Subversion Repositories planix.SVN

Rev

Rev 2 | Details | Compare with Previous | Last modification | View Log | RSS feed

Rev Author Line No. Line
2 - 1 
	TEXT	memset(SB),$0
2 
#define	BDNZ	BC	16,0,
3 
	MOVW R3, p+0(FP)		/* R3 is pointer */
4 
 
5 
/*
6 
 * performance:
7 
 *	about 100mbytes/sec (8k blocks) on a 603/105 without L2 cache
8 
 *	drops to 40mbytes/sec (10k blocks) and 28mbytes/sec with 32k blocks
9 
 */
10 
 
11 
	MOVW	n+8(FP), R4		/* R4 is count */
12 
	CMP	R4, $0
13 
	BLE	ret
14 
	MOVW	c+4(FP), R5		/* R5 is char */
15 
 
16 
/*
17 
 * create 16 copies of c in R5 .. R8
18 
 */
19 
	RLWNM	$0, R5, $0xff, R5
20 
	RLWMI	$8, R5, $0xff00, R5
21 
	RLWMI	$16, R5, $0xffff0000, R5
22 
	MOVW	R5, R6
23 
	MOVW	R5, R7
24 
	MOVW	R5, R8
25 
 
26 
/*
27 
 * let STSW do the work for 16 characters or less; aligned and unaligned
28 
 */
29 
	CMP	R4, $16
30 
	BLE	out
31 
 
32 
/*
33 
 * store enough bytes to align pointer
34 
 */
35 
	ANDCC	$7,R3, R9
36 
	BEQ	l2
37 
	SUBC	R9, $8, R9
38 
	MOVW	R9, XER
39 
	STSW	R5, (R3)
40 
	ADD	R9, R3
41 
	SUB	R9, R4
42 
 
43 
/*
44 
 * store 16 at a time while there's room
45 
 * STSW was used here originally, but it's `completion serialised'
46 
 */
47 
l2:
48 
	SRAWCC	$4, R4, R9
49 
	BLE	out
50 
	MOVW	R9, CTR
51 
l3:
52 
	MOVW	R5, 0(R3)
53 
	ADD	$8, R3, R10
54 
	MOVW	R6, 4(R3)
55 
	MOVW	R7, 0(R10)
56 
	ADD	$8, R10, R3
57 
	MOVW	R8, 4(R10)
58 
	BDNZ	l3
59 
	RLWNMCC	$0, R4, $15, R4	/* residue */
60 
	BEQ	ret
61 
 
62 
/*
63 
 * store up to 16 bytes from R5 .. R8; aligned and unaligned
64 
 */
65 
 
66 
out:
67 
	MOVW	R4, XER
68 
	STSW	R5, (R3)
69 
 
70 
ret:
71 
	MOVW	0(FP), R3
72 
	RETURN
73 
	END