1dnl  AMD64 mpn_sublshC_n -- rp[] = up[] - (vp[] << C), optimised for Core 2 and
2dnl  Core iN.
3
4dnl  Contributed to the GNU project by Torbjorn Granlund.
5
6dnl  Copyright 2008, 2010-2012 Free Software Foundation, Inc.
7
8dnl  This file is part of the GNU MP Library.
9dnl
10dnl  The GNU MP Library is free software; you can redistribute it and/or modify
11dnl  it under the terms of either:
12dnl
13dnl    * the GNU Lesser General Public License as published by the Free
14dnl      Software Foundation; either version 3 of the License, or (at your
15dnl      option) any later version.
16dnl
17dnl  or
18dnl
19dnl    * the GNU General Public License as published by the Free Software
20dnl      Foundation; either version 2 of the License, or (at your option) any
21dnl      later version.
22dnl
23dnl  or both in parallel, as here.
24dnl
25dnl  The GNU MP Library is distributed in the hope that it will be useful, but
26dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
27dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
28dnl  for more details.
29dnl
30dnl  You should have received copies of the GNU General Public License and the
31dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
32dnl  see https://www.gnu.org/licenses/.
33
34C	     cycles/limb
35C AMD K8,K9	 4.25
36C AMD K10	 ?
37C Intel P4	 ?
38C Intel core2	 3
39C Intel NHM	 3.1
40C Intel SBR	 2.47
41C Intel atom	 ?
42C VIA nano	 ?
43
44C INPUT PARAMETERS
45define(`rp',`%rdi')
46define(`up',`%rsi')
47define(`vp',`%rdx')
48define(`n', `%rcx')
49
50ASM_START()
51	TEXT
52	ALIGN(8)
53PROLOGUE(func)
54	FUNC_ENTRY(4)
55	push	%rbx
56	push	%r12
57
58	mov	R32(%rcx), R32(%rax)
59	lea	24(up,n,8), up
60	lea	24(vp,n,8), vp
61	lea	24(rp,n,8), rp
62	neg	n
63
64	xor	R32(%r11), R32(%r11)
65
66	mov	-24(vp,n,8), %r8	C do first limb early
67	shrd	$RSH, %r8, %r11
68
69	and	$3, R32(%rax)
70	je	L(b0)
71	cmp	$2, R32(%rax)
72	jc	L(b1)
73	je	L(b2)
74
75L(b3):	mov	-16(vp,n,8), %r9
76	shrd	$RSH, %r9, %r8
77	mov	-8(vp,n,8), %r10
78	shrd	$RSH, %r10, %r9
79	mov	-24(up,n,8), %r12
80	ADDSUB	%r11, %r12
81	mov	%r12, -24(rp,n,8)
82	mov	-16(up,n,8), %r12
83	ADCSBB	%r8, %r12
84	mov	%r12, -16(rp,n,8)
85	mov	-8(up,n,8), %r12
86	ADCSBB	%r9, %r12
87	mov	%r12, -8(rp,n,8)
88	mov	%r10, %r11
89	sbb	R32(%rax), R32(%rax)	C save cy
90	add	$3, n
91	js	L(top)
92	jmp	L(end)
93
94L(b1):	mov	-24(up,n,8), %r12
95	ADDSUB	%r11, %r12
96	mov	%r12, -24(rp,n,8)
97	mov	%r8, %r11
98	sbb	R32(%rax), R32(%rax)	C save cy
99	inc	n
100	js	L(top)
101	jmp	L(end)
102
103L(b2):	mov	-16(vp,n,8), %r9
104	shrd	$RSH, %r9, %r8
105	mov	-24(up,n,8), %r12
106	ADDSUB	%r11, %r12
107	mov	%r12, -24(rp,n,8)
108	mov	-16(up,n,8), %r12
109	ADCSBB	%r8, %r12
110	mov	%r12, -16(rp,n,8)
111	mov	%r9, %r11
112	sbb	R32(%rax), R32(%rax)	C save cy
113	add	$2, n
114	js	L(top)
115	jmp	L(end)
116
117	ALIGN(16)
118L(top):	mov	-24(vp,n,8), %r8
119	shrd	$RSH, %r8, %r11
120L(b0):	mov	-16(vp,n,8), %r9
121	shrd	$RSH, %r9, %r8
122	mov	-8(vp,n,8), %r10
123	shrd	$RSH, %r10, %r9
124	mov	(vp,n,8), %rbx
125	shrd	$RSH, %rbx, %r10
126
127	add	R32(%rax), R32(%rax)	C restore cy
128
129	mov	-24(up,n,8), %r12
130	ADCSBB	%r11, %r12
131	mov	%r12, -24(rp,n,8)
132
133	mov	-16(up,n,8), %r12
134	ADCSBB	%r8, %r12
135	mov	%r12, -16(rp,n,8)
136
137	mov	-8(up,n,8), %r12
138	ADCSBB	%r9, %r12
139	mov	%r12, -8(rp,n,8)
140
141	mov	(up,n,8), %r12
142	ADCSBB	%r10, %r12
143	mov	%r12, (rp,n,8)
144
145	mov	%rbx, %r11
146	sbb	R32(%rax), R32(%rax)	C save cy
147
148	add	$4, n
149	js	L(top)
150
151L(end):	shr	$RSH, %r11
152	pop	%r12
153	pop	%rbx
154	sub	R32(%r11), R32(%rax)
155	neg	R32(%rax)
156	FUNC_EXIT()
157	ret
158EPILOGUE()
159