mirror of
				https://sourceware.org/git/glibc.git
				synced 2025-10-30 10:45:40 +03:00 
			
		
		
		
	
		
			
				
	
	
		
			128 lines
		
	
	
		
			3.6 KiB
		
	
	
	
		
			ArmAsm
		
	
	
	
	
	
			
		
		
	
	
			128 lines
		
	
	
		
			3.6 KiB
		
	
	
	
		
			ArmAsm
		
	
	
	
	
	
| /* Copyright (C) 1996-2016 Free Software Foundation, Inc.
 | |
|    Contributed by Richard Henderson (rth@tamu.edu)
 | |
|    This file is part of the GNU C Library.
 | |
| 
 | |
|    The GNU C Library is free software; you can redistribute it and/or
 | |
|    modify it under the terms of the GNU Lesser General Public
 | |
|    License as published by the Free Software Foundation; either
 | |
|    version 2.1 of the License, or (at your option) any later version.
 | |
| 
 | |
|    The GNU C Library is distributed in the hope that it will be useful,
 | |
|    but WITHOUT ANY WARRANTY; without even the implied warranty of
 | |
|    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 | |
|    Lesser General Public License for more details.
 | |
| 
 | |
|    You should have received a copy of the GNU Lesser General Public
 | |
|    License along with the GNU C Library.  If not, see
 | |
|    <http://www.gnu.org/licenses/>.  */
 | |
| 
 | |
| /* Fill a block of memory with a character.  Optimized for the Alpha
 | |
|    architecture:
 | |
| 
 | |
|    - memory accessed as aligned quadwords only
 | |
|    - destination memory not read unless needed for good cache behaviour
 | |
|    - basic blocks arranged to optimize branch prediction for full-quadword
 | |
|      aligned memory blocks.
 | |
|    - partial head and tail quadwords constructed with byte-mask instructions
 | |
| 
 | |
|    This is generally scheduled for the EV5 (got to look out for my own
 | |
|    interests :-), but with EV4 needs in mind.  There *should* be no more
 | |
|    stalls for the EV4 than there are for the EV5.
 | |
| */
 | |
| 
 | |
| 
 | |
| #include <sysdep.h>
 | |
| 
 | |
| 	.set noat
 | |
| 	.set noreorder
 | |
| 
 | |
| 	.text
 | |
| 	.type	memset, @function
 | |
| 	.globl	memset
 | |
| 	.usepv	memset, USEPV_PROF
 | |
| 
 | |
| 	cfi_startproc
 | |
| 
 | |
| 	/* On entry to this basic block:
 | |
| 	   t3 == loop counter
 | |
| 	   t4 == bytes in partial final word
 | |
| 	   a0 == possibly misaligned destination pointer
 | |
| 	   a1 == replicated source character  */
 | |
| 
 | |
| 	.align 3
 | |
| memset_loop:
 | |
| 	beq	t3, $tail
 | |
| 	blbc	t3, 0f		# skip single store if count even
 | |
| 
 | |
| 	stq_u	a1, 0(a0)	# e0    : store one word
 | |
| 	subq	t3, 1, t3	# .. e1 :
 | |
| 	addq	a0, 8, a0	# e0    :
 | |
| 	beq	t3, $tail	# .. e1 :
 | |
| 
 | |
| 0:	stq_u	a1, 0(a0)	# e0    : store two words
 | |
| 	subq	t3, 2, t3	# .. e1 :
 | |
| 	stq_u	a1, 8(a0)	# e0    :
 | |
| 	addq	a0, 16, a0	# .. e1 :
 | |
| 	bne	t3, 0b		# e1    :
 | |
| 
 | |
| $tail:	bne	t4, 1f		# is there a tail to do?
 | |
| 	ret			# no
 | |
| 
 | |
| 	.align 3
 | |
| 1:	ldq_u	t0, 0(a0)	# e1    : yes, load original data
 | |
| 	mskql	a1, t4, t1	# .. e0 :
 | |
| 	mskqh	t0, t4, t0	# e0    :
 | |
| 	or	t0, t1, t0	# e1 (stall)
 | |
| 	stq_u	t0, 0(a0)	# e0    :
 | |
| 	ret			# .. e1 :
 | |
| 
 | |
| memset:
 | |
| #ifdef PROF
 | |
| 	ldgp	gp, 0(pv)
 | |
| 	lda	AT, _mcount
 | |
| 	jsr	AT, (AT), _mcount
 | |
| #endif
 | |
| 
 | |
| 	and	a1, 0xff, a1	# e0    : zero extend input character
 | |
| 	mov	a0, v0		# .. e1 : move return value in place
 | |
| 	sll	a1, 8, t0	# e0    : begin replicating the char
 | |
| 	beq	a2, $done	# .. e1 : early exit for zero-length store
 | |
| 	or	t0, a1, a1	# e0    :
 | |
| 	and	a0, 7, t1	# .. e1 : dest misalignment
 | |
| 	sll	a1, 16, t0	# e0    :
 | |
| 	addq	a2, t1, a2	# .. e1 : add dest misalignment to count
 | |
| 	or	t0, a1, a1	# e0    :
 | |
| 	srl	a2, 3, t3	# .. e1 : loop = count >> 3
 | |
| 	sll	a1, 32, t0	# e0    :
 | |
| 	and	a2, 7, t4	# .. e1 : find number of bytes in tail
 | |
| 	or	t0, a1, a1	# e0    : character replication done
 | |
| 
 | |
| 	beq	t1, memset_loop	# .. e1 : aligned head, jump right in
 | |
| 
 | |
| 	ldq_u	t0, 0(a0)	# e1    : load original data to mask into
 | |
| 	mskqh	a1, a0, t1	# .. e0 :
 | |
| 
 | |
| 	cmpult	a2, 8, t2	# e0    : is this a sub-word set?
 | |
| 	bne	t2, $oneq	# .. e1 (zdb)
 | |
| 
 | |
| 	mskql	t0, a0, t0	# e0    : we span words.  finish this partial
 | |
| 	subq	t3, 1, t3	# .. e1 :
 | |
| 	addq	a0, 8, a0	# e0    :
 | |
| 	or	t0, t1, t0	# .. e1 :
 | |
| 	stq_u	t0, -8(a0)	# e0    :
 | |
| 	br 	memset_loop	# .. e1 :
 | |
| 
 | |
| 	.align 3
 | |
| $oneq:
 | |
| 	mskql	t1, a2, t1	# e0    : entire operation within one word
 | |
| 	mskql	t0, a0, t2	# e0    :
 | |
| 	mskqh	t0, a2, t3	# e0    :
 | |
| 	or	t1, t2, t0	# .. e1 :
 | |
| 	or	t0, t3, t0	# e1    :
 | |
| 	stq_u	t0, 0(a0)	# e0 (stall)
 | |
| 
 | |
| $done:	ret
 | |
| 
 | |
| 	cfi_endproc
 | |
| libc_hidden_builtin_def (memset)
 |