arm/crypto/chacha-neon-core.S

96 	// x1 = shuffle32(x1, MASK(0, 3, 2, 1))
98 	// x2 = shuffle32(x2, MASK(1, 0, 3, 2))
100 	// x3 = shuffle32(x3, MASK(2, 1, 0, 3))
126 	// x1 = shuffle32(x1, MASK(2, 1, 0, 3))
128 	// x2 = shuffle32(x2, MASK(1, 0, 3, 2))
130 	// x3 = shuffle32(x3, MASK(0, 3, 2, 1))
147 	add		ip, r0, #0x20
158 	add		ip, r2, #0x20
178 	add		ip, r1, #0x20
204 .Lctrinc:	.word	0, 1, 2, 3
205 .Lrol8_table:	.byte	3, 0, 1, 2, 7, 4, 5, 6
211 	sub		ip, sp, #0x20		// allocate a 32 byte buffer
212 	bic		ip, ip, #0x1f		// aligned to 32 bytes
228 	// x0..15[0-3] = s0..15[0-3]
229 	add		ip, r0, #0x20
235 	vdup.32		q14, d7[0]
238 	vdup.32		q12, d6[0]
240 	vdup.32		q10, d5[0]
241 	vadd.u32	q12, q12, q4		// x12 += counter values 0-3
243 	vdup.32		q8, d4[0]
245 	vdup.32		q6, d3[0]
247 	vdup.32		q4, d2[0]
249 	vdup.32		q2, d1[0]
251 	vdup.32		q0, d0[0]
260 	// x1 += x5, x13 = rotl32(x13 ^ x1, 16)
279 	// x9 += x13, x5 = rotl32(x5 ^ x9, 12)
304 	// x1 += x5, x13 = rotl32(x13 ^ x1, 8)
330 	// x9 += x13, x5 = rotl32(x5 ^ x9, 7)
358 	// x2 += x7, x13 = rotl32(x13 ^ x2, 16)
377 	// x8 += x13, x7 = rotl32(x7 ^ x8, 12)
402 	// x2 += x7, x13 = rotl32(x13 ^ x2, 8)
428 	// x8 += x13, x7 = rotl32(x7 ^ x8, 7)
454 	// x0..7[0-3] are in q0-q7, x10..15[0-3] are in q10-q15.
455 	// x8..9[0-3] are on the stack.
458 	// Also add the counter values 0-3 to x12[0-3].
459 	  vld1.32	{q8}, [lr, :128]	// load counter values 0-3
460 	vzip.32		q0, q1			// => (0 1 0 1) (0 1 0 1)
464 	  vadd.u32	q12, q8			// x12 += counter values 0-3
477 	// x0..3[0-3] += s0..3[0-3]	(add orig state to 1st row of each block)
483 	// x4..7[0-3] += s4..7[0-3]	(add orig state to 2nd row of each block)
511 	// x8..11[0-3] += s8..11[0-3]	(add orig state to 3rd row of each block)
517 	// x12..15[0-3] += s12..15[0-3] (add orig state to 4th row of each block)
636 	.byte		0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07
637 	.byte		0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f
638 	.byte		0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17
639 	.byte		0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f
640 	.byte		0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07
641 	.byte		0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f
642 	.byte		0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17
643 	.byte		0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f