1dnl  AMD64 mpn_rshift -- mpn right shift, optimised for Atom.
2
3dnl  Contributed to the GNU project by Torbjorn Granlund.
4
5dnl  Copyright 2011, 2012 Free Software Foundation, Inc.
6
7dnl  This file is part of the GNU MP Library.
8dnl
9dnl  The GNU MP Library is free software; you can redistribute it and/or modify
10dnl  it under the terms of either:
11dnl
12dnl    * the GNU Lesser General Public License as published by the Free
13dnl      Software Foundation; either version 3 of the License, or (at your
14dnl      option) any later version.
15dnl
16dnl  or
17dnl
18dnl    * the GNU General Public License as published by the Free Software
19dnl      Foundation; either version 2 of the License, or (at your option) any
20dnl      later version.
21dnl
22dnl  or both in parallel, as here.
23dnl
24dnl  The GNU MP Library is distributed in the hope that it will be useful, but
25dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
26dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
27dnl  for more details.
28dnl
29dnl  You should have received copies of the GNU General Public License and the
30dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
31dnl  see https://www.gnu.org/licenses/.
32
33include(`../config.m4')
34
35C	     cycles/limb
36C AMD K8,K9	 ?
37C AMD K10	 ?
38C Intel P4	 ?
39C Intel core2	 ?
40C Intel NHM	 ?
41C Intel SBR	 ?
42C Intel atom	 4.5
43C VIA nano	 ?
44
45C TODO
46C  * Consider using 4-way unrolling.  We reach 4 c/l, but the code is 2.5 times
47C    larger.
48
49C INPUT PARAMETERS
50define(`rp',	`%rdi')
51define(`up',	`%rsi')
52define(`n',	`%rdx')
53define(`cnt',	`%rcx')
54
55ABI_SUPPORT(DOS64)
56ABI_SUPPORT(STD64)
57
58ASM_START()
59	TEXT
60	ALIGN(16)
61PROLOGUE(mpn_rshift)
62	FUNC_ENTRY(4)
63	shr	R32(n)
64	mov	(up), %rax
65	jnc	L(evn)
66
67	mov	%rax, %r11
68	shr	R8(cnt), %r11
69	neg	R8(cnt)
70	shl	R8(cnt), %rax
71	test	n, n
72	jnz	L(gt1)
73	mov	%r11, (rp)
74	FUNC_EXIT()
75	ret
76
77L(gt1):	mov	8(up), %r8
78	mov	%r8, %r10
79	shl	R8(cnt), %r8
80	jmp	L(lo1)
81
82L(evn):	mov	%rax, %r10
83	neg	R8(cnt)
84	shl	R8(cnt), %rax
85	mov	8(up), %r9
86	mov	%r9, %r11
87	shl	R8(cnt), %r9
88	neg	R8(cnt)
89	dec	n
90	lea	-8(rp), rp
91	lea	8(up), up
92	jz	L(end)
93
94	ALIGN(8)
95L(top):	shr	R8(cnt), %r10
96	or	%r10, %r9
97	shr	R8(cnt), %r11
98	neg	R8(cnt)
99	mov	8(up), %r8
100	mov	%r8, %r10
101	mov	%r9, 8(rp)
102	shl	R8(cnt), %r8
103	lea	16(rp), rp
104L(lo1):	mov	16(up), %r9
105	or	%r11, %r8
106	mov	%r9, %r11
107	shl	R8(cnt), %r9
108	lea	16(up), up
109	neg	R8(cnt)
110	mov	%r8, (rp)
111	dec	n
112	jg	L(top)
113
114L(end):	shr	R8(cnt), %r10
115	or	%r10, %r9
116	shr	R8(cnt), %r11
117	mov	%r9, 8(rp)
118	mov	%r11, 16(rp)
119	FUNC_EXIT()
120	ret
121EPILOGUE()
122