xtensa/lib/memcopy.S

2  * arch/xtensa/lib/hal/memcopy.S -- Core HAL library functions
9  * Copyright (C) 2002 - 2012 Tensilica Inc.
24  * 32-bit load and store instructions (as required for these
34  *   If source is aligned,
39  *   This code tries to use fall-through branches for the common
40  *     case of aligned source and destination and multiple
44  *	a0/ return address
71 	add	a7, a3, a4	# a7 = end address for source
89 .Ldst1mod2:	# dst is only byte aligned
95 	addi	a4, a4, -1
98 	_bbci.l	a5, 1, .Ldstaligned	# if dst is now aligned, then
100 .Ldst2mod4:	# dst 16-bit aligned
106 	addi	a4, a4, -2
110 	j	.Ldstaligned	# dst is now aligned, return to main algorithm
121 .Ldstaligned:	# return here from .Ldst?mod? once dst is aligned
124 	movi	a8, 3		# if source is not aligned,
127 	 * Destination and source are word-aligned, use word copy.
129 	# copy 16 bytes per iteration for word-aligned dst and word-aligned src
189  * Destination is aligned, Source is unaligned
194 	_beqz	a4, .Ldone	# avoid loading anything for zero-length copies
195 	# copy 16 bytes per iteration for word-aligned dst and unaligned src
287  * 32-bit load and store instructions (as required for these
299  *	a0/ return address
324 	sub	a7, a3, a4	# a7 = start address for source
327 	addi	a3, a3, -1
329 	addi	a5, a5, -1
343 .Lbackdst1mod2:	# dst is only byte aligned
347 	addi	a3, a3, -1
349 	addi	a5, a5, -1
351 	addi	a4, a4, -1
352 	_bbci.l	a5, 1, .Lbackdstaligned	# if dst is now aligned, then
354 .Lbackdst2mod4:	# dst 16-bit aligned
357 	addi	a3, a3, -2
360 	addi	a5, a5, -2
363 	addi	a4, a4, -2
364 	j	.Lbackdstaligned	# dst is now aligned,
382 .Lbackdstaligned:	# return here from .Lbackdst?mod? once dst is aligned
385 	movi	a8, 3		# if source is not aligned,
388 	 * Destination and source are word-aligned, use word copy.
390 	# copy 16 bytes per iteration for word-aligned dst and word-aligned src
399 	addi	a3, a3, -16
402 	addi	a5, a5, -16
415 	addi	a3, a3, -8
418 	addi	a5, a5, -8
428 	addi	a3, a3, -4
430 	addi	a5, a5, -4
437 	addi	a3, a3, -2
439 	addi	a5, a5, -2
445 	addi	a3, a3, -1
447 	addi	a5, a5, -1
452  * Destination is aligned, Source is unaligned
457 	_beqz	a4, .Lbackdone	# avoid loading anything for zero-length copies
458 	# copy 16 bytes per iteration for word-aligned dst and unaligned src
476 	addi	a3, a3, -16
479 	addi	a5, a5, -16
496 	addi	a3, a3, -8
499 	addi	a5, a5, -8
508 	addi	a3, a3, -4
510 	addi	a5, a5, -4
524 	addi	a3, a3, -2
527 	addi	a5, a5, -2
534 	addi	a3, a3, -1
535 	addi	a5, a5, -1