1;! HP-PA-1.1 __mpn_addmul_1 -- Multiply a limb vector with a limb and
2;! add the result to a second limb vector.
3
4;! Copyright (C) 1992-2021 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 2.1 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
20;! <https://www.gnu.org/licenses/>.
21
22
23;! INPUT PARAMETERS
24;! res_ptr	r26
25;! s1_ptr	r25
26;! size		r24
27;! s2_limb	r23
28
29;! This runs at 11 cycles/limb on a PA7000.  With the used instructions, it
30;! can not become faster due to data cache contention after a store.  On the
31;! PA7100 it runs at 10 cycles/limb, and that can not be improved either,
32;! since only the xmpyu does not need the integer pipeline, so the only
33;! dual-issue we will get are addc+xmpyu.  Unrolling could gain a cycle/limb
34;! on the PA7100.
35
36;! There are some ideas described in mul_1.s that applies to this code too.
37
38	.text
39	.export		__mpn_addmul_1
40__mpn_addmul_1:
41	.proc
42	.callinfo	frame=64,no_calls
43	.entry
44
45	ldo		64(%r30),%r30
46	fldws,ma	4(%r25),%fr5
47	stw		%r23,-16(%r30)		;! move s2_limb ...
48	addib,=		-1,%r24,L$just_one_limb
49	 fldws		-16(%r30),%fr4		;! ... into fr4
50	add		%r0,%r0,%r0		;! clear carry
51	xmpyu		%fr4,%fr5,%fr6
52	fldws,ma	4(%r25),%fr7
53	fstds		%fr6,-16(%r30)
54	xmpyu		%fr4,%fr7,%fr8
55	ldw		-12(%r30),%r20		;! least significant limb in product
56	ldw		-16(%r30),%r28
57
58	fstds		%fr8,-16(%r30)
59	addib,=		-1,%r24,L$end
60	 ldw		-12(%r30),%r1
61
62;! Main loop
63L$loop:
64	ldws		0(%r26),%r29
65	fldws,ma	4(%r25),%fr5
66	add		%r29,%r20,%r20
67	stws,ma		%r20,4(%r26)
68	addc		%r28,%r1,%r20
69	xmpyu		%fr4,%fr5,%fr6
70	ldw		-16(%r30),%r28
71	fstds		%fr6,-16(%r30)
72	addc		%r0,%r28,%r28
73	addib,<>	-1,%r24,L$loop
74	 ldw		-12(%r30),%r1
75
76L$end:
77	ldw		0(%r26),%r29
78	add		%r29,%r20,%r20
79	stws,ma		%r20,4(%r26)
80	addc		%r28,%r1,%r20
81	ldw		-16(%r30),%r28
82	ldws		0(%r26),%r29
83	addc		%r0,%r28,%r28
84	add		%r29,%r20,%r20
85	stws,ma		%r20,4(%r26)
86	addc		%r0,%r28,%r28
87	bv		0(%r2)
88	ldo		-64(%r30),%r30
89
90L$just_one_limb:
91	xmpyu		%fr4,%fr5,%fr6
92	ldw		0(%r26),%r29
93	fstds		%fr6,-16(%r30)
94	ldw		-12(%r30),%r1
95	ldw		-16(%r30),%r28
96	add		%r29,%r1,%r20
97	stw		%r20,0(%r26)
98	addc		%r0,%r28,%r28
99	bv		0(%r2)
100	 ldo		-64(%r30),%r30
101
102	.exit
103	.procend
104