1; mc88100 mpn_add_n -- Add two limb vectors of the same length > 0 and store
2; sum in a third limb vector.
3
4; Copyright 1992, 1994, 1995, 2000 Free Software Foundation, Inc.
5
6; This file is part of the GNU MP Library.
7
8; The GNU MP Library is free software; you can redistribute it and/or modify
9; it under the terms of the GNU Lesser General Public License as published by
10; the Free Software Foundation; either version 3 of the License, or (at your
11; option) any later version.
12
13; The GNU MP Library is distributed in the hope that it will be useful, but
14; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15; or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
16; License for more details.
17
18; You should have received a copy of the GNU Lesser General Public License
19; along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
20
21
22; INPUT PARAMETERS
23; res_ptr	r2
24; s1_ptr	r3
25; s2_ptr	r4
26; size		r5
27
28; This code has been optimized to run one instruction per clock, avoiding
29; load stalls and writeback contention.  As a result, the instruction
30; order is not always natural.
31
32; The speed is about 4.6 clocks/limb + 18 clocks/limb-vector on an 88100,
33; but on the 88110, it seems to run much slower, 6.6 clocks/limb.
34
35	text
36	align	 16
37	global	 ___gmpn_add_n
38___gmpn_add_n:
39	ld	r6,r3,0			; read first limb from s1_ptr
40	extu	r10,r5,3
41	ld	r7,r4,0			; read first limb from s2_ptr
42
43	subu.co	r5,r0,r5		; (clear carry as side effect)
44	mak	r5,r5,3<4>
45	bcnd	eq0,r5,Lzero
46
47	or	r12,r0,lo16(Lbase)
48	or.u	r12,r12,hi16(Lbase)
49	addu	r12,r12,r5		; r12 is address for entering in loop
50
51	extu	r5,r5,2			; divide by 4
52	subu	r2,r2,r5		; adjust res_ptr
53	subu	r3,r3,r5		; adjust s1_ptr
54	subu	r4,r4,r5		; adjust s2_ptr
55
56	or	r8,r6,r0
57
58	jmp.n	r12
59	 or	r9,r7,r0
60
61Loop:	addu	r3,r3,32
62	st	r8,r2,28
63	addu	r4,r4,32
64	ld	r6,r3,0
65	addu	r2,r2,32
66	ld	r7,r4,0
67Lzero:	subu	r10,r10,1		; add 0 + 8r limbs (adj loop cnt)
68Lbase:	ld	r8,r3,4
69	addu.cio r6,r6,r7
70	ld	r9,r4,4
71	st	r6,r2,0
72	ld	r6,r3,8			; add 7 + 8r limbs
73	addu.cio r8,r8,r9
74	ld	r7,r4,8
75	st	r8,r2,4
76	ld	r8,r3,12		; add 6 + 8r limbs
77	addu.cio r6,r6,r7
78	ld	r9,r4,12
79	st	r6,r2,8
80	ld	r6,r3,16		; add 5 + 8r limbs
81	addu.cio r8,r8,r9
82	ld	r7,r4,16
83	st	r8,r2,12
84	ld	r8,r3,20		; add 4 + 8r limbs
85	addu.cio r6,r6,r7
86	ld	r9,r4,20
87	st	r6,r2,16
88	ld	r6,r3,24		; add 3 + 8r limbs
89	addu.cio r8,r8,r9
90	ld	r7,r4,24
91	st	r8,r2,20
92	ld	r8,r3,28		; add 2 + 8r limbs
93	addu.cio r6,r6,r7
94	ld	r9,r4,28
95	st	r6,r2,24
96	bcnd.n	ne0,r10,Loop		; add 1 + 8r limbs
97	 addu.cio r8,r8,r9
98
99	st	r8,r2,28		; store most significant limb
100
101	jmp.n	 r1
102	 addu.ci r2,r0,r0		; return carry-out from most sign. limb
103