a/arch/arm/crypto/sha1-armv7-neon.S b/arch/arm/crypto/sha1-armv7-neon.S
new file mode 100644
index 000000000000..50013c0e2864
--- /dev/null
+++ b/arch/arm/crypto/sha1-armv7-neon.S
@@ -0,0 +1,634 @@
+/* sha1-armv7-neon.S - ARM/NEON accelerated SHA-1 transform function
+ *
+ * Copyright © 2013-2014 Jussi Kivilinna <jussi.kivilinna@iki.fi>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ */
+
+#include <linux/linkage.h>
+
+
+.syntax unified
+.code   32
+.fpu neon
+
+.text
+
+
+/* Context structure */
+
+#define state_h0 0
+#define state_h1 4
+#define state_h2 8
+#define state_h3 12
+#define state_h4 16
+
+
+/* Constants */
+
+#define K1  0x5A827999
+#define K2  0x6ED9EBA1
+#define K3  0x8F1BBCDC
+#define K4  0xCA62C1D6
+.align 4
+.LK_VEC:
+.LK1:	.long K1, K1, K1, K1
+.LK2:	.long K2, K2, K2, K2
+.LK3:	.long K3, K3, K3, K3
+.LK4:	.long K4, K4, K4, K4
+
+
+/* Register macros */
+
+#define RSTATE r0
+#define RDATA r1
+#define RNBLKS r2
+#define ROLDSTACK r3
+#define RWK lr
+
+#define _a r4
+#define _b r5
+#define _c r6
+#define _d r7
+#define _e r8
+
+#define RT0 r9
+#define RT1 r10
+#define RT2 r11
+#define RT3 r12
+
+#define W0 q0
+#define W1 q1
+#define W2 q2
+#define W3 q3
+#define W4 q4
+#define W5 q5
+#define W6 q6
+#define W7 q7
+
+#define tmp0 q8
+#define tmp1 q9
+#define tmp2 q10
+#define tmp3 q11
+
+#define qK1 q12
+#define qK2 q13
+#define qK3 q14
+#define qK4 q15
+
+
+/* Round function macros. */
+
+#define WK_offs(i) (((i) & 15) * 4)
+
+#define _R_F1(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	      W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	ldr RT3, [sp, WK_offs(i)]; \
+		pre1(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	bic RT0, d, b; \
+	add e, e, a, ror #(32 - 5); \
+	and RT1, c, b; \
+		pre2(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	add RT0, RT0, RT3; \
+	add e, e, RT1; \
+	ror b, #(32 - 30); \
+		pre3(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	add e, e, RT0;
+
+#define _R_F2(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	      W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	ldr RT3, [sp, WK_offs(i)]; \
+		pre1(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	eor RT0, d, b; \
+	add e, e, a, ror #(32 - 5); \
+	eor RT0, RT0, c; \
+		pre2(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	add e, e, RT3; \
+	ror b, #(32 - 30); \
+		pre3(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	add e, e, RT0; \
+
+#define _R_F3(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	      W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	ldr RT3, [sp, WK_offs(i)]; \
+		pre1(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	eor RT0, b, c; \
+	and RT1, b, c; \
+	add e, e, a, ror #(32 - 5); \
+		pre2(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	and RT0, RT0, d; \
+	add RT1, RT1, RT3; \
+	add e, e, RT0; \
+	ror b, #(32 - 30); \
+		pre3(i16,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28); \
+	add e, e, RT1;
+
+#define _R_F4(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	      W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	_R_F2(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	      W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28)
+
+#define _R(a,b,c,d,e,f,i,pre1,pre2,pre3,i16,\
+           W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	_R_##f(a,b,c,d,e,i,pre1,pre2,pre3,i16,\
+	       W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28)
+
+#define R(a,b,c,d,e,f,i) \
+	_R_##f(a,b,c,d,e,i,dummy,dummy,dummy,i16,\
+	       W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28)
+
+#define dummy(...)
+
+
+/* Input expansion macros. */
+
+/********* Precalc macros for rounds 0-15 *************************************/
+
+#define W_PRECALC_00_15() \
+	add       RWK, sp, #(WK_offs(0));			\
+	\
+	vld1.32   {tmp0, tmp1}, [RDATA]!;			\
+	vrev32.8  W0, tmp0;		/* big => little */	\
+	vld1.32   {tmp2, tmp3}, [RDATA]!;			\
+	vadd.u32  tmp0, W0, curK;				\
+	vrev32.8  W7, tmp1;		/* big => little */	\
+	vrev32.8  W6, tmp2;		/* big => little */	\
+	vadd.u32  tmp1, W7, curK;				\
+	vrev32.8  W5, tmp3;		/* big => little */	\
+	vadd.u32  tmp2, W6, curK;				\
+	vst1.32   {tmp0, tmp1}, [RWK]!;				\
+	vadd.u32  tmp3, W5, curK;				\
+	vst1.32   {tmp2, tmp3}, [RWK];				\
+
+#define WPRECALC_00_15_0(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vld1.32   {tmp0, tmp1}, [RDATA]!;			\
+
+#define WPRECALC_00_15_1(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	add       RWK, sp, #(WK_offs(0));			\
+
+#define WPRECALC_00_15_2(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vrev32.8  W0, tmp0;		/* big => little */	\
+
+#define WPRECALC_00_15_3(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vld1.32   {tmp2, tmp3}, [RDATA]!;			\
+
+#define WPRECALC_00_15_4(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32  tmp0, W0, curK;				\
+
+#define WPRECALC_00_15_5(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vrev32.8  W7, tmp1;		/* big => little */	\
+
+#define WPRECALC_00_15_6(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vrev32.8  W6, tmp2;		/* big => little */	\
+
+#define WPRECALC_00_15_7(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32  tmp1, W7, curK;				\
+
+#define WPRECALC_00_15_8(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vrev32.8  W5, tmp3;		/* big => little */	\
+
+#define WPRECALC_00_15_9(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32  tmp2, W6, curK;				\
+
+#define WPRECALC_00_15_10(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vst1.32   {tmp0, tmp1}, [RWK]!;				\
+
+#define WPRECALC_00_15_11(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32  tmp3, W5, curK;				\
+
+#define WPRECALC_00_15_12(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vst1.32   {tmp2, tmp3}, [RWK];				\
+
+
+/********* Precalc macros for rounds 16-31 ************************************/
+
+#define WPRECALC_16_31_0(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor      tmp0, tmp0;			\
+	vext.8    W, W_m16, W_m12, #8;		\
+
+#define WPRECALC_16_31_1(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	add       RWK, sp, #(WK_offs(i));	\
+	vext.8    tmp0, W_m04, tmp0, #4;	\
+
+#define WPRECALC_16_31_2(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor      tmp0, tmp0, W_m16;		\
+	veor.32   W, W, W_m08;			\
+
+#define WPRECALC_16_31_3(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor      tmp1, tmp1;			\
+	veor      W, W, tmp0;			\
+
+#define WPRECALC_16_31_4(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vshl.u32  tmp0, W, #1;			\
+
+#define WPRECALC_16_31_5(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vext.8    tmp1, tmp1, W, #(16-12);	\
+	vshr.u32  W, W, #31;			\
+
+#define WPRECALC_16_31_6(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vorr      tmp0, tmp0, W;		\
+	vshr.u32  W, tmp1, #30;			\
+
+#define WPRECALC_16_31_7(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vshl.u32  tmp1, tmp1, #2;		\
+
+#define WPRECALC_16_31_8(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor      tmp0, tmp0, W;		\
+
+#define WPRECALC_16_31_9(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor      W, tmp0, tmp1;		\
+
+#define WPRECALC_16_31_10(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32  tmp0, W, curK;		\
+
+#define WPRECALC_16_31_11(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vst1.32   {tmp0}, [RWK];
+
+
+/********* Precalc macros for rounds 32-79 ************************************/
+
+#define WPRECALC_32_79_0(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor W, W_m28; \
+
+#define WPRECALC_32_79_1(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vext.8 tmp0, W_m08, W_m04, #8; \
+
+#define WPRECALC_32_79_2(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor W, W_m16; \
+
+#define WPRECALC_32_79_3(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	veor W, tmp0; \
+
+#define WPRECALC_32_79_4(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	add RWK, sp, #(WK_offs(i&~3)); \
+
+#define WPRECALC_32_79_5(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vshl.u32 tmp1, W, #2; \
+
+#define WPRECALC_32_79_6(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vshr.u32 tmp0, W, #30; \
+
+#define WPRECALC_32_79_7(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vorr W, tmp0, tmp1; \
+
+#define WPRECALC_32_79_8(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vadd.u32 tmp0, W, curK; \
+
+#define WPRECALC_32_79_9(i,W,W_m04,W_m08,W_m12,W_m16,W_m20,W_m24,W_m28) \
+	vst1.32 {tmp0}, [RWK];
+
+
+/*
+ * Transform nblks*64 bytes (nblks*16 32-bit words) at DATA.
+ *
+ * unsigned int
+ * sha1_transform_neon (void *ctx, const unsigned char *data,
+ *                      unsigned int nblks)
+ */
+.align 3
+ENTRY(sha1_transform_neon)
+  /* input:
+   *	r0: ctx, CTX
+   *	r1: data (64*nblks bytes)
+   *	r2: nblks
+   */
+
+  cmp RNBLKS, #0;
+  beq .Ldo_nothing;
+
+  push {r4-r12, lr};
+  /*vpush {q4-q7};*/
+
+  adr RT3, .LK_VEC;
+
+  mov ROLDSTACK, sp;
+
+  /* Align stack. */
+  sub RT0, sp, #(16*4);
+  and RT0, #(~(16-1));
+  mov sp, RT0;
+
+  vld1.32 {qK1-qK2}, [RT3]!; /* Load K1,K2 */
+
+  /* Get the values of the chaining variables. */
+  ldm RSTATE, {_a-_e};
+
+  vld1.32 {qK3-qK4}, [RT3]; /* Load K3,K4 */
+
+#undef curK
+#define curK qK1
+  /* Precalc 0-15. */
+  W_PRECALC_00_15();
+
+.Loop:
+  /* Transform 0-15 + Precalc 16-31. */
+  _R( _a, _b, _c, _d, _e, F1,  0,
+      WPRECALC_16_31_0, WPRECALC_16_31_1, WPRECALC_16_31_2, 16,
+      W4, W5, W6, W7, W0, _, _, _ );
+  _R( _e, _a, _b, _c, _d, F1,  1,
+      WPRECALC_16_31_3, WPRECALC_16_31_4, WPRECALC_16_31_5, 16,
+      W4, W5, W6, W7, W0, _, _, _ );
+  _R( _d, _e, _a, _b, _c, F1,  2,
+      WPRECALC_16_31_6, WPRECALC_16_31_7, WPRECALC_16_31_8, 16,
+      W4, W5, W6, W7, W0, _, _, _ );
+  _R( _c, _d, _e, _a, _b, F1,  3,
+      WPRECALC_16_31_9, WPRECALC_16_31_10,WPRECALC_16_31_11,16,
+      W4, W5, W6, W7, W0, _, _, _ );
+
+#undef curK
+#define curK qK2
+  _R( _b, _c, _d, _e, _a, F1,  4,
+      WPRECALC_16_31_0, WPRECALC_16_31_1, WPRECALC_16_31_2, 20,
+      W3, W4, W5, W6, W7, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F1,  5,
+      WPRECALC_16_31_3, WPRECALC_16_31_4, WPRECALC_16_31_5, 20,
+      W3, W4, W5, W6, W7, _, _, _ );
+  _R( _e, _a, _b, _c, _d, F1,  6,
+      WPRECALC_16_31_6, WPRECALC_16_31_7, WPRECALC_16_31_8, 20,
+      W3, W4, W5, W6, W7, _, _, _ );
+  _R( _d, _e, _a, _b, _c, F1,  7,
+      WPRECALC_16_31_9, WPRECALC_16_31_10,WPRECALC_16_31_11,20,
+      W3, W4, W5, W6, W7, _, _, _ );
+
+  _R( _c, _d, _e, _a, _b, F1,  8,
+      WPRECALC_16_31_0, WPRECALC_16_31_1, WPRECALC_16_31_2, 24,
+      W2, W3, W4, W5, W6, _, _, _ );
+  _R( _b, _c, _d, _e, _a, F1,  9,
+      WPRECALC_16_31_3, WPRECALC_16_31_4, WPRECALC_16_31_5, 24,
+      W2, W3, W4, W5, W6, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F1, 10,
+      WPRECALC_16_31_6, WPRECALC_16_31_7, WPRECALC_16_31_8, 24,
+      W2, W3, W4, W5, W6, _, _, _ );
+  _R( _e, _a, _b, _c, _d, F1, 11,
+      WPRECALC_16_31_9, WPRECALC_16_31_10,WPRECALC_16_31_11,24,
+      W2, W3, W4, W5, W6, _, _, _ );
+
+  _R( _d, _e, _a, _b, _c, F1, 12,
+      WPRECALC_16_31_0, WPRECALC_16_31_1, WPRECALC_16_31_2, 28,
+      W1, W2, W3, W4, W5, _, _, _ );
+  _R( _c, _d, _e, _a, _b, F1, 13,
+      WPRECALC_16_31_3, WPRECALC_16_31_4, WPRECALC_16_31_5, 28,
+      W1, W2, W3, W4, W5, _, _, _ );
+  _R( _b, _c, _d, _e, _a, F1, 14,
+      WPRECALC_16_31_6, WPRECALC_16_31_7, WPRECALC_16_31_8, 28,
+      W1, W2, W3, W4, W5, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F1, 15,
+      WPRECALC_16_31_9, WPRECALC_16_31_10,WPRECALC_16_31_11,28,
+      W1, W2, W3, W4, W5, _, _, _ );
+
+  /* Transform 16-63 + Precalc 32-79. */
+  _R( _e, _a, _b, _c, _d, F1, 16,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 32,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _d, _e, _a, _b, _c, F1, 17,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 32,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _c, _d, _e, _a, _b, F1, 18,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 32,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _b, _c, _d, _e, _a, F1, 19,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 32,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+
+  _R( _a, _b, _c, _d, _e, F2, 20,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 36,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _e, _a, _b, _c, _d, F2, 21,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 36,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _d, _e, _a, _b, _c, F2, 22,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 36,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _c, _d, _e, _a, _b, F2, 23,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 36,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+
+#undef curK
+#define curK qK3
+  _R( _b, _c, _d, _e, _a, F2, 24,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 40,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _a, _b, _c, _d, _e, F2, 25,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 40,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _e, _a, _b, _c, _d, F2, 26,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 40,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _d, _e, _a, _b, _c, F2, 27,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 40,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+
+  _R( _c, _d, _e, _a, _b, F2, 28,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 44,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _b, _c, _d, _e, _a, F2, 29,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 44,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _a, _b, _c, _d, _e, F2, 30,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 44,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _e, _a, _b, _c, _d, F2, 31,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 44,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+
+  _R( _d, _e, _a, _b, _c, F2, 32,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 48,
+      W4, W5, W6, W7, W0, W1, W2, W3);
+  _R( _c, _d, _e, _a, _b, F2, 33,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 48,
+      W4, W5, W6, W7, W0, W1, W2, W3);
+  _R( _b, _c, _d, _e, _a, F2, 34,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 48,
+      W4, W5, W6, W7, W0, W1, W2, W3);
+  _R( _a, _b, _c, _d, _e, F2, 35,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 48,
+      W4, W5, W6, W7, W0, W1, W2, W3);
+
+  _R( _e, _a, _b, _c, _d, F2, 36,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 52,
+      W3, W4, W5, W6, W7, W0, W1, W2);
+  _R( _d, _e, _a, _b, _c, F2, 37,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 52,
+      W3, W4, W5, W6, W7, W0, W1, W2);
+  _R( _c, _d, _e, _a, _b, F2, 38,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 52,
+      W3, W4, W5, W6, W7, W0, W1, W2);
+  _R( _b, _c, _d, _e, _a, F2, 39,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 52,
+      W3, W4, W5, W6, W7, W0, W1, W2);
+
+  _R( _a, _b, _c, _d, _e, F3, 40,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 56,
+      W2, W3, W4, W5, W6, W7, W0, W1);
+  _R( _e, _a, _b, _c, _d, F3, 41,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 56,
+      W2, W3, W4, W5, W6, W7, W0, W1);
+  _R( _d, _e, _a, _b, _c, F3, 42,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 56,
+      W2, W3, W4, W5, W6, W7, W0, W1);
+  _R( _c, _d, _e, _a, _b, F3, 43,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 56,
+      W2, W3, W4, W5, W6, W7, W0, W1);
+
+#undef curK
+#define curK qK4
+  _R( _b, _c, _d, _e, _a, F3, 44,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 60,
+      W1, W2, W3, W4, W5, W6, W7, W0);
+  _R( _a, _b, _c, _d, _e, F3, 45,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 60,
+      W1, W2, W3, W4, W5, W6, W7, W0);
+  _R( _e, _a, _b, _c, _d, F3, 46,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 60,
+      W1, W2, W3, W4, W5, W6, W7, W0);
+  _R( _d, _e, _a, _b, _c, F3, 47,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 60,
+      W1, W2, W3, W4, W5, W6, W7, W0);
+
+  _R( _c, _d, _e, _a, _b, F3, 48,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 64,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _b, _c, _d, _e, _a, F3, 49,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 64,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _a, _b, _c, _d, _e, F3, 50,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 64,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+  _R( _e, _a, _b, _c, _d, F3, 51,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 64,
+      W0, W1, W2, W3, W4, W5, W6, W7);
+
+  _R( _d, _e, _a, _b, _c, F3, 52,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 68,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _c, _d, _e, _a, _b, F3, 53,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 68,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _b, _c, _d, _e, _a, F3, 54,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 68,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+  _R( _a, _b, _c, _d, _e, F3, 55,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 68,
+      W7, W0, W1, W2, W3, W4, W5, W6);
+
+  _R( _e, _a, _b, _c, _d, F3, 56,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 72,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _d, _e, _a, _b, _c, F3, 57,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 72,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _c, _d, _e, _a, _b, F3, 58,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 72,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+  _R( _b, _c, _d, _e, _a, F3, 59,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 72,
+      W6, W7, W0, W1, W2, W3, W4, W5);
+
+  subs RNBLKS, #1;
+
+  _R( _a, _b, _c, _d, _e, F4, 60,
+      WPRECALC_32_79_0, WPRECALC_32_79_1, WPRECALC_32_79_2, 76,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _e, _a, _b, _c, _d, F4, 61,
+      WPRECALC_32_79_3, WPRECALC_32_79_4, WPRECALC_32_79_5, 76,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _d, _e, _a, _b, _c, F4, 62,
+      WPRECALC_32_79_6, dummy,            WPRECALC_32_79_7, 76,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+  _R( _c, _d, _e, _a, _b, F4, 63,
+      WPRECALC_32_79_8, dummy,            WPRECALC_32_79_9, 76,
+      W5, W6, W7, W0, W1, W2, W3, W4);
+
+  beq .Lend;
+
+  /* Transform 64-79 + Precalc 0-15 of next block. */
+#undef curK
+#define curK qK1
+  _R( _b, _c, _d, _e, _a, F4, 64,
+      WPRECALC_00_15_0, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F4, 65,
+      WPRECALC_00_15_1, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _e, _a, _b, _c, _d, F4, 66,
+      WPRECALC_00_15_2, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _d, _e, _a, _b, _c, F4, 67,
+      WPRECALC_00_15_3, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+
+  _R( _c, _d, _e, _a, _b, F4, 68,
+      dummy,            dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _b, _c, _d, _e, _a, F4, 69,
+      dummy,            dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F4, 70,
+      WPRECALC_00_15_4, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _e, _a, _b, _c, _d, F4, 71,
+      WPRECALC_00_15_5, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+
+  _R( _d, _e, _a, _b, _c, F4, 72,
+      dummy,            dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _c, _d, _e, _a, _b, F4, 73,
+      dummy,            dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _b, _c, _d, _e, _a, F4, 74,
+      WPRECALC_00_15_6, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _a, _b, _c, _d, _e, F4, 75,
+      WPRECALC_00_15_7, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+
+  _R( _e, _a, _b, _c, _d, F4, 76,
+      WPRECALC_00_15_8, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _d, _e, _a, _b, _c, F4, 77,
+      WPRECALC_00_15_9, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _c, _d, _e, _a, _b, F4, 78,
+      WPRECALC_00_15_10, dummy, dummy, _, _, _, _, _, _, _, _, _ );
+  _R( _b, _c, _d, _e, _a, F4, 79,
+      WPRECALC_00_15_11, dummy, WPRECALC_00_15_12, _, _, _, _, _, _, _, _, _ );
+
+  /* Update the chaining variables. */
+  ldm RSTATE, {RT0-RT3};
+  add _a, RT0;
+  ldr RT0, [RSTATE, #state_h4];
+  add _b, RT1;
+  add _c, RT2;
+  add _d, RT3;
+  add _e, RT0;
+  stm RSTATE, {_a-_e};
+
+  b .Loop;
+
+.Lend:
+  /* Transform 64-79 */
+  R( _b, _c, _d, _e, _a, F4, 64 );
+  R( _a, _b, _c, _d, _e, F4, 65 );
+  R( _e, _a, _b, _c, _d, F4, 66 );
+  R( _d, _e, _a, _b, _c, F4, 67 );
+  R( _c, _d, _e, _a, _b, F4, 68 );
+  R( _b, _c, _d, _e, _a, F4, 69 );
+  R( _a, _b, _c, _d, _e, F4, 70 );
+  R( _e, _a, _b, _c, _d, F4, 71 );
+  R( _d, _e, _a, _b, _c, F4, 72 );
+  R( _c, _d, _e, _a, _b, F4, 73 );
+  R( _b, _c, _d, _e, _a, F4, 74 );
+  R( _a, _b, _c, _d, _e, F4, 75 );
+  R( _e, _a, _b, _c, _d, F4, 76 );
+  R( _d, _e, _a, _b, _c, F4, 77 );
+  R( _c, _d, _e, _a, _b, F4, 78 );
+  R( _b, _c, _d, _e, _a, F4, 79 );
+
+  mov sp, ROLDSTACK;
+
+  /* Update the chaining variables. */
+  ldm RSTATE, {RT0-RT3};
+  add _a, RT0;
+  ldr RT0, [RSTATE, #state_h4];
+  add _b, RT1;
+  add _c, RT2;
+  add _d, RT3;
+  /*vpop {q4-q7};*/
+  add _e, RT0;
+  stm RSTATE, {_a-_e};
+
+  pop {r4-r12, pc};
+
+.Ldo_nothing:
+  bx lr
+ENDPROC(sha1_transform_neon)